HHI AT A GLANCE HEINRICH-HERTZ-INSTITUT Areas of Research … · HHI REPORT 2001 5 TABLE OF...

HHI AT A GLANCE

Government research institute (FederalRepublic of Germany and State of Berlin)Total staff at end of 2001: 272 employees

Areas of Research andDevelopment

Photonic Networks• Design, development and demonstration

of optical communication networks and subsystems (access and customer networks, core networks)

• Investigation and development of WDMand high-speed OTDM techniques forhigh capacity transmission and routing

• Exploration of high speed transmissionperformance of photonic networks

• Development of techniques for networkoperation and maintenance

• Development and fabrication of photonicdevices and integrated circuits (lasers,modulators, switches, optical amplifiers,filters, multiplexers and demultiplexers,signal regenerators, transceivers, receiverfrontends) based on InP, for passive components on SiO2/Si and polymers

• Development of new concepts for carrying IP traffic over WDM optical networks and wireless networks in an efficient manner (TransiNet)

• Switchable network design and testing

Mobile Broadband Systems• Development of space-time-receiver for

the uplink of mobile communication systems

• Design of downlink beamforming for CDMA-based mobile communication systems

• Development of calibration algorithms for smart antennas

• Teletraffic engineering for mobile communication systems

• Development of signal processing algorithms for OFDM systems

• Development of sequences for CDMA applications

• Development of optical microwave generation and transmission systems forcellular mobile communication systems

• Implementation of MIMO systems• Development of signal processing

algorithms for MIMO systems

ISSN 1435-1587

H E I N R I C H - H E R T Z - I N S T I T U T

R E P O R T 2 0 0 1

HE

IN

RI

CH

-H

ER

TZ

-I

NS

TI

TU

T•

RE

PO

RT

20

01

Electronic Imaging Technology for Multimedia• Development of algorithms and hardware

architectures for video and audio compression

• Development of algorithms and hardwarearchitectures for image analysis and synthesis

• 3D signal processing for tele-immersion• Image processing for studio applications• Video-streaming over IP and mobile

networks• Design of integrated circuits for image

processing• Development of 3-D display technologies• Man-machine-interaction for future

3D-desktop applications• Virtual reality for telework and

telecommunication• Development of user interfaces for

multimedia applications• Analysis and optimization of

communication services• Image and video retrieval systems

HEINRICH-HERTZ-INSTITUT FÜR NACHRICHTENTECHNIK BERLINGMBH

Einsteinufer 37, 10587 BerlinGermanyPhone: +49 (0)30 310 02-0Fax: +49 (0)30 310 02-213Email: [email protected]://www.hhi.de

Scientific Managing Director:Prof. Dr. Clemens BaackAdministrative Managing Director:Dr. Wolfgang Grunow

indoor-multipath channel

data

input

T1

Tn

R1

Rm

transmitter receiver

data

output

Multiple Input • Multiple OutputM iM o

Radio System for Wireless Local Area Networks


R E P O R T 2 0 0 1

HHI REPORT 2001

Editorial coordination: Regina KiesewetterLanguage consultants:Ingeborg and Harvey HolmesLayout and typesetting:Zühlke Scholz & PartnerWerbeagentur, BerlinPrinting: Druckhaus Mitte, BerlinISSN 1435-1587

Reproduction is allowed only after permission from HHI and with a complete citation

Heinrich-Hertz-Institut für Nachrichtentechnik Berlin GmbH, Einsteinufer 37, 10587 Berlin

Cover illustration: The multiple reflected radio wavesin indoor environments arrive fromall directions like light, which isscattered in a cloud. Informationtheory shows that such radio chan-nels offer an enormous transmissi-on capacity, which is advantage-ously exploited by multiple-imputmultiple output (MIMO) radio systems.

New mathematical techniques forthe design of multiuser MIMO systems were developed at HHI.Prof. Dr Dr Holger Boche receivedthe PhD-award of the “Wissen-schaftsgemeinschaft GottfriedWilhelm Leibniz” for his mathema-tical contribution in this researcharea.

5H H I R E P O R T 2 0 0 1

TABLE OF CONTENTS

Foreword 9

Profile of the Institute 13Mission, Activities, Personnel and Financing 15Corporate Bodies 17Organisation and Contact Persons 18

R & D Activities 19R & D Fields 21Photonic Networks 21Mobile Broadband Systems 29Electronic Imaging Technology for Multimedia 31R & D Projects 37

Selected Contributions 45KomNet – on the Way to the Next Generation Internet 47TransiNet – Innovative Transport Networks for the Broadband Internet 48Fast Optical Switching in Future IP Networks 49Optical Amplifiers in Wavelength- and Burst-Switched Optical Networks 50Control Modulation Technique for Client-Independent Performance Monitoring and Channel Identification in Transparent Optical Networks 51Potential and Implementation of Non-Zero Dispersion Shifted Fibres (NZDSF) in Long Haul Transmission 5210 Gb/s NRZ Transmission over 4000 km Using All-Distributed Raman-Amplified Transmission Links Without Lumped Amplifiers 53Application of 80 GHz All-Optical Clock Recovery in a 160 km Transmission Experiment 54Monolithically Integrated All-Optical Demultiplexer for 160 to 10 Gbit/s 55Optical Sampling of 160 Gb/s Eye Diagrams in an 80 km Transmission Experiment 56160 Gbit/s Optical Clock Recovery Device and its Application in an Optical Demultiplexer 57Automatic Polarization-Mode Dispersion Compensation for 80 Gbit/s Data Transmission over Installed Single Mode Fibre 58Development of InP Modulators Based on a Travelling Wave Electrode Design for 40 Gbit/s and Beyond 59Ultrafast Optical Nonlinearity at Wavelengths around 1.55 µm of GaInAs/AlInAs Quantum Wells Grown at Low Temperatures 60High-Power 1.55 µm Twin-Photodetector Modules for 40 Gbit/s TDM Field Trials 6150 GHz Photoreceiver Modules With InP OEICs for FEC 40 Gbit/s Systems 62Tapered 1550 nm InGaAsP-BH-FP Lasers 63Trade-Off Between High Single-Mode Yield and Low Feedback Sensitivity in DFB RW Laser Arrays 64Optical Filter and Laser Using GaInAsP/InP Ring Resonators 65Fabrication of InGaAsP/InP Ridge Waveguide Lasers with Dry Etched Facets 66Polymer-Based Athermal Arrayed-Waveguide Grating Multiplexer 67Silica-Based Beam-Forming Planar Lightwave Circuit (PLC) 68Peak Value Estimation for Multi-Carrier Signals 69Channel Estimation for OFDM MIMO Systems 70BER for MIMO Systems in Rayleigh and Rician Channels 71Eigenvalue and Inverse Eigenvalue Statistics for Rayleigh and Rician MIMO Channels 72A MIMO WLAN based on Linear Channel Inversion 73On Transmit Diversity with Imperfect Channel State Information 74Pilot Signal Design for Multiuser Channel Estimation Systems 75Sequence Design for Quasi-Synchronous CDMA Systems 76Resource Allocation in Asynchronous CDMA Systems 77Experimental Smart Antenna System for UMTS 78Optimal Multiuser Downlink Beamforming 79Joint Uplink/Downlink Beamforming Optimization 8060 GHz Optically Controlled Maximum Directivity Beamformer 81Silica-Based Beamformer in a 60 GHz Radio-Over-Fibre System 8260 GHz Microwave Broadband Wireless Access Demonstrator for the Next-Generation Mobile Internet 83Graph-Based Representation of Objects in Image Databases 84

7H H I R E P O R T 2 0 0 1

TABLE OF CONTENTS

Sensing People – a New Multimedia Project 85Hierarchical 3D Navigation in Image Databases 86Fast User-Adaptive Image Search Engines 87Group Awareness for Mobile Applications 88Supporting Group Awareness in Teleworking Applications – Human Factors Aspects 89Next Generation 3D Displays for Mixed-Reality Applications 90Tracking Technologies for Multimodal Human-Computer Interfaces 91Human Factors in Multimodal Human-Computer Interaction 92CABAC – Advanced Entropy Coding in the Draft ITU-T H.26L Video Compression Standard 93MPEG Coding Efficiency Tests 94Multimedia Systems based on MPEG-4 for Internet and Mobile Transmission 95MPEG-4 Streaming Solutions using the Delivery Multimedia Integration Framework 96Audio Player on DSP and RISC Processors 97VPS – a DSP-Based PCI Board for Real-Time 3D Video Processing 98E-Cinema – an Approach to Digital Cinema 99Real-Time Solution for Depth Estimation in Immersive Media Applications 100View Combining and Novel View Synthesis for Image-Based Rendering in 3D Video-Conferencing 101Virtual Tourism – Streaming of 360° Panoramas 102High-Resolution Video Mosaicing 103

Communications and Events 105Publications 107Reports 116Patent Applications 117Awards 118Doctorate Theses 118Diploma Theses 119Graduate Theses 120Oral Presentations 120Lectures 123Workshops Organised 124Contributions to Exhibitions 124Committee Activities 125Exchange Program 126Cooperations 127Start up Companies 130

9H H I R E P O R T 2 0 0 1

FOREWORD

The year 2001 was marked by preparationfor the Institute’s integration into theFraunhofer society (FhG). The most impor-tant prerequisites for the transfer of the own-ership of the HHI are now fulfilled, and thistransfer should be complete by the end ofMarch 2002, when the Fraunhofer Societywill become the sole shareholder of the HHI.

The starting point for the integration ofthe HHI into the FhG stems from the strategyof the German government for the domainof information and communication technolo-gy (ICT):

• It is the goal of the Federal Government tobring ICT research in Germany to a lead-ing position in Europe in order to createnew, sustainable employment opportuni-ties.

• The present high levels in basic researchand in the development of novel applica-tions of ICT shall be maintained and fur-ther expanded.

• To this end, government-sponsoredresearch in ICT will be concentrated in theFhG to facilitate the swift transformation ofresearch results into products.

The following reasons were decisive for theexecutive management of the HHI when de-ciding to approve and further this integra-tion:

• The goals of the Federal Government con-cerning ICT research are of great relevancefor the national economy and are a greatchallenge for science and research.

• The FhG will become the leading centre inICT R & D in Europe. As a member of thiscentre, the HHI will have a central role inthe creation of the future of ICT in Europe,whereas otherwise its role would be ofonly minor importance.

• The R & D fields pursued by the HHI are anideal complement to the work of the otherICT institutes in the FhG.

• It must be assumed that future FederalGovernment research funds will be mainlydirected towards the ICT centres in theFhG. The HHI would therefore have agood chance of financial survival inside theFhG.

• Under the policy of reducing the financialdeficit of the State of Berlin the institution-al funding of the institute was cut. HHIhas suffered a reduction in its annual bud-

get by one million ¤ in 1996 and the fol-lowing years until today, despite of thefact that, in the meantime, it has very suc-cessfully passed two evaluations by theScience Council (Wissenschaftsrat) andtwo more by the ElectrotechnicalIndustry's Association (ZVEI). The deficit inthe State budget prevails – there has beenan additional cut of 300 k¤ in the 2001annual budget in excess of the onemillion ¤ mentioned above –, and it isuncertain whether the funding of the insti-tute will recover during the next years.

The ICT centre in the FhG consists of thesoftware oriented “Group of Information andCommunication Technology” (VIK) and thehardware oriented “Microelectronics Group”(VµE). The HHI will be primarily a member ofthe VµE and a secondary member of the VIK.The Federal Government has provided anIntegration Fund of e 69 million over threeyears, which aims to strengthen basic re-search and create synergies between thepartner institutes. The HHI participates inseven projects in the Integration Fund.

In April 2001 Lucent Technologies and theHHI opened the joint laboratory “Tera LabNuremberg”, in which, under the frameworkof a longer term collaboration, photonic highspeed networks for the next generation willbe developed. This is an area where LucentTechnologies and the HHI have already col-laborated intensively in the past.

In July 2001 the HHI reached a collabora-tive research agreement with Fujitsu LabEurope in the field of photonic networks andcomponents, following earlier joint researchprojects.

The division of Photonics was certified ascomplying with the quality standard ISO9001 in the year 2000. This certification wasconfirmed and renewed after the first controlaudit in July 2001.

In August 2001 the whole division ofElectronic Imaging Technology at the HHI,together with its industrial partners, partici-pated in the Internationale FunkausstellungBerlin (IFA), the international exhibition ofconsumer electronics. The presentations werevery successful and excited much interestamong many new companies and politicalinstitutions.

Following ten years of collaboration withAlcatel in several national and European pro-grams, the HHI became a member of theglobal Alcatel Research Partner Project inDecember 2001, a project which started in

10 H H I R E P O R T 2 0 0 1

October 2001. Under this agreement theHHI has started four research projects in thearea of photonic networks and components.

During the past years ten spin-off compa-nies have been created from the HHI. Two ofthose, u2t Innovative Optoelectronic Compo-nents and LKF Advanced Optics havemerged to become u2t Photonics GmbH.The company Vision Pearls Berlin did not sur-vive the difficult year 2001. The remainingeight spin-offs counted altogether 300 em-ployees in January 2002.

Over the last two years the HHI has estab-lished close links with the start-up companiesMergeOptics GmbH, QDI Germany GmbHand u2t Photonics GmbH, the work of whichis closely related to the research and devel-opment activities of the HHI. These compa-nies develop and produce high end compo-nents for photonic networks in close cooper-ation with the HHI, which also providessupport in R & D as well as in prototypingand small-scale wafer and chip supply. In ad-dition, through medium term agreements,the companies are granted access to thewafer technology and the measurementequipment of the HHI, which greatly facili-tates their establishment and growth duringthe difficult first phase. The three companiesare planning to increase their staff to 300 al-together by the year 2003. Thanks to theclose ties between research and industry,Berlin has become the leading German cen-tre of competency in the field of photonicsemiconductor components for the Internetof the future.

In November 2001 the KomNet projectwas presented at a press conference at theHHI. KomNet was started by the Federal Min-istry of Education and Research (BMBF) inMay 1998 as a joint project between Ger-man industry and research institutes, withthe aim of catching up with America in thefield of Internet infrastructure development.The program, which ends in April 2002, hasfunding of e 42,4 million from governmentplus e 34,8 million from industry. The mostsignificant testbed in the world for Internettechnologies has been set up by KomNet.According to the industrial partners, KomNethas made an important contribution towardsthe leading international position which theGerman industry has now reached. The HHIhas contributed a number of projects toKomNet, and also manages the System Inte-gration Office.

The still young Department of BroadbandMobile Communication Networks is now

firmly established in many national andEuropean framework programs. Cooperationin the field of space-time signal processingwith Stanford University, the MassachusettsInstitute of Technology and the University ofIllinois at Urbana-Champaign was further de-veloped, and in the meantime several scien-tists from these institutions have worked atthe HHI during longer research visits.

A breakthrough was achieved in the areaof space-time signal processing. Base-stationantenna arrays have the potential to drasti-cally increase the spectral efficiency of wire-less multi-user systems without increasingthe complexity of the hand-held terminals.The HHI has developed a novel and compu-tationally efficient optimisation scheme thatminimizes the total transmission power whilemaintaining the individual link qualities. Itcan also be used to test the feasibility of agiven scenario, and will thus prove useful forfuture real-time applications in which thesharing of resources must be optimised alongwith the dynamic channel management.

A historic partnership between ISO/IECand the ITU was established at the meetingsin December 2001 in Pattaya (Thailand) ofthe ISO/IEC MPEG (Moving Pictures ExpertsGroup) and ITU-T VCEG (Video CodingExperts Group). A new Joint Video Team(JVT) was formed. This action reunites thepowerful teams from the these organizationswho designed the earlier MPEG-2 video andsystems standards. The formation of the JVTfollows the demonstration, in tests conduct-ed by MPEG just prior to its July meeting, ofsignificant advances in video compressiontechnology by the VCEG H.26L project. TheJVT project will take over the previous H.26Lproject of the ITU-T and will create a singleinteroperable solution for the next genera-tion of standard video coding. The HHI madea significant contribution to this develop-ment by providing both the MPEG-4 refer-ence codec and the H.26L codec. These de-velopments were supervised by ThomasWiegand, head of the Image Communica-tion Group in the Image Processing Depart-ment of the HHI, who was also appointed asone of the two co-chairmen of the JVT.

Holger Boche, head of the Department ofMobile Broadband Communication Net-works, has accepted the chair for MobileCommunications in the Faculty of ElectricalEngineering at the Technische UniversitätBerlin. He will continue in his position at theHHI, thereby greatly benefiting the coopera-tion between the two institutions.

11H H I R E P O R T 2 0 0 1

Thomas Sikora, head of the Department ofInteractive Media – Human Factors, has ac-cepted the offer of the chair of Communica-tion Systems at the Faculty of Electrical En-gineering of the Technische Universität Berlin(successor to Prof. Noll).

Ch. Fehn, P. Kauff, O. Schreer and R.Schäfer, members of the Department ofImage Processing, were honoured in 2001with the International Broadcasting Con-vention (IBC) President’s Award for the besttechnical paper, entitled “Interactive virtualview video for immersive TV applications”.

Over recent years several procedures havebeen developed at the HHI to evaluate theprojects and departments of the Institute.We have now gained enough experiencewith these tools that it seemed appropriateto develop an integrated procedure. All pro-jects in every department were evaluatedwith this internal control system (ICS) at theend of 2001. Based on these results, individ-ual goals will be defined for all departmentsat the beginning of 2002. The goals for thedepartments during the coming years will bedefined in such a way that, after a transitionperiod of five years, the key parameters forthe Institute will match those of the FhG.This has special significance for the requiredamount of external project funding. Appro-

priate strategies for the transition period arecurrently being developed in all departmentsand will be finalised soon.

At the end of the year 2001 the HHI had272 employees, 19 more than the year be-fore. The overall budget was ¤ 29.6 million.The external funding was 52.7 %, an increaseof 8.9 % over the previous year. Externalfunding included 26.9 % from contracts withindustry.

In the year 2001 the HHI was able to sig-nificantly increase its cooperation with indus-try, despite the economic recession. The HHIwas very successful in competing for fundingfrom the Integration Fund of the FederalGovernment, so that collaboration withmany institutes of the FhG could be started.We wish to thank all members of theInstitute for these achievements.

The leading theme in 2001 was, of course,the forthcoming integration of the HHI intothe FhG. This important step, which is neces-sary for safeguarding the future of theInstitute, was discussed at all levels of theInstitute with fairness and objectivity, forwhich we express our respect and gratitude.

Berlin, January 2002Clemens Baack, Wolfgang Grunow

Development of IPPortfolio (Inland)

13H H I R E P O R T 2 0 0 1

Profile of the Institute

15H H I R E P O R T 2 0 0 1 · P R O F I L E O F T H E I N S T I T U T E

Mission, Activities, Personnel andFinancing

Information technology is one of the mostimportant parts of the infrastructure of mod-ern societies. The ever expanding global net-work will give virtually everybody access toimmense information resources, which opensthe door to the era of the knowledge based,or “information”, society.

The expansion of communication net-works has led to a huge market, currentlyworth $ 200 billion annually. Despite a mar-ket downturn of 10 – 20 % in the last year,when industrial capacities had to be readjust-ed, the user demand for broadband servicesis still growing, and the roadmaps for re-search and development in this field havehad only minor revisions.

The paramount task in information tech-nology is now the development of a seam-less global network, which will consist of anoptical Core Network, high capacity fibre op-tic Metro Networks, and Access Networks inwhich the end users are connected via fibre(or, as an intermediate solution, over copperwire), wireless or broadband mobile. Thenetwork must provide the bandwidth need-ed for the subscriber to access all existingand future communication and informationservices with high quality and low latencyand, above all, low cost.

HHI has taken an internationally recog-nised role in several key areas of informationtechnology: photonic networks, mobilebroadband systems, and electronic imagingtechnology for multimedia. Its aim is both toexpand the fundamentals in these areas andto develop new applications in close partner-ship with the industry.

In the research area of Photonic Networks,the ground is being laid for Terabit/s coreand regional backbone networks by ourwork on optical time-division (OTDM) andwavelength-division (WDM) multiplex tech-niques, methods for all-optical signal regen-eration, and ultra high speed fibre transmis-sion experiments. For metro and access net-works, dynamically flexible architectures withoptical switching in conjunction with WDMtechniques are also under investigation. Net-work management and protocol adaptationplay an increasing role in seamless networks.Fibre to the home has again become an ac-tive research area, while customer in-housenetworks with a fibre optic infrastructure andwireless access require extremely cost-effi-cient solutions.

Our research work on photonic devices fo-cuses on the next generation metro and corenetworks. Key components under develop-ment are semiconductor lasers for direct10 GHz modulation in the future Ethernet;lasers for WDM networks; picosecond pulselaser sources, fast modulators and demulti-plexer switches for OTDM networks; laser-based devices for optical 3R regenerators;and broadband receiver front ends for80 Gb/s and beyond.

Optical crossconnects are becoming animportant alternative to electronic switchesfor handling the multi-Terabit/s datathroughput at network nodes. Wavelengthconverters based on semiconductor opticalamplifiers have an important function in thisapplication. Another group of componentsare planar waveguide devices for the passiveoptical network using the SiO2/Si materialsystem and optical polymers. Photonic band-gap structures are an advanced concept foroptical elements and circuits, with great po-tential for realising ultracompact passive andactive devices in future photonic networks.

Mobile communication is at present stilllimited to narrowband applications, in con-trast to communication on the wired net-work. Because mobile networks, as an exten-sion of the fixed network or as (ad-hoc) clus-ters, are continually increasing in bothperformance and penetration, there is agrowing need for broadband mobile connec-tions. The challenge for research is to devel-op suitable system concepts and techniquesfor broadband mobile systems. To do this wemust reconsider the fundamentals of infor-mation and signal theory and show the feasi-bility of the new basic concepts in a demon-strator. This involves considerable effort toimprove coding schemes and algorithms, toestablish new technical principles, such assmart antenna systems for multiple-inputmultiple-output (MIMO) and space-divisionmultiplexing systems, and to drive the devel-opment of the standards. The HHI pursuesthese topics with projects in the area ofMobile Broadband Systems.

Data compression is needed to use thenetwork capacity economically, especially forhigh-rate video services. It is necessary tocontinue research in this area and to developcompression methods with not only im-proved performance, but also with new func-tionalities, especially for interactive applica-tions. Furthermore, efficient network solu-tions using media codecs is of crucialimportance due to their susceptibility to

transmission errors, especially for 3rd and 4th

generation mobile networks.Tele-Immersion is becoming increasingly

important. Therefore new methods for 3Dimage processing, 3D displays and man-ma-chine interaction must be developed. Thegoal is to enable the user to navigate andcommunicate in real and virtual worlds, andto use immersive services and applicationsover networks. The development of next ge-neration information systems, which will en-able the user to efficiently archive, searchand retrieve data, is becoming increasinglyimportant. Intelligent and user-adaptive sys-tems will be needed to make it easy for theuser to identify and access visual information,both natural and synthetic. It is important topursue research on various aspects of usabili-ty engineering to develop applications andservices that are user-oriented and user-friendly. The HHI is working on these tasks inthe area of Electronic Imaging Technologyfor Multimedia.

At the end of the reporting year, the insti-tute had 272 employees, 153 of whom werescientific staff and 119 technical and admin-istrative staff. Of these, 121 positions were inthe research area Photonic Networks, 80 inthe research area Electronic Imaging Techno-logy for Multimedia, 22 in the research areaMobile Broadband Systems, and 50 in cen-tral areas such as management, planning,administration and workshops. At the end ofthe year there were also four postgraduatedoctoral students, eight persons in training,92 student assistants and three guest scien-tists working at the HHI.

At the end of the year 2001 the overallbudget was e 29.6 million. The externalfunding was 52.7 % which included 26.9 %from contracts with industry. Cooperationagreements were concluded with the leadingIT companies. An outstanding event was theopening of the Joint Terabit Lab togetherwith Lucent Technologies Germany inNuremberg, where joint experiments on ul-tra high bitrate transmission are performed.Equally important is the new membership ofHHI in the Alcatel Research PartnershipProgram, under which advanced researchprojects in all-optical signal processing, ultra-high bitrate transmission, semiconductor ma-terials technology, and optical polymer de-vices have been started. A highlight of thejoint national project KomNet (sponsored byBMBF and coordinated by HHI) was on21 November, when the 1000 km link fromStuttgart to the Berlin metro ring network

was switched on. The KomNet pilot systemnow carries data rates of up to 1.2 Terabit/s(32 channels at 40 Gb/s).

HHI has also engaged in strategic partner-ships with innovative SMEs working on pho-tonic components. These companies run col-laborative R & D projects in the areas of highend components for optical telecommunica-tion networks, such as fast lasers, WDMsources, modulators, and detector frontends. They also carry out system develop-ment and system testbed evaluation. HHI al-so serves these partners as a chip and wafersource for prototype series of new devices,thereby creating the opportunity for flexibleand fast introduction to the market. Throughthese initiatives HHI has participated in creat-ing some 100 highly skilled working posi-tions in the Berlin area. Many of these collab-orations with innovative SMEs are embeddedin regional research framework programssupported by the Senate of Berlin.

We have continued our policy of increas-ing public relations and marketing activities,including presentations of our research activ-ities at the technical exhibitions of the majorconferences (OFC, ECOC) and at prominenttrade fairs (CeBIT, Hannover Messe, Interna-tionale Funkausstellung, etc.).

Last but not least, services in the areas oftesting and measurements, feasibility studies,design work, e-beam mask plate manufactur-ing, and others, have contributed to our in-dustrial relations.

HHI has further developed its quality man-agement system to give increased customerorientation, and ISO 9001 was recertified inJuly.

16 H H I R E P O R T 2 0 0 1 · P R O F I L E O F T H E I N S T I T U T E

17H H I R E P O R T 2 0 0 1 · P R O F I L E O F T H E I N S T I T U T E

Corporate Bodies

The corporate bodies of the HHI are theGeneral Meeting, the Supervisory Board, theManaging Directors and the Scientific-Tech-nical Committee.

Members of the Supervisory Board forthis report period are:

The Scientific-Technical Committee iscomprised of heads of departments and anequal number of elected members from theInstitute, and advises the Supervisory Boardand the Managing Directors on all importantscientific and technical matters.Further, the HHI has appointed a ScientificAdvisory Committee of experts from indu-stry, the Deutsche Telekom AG and the aca-demic sector. Members and permanentguests of the Scientific Advisory Committeefor this report period are:

Prof. Dr. G. Litfin, (chair), LINOS AG, GöttingenMinDirig Dr. K. Rupf, (1st vice-chair), Bundesministerium für Bildung und Forschung, BonnORR Dr. B. Colditz, (2nd vice-chair), Senatsverwaltung für Wissenschaft, Forschung und Kultur,Berlin (till Sept. 2001)ORR B. D. Lietzau, (2nd vice-chair), Senatsverwaltung für Wissenschaft, Forschung und Kultur,Berlin (from Sept. 2001)M. Ferstl, HHI, BerlinProf. Dr. P. Noll, Technische Universität BerlinMinR W.-P. Ottenbreit, Deutsche Telekom AG, BonnRD K. Trantow, Senatsverwaltung für Finanzen, BerlinDr. H. Venghaus, HHI, BerlinMinDirig M. Cronenberg (guest), Bundesministerium für Wirtschaft und Technologie, Berlin

Prof. Dr. J. Eberspächer, (chair), Technische Universität MünchenDr. R. Fechner, Lucent Technologies Network Systems GmbH, NürnbergProf. Dr. G. Fettweis, Technische Universität DresdenDr. H.-J. Grallert, Marconi Communications ONDATA, BacknangDr. R. Gossink, Philips GmbH, AachenProf. Dr. W. Klimek, Deutsches Zentrum für Luft- und Raumfahrt e.V., KölnProf. Dr. P. Leuthold, Eidgenössische Technische Hochschule ZürichDr. M. Rocks, T-Nova Deutsche Telekom, BerlinDr. G. Ruopp, Marconi Communications GmbH, BacknangDr. B. Schwaderer, Tesat-Spacecom GmbH & Co. KG, BacknangProf. Dr. G. Siegle, Robert-Bosch-GmbH, BonnDr. E. Zielinski, Alcatel SEL AG, Stuttgart

18 H H I R E P O R T 2 0 0 1 · P R O F I L E O F T H E I N S T I T U T E

Directors

Prof. Dr. Clemens Baack +49-30-31 002-200 [email protected]. Wolfgang Grunow +49-30-31 002-300 [email protected]

Departments

Optical Networks Dipl.-Ing. +49-30-31 002-455 [email protected] Walf

Optical Signal Processing Dr. Hans-Peter Nolting +49-30-31 002-427 [email protected]

Broadband Mobile Prof. Dr. Dr. +49-30-31 002-540 [email protected] Networks Holger Boche

Materials Technology Dr. Norbert Grote +49-30-31 002-431 [email protected]

Micro Fabrication Dr. Udo Niggebrügge +49-30-31 002-550 niggebrueggeTechnology @hhi.de

Integration Technology Dr. Herbert Venghaus +49-30-31 002-555 [email protected]

Image Processing Dr. Ralf Schäfer +49-30-31 002-560 [email protected]

Interactive Media – Prof. Dr. Thomas Sikora +49-30-31 002-210 [email protected] Factors

Central Services

Administration Dipl.-Ing. +49-30-31 002-310 [email protected] Mrowka

Planning + Marketing Dr. Walter Döldissen +49-30-31 002-253 [email protected] Coordination

Organisation and Contact Persons

19H H I R E P O R T 2 0 0 1

R & D Activities

21H H I R E P O R T 2 0 0 1 · R & D A C T I V I T I E S · P H O T O N I C N E T W O R K S

R & D FIELDS

Photonic Networks

Topics and Results

Data transmission traffic is now growing atrates of 100 to 200 % per year. Telephonetraffic, which is expected to grow at a rate ofonly 5 % per year, will soon be only a minorpart of the total traffic. Worldwide telecom-munication is therefore experiencing an ex-plosive increase in the demand for transmis-sion capacity. The main driving force behindthis development is the Internet, which willevolve over the next few years into a Broad-band Internet with an even higher demandfor bandwidth. The Internet will be the back-bone of the future information society andthe main driving force of commerce andbusiness.

Apart from microelectronics, the founda-tion for this development is optical fibrecommunication, which utilises photonic net-works. The enormous capacity of an opticalfibre enables the design and construction ofhigh-capacity communication networks. Mul-tiplexing techniques such as optical wave-length-division multiplexing (WDM) and op-tical time-division multiplexing (OTDM) willbe further developed to exploit the largebandwidths of optical fibres. Moreover, in aphotonic network the data are not onlytransmitted over the optical pathway, butthey are also routed optically. Thus, the pho-tonic network will be service-independent,offering high capacity pathways for all cur-rent and future services.

These developments will also increasinglyaffect the customer networks, i.e. the com-munication networks located in the premisesof private or commercial users. As the infor-mation must be transmitted to the user's ter-minals, the customer network must linkthese multimedia terminals with the broad-band Internet. The prognoses for the year2010 are data rates of 150 Mbit/s for privateand 1 Gbit/s for professional end terminalusers.

Technologies for the future BroadbandInternet are being developed and tested in17 R & D projects under the framework of theGerman Federal Ministry of Education andResearch (BMBF) funding initiative KomNet –Innovative Communications Networks forthe Knowledge-Based Society. As well as the-oretical work and experiments at the variouslaboratory sites, an experimental DWDM

communication network connecting thecities of Berlin, Darmstadt, Mannheim andStuttgart has been implemented. Initial seg-ments of this testbed have already been setup by the individual project teams and havebeen put into operation. This network servesas the national-scale infrastructure to assessand to field-trial optical networking tech-nologies, network elements, protocols andsubscriber line approaches for the NextGeneration Internet.

The HHI contributes to the KomNet pro-jects through several subcontracts with theGerman telecommunication industry, includ-ing work on the design and simulation of op-tical networks, a loop testbed for WDM net-work elements, ultra-high-speed transmissionusing time-division multiplexing, 40 Gbit/sphotodetectors, 40 Gbit/s receivers, andtransceiver photonic integrated circuits. TheHHI Systems Integration Office is also re-sponsible for coordinating the network con-struction and the field trials.

The R & D Program KomNet will be fin-ished in April 2002. The BMBF has now an-nounced a new Program called MultiTera-Net, in which the HHI is also involved withmany proposals.

The HHI is a partner in the project “OpticalTechnologies in Motion for the IST Program”(OPTIMIST). This project aims to facilitateconcerted action in the general area of pho-tonics across the whole of the European R&Dprogram IST (Information Society Technolo-gies). In particular it will encourage impor-tant synergies and interchanges, for examplebetween systems projects and componentsprojects, and improve awareness between in-dustry and academia of new ideas and theresults of longer-term research in the area ofphotonic technologies. These objectives willbe achieved by organising topical workshopsand meetings and a range of other dissemi-nation activities.

The main objective of photonic research atthe HHI is to make substantial contributionsto the development of photonic networks.The following main areas are being ad-dressed:• Development of network concepts for

photonic core, access and customer net-works

• Investigation of the potential of opticalmultiplexing techniques such as OFDMand OTDM. The optimum combination ofthese techniques is of paramount impor-tance.

• Determination of the limits of optical trans-

parency, i.e. the estimation of the trans-parency path lengths in photonic networks

• Determination of the requirements of thenetwork concepts and photonic compo-nents, including the transmission lines androuting subsystems

• Investigations into the supervision (opera-tion, administration and maintenance –OAM) and control (telecommunicationmanagement network – TMN) of photonicnetworks

• Investigation of functional linking methodsbetween the physical and protocol layers

• Investigation of protocols for IP over WDMnetworks

• Development of methods and devices forall-optical signal processing for ultra highspeed packet routing, signal monitoringand control

• Development and fabrication of photoniccomponents and subsystems

The overall research area of photonic net-works is divided into the two specialist areasCore Network and Access and CustomerNetwork.

Core Network

Work in this research area concentrates ondense wavelength-division multiplexing(DWDM) and high bitrate time-division mul-tiplexing (TDM) techniques, optical cross-connects, and optically transparent net-works. Multiplexing and demultiplexing maybe performed either electrically (ETDM) oroptically (OTDM) in high bitrate TDM. Thegeneration, modulation, detection and syn-chronization of the optical signals are impor-tant issues being studied. The effects of thephotonic components and of fibre nonlinear-ities and dispersion on the quality of the op-tical signals are also under investigation.

Optical time division multiplexing(OTDM): The optical time division multiplex-ing (OTDM) work includes the investigationand realization of optical subsystems, and al-so transmission experiments at bitrates ashigh as 160 Gbit/s. An OTDM transmissionsystem comprises several components andsubsystems, as follows. On the transmitterside an optical pulse source generates a pulsetrain with a repetition frequency at the baserate, which in our case is 40 GHz, the highestelectronic data rate available at present. Thisoptical pulse train is coupled into four optical

branches, in which the 40 Gbit/s electricalsignals are modulated onto the optical pulsestreams. These four optical signals are bit-in-terleaved by a delay multiplexer to generatea 160 Gbit/s bit-stream, which is transmittedvia mono-mode fibres from various suppliers.At the receiver side, an optical demultiplexerseparates the four base signals for subse-quent electronic detection and processing.The demultiplexer must be very fast for bit-wise switching. The RZ data source was real-ized by a mode-locked laser. This data sourcehas been demonstrated in 160 Gbit/s trans-mission experiments over unrepeatered (nooptical amplifiers in the transmission line)transmission spans of up to 160 km of stan-dard fibre. To compensate for fibre disper-sion, we applied passive dispersion compen-sation using dispersion-compensating fibre.

We reported the first 160 Gbit/s OTDMsystem, which used optical demultiplexing tothe base data rate of 40 Gbit/s, electrical sig-nal processing at 40 Gbit/s, and transmissionover standard single-mode fibre (SMF). Thisis the first demonstration of the next genera-tion OTDM systems, which will use opticaldemultiplexing to the highest electrical pro-cessing speed available.

In-service quality monitoring of high bi-trate optical transmission systems working at160 Gbit/s has been achieved at the full linerate by using optical sampling systems.These systems provide the picosecond tim-ing resolution needed to observe eye dia-grams at this speed. Key devices are theclock recovery device and the fast switch forsampling. The clock recovery circuit containsan optical phase locked loop (PLL) in whichthe ultra-fast optical mixer is realised by asemiconductor laser amplifier in a loop mir-ror (SLALOM). The sampling gate is a non-linear optical loop mirror (NOLM) containing320 m of highly nonlinear fibre (HNLF, (λ0 =1547 nm).

Monolihically integrated picosecondpulse source: One of HHI´s central activitiesis the development of components and sub-systems for OTDM/WDM systems for Tbit/scapacity networks. One of the key compo-nents is an optical picosecond pulse source.After several years of experience in setting uphybrid modules, HHI has started the devel-opment of a monolithically integrated deviceat a repetition rate of 40 GHz. Major chal-lenges are to guarantee sufficiently low tim-ing jitter and sufficiently high stability for usein 160 GHz OTDM systems. The opto-elec-

22 H H I R E P O R T 2 0 0 1 · R & D A C T I V I T I E S · P H O T O N I C N E T W O R K S


tronic integrated circuit (OEIC) architecturewill include a biased saturable absorber and again-tunable and phase-tunable distributedBragg reflector (DBR) within a cavity lengthof about 1.1 mm. The theoretical investiga-tions and experimental analysis of the sub-el-ements have now lead to an OEIC fabricationprocess.

Monolithically integrated nonlinear in-terferometer switches: We have fabricatedand investigated nonlinear Mach-Zehnder in-terferometers for use as demultiplexers. Theseuse optical semiconductor amplifiers (SOAs)as nonlinear elements in the interferometerarms. With these all-optical switches, error-free demultiplexing from 160 to 10 Gbit/swith arbitrary choice of the switched channelwas achieved in system experiments.

Further investigations were focussed on adifferent switching architecture, in which theworking wavelength of the integrated SOA is1500 nm while the data wavelength is1550 nm. This is called the “band gap shift-ed” (GS) switching scheme. Its advantagesare better linearity, enhanced extinction ratioand greater phase shift efficiency.

In cooperation with an industrial partner,we are developing a specially designed all-optical wavelength converter for 10 and40 Gbit/s signals.

All-optical signal regeneration: Scalablenetworks need full “3R” (re-amplification, re-shaping, re-timing) signal regeneration.Components and methods for this functionare under development. Optical clock recov-ery is performed using self-pulsating DFBfeedback lasers. These three-section devicesare controlled by three dc currents whichswitch the self-pulsation on and off and tuneits frequency. Very large (two octave) tuningranges and frequencies of more than 80 GHzhave now been achieved using the newPhaseCom architecture based on a tandemDFB laser structure. The self-pulsation fre-quency synchronizes to the optically injecteddata signals, and thus realizes all-opticalclock recovery.

In cooperation with Alcatel, a full 3R re-generator has been assembled using the self-pulsating laser for 40 GHz clock recovery anda synchronous modulation/soliton compres-sion scheme for regeneration. This regenera-tor was tested at 40 Gb/s in loop experi-ments, with the loop comprising 160 km ofdispersion-managed fibre and several EDFAs.The eye diagrams and the bit error rate (BER)

were measured after 65 laps, correspondingto transmission over 10,000 km. The eyes wereclearly open, and the BER was below 10-9.Error-free transmission of 40 Gb/s signals over10,000 km demonstrates that the all-optical3R regenerator is very effective at suppress-ing the accumulation of noise and jitter.

To improve the yield of chirp-free, low jit-ter and high extinction ratio self-pulsatinglasers, the influence of internal reflections hasbeen investigated. Automatic characteriza-tion of multi-section lasers has also been in-troduced.

High-speed modulators: 40 Gbit/s opticalmodulators are key components for next-generation optical communication systems.LiNbO3 modulators for this task are commer-cially available, but require relatively large RFdriving voltages (> 6 V), and as a conse-quence very expensive driving electronics.

GaAs devices are more favourable in thisrespect, but even better performance andlower driving voltages are expected fromInP-based modulators. The Mach-Zehndermodulator structure we developed is basedon a travelling wave electrode design to en-sure proper high speed characteristics. It ismuch smaller (< 3 mm2) and needs less dri-ving voltage (VPP ~ 2V) than comparable de-vices made with GaAs or lithium niobate.This inherently low driving voltage is one ofthe key issues for future transmission rates of80 Gbit/s and more.

High speed photodetectors and receivers:Fast photodetectors for the reception of highbitrate data streams at 1.55 µm wavelengthhave been developed under the frameworkof the KomNet project. These photodetec-tors were required to convert large opticalpowers with effective noise suppression. Toachieve this, two waveguide-fed PIN photo-diodes in a differential configuration weremonolithically integrated, together with aspot-size converter at the optical input port,an optical 3 dB splitter, and biasing circuitry.Packaged as modules, the twin photodiodeshad highly symmetrical output behaviourand a 3 dB bandwidth of 45 GHz. In systemexperiments and field trials, performed by anindustrial partner, a differential polarization-dependent loss of only 0.34 dB was ob-tained, which is a 3 dB improvement of theoptical signal-to-noise ratio compared to sin-gle-ended detector schemes. Data transmis-sion at 43 Gbit/s with forward error correc-tion (FEC) was successful.

Optical receiver front ends, monolithicallyintegrated on InP, for high speed TDM trans-mission systems are also being developed asa part of the KomNet project. In addition toa waveguide-fed PIN photodiode with aspot-size converter, they also contain amonolithic microwave integrated circuit(MMIC) -like electrical preamplifier usinghigh electron mobility transistors (HEMTs).Further improvement of device design, ma-terial properties and processing technologyincreased the 3 dB bandwidth to 50 GHz, asmeasured in completely assembled modules,at the same time maintaining the conversiongain (larger than 1 A/W) and the flat transfercharacteristic. With a newly developed am-plifier stage of cascode type using eightHEMTs, a conversion gain of 1.7 A/W wasachieved. The receiver modules proved ableto transmit a TDM signal with FEC at43 Gbit/s during in-system experiments car-ried out by an industrial partner.

Ultra-fast switching devices: Followingprognoses that an OTDM transmission bi-trate of 1 Tbit/s may be achieved around2010, the search for photonic materials withless than 1 ps switching capability is a keytask. Hence the investigations into InGaAs/InAlAs structures grown at low-temperaturewere continued. An optical pulse train re-sponse with a characteristic time constant ofthe order of 250 fs could be demonstrated.As an alternative approach, work has beencommenced to exploit ultrafast intersubbandtransitions in Sb-based quantum well struc-tures.

WDM system testbed for novel networkapplications: The HHI operates a powerfulmeasurement system for the investigation ofvarious aspects of future DWDM networks.WDM systems with up to 16 wavelengthchannels, each up to 10 Gb/s, can be tested.This is being extended to 40 Gbit/s channelrates.

Long distances are simulated using multi-ple round trips in a WDM loop testbed.Multiple round trips correspond well withthe ITU-T proposals to build networks withidentical sections. The optical transmissionsections (OTSs) include transmission fibres,optical amplifiers, and means for chromaticdispersion compensation. The transmissionfibre for the 10 Gb/s channels is 80 km ofstandard SMF, and is compensated for dis-persion using dispersion compensating fibers(DCFs).

The reason for testing long distances is tofind the optimum configuration and parame-ters for dispersion management of so-callednormalized sections when a large number ofOTSs is cascaded. Such normalized sectionsare being investigated under a contract witha network operator. Different types of non-zero dispersion shifted fiber (NZ-DSF) sec-tions are being tested. Optimization with re-spect to dispersion compensation must allowfor the complex interplay between fibre non-linearities and local signal dispersion. Variousconfigurations, with pre-compensation, post-compensation or symmetrical compensation,have been tested with both NRZ and RZ sig-nals and with ASK and DPSK modulation for-mats. The outcome was a series of chartsshowing the maximum transmission dis-tances and the corresponding tolerances forthe compensation ratio (CR) and the channelpower levels. A maximum distance of 4000km over standard fibres was achieved withNRZ signals, symmetrical compensation ofthe chromatic dispersion, and a CR of 99 %.

In consideration of future switched opticalWDM networks with varying numbers ofchannels, the crosstalk penalty due to crossphase modulation (XPM) was measured in a16 channel 10 Gb/s ASK NRZ long distancetransmission system with 100 GHz channelspacing. Up to 3200 km the maximum XPMcrosstalk penalty was below 2.5 dB, com-pared with single channel transmission. Thisresult emphasizes the benefits of standardSMF systems for future flexible and switchedWDM networks.

All tests and measurements performed inthe WDM testbed are accompanied by nu-merical simulations. The agreement betweenmeasurements and simulations gives confi-dence in the correctness of both the mea-surement procedure and the numerical mod-els.

Distributed Raman amplification: Distrib-uted Raman amplification uses the transmis-sion fibre as a gain medium and promisesoptical links with virtually zero attenuationover large wavelength ranges. Optimumpump configurations are being investigatedfor possible replacement of the standard er-bium-doped fibre amplifiers in ring networks.A maximum distance of 4000 km wasachieved with pure Raman amplification. Anovel emergency switching system will beimplemented to protect eyes from the highpump power levels if the fibre breaks.



Channel identification and performancemonitoring: A problem not solved so far ischannel identification and performance con-trol in a transparent WDM network. This ismost important in switched WDM networks.Novel control modulation techniques are un-der test that allow simultaneous identifica-tion and performance control. Two conceptswith a combination of a 50 Mb/s controlchannel and a 10 Gb/s payload channel werefound to have good quality correlation be-tween the two channels, as required for per-formance control, and to give error-freetransmission of both channels up to dis-tances of 2500 km.

Design and analysis of WDM networksfor IP: Optical networking is based on twocomplementary views, which are reflected inrecent ITU-T recommendations and the cur-rent activities of the study groups. The func-tional view is mostly related to the mana-gement aspects, and the physical view to thetransmission issues of the different topolo-gies. The problem of designing transparentnetwork domains is due to the analoguecharacter of the channels. This complicatesnetwork planning because of the limited cas-cadability of the network elements and thedifficulties of controlling the channels, espe-cially in switched networks.

Ring networks are of great practical impor-tance because of simple protection switchingschemes. Various topologies for unidirection-al and bidirectional rings with two and fourfibres have been analysed for normal andprotection-mode operation. The simulationresults of a two-fibre ring with four add-dropnodes fully meshed through 40 Gb/s WDMchannels using standard SMF indicates itssuitability for metro rings up to 60 kmperimeter without in-line amplifiers. A goodapproach to large meshed networks is seento be a multiple connection of ring topolo-gies using optical crossconnects.

Upgrading of such basic network topolo-gies is planned towards switchable WDMnetworks and optical packet-switched net-works at the IP level. This requires the appli-cation of wavelength-switchable transmit-ters, the implementation of optical crosscon-nects and the development of burst-modereceivers with fast clock recovery. Novel la-belling techniques and routing measures willbe simulated and tested in the WDM test-bed.

Transinet – Innovative Transport Net-works for the Broadband Internet: The ob-jective of this joint project, which was begunin June 2000, is to develop new networkingtechnologies to carry IP traffic over WDM op-tical networks in an efficient manner.

The main research areas for the next-gen-eration optical Internet in the TransiNet pro-ject are new concepts for network and nodearchitectures, the corresponding protocols,standards and technologies, traffic aggrega-tion, scalability of WDM networks, Quality ofService (QoS), network management, re-silience, and the interworking and integra-tion of high-speed photonic and mobilecommunication systems.

At the HHI advanced optical subsystemssuch as WDM transmission, add-drop multi-plexers, optical crossconnects and wave-length router/switches, together with theirinterworking with higher-layer protocols,mainly IP, are being investigated.

Parts of an IP-optimized air interface willbe developed. The main research areas forwireless communications are new signal pro-cessing algorithms, estimation of the radiochannel, resource allocation, and teletraffictechniques for the planning process.

Network and subsystem modelling: Meth-ods for simulating optical components andsignal propagation in an optical path are be-ing developed, and are being checked bycomparison with the experimental measure-ments. As well as the programs we have de-veloped ourselves, we also use the commer-cially available program package “VPITrans-missionMaker & VPIComponentMaker” fromVirtual Photonics Inc. (a spin-off from HHI).With the transmission-line laser model it ispossible to optimize photonic componentsand subsystems, such as the self pulsatinglaser and the nonlinear Mach-ZehnderInterferometer (used as a very fast opticalgate), which can be connected to an all-opti-cal circuit for 3R regeneration. The work to-wards high bitrate TDM and the combina-tion of TDM and WDM technologies is alsosupported by analytical and numerical inves-tigations, the main question being how toset up the transmission line to achieve opti-mum transmission capacity.

Access and Customer Network

Research and development activities in thisarea are concentrated on the use of WDM

techniques in the access network. Specifi-cally, the goal is to develop and investigatesystem concepts and components for WDMapplications in the access network.

Monolithically integrated laser compo-nent for optical transceivers: The bidirec-tional optical transceiver is a key componentfor the realisation of optical subscriber links.As the crucial crosstalk requirement of atleast 50 dB optical crosstalk suppression,which was set by the Full Services AccessNetwork (FSAN) group, is not met by smallmonolithically integrated transceivers, we fo-cussed on the development of Fabry-Perotlaser OEICs with integrated laser resonatorsand integrated 45° mirrors for the subassem-bly setup. The lasers include field transform-ers to reduce the numerical aperture. Thefull-wafer fabrication of the laser cavities bydry etching and of the 45° mirror by wetetching allows on-wafer testability of theOEICs and has the potential to reduce themodule fabrication cost. The verification ofthe transceiver modules will be performed bythe industrial partner who funded this devel-opment.

Laser devices: Multi-wavelength singlemode sources are of great interest for futureWDM networks. One prerequisite for obtain-ing high yield for laser arrays is to have ahigh single-mode yield for the individual de-vices. This topic has received particular atten-tion during the last year. Different theoreticalmodels suggested either to raise or to reducethe κL product of our DFB lasers. Finally anappropriate range of target values for κL, to-gether with other design and technologicalparameters, was determined, and this en-abled the fabrication of lasers and laser arrayswith substantially enhanced single-mode yields.

Another activity is aimed at hybrid DBRlasers, which are comprised of an InP gainsection and a fibre Bragg grating. Realizingthe Bragg grating in a fibre (instead of InP)reduces the temperature-dependent varia-tion of the emission wavelength of the hy-brid laser by an order of magnitude. Lowthreshold current and high optical outputhybrid lasers require efficient coupling be-tween the InP chip and the fibre, and this inturn requires an InP chip with an integratedspot size transformer. Such devices havebeen developed and prototypes have beentransferred to a company which will soonproduce chips in larger volumes and com-mercialize the hybrid laser.

With the expansion of optical networks in-to the metropolitan and access areas, the de-mand for inexpensive laser sources with highperformance will steadily increase. A modula-tion speed of 10 Gbit/s, uncooled operationup to 85° C, and a high optical laser-fibrecoupling efficiency are of crucial importance.Various research activities were carried out inpursuit of these goals. Tapered laser diodeswith a burried hetero (BH)-type structureand 10 Gbit/s modulation capability havebeen developed. To improve the high tem-perature performance, gain material basedon InGaAlAs has been optimized and testedin ridge waveguide- (RW) -type lasers. Dryetching was also successfully applied to formthe facets. Vertical cavity lasers are anotherapproach to achieving low cost light sources.A design is being investigated for the 1.3 µmwavelength that uses oxidized Bragg mirrorsand a GaInAsN active layer. Work on this de-manding structure has concentrated on itstechnological problems.

Micro-ring filters: Miniaturisation of wave-guide circuits is inherently connected withthe implementation of highly confined wave-guides, which opens the potential for realis-ing optical micro-ring circuits. Integrated op-tical ring resonators are promising candi-dates for compact optical filters andwavelength multiplexers and demultiplexers.Their realisation in active semiconductor ma-terial could lead to “lossless” filter devices aswell as novel laser components with out-standing performance. As ring resonators donot require facets or gratings for opticalfeedback, they are particularly suited formonolithic integration. We followed ourstrategy of learning from passive devices,then improving their performance by inte-grating semiconductor optical amplifiers(SOAs), and finally extending the functionali-ty in all-active structures. Passive single andmultiple ring filters as well as add/drop mul-tiplexers could be demonstrated, includingfine tuning of the filter response by integrat-ed platinum heaters. First results on fabricat-ed SOA integrated ring resonators and all-ac-tive resonators in the transmission and emis-sion mode have been obtained.

WDM filter components: As well as ar-rayed waveguide gratings (AWG) and specialcomponents associated with them, addition-al waveguide components based on estab-lished silica technology have been devel-oped, especially OTDM multiplexers and



beam forming networks for broadband mo-bile communications.

The beam forming network uses the inter-ference between two neighbouring opticalfrequencies to generate an intermediate fre-quency in the 60 GHz range. These signalsare split into 4 or 8 outputs, which are indi-vidually phase shifted to change the beamcharacteristic of a microwave antenna. Twodifferent silica-based planar lightwave circuit(PLC) devices designed for optical phasecontrol and amplitude control were imple-mented and assessed.

The AWG activities include a special AWGfor passive optical networks. The idea is tooverlay a broadcast wavelength onto WDMchannels. Following previous work on such acomponent, this concept was successfully ex-tended to bidirectional WDM transmission.Prototype devices with excellent perfor-mance were demonstrated.

Another important AWG activity was basedon the use of polymer materials to form notonly the optical waveguide but also the sub-strate. By properly adjusting the thermal ex-pansion coefficient of the substrate, consist-ing of polymer materials developed by FhG-IZM, a novel scheme to achieve athermalbehaviour, i.e. channel wavelength indepen-dent of temperature, was successfully devel-oped.

Photonic crystals: Photonic crystals are ex-citing new materials in photonics, still in thestage of fundamental research. HHI has re-cently entered this area using silicon-on insu-lator (SOI) as the most promising material.So far, basic design work on waveguides andetching studies have been carried out.

Competencies

Investigation and development of architec-tures for photonic networks, development ofplanning guidelines, and studies of specificproblems of photonic networks

Characterization and testing of optical net-works and network components, includingexperiments on fibre loops, transmission ex-periments over large distances, and field tri-als

Development of high rate optical WDM sys-tems (10 Gbit/s per channel) and the corre-sponding measurement methods

Investigation, design and development ofoptical WDM LANs/MANs, optical accessnetworks and passive optical networks

Development of methods for the control andsupervision of networks

Development, application and verification ofsimulation software for optical transport net-works

Design and development of optical systemsusing heterodyning

Development of wavelength conversionmethods for WDM systems

Investigation and development of high rateOTDM subsystems (40 to 160 Gbit/s andabove), including multiplex/demultiplex andadd/drop techniques

Development of methods for purely opticalclock recovery and signal regeneration (2Rand 3R) for bitrates of 10 to 160 Gbit/s

Development of methods for the control,maintenance and supervision of high rateOTDM subsystems

Development of methods for the generationof ultra-short optical pulses

Development of optical sampling methodsfor network monitoring

Optimization of methods of dispersion man-agement

Design and development of optical frequen-cy reference equipment

Design and fabrication of transponders andoptical synchronous digital hierarchie (SDH)front ends

Development, fabrication and characteriza-tion of fibre optic components

Development, fabrication and characteriza-tion of methods and devices for polarizationmode dispersion (PMD) compensation

Development, fabrication and characteriza-tion of opto-electrical components and pho-tonic integrated circuits based on InP:- Multi-wavelength laser arrays

- Optical amplifiers with spot-size transformers

- High-speed laser diodes- Wavelength converters- Demultiplexers and add/drop multiplexers

for OTDM applications- Components for optical clock recovery and

signal regeneration (self-pulsating lasers)- Integrated transceivers- Fast waveguide-fed, high-power photodi-

odes with optical spot-size converters- Differential mode photodetectors (twin

photodiodes) with integrated spot-size converters

- High bitrate optical receiver front ends with integrated spot-size converters

- High-speed monolithic integrated pulse sources

- Optic/microwave converters

Development and fabrication of polymer-based devices (thermo-optic switches andswitch matrices, AWGs)

Development and fabrication of planar wave-guide components on SiO2/Si

Development and fabrication of diffractiveoptical components

Modelling of photonic components, photon-ic subsystems and integrated circuits

Development, optimization and applicationof component technologies:- Clean room laboratories (class 10/1000)- Epitaxy (MOVPE, MBE) and characteriza-

tion of InP-based semiconductor materials- Secondary ion mass spectroscopy (SIMS)- CAD/CAM of photolithography masks- Electron beam lithography and optical

lithography- High resolution scanning electron

microscopy- Atomic force microscopy- Dry etch processes (reactive ion etching,

ion beam etching) with endpoint detection

- Rapid thermal processing- Deposition of metal and dielectric layers

(evaporation, sputtering, plasma deposition)- Optical coatings- Characterization and technology of optical

polymer materials- Chip mounting and connection techniques- Fibre-chip coupling (single and multiple

input/output ports)- High-frequency packaging


29H H I R E P O R T 2 0 0 1 · R & D A C T I V I T I E S · M O B I L E B R O A D B A N D …

Mobile Broadband Systems

Topics and Results

Mobile communication is rapidly growingand is now an important economic factor. Itwill grow even more and even faster with theimminent wireless internet access and theimplementation of wireless LANs (WLANs).The HHI activities in mobile communicationconcentrate on improvements for systems ofthe third generation, such as UMTS/IMT2000,and on the design of WLAN and fourth gen-eration systems.

Due to the extreme costs of spectral band-width, which became obvious at the UMTSlicence auctions in Great Britain and Ger-many, improvement of the spectral efficiencyof these systems is essential to increase theoverall system capacity and to give an ac-ceptable return on investment within reason-able time spans. Therefore smart antennasystems and multi-user detection principlesto improve the spectral efficiency of UMTSare being investigated at the Heinrich-Hertz-Institut. Strategies are also being developedthat allow for the strongly asymmetric trafficof future services such as IP over wireless.

New approaches are being investigated forefficient resource allocation in future mobilesystems, including consideration of the char-acteristics of new services. In particular, fun-damental principles for an IP-optimized mo-bile communication system are being investi-gated. Also, new multi-user receivers arebeing conceived and signal processing algo-rithms are being optimized. In addition, workis proceeding on the integration of mobileand photonic networks. New concepts forfourth generation systems are being devel-oped for multiple access control, interferencereduction and network planning. Thesemethods are based on the application of op-timization techniques to code division multi-ple access (CDMA) systems, orthogonal fre-quency division multiplex (OFDM) systems,and non-orthogonal multiple carrier systems.

Future cellular systems must be able tosupport high-rate multimedia services evenwith high user densities. This challenginggoal can be attained by using antenna arraysat the base station. One way to make effi-cient use of the spatial diversity provided bythe array is to steer the receive and transmitbeams towards the desired users, with theradiation pattern periodically updated de-pending on the instantaneous channel char-acteristics. In this way considerable carrier-to-

interference gains can be achieved, whichmay be directly traded for improved qualityof service, greater user density or greater da-ta rate. The HHI carries out theoretical andexperimental research in the field of beam-forming and applies this knowledge to thedesign of new algorithms for both up-linkand down-link space-time processing.

Compared with the conventional matchedfilter receiver, a multi-user receiver that uti-lizes the structure of the multi-access interfer-ence can substantially increase the spectralefficiency of CDMA using direct-sequencespread-spectrum methods. An important ob-servation is that both the performance andcomplexity of such receivers strongly dependon the choice of the spreading sequences.Moreover, if signals are transmitted via atime-variant multipath channel, careful de-sign of the training sequences is required toensure a small channel estimation error.Hence we have been concerned with the de-sign and allocation of spreading and trainingsequences so as to achieve the best perfor-mance of CDMA-based mobile communica-tion systems.

The antenna arrays and the related radiofrequency devices must be calibrated for cer-tain smart antenna applications. In particu-lar, highly accurate antenna calibration is re-quired for high resolution direction-of-arrivalestimation algorithms, for null-steering beam-forming and for optimal down-link beam-forming at the cellular base stations. At thesame time, in commercial mobile communi-cation systems the overhead in terms of ra-dio resources and system components that isrequired for calibration must be as low aspossible. Therefore we have developed opti-mized robust calibration schemes. A detailedanalysis of the physical device properties hasresulted in a highly accurate calibration mod-el. Using this model we successfully demon-strated a nullsteering beamforming applica-tion with an experimental array antenna sys-tem in a laboratory environment. For com-mercial system applications we are develop-ing calibration algorithms that are based onefficient mathematical and computationalmethods.

There is a new class of smart multi-elementantenna (MEA) systems that are promisingcandidates for wireless mobile systems withvery high spectral efficiencies and capacities.These are based on MEAs at both ends of thewireless link, with space-time encoding atthe transmitter, simultaneous transmission ofdifferent data streams at the same carrier fre-

quency, and signal reconstruction at the re-ceiver. Repeated estimation of the complexmatrix channel elements is required for sig-nal detection in the quasi-stationary flatRayleigh fading environment. First sets of al-gorithms and hardware architectures formultiple-input multiple-output (MIMO) chan-nel estimation and for signal recovery in dif-ferent MEA systems have been investigatedand compared with Lucent's VBLAST demon-strator. All systems have been designed tosupport high-speed real-time indoor WLANapplications.

If channel state information (CSI) is avail-able at the transmitter, then it is possible todo signal processing before data transmis-sion. In order to calculate the required trans-mit power, the probability density functionsfor the eigenvalues and inverse eigenvaluesof Rayleigh and Rician MIMO channels havebeen investigated and analytic expressionsfor the moments of these distributions havebeen derived. Since the radio channel is reci-procal, CSI can be made available at thetransmitter by taking it from the CSI used atthe receiver for signal detection. Linear chan-nel inversion (LCI) is a simple processingscheme using CSI at the transmitter. In thedown-link, data are multiplied at the trans-mitter by the pseudo-inverse of the trans-posed channel matrix HT, while slightly morecomplex processing is needed in the up-link.Like VBLAST, LCI offers significantly betterperformance than zero forcing. However, LCIneeds less signal processing than VBLASTwhen the number of data streams is large.

The combination of OFDM and smart an-tennas (MIMO-OFDM) turns out to be apromising candidate for future fourth gener-ation mobile systems, since it offers highspectral efficiency and low-complexity equal-ization in the most demanding wireless envi-ronments. The capacity of this system con-cept was investigated, starting with a physi-cal description of the underlying broadbandspace-time channel, and gave very encour-aging results. Channel estimation is crucialfor exploiting the capacity, and highly effi-cient estimation schemes were proposed.The design of the space-frequency code iscurrently under investigation, to round outthe complete system concept.

Smart antennas in the 60 GHz range willplay an important role in future mobile com-munication systems at data rates of155 Mbit/s. The generation and distributionof the microwave signals, and also beam-forming, will be carried out using optical mi-

crowave technology, which allows the reali-sation of simple base stations in the picocel-lular network. The required microwave sig-nals, with quartz stability and low phasenoise, are generated using the optical het-erodyne principle with modulation sidebandinjection locking. Beamforming is accom-plished by a newly developed a SiO2/Si-based beamforming network. The beam-former components and the beam steeringalgorithms will be tested in an experimentalsystem, and first experiments have been suc-cessfully carried out. The RF properties at60 GHz of the phase and amplitude sectionsof a 4-channel beamformer were measuredusing an antenna test setup with a 1 x 4phased array antenna. Error-free 155 Mbit/sdata transmission was achieved at variouslook directions of the antenna. This systempresently uses the maximum-directivity beam-forming algorithm.

Optical millimetre-wave techniques forbeamforming array antennas are being in-vestigated and specific components are be-ing developed as part of the EuropeanOBANET project. This project is studying var-ious coverage-area management strategiesand the technologies required to optimisetheir performance. The scenarios of fixed andmobile broadband radio access in the 40 GHzband are both being considered. An impor-tant issue for beamforming and beam steer-ing is the estimation of the direction of ar-rival. To investigate this problem, methodsusing pilot signals in a 40 GHz system are be-ing implemented at HHI, and DSP is beingused to evaluate them.

The beamforming and MIMO techniquescurrently being investigated lead to a drasticreduction of transmitter power, which great-ly eases the electromagnetic interferenceproblem. The reason for this is that the trans-mitter power with beamforming is concen-trated in the strongest paths between thetransmitter and receiver, which saves a con-siderable amount of RF power compared tonondirected emission. MIMO systems offerboth diversity gain and multiplexing gainover single-input single-output (SISO) sys-tems. These gains translate into drastically in-creased capacity, which can be used eitherto transmit more information without in-creased power or to save power at constantdata rates.

30 H H I R E P O R T 2 0 0 1 · R & D A C T I V I T I E S · M O B I L E B R O A D B A N D …

31H H I R E P O R T 2 0 0 1 · R & D A C T I V I T I E S · E L E C T R O N I C I M A G I N G …

Electronic Imaging Technology forMultimedia

Topics and Results

The R&D contributions of the HHI to thisresearch area are in the domains of ImageProcessing and Interactive Multimedia Appli-cations.

Image Processing

The HHI activities in image processing areconcentrated on signal processing and cod-ing for video services in various applications,on the design of VLSI components, and onsystem integration. A wide range of imageformats, from very low resolution for narrow-band video communication up to high reso-lution for multimedia services, are supported.Typical applications include VLBV (very lowbitrate video) for multimedia communicationat low bit rates over the Internet or mobilenetworks, interactive multimedia services,broadcast and communication services, andstudio applications.

As far as hardware projects are concerned,the development of architectures for futuremultimedia terminals is the central activity,with the focus on key components for 3Dgraphics, compositing, MPEG encoding anddecoding, and format conversion.

One emphasis is on the collaborative pro-ject Interoperability for Video Communica-tion over Distributed Networks (INVINET),which is sponsored by the Federal Ministry ofEducation and Research (BMBF). The mainaims of this project are the development ofbasic technologies for IP-based video com-munication, especially for mobile networks,and of interactive services over these net-works, particularly those using virtual envi-ronments. Important topics are the newcompression standard H.26L that is beingdeveloped in ITU-T, the development of anauthoring tool based on MPEG-4 for the cre-ation of 3D scenes, and a 3D compositor forMPEG-4.• As a result of its work on H.26L coding,HHI submitted a proposal to the ITU-T forstandardization, which has been accepted.The proposed scheme is a new and efficientmethod for context-based adaptive binaryarithmetic coding (CABAC), which gives bit-rate savings of up to 40 %.• HHI has taken over the responsibility forthe integration of the H.26L reference soft-ware.

• HHI has delivered two sets of test se-quences for comparison tests organised byMPEG. One set has been encoded with anMPEG-4 Advanced Simple codec optimisedfor rate distortion, and the other with anH.26L codec, also optimised for rate distor-tion. Both codecs were developed at HHIand are the most efficient MPEG-4 andH.26L codecs known. Based on these tests,MPEG has selected H.26L as the best codecamong seven proposals, and it will be usedas the basis for the development of a newMPEG coding standard.• The development of an MPEG-4 playercontaining a 3D compositor has been com-pleted. This player was used to demonstratea traffic information system based on MPEG-4 at IFA 2001.

The development and verification of tech-nologies that are required for new function-alities in mobile multimedia and Internet sys-tems is of crucial importance. Such function-alities have been identified in ISO-MPEG, andthe MPEG-4 standard will be the basis for im-age and sound compression in future multi-media systems. These new functionalities, in-cluding content manipulation, content-based scalability and content-based access,are obtained with coding algorithms that areboth very efficient and robust against trans-mission errors.• Hardware architectures for a completeMPEG-4 mobile multimedia terminal , basedon embedded processor cores and state-of-the-art memory technology, have been fur-ther developed, and MPEG-4 decoders havebeen ported to different platforms (Trimedia,ARM, TI).• A video streaming application over WLANhas been implemented and was demonstrat-ed at IFA 2001.• A final version of an e-commerce systembased on MPEG-4 software components hasbeen developed, in cooperation with T-Novaand Blaxxun Interactive, for the IST projectSoNG (Portals of Next Generation). The clientpart of this e-commerce system is Web-basedand consists of a generic, application-un-aware MPEG-4 compositor that can be em-bedded in an HTML page, along with appli-cation-specific Java code and MPEG-4 videoencoder and decoder modules. The core ofthe application logic is implemented indownloadable Java code modules, whichcommunicate with the MPEG-4 compositorusing the standard compliant MPEG-J API. Acontent provider can update the contentand logic of the e-commerce system by

changing BIFS, Java or JavaScript code on theHTML/MPEG-4 server. The video encoderand decoder facilities of the system allowcustomers to contact sales and support staffusing audio or video communication. Thissystem was demonstrated at IBC 2001.

Audio is an important feature of manyMPEG-4 application projects. Due to itsmany refinements, MPEG-4 AAC (AdvancedAudio Coding) is the most efficient audiocompression technique known at present.This standard supports sampling rates from8 kHz up to 96 kHz and provides audio ofnear CD quality at a very low bitrate (64 kbper channel). The AAC standard handles upto 48 channels in one bit stream, and istherefore suitable for any multichannel appli-cation. MPEG-4 AAC can be used in a widerange of applications such as home cinema,internet audio and hand-held players/recor-ders.• Audio encoders and decoders based onthe AAC standard have been further devel-oped. The implementations on PCs and DSPsare compliant with the ISO/IEC 14496-3 AACMAIN and LC (Low Complexity) profiles, anddeliver very high audio quality, almost indis-tinguishable from the original PCM. The soft-ware version for PCs is able to encode 4channels of audio data and decode up to 17channels in real time. The ANSI-C implemen-tation allows the algorithm to be transferredto other hardware platforms without exten-sive modifications. The TMS320 C6x DSP se-ries from Texas Instruments has been chosenfor the DSP implementation. These imple-mentations were all exhibited at IFA 2001and IBC 2001.

Image analysis and feature extraction arebasic technologies for intelligent network as-sistance systems. They facilitate access tomultimedia information over networks.Methods of processing for image analysisand classification are being developed forthis purpose as part of the MPEG-7 standard-ization process. In addition, object-basedand model-based methods are being investi-gated and further developed for the extrac-tion of useful visual parameters for such sys-tems. These techniques will enable the userto search for information among the increas-ingly overwhelming choice of programs thatare available over hundreds of DVB channelsand the Internet. To achieve this, the audioand video streams must be provided withstandardized index information to enablesearch engines using MPEG-7 to browse forspecial features. The selection of a program

with given content can then be either usercontrolled or event driven (e.g. a sportschannel could be automatically switched onif a goal is scored in a football match).• Several components for the MPEG-7 stan-dard have been specified and basic descrip-tors for low-level signal-based features and allvisual features (colour, texture, shape, mo-tion) have been developed.• HHI descriptors have been accepted aspart of the MPEG-7 standard.• Multimedia Description Schemes (Mid-Level Description of Multimedia Content)have been developed by HHI and contribu-tions have been made to the official MPEG-7Reference Software, to the definition of its bi-nary syntax, and to the overall MPEG-7 sys-tem design.

Computer graphics and image processingtechnologies have reached a state that allowsthe development of novel multimedia appli-cations and services. The new challenge inthe field of television and multimedia appli-cations is called immersive telepresence, inwhich users should feel that they are im-mersed in the scene. Large displays withhigh resolution are required. Two major ap-plications are 3D video conferencing and im-mersive television. The increased resolution,in particular, requires new approaches in thewhole image processing chain. Major contri-butions to 3D video processing have beenmade:• For 3D video conferencing, the completeprocessing chain, including segmentation,disparity estimation, view combining, MPEGencoding and decoding, 3D warping andnovel view synthesis, has been tested suc-cessfully using computer simulations underreal-time conditions.• The development of a DSP-based PCIboard called VPS (VIRTUE Processor Station)has been finalised und is currently being test-ed under real operating conditions. It con-sists of a multi-processor board and an inter-face board, and supports flexible digitalvideo input and output. The objective of theVPS is to assist real-time 3D video processingin a state-of-the-art PC. It has a standard PCIinterface and offers flexibility of software im-plementation and scalability in terms of fu-ture extensions. It is the core of a prototypeterminal for immersive 3D video conferenc-ing.• Two slightly different earlier versions of thisdevice had already been made as part of theEuropean IST project VIRTUE at Sony UK andof the BMBF project ITI at HHI. The VIRTUE

32 H H I R E P O R T 2 0 0 1 · R & D A C T I V I T I E S · E L E C T R O N I C I M A G I N G …


version was successfully demonstrated inDecember 2001 at the IST exhibition Event2001 in Duesseldorf.

Immersive Television (ImTV) is currentlygaining increased attention. The philosophyof ImTV is to deliver a combination of differ-ent depth cues, such as wide field-of-viewimagery and head motion parallax or stereovision, in order to evoke in watchers the sen-sations and emotions of being present at alive event. As a first step in the realization ofthis objective, a hardware system consistingof three fixed-view HD cameras (centre viewand side panels) for capturing the live eventshas been developed. A home receiver unit isable to combine these images into a pano-ramic Ultra-High Definition (UHDTV) view.The UHDTV images are displayed using ei-ther flat or curved screen projection systemsor a head-tracked, look-around head-mount-ed display (HMD) system.

The HMD feature is a further step of anevolutionary strategy towards a future ImTVservice. The ultimate goal behind this strate-gy is a visionary TV system, called InteractiveVirtual Viewpoint Video (IVVV), that allowsthe user to walk freely around in a naturalvideo scene, as known from virtual reality ap-plications. The following results have beenachieved so far:• A first prototype version of a UHDTV re-ceiver, based on MPEG-2, was demonstratedat the IFA fair in Berlin in September 2001.Further extensions supporting the HMD op-tion are in development and will be testedvery soon.• First extensions towards an IVVV systemhave been designed and investigated usingcomputer simulations. This system extensionis based on MPEG-4 and includes head-mo-tion parallax viewing to stimulate depth per-ception. An IVVV-related contribution to theIBC conference in Amsterdam in September2001 was honoured by the President’sAward for Best Technical Paper.• A concept of a hybrid MPEG-2/4 transmis-sion system for the evolutionary introductionof scalable and backward-compatible IVVVsystems into the existing DVB framework hasbeen developed for future integration work.

The integration of video decoding, formatconversion and graphics is one of the chal-lenges in the design of future multimedia ter-minals. A multi-standard Display Unit forMPEG Applications (DUMA), consisting ofhighly integrated components, has been de-signed in the KOKON project. The system,which is based on FPGAs, DSPs and micro-

controllers, combines video and graphicsoutputs from various decoder sources(MPEG-2 HDTV, MPEG-2 SDTV, MPEG-4) in-to one high-resolution video output. DUMAincludes the main functions of the types ofvideo output units that will become increas-ingly essential in future multimedia systems.• Various components of a multi-standarddisplay unit for MPEG-4 applications havebeen developed. This demonstrator will in-clude the main functions of a display unitwhich combines the video and graphic sig-nals from various sources (MPEG-2 decoder,MPEG-4 decoder, PC graphic system, etc.)into a high-resolution video output (HDTV).It will provide several interaction options.This interaction can be controlled by soft-ware running on an onboard RISC processor(StrongARM), which can also be used to de-code various audio standards (AAC, CELP,MP3). In order to reduce the amount ofhardware, an FPGA-based solution has beenchosen. All functions are implemented inVHDL and can be used as modules in subse-quent chip designs. The board has severalstandard interfaces for video (FBAS for TV,DVI), audio (I2S) and control (EPP). This hard-ware unit allows verification under real-timeconditions, and will also be used as an emu-lator for future ASIC developments.

Interactive Media – Human Factors

The HHI carries out R&D in the area of in-teractive multimedia services and new me-dia, concentrating on user friendly multime-dia applications, interaction technologiesand multimedia terminals.

The focus of the work is on autostereo-scopic display technologies, novel and trend-setting technologies for 3D desktop comput-ers, the development of innovative 2D and3D interaction techniques for man-machinecommunication, virtual telepresence andtele-immersion applications, methods for da-ta search and data visualisation, and intelli-gent agent-based information managementand user guidance. These main themes areextended by both fundamental and appliedwork in the area of usability engineering,with special emphasis on human factors con-straints.

The autostereoscopic display develop-ments concentrate on approaches that donot require the user to wear special glasseswhen viewing natural or computer generat-ed 3D images or videos. Applications of this

technology are in the areas of 3D telephonyand video conferencing, 3D TV, 3D multime-dia desktop computing, 3D virtual worlds,telepresence, telework, telesupervision, ve-hicular technology, CAD/CAM, 3D computergames, medicine, and biology. The work inthese areas includes the development ofsmall to large flat panel autostereoscopic dis-plays for one or more viewers, as well aslarge format front or back projection displaysusing either lenticular array screens or fieldlens technology. Prototypes of these devel-opments were introduced to the public thisyear at the IFA 2001 (Internationale Funkaus-stellung) in Berlin and last year at many tradefairs and exhibitions.• High resolution single-person autostereo-scopic 15.1" and 20" flat panel displays((2x512)x768 and (2x512)x1280, respective-ly) for use in the CAD/CAM area were devel-oped. These do not require mechanicaltracking of the lenticular screen – instead,the monitor is mounted on a mechanicaltracking platform so that the image planecan track the user’s head movements.• An autostereoscopic 15.1" notebook dis-play ((2x512)x768) for use in mobile applica-tions was developed.• High-resolution single-person autostereo-scopic 20" flat panel displays (both(2x512)x768 and (2x512)x1280) for use inthe CAD/CAM area were also developed.This particular technology consists of a 20"NEC display and a removable lens plate con-nected to the NEC device. IR head trackingmeasurements in combination with pixelswitching technologies give the users track-ing comfort when moving in front of thescreen.• A high-resolution autostereoscopic single-person display using field lens technologywas developed for applications in video com-munication. In order to show figures in nat-ural size for telepresence applications, thedisplay integrates a collimation lens systemto create a virtual image several metres be-hind the display surface. The display has out-standing resolution and brightness and mini-mal crosstalk.• A large format 40" autostereoscopic backprojection display with extremely high reso-lution ((2x1000)x750 pixels) was developedfor applications in multimedia desktop com-puting. This monitor features mechanicaltracking of the lenticular screen.• A video-based head tracker was developedso that the HHI displays can be used withoutspecial glasses.

• A newly developed IR head tracker for thesame purpose provides the same functionali-ty at reduced cost and with reduced sensitivi-ty to lighting conditions.

In recognition of the need to develop newconcepts and technologies that point theway to the 3D desktop computer of the fu-ture, the work at the HHI concentrates onthe vision of a computer with intelligentman-machine interfaces. To make operationeasier for the user, prototypes of desktopcomputers were developed that present in-formation in a clearly arranged and sufficientform using a 3D display and suitable 3D visu-alisation methods. The wishes of the user areanticipated by innovative multimodal inter-action and agent technologies, thus enablingsimple and intuitive operation in interactiveapplications.

An advanced prototype of an intelligentand ‘seeing’ 3D desktop computer that over-comes the disadvantages of the commonwindows desktops and allows novel forms ofuser interaction was implemented andshown publicly at various trade fairs and ex-hibitions.• The development of a novel 3D visual op-erating system and suitable editors gives thedesktop computer a simple and clearlyarranged representation of multimedia infor-mation, together with simple means for theuser to interact with it. In contrast to con-ventional windows-based desktops, the newoperating system is object oriented and canlayer information in the depth dimension.• A 20" high-resolution autostereoscopic dis-play that was developed in the HHI was usedfor the prototype of a 3D computer that al-lows the visualisation of multimedia objectsarranged in the depth dimension. These ob-jects are generated and managed by the 3Doperating system. Because of the autostereo-scopic representation, the user can see theobjects and information stacked in depth lay-ers, and can view them from different sidesby moving his or her head (movement paral-lax).• An improved video-based method of mea-suring the user’s head movement and gazedirection that was developed at the HHI re-lieves the user of the need to wear eitherhead markings or special glasses.• A binocular camera-based hand trackingsystem was developed for novel man-ma-chine interaction applications. This device isable to track hands or fingers in front of ascreen in real time, so that the user can pointto and interact with 3D objects projected us-



ing 3D displays in front of a screen. Based onsoftware computer vision algorithms, thistechnology performs tracking in real time ongeneral purpose PCs.• An interface agent can recognise what theuser is viewing and can then autonomouslyinitiate appropriate actions (visually con-trolled graphics). In this way the representa-tions of the displayed objects can bechanged in a manner that is adapted to theuser (e.g. the depth representation of theobjects can be changed to model the depth-of-focus behaviour of the human visual sys-tem). Also, the user can initiate actions bymeans of gaze control.• A prototype system for teleworking in 3Dwas developed and presented to the publicat the IFA 2001 exhibition in Berlin. It allowsteleworkers to jointly develop and manipu-late 3D objects. The components of this sys-tem include an autostereoscopic 3D display,a visual 3D operating system, a camera thatsenses head and hand position, and a systemto determine the gaze direction (for gaze-controlled interaction).

Research in the area of algorithms for 2Dand 3D image processing was carried out inorder to equip future generations of multi-media terminals and desktop computers witha large degree of intelligence and new formsof man-machine interaction. Applications arein the areas of 3D desktop computing, tele-working and telepresence. The work in thisresearch area concentrates on user recogni-tion, object recognition and tracking, 3Ddepth estimation, 3D intermediate view gen-eration, and head and gaze tracking.• A video-based head tracker was developedwith the aim of presenting users of an au-tostereoscopic display with views of 3D im-ages from various directions. The algorithmthat was developed responds reliably to eventhe smallest changes of head position, evenin difficult lighting conditions, and passes themeasurements to a head tracking display.This algorithm works in real time and wasimplemented under IRIX on O2 and underWindows NT.• The video-based head tracker was extend-ed, using infrared techniques, so that it canalso measure gaze direction. Gaze directionmeasurements are also being investigated forpossible use as input variables for interactivepurposes in future desktop computers (visu-ally controlled graphics).• A new work item dedicated to peoplesensing technologies was initiated in 2001.The purpose of this research effort is to equip

systems and devices with the ability to detectpeople in front of devices or in their sur-roundings. Of particular interest is the identi-fication of users, the sensing of their interac-tions with others, the sensing of their emo-tions or communication abilities, and therecognition of their gestures or general in-tentions.

Informal communication is an importantaspect of human communication. This in-cludes spontaneous and mostly confidentialconversations, e.g. on the fringes of confer-ences, during coffee breaks or at otherchance meetings. Since this form of commu-nication has a positive significance for the in-dividual worker as well as for the organiza-tion, means of providing technical supportfor informal communication among distrib-uted work groups (for either telework or tele-cooperation) are being investigated. A ques-tionnaire to potential users of telecoopera-tion systems showed that informalcommunication in many areas connectedwith work is viewed as absolutely essential.• To determine the essential features neededfor telecooperation systems to support infor-mal communication, various field studieswere carried out using a chat system, a chatsystem extended by still images, and a virtualenvironment. The results showed that infor-mal communication is possible with text-based communication systems, but that theyshould be augmented by indicators of thedegree of communication readiness of po-tential communication partners.• For telework applications, the telepresenceof a work group was simulated using a virtu-al office environment generated by comput-er graphics. Each participant of the teleworkgroup, including his or her position and ac-tions in space, was given a computer graphicrepresentation (avatar). This concept ap-peared to be promising, especially for infor-mal communication among teleworkers. Itdelivered a high degree of telepresence witha sufficient degree of privacy protection.• A PreMon Awareness Monitor was devel-oped that senses whether a teleworker is pre-sent in front of the monitor, and whether heor she is working or participating in oralcommunication. Speech and vision sensorsare used in this monitor.• In view of the increasing importance ofmobile teleworkers, the PreMon Awareness vis-ualization system was also adapted for use inmobile phones based on the WAP protocol.

Technology that allows users to search andnavigate through large image data bases is

attracting increasing attention in researchand development. Work was carried out inthe context of the MPEG-7 standardizationactivities, with particular focus on adaptiveand learning algorithms that are trained forthe preferences of the particular user andthat can search for images based on visualsimilarity measures. New philosophies andalgorithms were developed for clusteringand displaying image content based on hu-man visual criteria.• Human factors studies were initiated to un-derstand how human viewers perceive andcategorize information in images.• New 3D visualization tools were developedthat enable the user to navigate intuitivelythrough an image database.• A cognitive map of the database content,similar to a city street map, allows the user tonavigate through an image database using acoarse-to-fine concept.• A variety of MPEG-7 conformant descrip-tors that can be used to search for imagesbased on image content rather than text an-notations were developed and implemented.• Algorithms for the automatic classificationof images into pre-defined categories weredeveloped.

Competencies

Development of algorithms and hardware ar-chitectures for image and sound compres-sion (MPEG-2, MPEG-4, H.26L).

Development of algorithms and hardware ar-chitectures for 2D and 3D image analysisand synthesis using motion, stereo and mul-tiview information.

Image segmentation, feature extraction andclassification.

Development of applications based onMPEG-2/4/7 and JAVA for interactive servicesover the Internet, over DVB/DAB/DMB, overISDN, over xDSL, and over mobile networks(DECT, GPRS, HSCSD, UMTS).

Coding methods for videophones and videoconferencing (H.26x, MPEG-4).

MPEG and ITU-based signalling and trans-port protocols (H.32x, MPEG2-TS, DMIF).

Streaming technologies for video and audioover the Internet and wireless channels.

Development and design of integrated cir-cuits for image processing.

Design, integration and implementation ofprototypes and experimental systems forvideo-based applications in communicationand for tests and demonstrations of newcommunication technologies and hardwarearchitectures.

Modelling and development of integratedcircuits for image processing.

Development of 3D display technologies.

Conception and evaluation of user interfacesfor multimedia applications using VRML, JA-VA, MSDL and dVS.

Analysis and optimization of communicationterminals and services on the basis of humanfactors and usability criteria.

Development of video-based pattern recog-nition and photogrammetry methods.

Modelling and development of integratedcircuits for image processing.

Design and construction of experimental sys-tems for the development of video-basedcommunication applications and for testingand demonstrating new communicationtechnologies and hardware architectures.

Analysis of human sensory and sensorimotorfunctions in relation to communication appli-cations.

Expertise in desktop computer graphics de-sign.

Image and video retrieval systems.


37H H I R E P O R T 2 0 0 1 · R & D A C T I V I T I E S

Project ManagerphoneeMail

Godehard Walf+49(0)[email protected]

Erwin Patzak+49(0)[email protected]

Godehard Walf/Jürgen Saniter+49(0)30-31002-455/[email protected]/[email protected]

Ernst-Jürgen Bachus+49(0)[email protected]






Thomas Hermes+49(0)[email protected]

Hans-Georg Weber+49(0)[email protected]


R & D PROJECTS

Project

Photonic Networks

KomNet System Integration Office

Optical Technologies in Motion forthe IST Programme

Innovation of Photonic and MobileCommunication Network for theBroadband Internet (TansiNet ON)

Performance Monitoring of WDMChannels

DWDM and IP

Application of Distributed RamanAmplification for WDM Networks

Terabit Metro Ringnetworks –Application and Physical Limits ofWDM Techniques

System Aspects and Tests forCityring Networks

Design of Switchable OpticalNetworks

Frequency Selective IntegratedWDM Receiver

40 Gbit/s OTDM for KomNet

Optical Time Domain MultiplexTechnique for the Internet of theFuture

Provider of Grant/ ContractorPeriod

BMBF/Industry(KomNet)5/98 – 4/02

EU5/00 – 4/03

BMBF6/00 – 5/03

DFG6/99 – 12/01

HHI10/00 – 12/02

DFG3/00 – 5/01

Freunde des HHI10/00 – 9/01


BMBF(UMTS)4/01 – 3/03

DFG8/99 – 7/02

Industry /KomNet7/99 – 4/02

BMBF8/99 – 7/02

DFG5/00 – 4/02

DFG4/98 – 3/02

BMBF8/00 - 7/03

DFG3/00 – 2/02

HHI8/99 – 7/02

Industry4/00 – 12/02


BMBF2/01 – 7/03

EU9/01 – 8/04

BMBF7/01 – 6/04

HHI1/98 – 12/03


Industry11/00 – 10/01

DFG2/01 – 6/02








Bernd Sartorius+49(0)[email protected]



Michael Schlak/Bernd Sartorius+49(0)30-31002-407/[email protected]/[email protected]

Michael Schlak +49(0)[email protected]

Michael Schlak+49(0)[email protected]


Optical Sampling of High BitrateData Signals

Gain-Clamped Optical Amplifier inFibre-optical Transmission Systems

Active Resonant Grating-WaveguideStructures for Rapidly TuningSemiconductor Lasers with noMoving Elements

Saturable-absorber Switch

Technological Developments forHigh Speed OTDM TransmissionSystems

160 Gbit/s TransmissionTechniques

High-bitrate Optical Signal Processing

High-frequency Selfpulsation ofMulti-section DFB Laser

TOPRATE

Optical 3R Regeneration for Asynchronous IP Networks

Technical Development for OTDMDemultiplexers

160 Gbit/s 3R Regenerator

40 Gbit/s Wavelength-Converter

Design of Switching Characteristicsin Dispersive Q-switchSemiconductors

38 H H I R E P O R T 2 0 0 1 · R & D A C T I V I T I E S


MBI/HHI11/98 – 3/02

HHI4/99 – 12/01


Industry6/00 –3/01

Industry4/01 –10/01

BMBF/Industries(KomNet)6/98 – 4/02

BMBF4/00 – 3/03

BMBF10/00 – 9/03




HHI7/99 – 6/02

DFG9/00 – 8/03

State of Berlin1/00 – 6/01

Harald Künzel+49(0)[email protected]

Martin Möhrle+49(0)[email protected]

Martin Möhrle+49(0)[email protected]

Anagnostis Paraskevopoulos+49(0)[email protected]

Anagnostis Paraskevopoulos+49(0)[email protected]

Helmut Heidrich+49(0)[email protected]



Heinz-Gunter Bach+49(0)[email protected]





Peter Wolfram+49(0)[email protected]

Semiconductor Saturable AbsorberStructures for Femtosecond demul-tiplexer

Development of BH-Laser

Development of DFB-BH-Lasers

LED-Arrays

LED-Array Modules

Low-Cost Transceiver-PIC for theAccess Network

Innovative Lasers and Filters on InPUsing Ring Oscillators

ps-OEIC

Development of Twin Photodiode-Photoreceivers for a 40 Gbit/s TDMField Experiment in KomNet

40 Gbit/sPhotoreceiver ModuleDevelopment for the GIGATRANS-PORT Field Experiment in KomNet

40 Gbit/s Frontend Developmentfor an ETDM Long HaulTransmission in KomNet

GaAs-Based 1300 nm Laser

Lattice-matched MBE-growth of(In,Ga)As/(In,Al)As- and(In,Ga)As/Al(As,Sb)-structures onInP for Intersubband Emitters

Material Development for IR-LEDs

BMBF4/00 – 3/03

HHI1/01 – 6/03



HHI6/98 – 5/01

State of Berlin/HHI6/00 – 11/01

BMBF(UMTS)5/01 – 12/03

HHI1/01 – 12/02

BMBF6/01 – 5/04

BMBF(UMTS)4/01 – 12/03


BMBF6/00 – 5/03

HHI6/00 – 5/03

Berndt Kuhlow+49(0)[email protected]

Berndt Kuhlow+49(0)[email protected]

Margit Ferstl+49(0)[email protected]

Udo Niggebrügge+49(0)[email protected]

Udo Niggebrügge+49(0)[email protected]

Karl-Otto Velthaus+49(0)[email protected]

Karl-Otto Velthaus/Herbert Venghaus+49(0)30-31002-645/[email protected]/[email protected]

Norbert Keil+49(0)[email protected]

Norbert Keil+49(0)[email protected]

Thomas Rosin+49(0)[email protected]

Wolfgang Schlaak+49(0)[email protected]

Holger Boche+49(0)[email protected]


Micro-Filter in Silica-on-SiliconTechnology for OpticalCommunication

Planar integrated Waveguide-network on Silica

Development of ArrayedWaveguide Gratings

Development of Innovative andCompetitive Technologies as aPreliminary Step towards IndustrialFabrication of OptoelectronicComponents and CommunicationSystems in Berlin

Development of AdvancedTechnologies for UltrafastPhotoreceivers

Electro-optical Modulators

80 Gigabit Modulator Module

AWG Based on Polymer

Integrated Optical Add/DropMultiplexer Based on PolymerTechnology

Advanced Packaging Technologiesfor Highest Frequency Opto-Electronic Components

Development and Realization ofPhotodiodes


TransiNet – Innovative TransportNetworks for the BroadbandInternet (TV)

Teletraffic Engineering for PackedSwitched Services



HHI6/00 – 5/03

DFG7/00 – 6/02

Industry/Freundedes HHI9/99 – 12/01

HHI9/00 – 12/03

BMBF11/00 – 12/03

BMBF5/00 – 4/02

HHI5/00 – 4/02

EU12/00 – 11/03

Industry/Freundedes HHI4/00 – 12/01

DFN-Verein/Industry11/99 – 10/01

DFG6/00 – 6/02

DFG1/01 – 12/02

DFG6/99 – 5/01

DFG12/99 – 11/01



Martin Schubert+49(0)[email protected]

Clemens v. Helmolt+49(0)[email protected]

Clemens v. Helmolt+49(0)[email protected]

Gerd Großkopf+49(0)[email protected]



Andreas Kortke+49(0)[email protected]

Ralf Schäfer+49(0)[email protected]



Jens-Rainer Ohm+49(0)[email protected]


Multiuser Receiver for CDMASystems

Sequences in CDMA Systems

Uplink-Space-Time Array Processingfor 3G WB/CDMA

MULTIMODE: MIMO Techniquesand Antennas

Intelligent Multi-Element Antenna-and Non-Orthogonal Multicarrier-Systems for Modern BroadbandMobile Communication

Optically Controlled Array Antennasfor Wireless Cellular BroadbandNetworks

Optical Beamforming Network onSiO2/Si

Optical Beam-formed Antennas forAdaptive Broadband Fixed andMobile Wireless Access Networks

Design of Pilot-Assisted and BlindBeamforming Algorithms for Space-Time Rake Receivers

DFN Giga-Media-Services forCooperative Post-Production ofFilm and Video

Algorithms for a 3D-Realtime VideoConferencing System with highTelepresence

Multi View Synthesis of N ParallelStereo Rigs

Video Coding Using 2D and 3DObject and Motion Models

Feature Based Coding andSimilarity Analysis of 3D Objectsand Scenaries

Electronic Imaging Technology for Multimedia

DFG1/00 – 12/01

HHI1/00 – 12/02

BMBF7/00 – 6/03


EU4/00 – 3/02


HHI1/01 – 6/01

BMBF7/01 – 3/02

BMBF1/99 – 6/02

BMBF(UMTS)5/01 – 12/03



DFG1/00 – 12/01

EU1/00 – 12/01

EU1/00 – 12/02


Peter Kauff+49(0)[email protected]

Peter Kauff+49(0)[email protected]

Guido Heising+49(0)[email protected]

Guido Heising+49(0)[email protected]

Uli Höfker+49(0)[email protected]



Maati Talmi+49(0)[email protected]

Maati Talmi+49(0)[email protected]

Karsten Grüneberg+49(0)[email protected]

Karsten Grüneberg+49(0)[email protected]

Christoph Fehn+49(0)[email protected]

Peter Stammnitz+49(0)[email protected]

Oliver Schreer+49(0)[email protected]

Scalable Videocoding by Waveletsand Fractal Techniques

Technologies for Hybrid SceneComposition in Interaktive Tele-presence

Immersive Telepresence in theInternet

Interoperability for VideoCommunication over DistributedNetworks, INVINET

Dynamic Radio for IP Services inVehicular Environment

Immersive TV

Technologies for ImmersiveTelevision Systems

Digital Cinema/Distribution andProjection Technologies

Key Components for UserConfigurable Terminals, KOKON

Bluetooth- and MPEG-4-based KeyComponents for Mobile Systems

Hardware and SoftwareArchitectures for MultimediaTerminals, SIPROS

Programmable HYbrid TV Systemwith Integrated CoreS – PHYSICS

Virtual Views for Immersive TVApplications

Systems for Advanced MultimediaBroadcast and IT Services

Virtual Team User Environment



BMBF(UMTS)4/01 – 12/03

BMBF9/01 – 12/03

EU1/01 – 12/02

BMBF6/01 – 5/04

BMBF1/99 – 12/01

BMWi1/99 – 5/02



BMWi1/99 – 3/02

DFG1/99 – 12/01

BMBF(UMTS)6/01 – 12/03

Detlev Marpe+49(0)[email protected]

Aljoscha Smolic+49(0)[email protected]

Aljoscha Smolic+49(0)[email protected]

Thomas Wiegand+49(0)[email protected]

Siegmund Pastoor+49(0)[email protected]

Lothar Mühlbach+49(0)[email protected]

Lothar Mühlbach+49(0)[email protected]

Jens Faber+49(0)[email protected]

Thomas Meiers+49(0)[email protected]

Thomas Sikora+49(0)[email protected]

Thomas Ellerbrock+49(0)[email protected]

Open Multimedia StreamingArchitecture

Optical Information Systems forTraffic Analysis and Traffic Control/Videobased Traffic Monitoring

Metadata for Advanced ScalableVideo Coding Tools

Image Analysis and Recognition inInformation Systems

3D System with MultimodalInteractions

A Communication Platform forInformal and Computer SupportedCommunication in Telework andTelecooperation

Roomancer System: Evaluation ofthe KICK Browser in Terms ofUsability

EuroInfo@Berlin

Intelligent and User-adaptiveSystems for Navigation andRetrieval of Images

Image Recognition andSegmentation for MultimediaApplications

Sensing People – IntelligentCameras and Sensors

45H H I R E P O R T 2 0 0 1

Selected Contributions

47H H I R E P O R T 2 0 0 1 · S E L E C T E D C O N T R I B U T I O N S · P H O T O N I C N E T W O R K S

A consortium of industrial partners in Ger-many, consisting of Acterna, Alcatel SEL,Infineon Technologies, Lucent Technologies,Siemens, T-Systems Nova and VPI Systems,currently collaborates, together with 18 re-search institutes, universities and SMEs, ondeveloping and field-trialling optical net-working technologies, network elements,components, protocols and access networktechnologies [1]. A complete demonstrationcommunication network, using an in-placeStandard Single Mode Fibre infrastructureprovided by T-Systems Nova, has been im-plemented and demonstrated. The system,which comprises a long distance network, ametropolitan network and access networks,interconnects various locations in the cities ofBerlin, Darmstadt, Mannheim, and Stuttgart.

One part of the long-distance network isbased on N x 40 Gb/s DWDM systems (withN up to 32), equipped by different vendors.For example, foward error correction anddistributed Raman amplification have beenstudied. The work has also focussed on10 Gb/s systems. A single-fibre bidirectional16 x 10 Gb/s transmission system is used tobridge the 750 km distance betweenDarmstadt and Berlin. With an eye to themore distant future, field experiments withultra-high data rates up to 160 Gb/s (singlechannel) have been carried out successfullyon a 116 km network segment.

In the city of Berlin a managed optical me-tropolitan DWDM network has been imple-mented. This test bed consists of two trans-parently interconnected bidirectional line-switched rings that are equipped with

dynamically reconfigurable optical add-dropmultiplexers (OADMs, up to 80 channels at10 Gb/s) and fully managed opto-electronicIP-optimised nodes. In addition, burst-modetransmission and automatic switching of op-tical network nodes (ASON) have been field-trialled by the consortium.

For access networks, the work has focussedon innovative digital access techniques suit-able for transporting data signals at up to8 Mb/s via twisted pair lines (xDSL), provi-sion of high-quality multimedia services viahybrid fibre-coax networks, optical code divi-sion multiplexing feeder lines (up to1.25 Gb/sper channel, e.g. Gb-Ethernet), and packet-oriented hybrid-fibre-radio systems.

As described in more detail in the follow-ing contributions, HHI contributes to KomNetas a subcontractor to industry with severalprojects on the design and simulation of op-tical networks, evaluation of network ele-ments, ultra high-speed TDM transmissionup to 160 Gb/s, 40 Gb/s photodetectors andreceivers, all-polymer OADMs, optical signalregenerators (3R), transceiver photonic ICs,and picosecond opto-electronic ICs.

Also, the KomNet Systems Integration Of-fice at HHI oversees the implementation ofthe demonstration network and the field tri-als. The work is supported by the industrialpartners, and is partly funded by the GermanFederal Ministry of Education and Researchunder grant 01 BP 805.

For more information seehttp://www.hhi.de/komnet/

Klaus-Dieter Langer ([email protected])

KomNet – on the Way to the Next Generation Internet

Under the framework of the German R&D initiative KomNet, an industry-based consor-tium is developing and examining key technologies needed for the implementation of fu-ture multi-vendor multi-client dense wavelength division multiplexing (DWDM) commu-nication networks. The KomNet projects have made considerable progress in implement-ing and testing optical networking technologies, network elements, subscriber loops andprotocols under real-life operating conditions, and excellent results have been achieved.HHI contributes to KomNet with several R&D projects and the Systems Integration Office.

German KomNet testbedtopology. The entirenetwork has beendemonstrated byActerna, Alcatel SEL,Lucent Technologies,Infineon Technologies,Siemens and T-SystemsNova.

Darmstadt

Long - distancenetwork

Metro networkBerlin

107 km32 x 40 Gb/s

750 km2 x 8 x 10 Gb/s

16 x 2.5 Gb/s ...80 x 10 Gb/s

141 km40 Gb/s

15 km

15 km

29 km25 km

18 km

Stuttgart

Mannheim

Access network

48

Two important developments in opticalcommunications networks can be observed:• In recent years the bandwidth available on

optical fibres has increased by orders of magnitude, mainly due to wavelength-division multiplexing (WDM) technology.

• The exponential growth of traffic generat-ed by the Internet is causing a paradigm shift in optical networking away from circuit switching towards packet switching.

The objective of the joint project TransiNet(started in June 2000) is to develop new net-working technologies to carry IP traffic overWDM optical networks in an efficient manner.

The partners in the TransiNet research pro-ject are:• Heinrich-Hertz-Institut für

Nachrichtentechnik Berlin GmbH (HHI)• Universität Stuttgart, Institut für Nachrich-

tenvermittlung und Datenverarbeitung (IND)

• T-Systems Nova GmbH, Technologiezen-trum (T-Systems Nova)

• TU Berlin, Institut für Nachrichtentechnik und Theoretische Elektrotechnik, Fachge-biet Telekommunikationsnetze (TKN)

• TU München, Lehrstuhl für Kommunika-tionsnetze (LKN)

Cooperation between the partners is orga-nized into the working groups shown in thefigure.

The main research areas for the next-gen-eration optical Internet in the TransiNet pro-ject are new concepts for network and nodearchitectures and the corresponding proto-cols, standards and technologies, traffic ag-gregation, scalability of WDM networks,

quality of services (QoS), network manage-ment, resilience, and the interworking andintegration of high-speed photonic and mo-bile communication systems.

As there are still many technological prob-lems in optical packet switching, opticalwavelength-switching, and probably lateroptical label switching (OLS) or optical burstswitching (OBS), are considered to be thenext steps towards more dynamic opticalnetworks.

Advanced optical subsystems for opticallyswitched dynamic networks are being inves-tigated in the HHI Department of OpticalNetworks (ON), including WDM transmis-sion, add-drop multiplexers, optical cross-connects, and wavelength router/switches.Their interworking with higher-layer proto-cols, manly IP, is also being studied.

Parts of an IP-optimized air interface arebeing developed in the Department ofBroadband Mobile Communication Net-works (BM). The main research areas are newsignal processing algorithms, prediction ofthe radio channel, resource allocation, andteletraffic techniques for use in the planningprocess.

In October 2001 an open workshop wasorganized by the HHI together with allTransiNet partners. The title of the work-shop was “How dynamic must a future trans-port network be?”

This work is supported by the FederalMinistry of Education and Research undergrant 01 AK 020A.

For more information seewww.transinet.de.

Jürgen Saniter ([email protected])

H H I R E P O R T 2 0 0 1 · S E L E C T E D C O N T R I B U T I O N S · P H O T O N I C N E T W O R K S

TransiNet – Innovative Transport Networks for the Broadband Internet

The joint research project TransiNet aims to develop new concepts for carrying IP trafficover wavelength-division multiplexed (WDM) and wireless networks. This contributiongives a short overview of the TransiNet project.

Focal points inTransiNet

TransiNet

IP overWDM / Wireless

Architectures

Focal Points in TransNet

Wireless

Routing andResilience

NetworkOperation


Optical fibres can be operated to carry da-ta at several terabits per second. This leads toa mismatch with the current switching tech-nologies, which are only capable of switch-ing bit rates electronically in the order of sev-eral Gbit/s. The objective is to shift the switch-ing burden into the optical domain. Theintroduction of wavelength-routed opticalnetworks will be the first step. Later architec-tures using optical burst switching (OBS) oroptical packet switching (OPS) might follow.

Depending on the switching frequency,the requirements of the switches vary. Inwavelength-routed optical networks, trans-parent optical paths in the network areswitched, depending on the traffic, usingtunable optical add-drop multiplexers or op-tical cross-connects (OXC). Using wave-length-routed connections, every connectionwill last from a few seconds to several hours.Consequently, the optical switches or multi-plexers must have switching times in therange of 10 ms. Micro-electromechanicalsystems (MEMS) or acousto-optical tunablefilters (AOTFs) are capable of accomplishingthis and are commercially available.

However, in OBS or OPS networks the dataare transmitted in optically switched pack-ages. The average burst length using OBS isin the range of 50 µs, whereas the packagelength using OPS is in the range of several100 ns (see the figure). The switches in theoptical network nodes have to be fastenough for this high frequency switching.Currently, only semiconductor optical ampli-

fier (SOA) or LiNbO3 switches are capable ofachieving these speeds.

Switching times are important in opticalnetworks, but signal degradation also has tobe considered, since signal regeneration isexpensive and should be avoided wheneverpossible.

The TransiNet project is investigating theperformance of various optical node archi-tectures, including their influence on the sig-nal quality.

This work is supported by the FederalMinistry of Education and Research undergrant 01 AK 020A.

For more information seewww.transinet.de.

Jürgen Saniter ([email protected])

Fast Optical Switching in Future IP Networks

Network architectures and enabling technologies for future optical Internet networks areinvestigated in the TransiNet project.

Technologies and typicaltimes in optical switchednetworks

1 ns

Available Switching-Technologies:

Propagation Times:

OpticalSwitching:

10 ns 100 ns 1 µs 10 µs 100 µs 1 ms 10 ms 100 ms 1 s

t

SOAs LiNbOSwitches

1 - 10 nsec

3 TunableAOTFs

Typical transmission times @ 10 Gbit/s:

Burst:Packet: ~ 32 nsec; ~ 1.2 µsec

~ 32 µsec; ~ 0.8 msec40 KByte40 Byte 1500 Byte

1 MByte

MENS,Thermo - opt. SW

~ 10 msec

Burst SW

t ~ 5 msec(1000 km)

Prop

Packet SW Flow SWCircuit,

- SWλ

2 - wayE - to - E Setup

~ 20 msec - ? (t + t )Prop Proc

50

The first step in implementing opticalswitching is the introduction of opticaladd/drop multiplexers (OADM) and opticalcrossconnects (OXC), which dynamicallyswitch wavelength channels. In such a net-work the number of channels used in a fibremay vary randomly. To use the network moreefficiently, the next step might be to imple-ment optical burst and packet switching, sothat even the power of a single channel willno longer be constant but will change on atime scale corresponding to the lengths ofthe bursts and the gaps between them.

When EDFAs are used in networks withbursts or WDM channels that are randomlyturned on and off, the gain will change in re-sponse to the input power variation. Thisgives rise to dynamic power excursions ofthe channels used, thus leading to a degra-dation of the optical signal-to-noise ratiosand bit error rates of the channels.

Figure 1 shows a simulation result for theoutput power of one of eight WDM chan-nels, each carrying random bursts, after acascade of seven EDFAs. Crosstalk and largepower excursions are observed.

There are basically two methods of keep-ing the gain of the amplifiers in a link con-stant.

The first method is to adjust the pumppower. This can be done either by using afeedback control loop or by forward control.In the latter case the pump power ischanged according to the measured inputpower. In the feedback control loop case thegain is measured and the pump power is ad-justed.

The second method is to inject an addi-tional saturating optical signal into the EDFAand thereby adjust the gain. Once again, thepower in the additional channel can be ad-justed by either feedback or feedforwardcontrol. Feedback control can be implement-ed all-optically (gain-clamped optical amplifi-er). If the extra channel is within the flat gainregion of the EDFA, the feedforward controlmethod keeps the optical input power con-stant. In this case the method can be used tostabilise a whole link with a cascade of opti-cal amplifiers.

Our study lead to the following conclu-sions:• If some channels are operated continuous-

ly, very large power excursions are elimi-nated, but control is still necessary.

• At present the best method seems to be to use fast electronic gain control for each amplifier. The control should be a feed-back loop with the pump power adjusted according to measurements of the amplifi-er gain using suitable monitor outputs. The feedback signal can be filtered elec-tronically to optimise the total dynamic re-sponse and to avoid the oscillations of gain that are observed with the all-optical con-trol method. This method makes efficient use of the available pump power (unlike an optically stabilised EDFA where most of the pump power is converted into the lasing wavelength) and also allows fine control and adjustment of the EDFA gain without sacrificing any of the EDFA transmission bandwidth.

• The link control method, which uses an ex-tra channel to compensate for the power variations, is less expensive and might fulfil the quality requirements for moderately long amplifier cascades and a moderate number of channels.

Erwin Patzak ([email protected])


Optical Amplifiers in Wavelength- and Burst-Switched Optical Networks

Dynamic gain changes of erbium-doped fibre amplifiers (EDFAs), which occur if wave-length division multiplexed (WDM) signals or optical bursts are switched transparentlythrough an optical transport network, must be greatly reduced. In a study we have com-pared and assessed the different ways of stabilising the gains of EDFAs on a time scale ofmicroseconds.

Optical power of one ofeight WDM channels

after a cascade of sevenEDFAs

0 1 2 3Time [ms]

0

5

10

20

15

25

Out

put p

ower

[mW

]


Optical performance monitoring and chan-nel identification independent of the electri-cal client signal is a key issue for the manage-ment of future all-optical networks. Perform-ance monitoring is necessary to determinethe quality of a transparent light path inde-pendent of the client signal. Channel identifi-cation is essential for future switched trans-parent networks to ensure proper routing inoptical add-drop multiplexers and opticalcrossconnects.

The new control modulation technique canbe used to monitor both identity and qualityof the single wavelength channel. Unlike pre-viously known techniques, e.g. optical signal-to-noise ratio (OSNR) measurement or theQ-factor method, the control modulationtechnique predicts the quality of optical trans-mission independently of the electrical client,and also with a higher accuracy, since evendegradation due to pulse distortion is takeninto account.

The principle of the new method areshown in Figs. 1 and 2. The control signal isa low-bitrate digital signal upconverted tofrequencies above the payload modulationspectrum and is added to the payload signalusing an additional optical modulator. Tomonitor quality and identity at the networknodes, a small amount of the composite sig-nal is tapped off and fed to a control signalreceiver.

Figures 3 and 4 show experimental resultsfrom loop measurements with a 10 Gb/spayload and a 50 Mb/s control channel. Thequality correlation for optimum transmissionparameters up to transparent distances of2500 km is shown in Fig. 3. The transparentreach may be further extended by the use offorward error correction (FEC) in the controlchannel. While the payload channel remainserror free with a slight power penalty, themeasured BER in the control channel de-grades significantly. Figure 4 illustrates theearly warning function of the control channelfor transmission impairment due to selfphase modulation (SPM).

[1] M. Rohde, “HF-Prüfmodulation zur Sig-nalkontrolle in WDM-Netzen”, 2. ITG-Fach-tagung Photonische Netze, pp. 165-168, 12-13 March 2001, Dresden.

Michael Rohde ([email protected])

Control Modulation Technique for Client-Independent PerformanceMonitoring and Channel Identification in Transparent Optical Networks

A new in-channel monitoring technique for transparent networks using a high-frequencydigital control signal for optical performance monitoring and channel identification is in-vestigated by experiments and simulations. The overhead channel serves as an earlywarning function for client channel performance degradation due to transmission impair-ments.

Fig. 1: Application of the control modulation tech-nique in a transparentWDM network with optical crossconnects

Fig. 2: Insertion of the controlsignal and the opticalspectrum

Fig. 3: Quality correlation fortransmission parametersset to optimum.

Fig. 4: Quality correlation fortransmission impairmentdue to self phase modu-lation (SPM).

payloadsignals

payloadsignals

digitalcontrol signalon rf carrier

control signal receivers foridentification & quality monitoring

of individual WDM channels

Tx Rx

Tx Rx

Tx Rx

WDM-network withoptical Crossconnects

inter -ference

payloadsignal

controlsignal

λ

RF

digitalcontrol signal

electr. mod.

opt. mod.payload signalon opt. carrier

827 km

1E-3

1653 km

BTB

1E-4

1E-5

1E-6

1E-7

1E-8

0.01BER of control channel

2480 km2

1

0

Pen

alty

of p

aylo

ad c

h. [d

B]

1E-3

496 km

1157 km

992 km827 kmBTB

1E-4

1E-5

1E-6

1E-7

1E-8

0.01BER of control channel

2

1

0

Pen

alty

of p

aylo

ad c

h. [d

B]

52

New fibre types with different dispersionfeatures are currently being implemented inoptical networks (Fig. 1). Of special interestfor long-haul high bitrate applications isNZDSF (LEAF®, TeraLight™, TrueWave®).

We have investigated the potential of LEAFat 40 Gbit/s using the simulation toolTransmissionMaker from VPIsystems. The si-mulated transmission link consists of sym-metrically compensated LEAF sections, withthe section length equalling the amplifierspacing. A contour plot showing the trans-mission length for varying section power andsection length is given in Fig. 2. The largetransmission length can be achieved using ashort-period dispersion compensationscheme 10 km long. This reduces the pulsedispersion in NZDSF and also gives less irre-versible pulse distortion due to self phasemodulation.

Network providers are looking for a migra-

tion path from SSMF-only networks to a nextgeneration networks including NZDSF. Henceit is of major interest to evaluate the perfor-mance of a mixed setup of SSMF and NZDSFlinks. We have investigated the special case

of alternating SSMF and LEAF sections for 10Gbit/s NRZ signals. This link setup is shownin Fig. 3.

For this setup the dependence of the sys-

tem penalty on length is shown in Fig. 4.The single channel case reaches a transmis-sion length of 4000 km with a 2 dB penalty.This is actually slightly better than a SSMF-only link [1]. The 3-channel case with achannel spacing of 200 GHz reaches3000 km, and for the 5-channel case with100 GHz channel spacing a transmissionlength of 2000 km is reachable. The addi-tional penalty for the WDM transmissions iscaused by cross phase modulation (XPM), asshown in [1]. The SSMF-only setup and themixed SSMF and LEAF setup reach compara-ble transmissions lengths, thus demonstrat-ing the compatibility of SSMF and LEAF sec-tions in a link.

[1] C. Caspar et. al, “Penalties through

XPM crosstalk in a switched long haul stan-dard fiber WDM system based on normalizedtransmission sections”, OFC 2001, Anaheim,USA, 18.-22. March, Techn. Dig., paper WI5.

Kai Habel ([email protected])


Potential and Implementation of Non-Zero Dispersion Shifted Fibres(NZDSF) in Long Haul Transmission

New fibre types such as NZDSF are receiving more attention as the bit rates are rising.Simulation results for future 40 Gbit/s NRZ transmission systems are presented for disper-sion-optimised transmission links using NZDSF. To test the compatibility with existingstandard single mode fibre (SSMF) links, experiments are performed with a mixed setupof SSMF and LEAF® links at 10 Gbit/s.

Fig. 1: Dispersion parameters

of various fibre types including NZDSF

Fig. 3: Setup for mixed

standardised sections

Fig. 2: Contour plot of trans-

mission length as a function of the fibre

input power and sectionlength. The simulation isat 40 Gbit/s with disper-sion-optimised LEAF sec-

tions of up to 1500 kmtransmission lengths.

Fig. 4: Penalty of mixed LEAF

and SSMF links for singlechannel transmission of

10 Gbit/s NRZ signals.This shows the increas-

ing influence of XPM fordecreasing channel spac-

ing for WDM transmis-sion

1300 15001400

SSMFTeraLightTrueWave

LEAFMetroCor

1600Wavelength [nm]

-20

+20

0

Dis

pers

ion

para

met

er D

[ps

/ (nm

x k

m)]

Tx

SMF40 km

SMF - section LEAF - section

SMF40 km

LEAF40 km

LEAF40 km

DCF DCF

Rx

0 2000 4000Length [km]

0

8

4

Pen

alty

(B

ER

= 1

0)

[dB

]-9

5 ch.f=100GHz∆

3 ch.f=200GHz∆ 1 ch.

0 4

1500 km1250 km

2 6 8Power [dBm]

20

40

60

80

Sec

tion

Leng

th [k

m]

1000 km

800 km600 km


With increasing demand on transmissioncapacity, distributed Raman amplification(DRA) in conjunction with EDFAs has be-come a very important technique in broad-band high-speed wavelength division multi-plexed (WDM) transmission systems, includ-ing the L-band and S-band. Long-haultransmission over 8000 km with hybrid am-plification (DRA and EDFAs) and over 4000km with solitons and distributed Raman am-plification have been reported.

The work carried out at the HHI concen-trated on all-distributed Raman amplifiedlinks without EDFAs, forward error correction(FEC) or solitons [1]. The main objectives ofour investigations were to prove whetherlong transmission distances can be achievedwith Raman amplification only and to findwhat properties can be expected from theselinks, compared with those using EDFAs andhybrid amplification. The problems of imple-menting high-gain Raman amplification inthe system and of modelling the amplifica-tion process had to be solved. Simulations ofthe links show good agreement with the ex-periments (done in cooperation withVPIsystems Inc., Berlin).

To evaluate the properties of a WDMtransmission system using all-Raman amplifi-cation, we performed 10 Gb/s NRZ WDMtransmission experiments with up to 16channels in our loop testbed (Fig. 1). Theloop consisted of 171 km of sSMF and 28 kmof DCF and used pre and post dispersioncompensation. High-gain distributed Ramanamplification of 58 dB was applied to com-pensate for the losses of the fibres and com-ponents. To achieve this high gain and toavoid severe Rayleigh back-scattering of thepump power in the fibres, the sSMFs andDCFs were pumped at five points with a1455 nm Raman fibre laser of up to 3.4 Wand a 1447 nm diode laser at 0.24 W. Thepump power for the fibre laser was split bycouplers into four parts (30 %, 30 %, 20 %and 20 %).

Figure 2 shows the BER as a function of thereceived power for transmission over dis-tances from 1000 to 4000 km. The all-

Raman-amplified transmission over 4000 kmwithout EDFAs, FEC, and solitons is, to thebest of our knowledge, a record transmissionlength.

These transmission results show that longdistances can indeed be bridged with all-Raman amplification and that further optimi-sations might lead to similarly good results aswith EDFA and hybrid amplification. Ramanamplification has the advantage that it has amuch larger amplification band and betterdynamic behaviour when the system oper-ates in a switched or burst mode.

This work was funded by the DeutscheForschungsgemeinschaft under grant Ba1790/2-1.

[1] E. Schulze, R. Freund, M. Malach, F.Raub “10 Gb/s NRZ transmission over 1826km multiple pumped distributed Raman am-plified transmission link without lumped am-plifiers”, ECOC 2001.

Elmar Schulze ([email protected])

10 Gb/s NRZ Transmission over 4000 km Using All-Distributed Raman-Amplified Transmission Links Without Lumped Amplifiers

NRZ transmission at 10 Gb/s over the record length of 4000 km has been demonstratedusing all-distributed Raman-amplified links without the use of erbium-doped fibre ampli-fiers (EDFAs), FEC or solitons. Large Raman amplifications of 58 dB in long-haul transmis-sion links have been achieved. Link simulations show good agreement with the experi-ments.

Fig. 1: WDM loop testbed withall-distributed Ramanamplification (10 Gb/sWDM transmission system)

Fig. 2: BER for the 10 Gb/s NRZtransmission over 1000km to 4000 km

TransmitterTx

Switch

3 dB -coupler

SwitchFilter

Pump

SMF 284 km

DCF 37 km

DCF 18 km

SMF 187 km

DCF 213 km

30 %

20 %

20 %

30 %Filter

ReceiverRx

Powersplitter

ReceiverRx

LD1447 nm

- 40 - 36 - 28- 32Received power [dBm]

1E-10

1E-8

1E-6

1E-4

BE

R

Btb4000 km3000 km2000 km1000 km

54

Clock recovery following a transmission linkis an essential function for subsequent signalprocessing functions. With ever increasingdata rates, optical signal processing and opti-cal clock recovery covering the 40 to160 GHz range is of current interest, e.g. forfull optical 3R regeneration (Re-amplification,Re-timing, Re-shaping) [1]. A novel type ofself-pulsating laser, the PhaseCOMB (PhaseControlled Optical Mode Beating), has anextremely high speed potential beyond160 GHz, as predicted by modelling [2]. ThePhaseCOMB laser is based on a 1.55 µmmulti-section DFB laser structure. It consistsof two DFB sections (detuned by about thestop bandwidth) and an integrated phasetuning section. Self-pulsations are generatedwith frequencies defined by the detuning ofthe gratings and by the driving dc currents.For all-optical clock recovery the data signalsare simply injected into the device, whoseself-pulsation frequency is adjusted close tothe data rate. The synchronisation is insensi-tive to wavelength and polarisation, and thegood system performance of this clock re-covery device has been demonstrated at40 Gb/s.

For 80 GHz applications a new PhaseCOMBlaser was fabricated. The device is packagedinto a module for testing the system perfor-mance in a transmission experiment over160 km (Fig. 1). The data signal was gener-ated using a mode-locked laser emittingtransform-limited pulses with a width of1.1 ps at a repetition rate of 9.953 GHz anda wavelength of 1550 nm. A PRBS signal wasencoded onto RZ pulses using a LiNbO3

modulator and passively multiplexed into a80 Gb/s PRBS of length 27–1. It was launchedinto a fibre link consisting of 160 km of a sin-gle mode fiber and a matched length of dis-persion compensating fiber. At the receiverside the signal was injected into thePhaseCOMB laser. The output pulse trace ofthe recovered clock was analysed using anultra fast photodetector (u2t) and a 50 GHzRF analyser (Agilent), upgraded in band-width by using a selected mixer. Figure 2(left) shows the how the free running self-

pulsation synchronises to the data rate with anarrowed line width, indicating the lockingfunction. The synchronised clock pulse tracemeasured using an oscilloscope is shown inFig. 2 (right).

The most important parameter of a clock isthe timing jitter, which is analysed via phasenoise measurements with the RF analyser.The jitter analysis was performed for boththe optical clock and the data source in or-der to evaluate the measuring system and tocompare the jitter values. The rms jitter was238 fs for the data signal from the mode-locked laser and 282 fs for the recoveredclock, which is only a slight increase. Themode-locked laser has been already been ap-plied successfully in transmission experi-ments up to 160 Gb/s and thus represents avery high quality reference standard. Thesimilar jitter value from the 80 GHz opticalclock demonstrates an excellent system per-formance.

[1] B. Sartorius, OFC 2001, paper MG7,Anaheim, Cal., USA, March 2001.

[2] M. Möhrle, IEEE J. Sel. Topics QE, vol.7, no.2, March/April 2001, pp. 217-223.

Carsten Bornholdt ([email protected])


Application of 80 GHz All-Optical Clock Recovery in a 160 kmTransmission Experiment

Clock recovery is a key function for signal processing in OTDM systems, which have a the-oretically predicted speed potential of 160 Gb/s, and even higher for self-pulsatingPhaseCOMB lasers. A first optical clock recovery device is fabricated and tested in an80 Gb/s transmission experiment over 160 km. Jitter of less then 300 fs is measured, indi-cating the good system performance of the optical clock.

Fig .1: Setup for the 160 km

transmission experiment

Fig. 2: Analysis of locking

function: RF spectra(left) and pulse trace

(right), triggered by thedata signal

Transmitter Clock recovery160 kmSMF + DCF

80 Gbit/s

2 - 1 PRBS7

Phase COMB

80 GHz

79.6

locked atRZ PRBS

freeoscillation

79.8 80Frequency [GHz]

- 20

- 40

- 60

Amplitude [dBm]


Monolithically integrated interferometersrequire various types of optical elements:high gain SOAs, low-loss compact bends andcouplers, efficient couplers with small reflec-tions between active and passive sections,and low-loss fibre couplers. These have beenrealised by using buried hetero-structure am-plifiers with butt-coupled passive buriedwaveguides, leading to the worlds smallestintegrated MZI switch with dimensions 4.5 x1 mm2 (Fig. 1).

The MZI is used with a differential controlscheme, in which both amplifiers are saturat-ed separately with a controllable time delay,resulting in an adjustable switching window.

Key components of the MZI are the SOAsin a ‘band gap shifted’ (GS) configurationwith a gain maximum at a wavelength of1.5 µm, which is far from the data, which arein the 1550 nm wavelength range [1].

This configuration gives high phase shiftefficiency, good contrast ratio and good lin-earity. Figure 2 shows experimental resultsfor the switched output power as a functionof the data input power. Linear dependenceis observed over a range of 20 dB, which islimited only by the amplified spontaneousemission at low input powers and by gainsaturation of the SOAs at high input powers.The switching performance of the all-opticalswitch in the linear range is independent ofthe data input power.

The switching performance of the MZIswitch could be demonstrated in error-free160 to 10 Gbit/s all-optical demultiplexingexperiments for all channels (Fig. 3). MZI de-vice modules have been realised in coopera-tion with the packaging group of the HHI,and have been used in system testbeds atAlcatel and HHI.

Due to its potential, the monolithically in-tegrated MZI will play an important role as abasic building block in advanced all-opticalsignal processing functions such as decisions,comparisons, correlations, sampling and sig-nal conversion. One main future objectivewill be the application of the MZI switch as asynchronous modulator in a soliton support-ed 3R regeneration scheme at clock frequen-cies up to 160 GHz.

[1] T. Tekin, M. Schlak, W. Brinker, B. Mauland R. Molt, Proc. of 26th Europ. Conf. Opt.Commun., vol. 3, pp. 123-124, September3-7, 2000.

Michael Schlak ([email protected])

Monolithically Integrated All-Optical Demultiplexer for 160 to 10 Gbit/s

Monolithically integrated Mach-Zehnder interferometers (MZI) will be essential for ad-vanced all-optical signal processing in the ultrafast optical communication networks thatwill be used by the coming information society. Error-free all-optical demultiplexing from160 Gbit/s to 10 Gbit/s is performed here using a MZI with monolithically integratedband-gap-shifted semiconductor optical amplifiers (GS-SOAs).

Fig. 1: The world’s smallest all-optical demultiplexer: A symmetric Mach-Zehnder interferometerwith monolithically inte-grated band-gap-shiftedsemiconductor opticalamplifiers

Fig. 3: Error-free all-optical demultiplexing perfor-mance with a GS-MZI

Fig. 2: Linearity of monolithical-ly integrated MZI

- 20 - 10 0 + 10Data input power [dBm]

- 20

- 30

- 40

Dat

a ou

tput

pow

er [d

Bm

]

- 40 - 35 - 30 - 25Receiver input power [dBm]

4

6

10

8

- lo

g (B

ER

)

back to back80 to 10 Gbit/s160 to 10 Gbit/s

56

In-service quality monitoring of high bit-rate optical transmission systems can beachieved at the full line rate by using opticalsampling systems. These systems provide thepicosecond timing resolution needed to re-cord eye diagrams of 160 Gb/s data signals.

Figure 1 shows the experimental setup ofthe transmission experiment. A 160 Gb/s op-tical RZ data signal is transmitted over an80 km unrepeatered fibre span of standardsingle mode fibre (SSMF). The optical sam-pling system, which is used to measure theeye diagram of the optical data signal, isshown in the enlarged detail.

Part of the signal is used for clock recoveryat the input of the sampling system. Theclock recovery device contains an optical PLLwith a semiconductor laser amplifier in anoptical loop mirror (SLALOM) gate as an ul-trafast optical mixer [1]. The extracted10 GHz base rate of the optical time-divisionmultiplexed (OTDM) data signal is frequencyshifted by ∆f = 10 kHz to the frequency f1 us-ing the RF carrier-suppressed single-sidebandmodulation technique. The sampling pulsesource contains a tunable mode-locked semi-conductor laser (TMLL). It operates at thepulse repetition rate f1 and delivers 1.5 pssech2 pulses with an rms timing jitter of0.3 ps (λ = 1560 nm). The sampling pulserate is reduced to 310 MHz by a LiNbO3

Mach-Zehnder modulator (MOD) driven bya frequency divider (PPG, pulse pattern gen-erator) because of the limited clock speed ofthe digital oscilloscope. The sampling gate isa nonlinear optical loop mirror (NOLM) con-

taining 320 m of highly nonlinear fibre(HNLF, λ0 = 1547 nm) [2]. The average inputpowers of the sampling pulses and data sig-nal are about –5 dBm and +14 dBm, respec-tively. The switched output signal of theNOLM (i.e. the optical samples) is fed into alow bandwidth receiver. The peak powers ofthe optical samples are detected and dis-played as an eye diagram on a digital oscillo-scope with an external clock input.

Figure 2 shows a typical 160 Gb/s eye dia-gram measured by the optical sampling sys-tem. Due to the excellent timing resolutionof the system, the different channels in thedata signal are clearly resolved. The switch-ing window of the sampling gate determinesthe shape of the eye diagram. Thus there isno visible ringing. The use of a clock recov-ery device in the sampling system gives per-sistence times that are large enough to en-able the collection of sufficient statistical datafor a quantitative evaluation of the eye dia-gram, by Q factors for example.

[1] T. Yamamoto et al., “Clock recoveryfrom 160 Gbit/s data signals using a phase-locked loop”, Electron. Lett., 37(8), 509-510,(2001).

[2] S. Watanabe et al., “All-optical signalprocessing using highly-nonlinear opticalfibers”, IEICE Trans. Electron., E84-C(5), pp.553, (2001).

Carsten Schmidt([email protected])


Optical Sampling of 160 Gb/s Eye Diagrams in an 80 km TransmissionExperiment

We measured 160 Gb/s eye diagrams after 80 km fibre transmission using a complete op-tical sampling system. The system comprises an all-optical clock recovery device and asampling gate based on a nonlinear optical loop mirror (NOLM).

Fig. 1: Schematic of the opticalsampling system and its

application for eye diagram monitoring inan 80 km transmission

experiment

Fig. 2: Measured eye diagram

of an optical RZ data signal at 160 Gb/s

10 Gb/sRx

Data signal INPUTReceiver

Sampling Pulse Source

Clock Recovery

Sampling Gate

electr. amp.DC - 1.1 GHz

MOD

300 mHNLF

NOLM

Trigger : f∆

Clock : f / 321

f / 321f1

λ = 1560 nm

DigitalOscillosope

f0

λdata

λdata

DEMUX160 Gb/s10 Gb/s

OpticalSamplingSystem

MUX10 Gb/s160 Gb/s

10 Gb/sTx

PPG

pulse sourcebased

on TMLL

ff =f + f

0

1 0 ∆

opticalPLL

SS

MF

Time [1.6 ps / division]

Am

plitu

de [a

.u.]

6.25 ps


Clock recovery is an essential operation inboth receivers and regenerators in transmis-sion systems. We succeeded in developing aPLL-based clock recovery device for 160 Gb/sdata signals using a semiconductor laser am-plifier in a loop mirror (SLALOM) as the opti-cal phase comparator [1]. We examined itsperformance in a 160 Gbit/s OTDM demulti-plexer in a fibre transmission experimentover a link length of 120 km.

Figure 1 shows the setup, including theclock recovery device, of the demultiplexerfrom a 160 Gbit/s OTDM signal to a 10 Gbit/ssignal. In the clock recovery unit, the SLALOMacts as a fast phase comparator between the160 Gbit/s data signal and the 10 GHz opti-cal clock pulses generated locally by a tun-able mode-locked laser (TMLL). The feed-back loop locks the frequency of the opticalclock to the base-rate frequency of the datasignal. The timing jitter of the recovered op-tical clock was less than 0.3 ps. Owing to thetwo Erbium-doped fiber amplifiers (EDFAs) infront of the clock recovery unit, whichworked in the saturated region, the clock re-covery device locked to 160 Gbit/s data sig-nals over the wide power range -30 dBm to+5 dBm at the input of the demultiplexer,without readjustment of the unit. The opticalswitch for demultiplexing was a gain-trans-parent ultrafast nonlinear interferometer (GT-UNI) – i.e. the SOA in the GT-UNI had itsgain peak in the 1300 nm wavelength range,so that it was transparent for the data sig-

nals. The switching contrast was more than20 dB. The width of the switching windowwas about 5 ps.

We confirmed that the PLL did not cause apower penalty in the demultiplexer, and suc-ceeded in error-free transmission over a 120km fibre (see Fig. 2). This demultiplexer isthus a promising candidate for future ultra-high speed OTDM systems.

[1] T. Yamamoto, L.K. Oxenløwe, C.Schmidt, C. Schubert, E. Hilliger, U. Feiste, J.Berger, R. Ludwig and H. G. Weber, “Clockrecovery from 160 Gb/s data signals using aphase-locked loop”, Electron. Lett., vol. 37,no. 8, pp. 509-510, 2001.

Hans-Georg Weber ([email protected])

160 Gbit/s Optical Clock Recovery Device and its Application in anOptical Demultiplexer

We report on an optical clock recovery device, based on an optical phase-locked loop(PLL), for 160 Gbit/s optical time-division multiplexed (OTDM) data signals. This devicewas successfully operated in a 160 to 10 Gbit/s optical demultiplexer in a 120 km fibretransmission experiment.

GT - UNI switch

DelayPLL with SLALOMclock recovery unit

10 Gbit/s signal

160 Gbit/s signal = 1550 nmλ

PC

Amp. LPF

VCO

PBS

PMF

λ = 1540 nm

λ =1540 nm

PDAtt.

PD

TMLL

TMLL

SOA SO

A

+-

120 km TrueWave fiber

received power

160 Gbit/sreceiver

OTDMdemultiplexer

post -compensation

160 Gbit/stransmitter

EDFA

- 280 ps /nmEHSDK

6 nm

- 280 ps /nmEHSDK

2.3 kmSMF

dispersion compensator

EDFA EDFA

Fig. 2: Setup of the 160 Gbit/stransmission experimentusing a demultiplexerwith clock recovery device

Fig. 1: Setup of the 160 Gbit/sdemultiplexer with clockrecovery device usingtwo SOA-based interfer-ometric switches

58

Polarization-mode dispersion causes severesignal impairments in high bit-rate opticaltransmission systems. For bitrates higher than40 Gbit/s automatic PMD compensation isrequired. A critical issue is the response timeof the differential group delay (DGD) ele-ment of the PMD compensator. The groupat the Technical University Hamburg-Harburg applied open loop feedforward con-trol of the compensator, in which the linkDGD is continuously measured and fed for-ward to a variable DGD element in order toset the appropriate DGD of the compensatorin a single step. Continuous DGD monitoringwas realized by a polarization-resolved mea-surement of the degree of polarization (DOP)combined with a polarization scrambler atthe fibre input. This allowed the DGD of thelink to be continuously monitored, and thedependence on the input polarization waseliminated.

Using this PMD compensator, we per-formed 80 Gbit/s RZ data transmission overa PMD-compensated installed fibre link inthe city of Berlin of 45 km length with an av-erage DGD of 6 ps [1]. The PMD compen-sator comprised two controllers, one for themotor-driven variable DGD element and onefor the polarization controller, which wasbased on piezo-driven fibre squeezers. Thevariable DGD element was controlled bycontinuously measuring the minimum DOPat the input of the delay line. A stored re-sponse curve was used to calculate the linkDGD. This value directly controlled the DGDof the compensator in the feedforward struc-ture. The second controller used the mea-sured DOP at the delay line output for feed-back control.

The figure shows the bit-error rate (BER)performance of the system with patternlength 27-1.

At the operating wavelength the instanta-neous DGD was about 3.7 ps. Without PMDcompensation, an error floor occurs at a BERof 10-6 and error-free transmission was notpossible.With PMD compensation, we ob-tained error-free transmission with a penaltyof less than 1 dB, compared to the back-to-

back measurements. To our knowledge, thisis the first demonstration of a PMD compen-sator at 80 Gbit/s.

We have demonstrated that a feedforwardapproach for PMD compensation avoids theneed for dithering and allows the compen-sator DGD to be adjusted in a single step.DOP evaluation combined with polarizationscrambling provides access to link parame-ters such as DGD and principles states of po-larization.

[1] H. Rosenfeldt, R. Ulrich, E. Brinkmeyer,U. Feiste, C. Schubert, J. Berger, R. Ludwig,H. G. Weber and A. Ehrhardt, “Feed-forwardapproach for automatic PMD-compensationat 80 Gbit/s over 45 km installed singlemode fiber”, 27th European Conference onOptical Communication (ECOC), 30 Sep – 4Oct 2001, post-deadline.

Hans-Georg Weber ([email protected])


Automatic Polarization-Mode Dispersion Compensation for 80 Gbit/sData Transmission over Installed Single Mode Fibre

We report on 80 Gbit/s transmission over 45 km of installed fibre using automatic polar-ization-mode dispersion (PMD) compensation. To our knowledge, this is the first demon-stration of a PMD compensator at this speed. This work is a cooperation between HHIand the Technical University Hamburg-Harburg FSP 2-03 Optical Communications (E.Brinkmeyer, H. Rosenfeldt and R. Ulrich).

BER performance

- 35 - 30 - 25 - 15 - 10- 20Receiver input power [dBm]

- 9

- 7

- 3

- 5

log

(BE

R)

back to backcompensateduncompensated


Over the past decade, as the demand fortelecommunication services and bandwidthhas boomed, the advantages of external mod-ulation in fibre-optic transmission systemshave been firmly established. Indium phos-phide (InP) modulators based on a Mach-Zehnder interferometer (MZI) structure witha travelling wave electrode (TWE) design cangive high data rates of 40 Gbit/s and be-yond, as well as zero-chirp designs and anoptical bandwidth of 30 nm. In addition, dueto the highly efficient quantum-confinedStark effect (QCSE) used by the modulator,its structure is much smaller (< 3 mm2) andneeds less driving voltage (VPP ~ 2V) thancomparable devices made with GaAs or lithi-um niobate. This inherently low driving volt-age is one of the key issues for future trans-mission rates of 80 Gbit/s and higher.

The schematic layout of the modulator isshown in Fig. 1, which shows the two TWEelectrodes, which are designed as microstriplines. Their overall impedance can bematched to 50 Ω by the capacitive load ofthe distributed electrodes on the MZ arms.So far an electrical bandwidth of 50 GHz hasbeen achieved [1] with this design, but fur-ther improvements up to modulation speedsof 160 Gbit/s should be possible in the nearfuture.

The cross section of an MZI is shown inFig. 2. The layers with various material se-quences and the electrical equivalent circuitare indicated schematically. The reverse-bi-ased PN junction of the MZI is shown as thecapacitive load of the TWE electrode.

In the scanning electron microscope (SEM)photo of Fig. 3 one can see parts of this TWEelectrode and the MZI waveguides (WG)with two active sections, including wave-

guide electrodes and connecting strips as airbridges spanning the isolation ditches.

The optical design of this modulator alsoincludes an optical mode transformer forcleaved fibre coupling with low insertionloss.

This 40 Gbit/s modulator concept is nowbeing developed as an industrial processwith good reproducibility. The manufactur-ing partner for this work is u2t PhotonicsGmbH.

Future developments will include a hybridintegration of the modulators together withtheir driver electronics. This will be the basisof new TWE and modulator designs thatcould achieve transmission rates up to 80-160 Gbit/s.

This work was supported by the State ofBerlin (874700001) and the Federal Ministryof Education and Research under grant 01AK 936B.

[1] L. Mörl, D. Hoffmann, K. Matzen, C.Bornholdt, G. G. Mekonnen and F. Reier,“Travelling wave electrodes for 50 GHz oper-ation of opto-electronic devices based onInP”, Proc. IPRM, 1999, pp. 385-388.

Karl-Otto Velthaus ([email protected])

Development of InP Modulators Based on a Travelling Wave ElectrodeDesign for 40 Gbit/s and Beyond

An InP-based modulator with an electrical bandwidth above 50 GHz has been developed.Current research activities are directed towards an industrial process with good repro-ducibility and high yield. Currently under development is the integration in hybrid formof a driver IC into a modulator module.

Fig. 1: Schematic layout of anMZI-based modulatorwith capacitively loadedtravelling wave electrodes

Fig. 2: Cross-sectional view ofan MZI and the TWEswith schematic electricalequivalent circuit

Fig. 3: SEM Photo showingparts of the TWE electrode; a pair of airbridge connections between TWEs, and twowaveguide diodes

MMI

activeWGRF input 50 Ω

Z = 50 Ω Z = 50 ΩZ = 65

without loadingΩ

input

Y -branch

output

B

A

semi - insulatingInP substrate

TWEMZI

MQW stack N - doped P - doped

air - bridge metal TWE

60

The availability of III/V materials with ultra-fast optical response times and large opticalnonlinearities is crucial for future ultrahighbitrate fibre-based communication systems.To this end various types of optical recombi-nation scenarios may be exploited, one ofwhich, conventional interband transitions indefect-rich bulk or low-dimensional III/V ma-terials, is regarded as attractive. These mate-rials can be achieved with high quality by us-ing molecular beam epitaxy at low growthtemperatures (LT), down to 100 °C (cf. HHIannual report 1999). In contrast to the well-established LT GaAs/AlAs material system, theabsorption edge of LT GaInAs/AlInAs multi-ple quantum wells (MQWs) can be adjustedto the 1.55 µm wavelength range used inoptical fibre transmission.

Uniform p-doping using Be was demon-strated recently as a method for reducing theoptical response times in GaInAs/AlInAsMQWs. Additionally, annealing was consid-ered relevant as for GaAs-based structuresbecause of the formation of clusters of excessAs. This effect was found also to occur inGaInAs/AlInAs MQWs, as demonstrated bythe transmission electron microscopy (TEM)pictures in Fig. 1.

Significant differences have been found be-tween the absorption recovery times of as-grown and annealed samples. The normal-ized change of transmission of annealedsamples decays with a time constant of ap-proximately 1.5 ps, whereas as-grown sam-ples decay with the much smaller time con-stant of 230 fs, which reflects the decrease ofthe excess carrier density due to trapping

from conduction and valence band states in-to localized defect levels.

Pump-probe experiments with pairs ofpump pulses with widths of 140 fs and sepa-rated by 1.5 ps were carried out to simulateoptical switching conditions at very high da-ta repetition rates. As shown in Fig. 2, the se-quence of these pump pulses is clearly re-solved only in the case of the as-grown Be-doped sample. In the annealed Be-dopedsample we observe an accumulation of carri-ers during the second pump pulse and asubstantially slower decay of the transmissionchange after the second pump pulse.

Long-term accumulation effects after longpulse sequences are a problem in opticalswitches. To further study such phenomena,we performed experiments with simultane-ous pulsed and cw excitation at 50 mW inci-dent average power. The annealed Be-dopedsample clearly shows accumulation effects,whereas such effects are virtually absent inthe as-grown Be-doped sample.

In conclusion, as-grown Be-doped LTGaInAs/AlInAs MQW material is consideredto be a highly promising material for all-opti-cal switching, even in the Tbit/s domain.

This work was performed as a joint re-search effort with the Max-Born-Institut,Berlin, and the Humboldt University ofBerlin.

Harald Künzel ([email protected])


Ultrafast Optical Nonlinearity at Wavelengths around 1.55 µm ofGaInAs/AlInAs Quantum Wells Grown at Low Temperatures

The ultrafast optical response of GaInAs/AlInAs multiple quantum wells grown at lowtemperatures was studied in pump-probe experiments. Experiments with pairs of ultra-short pulses separated by 1.5 ps demonstrate fast modulation of material transmissionand very small accumulation effects.

Fig. 2: Ultrafast transmission

changes after excitationby a pair of femtosecond

pulses separated by

1.5 ps in time

Fig. 1: TEM pictures of the

cleavage plane of 100period GaInAs/AlInAs

MQWs grown at 200°C

0- 2 + 2 + 4 + 6

annealed

as - grown

Time delay [ps]

Tra

nsm

issi

on c

hang

e [a

.u.]

annealed (ex - situ):30 min @ 700° C

as grown:

(dark field image) (bright field image)


For compact integration of 40 Gbit/s opti-cal front ends it is desirable to have a distor-tion-free and robust detector type that effec-tively suppresses all electromagnetically cou-pled interference signals from digital circuitrymounted in the same housing. Our ap-proach is an integrated twin-photodetector,comprising a tapered spot-size converter forefficient fibre to chip coupling, a multimodeinterference coupler for optical 3 dB powersplitting, and two waveguide-fed evanes-cently coupled high-speed p-i-n photodiodesdelivering two symmetrical electrical signalswith inverse polarities to the differential in-puts of a subsequent electronic circuit.

For the 40 Gbit/s field trials, this photonicintegrated circuit was packaged robustly bymounting the photodetector on a ceramicsubstrate and connecting it with bondingwires to the coplanar stripe lines leading tothe two V-connector output ports. A stan-dard single-mode fibre was coupled to theTiO2-coated input facet of the chip (Fig. 1),which ensures an optical return loss betterthan 27 dB for the module.

Excellent electrical and optical symmetry,necessary for differential operation, wasachieved because of our compact integrationconcept. The ratio of the photocurrents ofthe two photodiodes in the detector is lessthan 0.5 dB for all states of polarization. Thetotal responsivity of the photodetector is0.41 A/W, with a polarization-dependent lossof only 0.5 dB.

The high-frequency behaviour of the mod-ule was studied by injecting high power opti-cal picosecond pulses. Figure 2 shows the re-ceived symmetric electrical pulses from the

two photodiodes (pulse width 14 ps). Thesehave 0.5 V peak voltages, which are largeenough to drive the demultiplexer stage di-rectly. The 3 dB bandwidth of the twin-pho-todetector module was determined to be45 GHz with an optical heterodyne setup [1].

Eye diagram measurements were per-formed by the industry partner and showedclearly opened eyes (Fig. 3). The detectormodules were successfully employed in theKomNet project in a 43 Gbit/s (forward errorcorrection) field trial.

[1] A. Beling, D. Schmidt, H.-G. Bach, G.G. Mekonnen, R. Ziegler, V. Eisner, M.Stollberg, G. Jacumeit, E. Gottwald, C.-J.Weiske, A. Umbach, “High power 1550 nmtwin-photodetector modules with 45 GHzbandwidth based on InP”, Proc. Optical FiberCommun. (OFC 2002) , March 17-22, 2002,Anaheim, CA, USA, paper WN4.

Andreas Beling ([email protected])

High-Power 1.55 µm Twin-Photodetector Modules for 40 Gbit/s TDMField Trials

Monolithically integrated twin-photodetectors based on InP were fabricated and mount-ed into butt-coupled modules. The detectors, with large bandwidths and high optical andelectrical symmetry, were successfully used in a high-speed field trial.

Fig. 1: Top view of the twin-photodetector moduleFig. 3: 43 Gbit/s eye diagramsof both photodiodes

Fig. 2: Pulse responses of thephotodiodes PD 1 andPD 2 in the detectormodule

0 50 150

PD 1

PD 2

100Time [ps]

- 0.4

0.0

+ 0.4

Pul

se a

mpl

itude

[V]

62

The upgrading of high-speed fibre trans-mission systems to a bit rate of 40 Gbit/s isnow in progress. The need for forward errorcorrection (FEC) in data traffic and the de-mand for faster transmission will require afurther extension of the bit rate in eachchannel. Broadband high-speed photoreceiv-ers are one of the key elements in presentand future optical communication systems.This work focuses on the robust and cost ef-fective packaging of single-chip photoreceiv-er modules with 50 GHz bandwidths.

The ultra-broadband photoreceiver con-sists of a wave-guide integrated photodiodewith a passive optical chip waveguide and atravelling wave amplifier (TWA) using high-electron mobility transistors (HEMT). The ad-dition of an integrated spot-size converterand a redesign of the distributed amplifiercircuit improved the receiver bandwidth to50 GHz (Fig. 2). The overall opto-electricconversion efficiency also increased to1.1 A/W and the amplifier gain flatness im-proved to ± 0.8 dB [1]. The transimpedanceof the redesigned amplifier, consisting of 4HEMTs (with power consumption less than100 mW), is 110 Ω. For antireflection pur-poses when using a standard single-mode fi-bre, the optical input facet of the photore-ceiver OEIC was coated with TiO2/SiO2.Optical coupling is done by fixing the butt fi-bre directly at the tapered waveguide facetof the OEIC with UV-curable resin.

The optical return loss was better than30 dB. This coupling type has high robust-ness against vibration and thermal cycling

under field operating conditions. The cou-pling efficiency, measured in terms of the re-sponsivity of about 0.6 A/W, remained un-changed after a 10 minute vibration test at15 g acceleration. The bandwidth of thepackaged OEIC is still as high as for the chipalone (50 GHz), and the low ripple behav-iour is also unaffected (Fig. 2). The moduleswere supplied to industry partners in theKomNet program for system characteriza-tion. Figure 3 shows a widely opened eyepattern of this photoreceiver in a 43 Gbit/selectrical time division multiplex (ETDM) testbed at Alcatel SEL AG.

[1] H.-G. Bach et al., “50 GHz photoreceiv-er modules for RZ and NRZ modulation for-mat comprising InP-OEICs”, Proc. 27th Euro-pean Conf. On Optical Communication(ECOC2001), 30 September – 4 October2001, Amsterdam, The Netherlands, paperTh. M.2.5.

Heinz-Gunter Bach ([email protected])


50 GHz Photoreceiver Modules With InP OEICs for FEC 40 Gbit/s Systems

Photoreceivers are key components in high bitrate telecommunication systems. This workdescribes an InP-based pin diode with travelling wave amplifier (pinTWA) photoreceiveroptoelectronic integrated circuit (OEIC) with a monolithically integrated spot-size con-verter, including its packaging and module performance. Flat gain characteristics, a band-width extended to 50 GHz and a fully open eye pattern at 43 Gbit/s are achieved.

Fig. 1: 50 GHz single-chip

pigtailed photoreceivermodule with a pinTWA

InP receiver OEIC

Fig. 3: Measured eye pattern ofa photoreceiver moduleat 42.6 Gbit/s with NRZ

signalling. x: 5 ps/div, y: 20 mV/div.

Fig. 2: Optical heterodyne

measurement of the op-to-electric power trans-

fer characteristics of thephotoreceiver OEIC and

of the receiver modulewith the same OEIC

0 20 30 40 5010Frequency [GHz]

- 12

- 9

- 6

- 3

0

+ 3

RF

pow

er [n

orm

. dB

]e

l

OEICmodule C04 #D


In this project, InGaAsP-BH-FP lasers with atapered active stripe have been developedand optimised. The investigations focussedon the realisation of lasers with cavity lengthssmaller than 500 µm in order to achieve alarge number of devices per processed wafer.A second requirement was that the lasersshould be able to operate without an addi-tional high-reflection coating of the backfacet.

To meet these specifications, the activelayer, the waveguide design and the lengthof the lasers have been optimised [1].

The active layer consists of 6 compressivelystrained InGaAsP quantum wells separatedby tensilely strained barriers. The FP lasersare processed using three MOVPE epitaxyprocessing steps. Current blocking is a-chieved using conventional pnp InP blockinglayers. The length of the optimised lasers is400 µm.

Upside-up mounted devices have thresh-old currents of about 6 mA and more than40 mW CW output power (Fig. 1). It shouldbe noted that a facet coating was not ap-plied.

At 90 °C the threshold current is only about32 mA, but more than 10 mW output poweris still achieved at the front facet. These de-vices are hence suitable for uncooled opera-tion.

The devices have virtually circular FWHMfar field angles of about 17 ° laterally and 18 °vertically. Figure 2 shows the fibre-coupledoptical powers of tapered and untaperedlasers with coupling to a flat-end single-mode fibre.

The tapered devices achieve 10 mW cou-pled optical power. The coupling loss of thedevices to a flat-end single-mode fibre is-3.5 dB.

The reliability of these devices was con-firmed with ageing tests at 85 °C and withan operating current of 100 mA (Fig. 3).

Investigations are now concentrating onoptimising the devices for direct modulationat 10 Gb/s.

[1] M. Möhrle, A. Sigmund, A. Suna, H.Roehle, L. Mörl and F. Reier, “High perfor-mance all-active tapered 1550 nm InGaAsP-BH-FP lasers”, European Semiconductor LaserWorkshop, Ghent, Belgium, 28-29 Septem-ber 2001.

Martin Möhrle ([email protected])

Tapered 1550 nm InGaAsP-BH-FP Lasers

Large spot-size laser diodes are required to realise low cost optical transmitter modules.The all-active tapered buried heterostructure (BH) FP lasers developed here combine lowthreshold currents and high output powers with low fibre coupling losses.

0 100 150 200 25050Current [mA]

- 12

10

20

30

40

50

CW

opt

ical

pow

er [m

W]

T = 20 °CL = 0.4 mmas - cleaved

Fig. 3: Ageing characteris-tics of 4 taperedlasers

Fig. 1: CW optical outputpower at front facet

Fig. 2: Optical power cou-pled to a flat-endsingle-mode fibre

0 1000 1500 2000500Ageing time [h]

0.2

0.6

1.0

Rel

. opt

ical

pow

er [a

.u.]

Ageing conditions :85 °C 100 mA

4 Lasers

0 40 60

L = 0.4 mml = 100 mAas - cleaved

80 10020Distance fiber - facet [µm]

0

4

8

12

Fib

er c

oupl

ed o

ptic

al p

ower

[mW

]

tapered laseruntapered laser

64

Multiple-wavelength sources are key com-ponents of dense wavelength-division multi-plexed communication systems. The twomain concepts are tunable lasers and laserarrays. Laser arrays require less complex pro-cessing, but for statistical reasons an extraor-dinarily high individual laser yield is requiredfor an adequate array yield. In our project wehave applied the phase-shifted distributed-feedback (DFB) ridge waveguide (RW) laserdesign with index grating [1], which givesgood accuracy for the individual wave-lengths. The specifications for the 4-λ arraysunder investigation are: 400 GHz line spac-ing, single-mode operation with a side-modesuppression ratio (SMSR) better than 30 dB,and a minimum dc power output of 10 mWat 100 mA driving current.

In a first working package the single laseryield for λB/4-shifted gratings (where λB is theBragg wavelength) at various couplingstrengths κL has been optimised (where κ isthe coupling coefficient and L the laserlength). Figure 1b shows the high repro-ducibility of our grating process. Typical ar-ray spectra with very good single-mode be-haviour are shown in Fig. 1c. Based on the

characterization of more than 600 arrays, thedependence of the individual laser yield ofthe λB/4-shifted lasers on the couplingstrength κL is given by the open circles inFig. 2. The distinct optimum value of κL liesbetween 1.2 and 1.5. Away from this opti-mum the DFB laser yield decreases either be-cause the threshold current increases and theSMSR becomes more sensitive to imperfec-tions (on the lower coupling side), or be-cause, in the stronger coupling regime, thelasers start to degrade due to longitudinalspatial hole burning (LSHB).

As well as the specifications describedabove, low feedback sensitivity is also re-quired in system applications. One way toachieve this is to increase κL. To maintain thehigh yield for κL values above 1.5 we havereplaced the λB/4 phase shift by two tailoredλB/8 phase shifts. For suitable separation dis-tances (17-30 µm) the LSHB effects could bereduced, giving increased yield compared tothe standard λB/4-shifted lasers (Fig. 2, filledcircles).

[1] J. Kreissl, U. Troppenz, W. Rehbein, B.Hüttl, E. Lenz, H. Venghaus, F. Fidorra, “Highsingle-mode-yield multiple-wavelength DFBlaser arrays in the 1.55 µm range”, Proc.ECOC 2001, vol. 2, pp. 118-119.

Jochen Kreissl ([email protected])


Trade-Off Between High Single-Mode Yield and Low Feedback Sensitivityin DFB RW Laser Arrays

We have developed a four-wavelength laser array with excellent single-mode operationand high power output. By optimising the coupling strength, the single laser yield couldbe routinely stabilized at about 90 %, giving array yields better than 60 %. The tailoredλB/8 phase shift design also helps to maintain this high yield while at the same time satis-fying the commercial demand for low feedback sensitivity.

Fig. 2: Laser yield versus cou-pling strength for λB/4

and tailored λB/8 phase-shifted grating designs.

The insets show the twokinds of gratings.

2 4 6Reactive ion etching time [min]

40

80

120

κ [c

m]

-1

a)

1.561.54 1.55Wavelength [µm]

- 60

- 40

- 20

0

Opt

ical

pow

er [d

B]

b)

Fig. 1: (a) The linear depen-

dence of the couplingcoefficient k on the dry

etching time; (b) The spectrum from a

four-laser array.

0 1 32κ*L

60

80

100

Sin

gle

lase

r yi

eld

[%]

50 : 50

λ / 4 λ / 8 λ / 8

∆


We present typical results with microringdevices used for the realization of compactoptical filters. Following the developmentstrategy of first learning from passive devices,then improving their performance by Semi-conductor Optical Amplifier (SOA) integra-tion, and finally extending the functionalityto all-active structures, we first fabricated apassive double ring device with add/dropfunctionality (Fig. 1). Two racetrack rings arecoupled by a directional coupler (couplingcoefficient -13 dB). Input and output cou-pling are performed by -3 dB multimode in-terference couplers. The free spectral range(FSR) is 50 GHz (radii 200 µm, ring loss1.6 dB/mm), and the on/off ratio is 18 dB.Platinum resistors (100 Ω) are integrated tofine tune the temperature. We then integrat-ed an SOA section for loss compensation,since passive devices cannot satisfy the re-quirement for strong optical crosstalk sup-pression (> 20 dB) in optical communicationsystems because of their residual losses. Thebest on/off ratio so far is 25 dB.

The response of a typical all-active deviceis shown in Fig. 2. The device can operate ina filter mode (e.g. with IRing = 50 mA) or in acomb laser emission mode (e.g. with IRing =90 mA). In both cases the feeding wave-guide and coupler sections are transparentby a current injection Itr. After antireflectioncoating Itr can be increased so the device canbe used as an inline SOA.

Ongoing investigations are focusing on ex-ploiting nonlinear effects in microrings. Byproper choice of the injection currents in thefeeding waveguide, coupler and ring sectionit is planned to demonstrate all-optical non-linear switches and tunable dispersion com-pensators. It is also planned to realize Wave-length Division Multiplexing (WDM) inter-leaver multiplexers and demultiplexers bycascading rings of different FSRs.

This work was supported by the FederalMinistry of Education and Research undergrant 01 BC 925.

[1] D. Rabus and M. Hamacher, “MMI cou-pled microring resonators on GaInAsP/InP”,IEEE Photon. Technol. Lett., vol. 13, no. 8,pp. 812-814, 2001.

[2] D. Rabus, M. Hamacher and H.Heidrich, “Active and passive microring res-onator filter applications on GaInAsP/InP”,Proc. IPRM 2001, May 2001, Nara, Japan,pp. 477-480, paper ThA1-3, ISBN 0-7803-6700-6.

Helmut Heidrich ([email protected])

Optical Filter and Laser Using GaInAsP/InP Ring Resonators

Network engineers need more components than are available today, especially ones withbetter performance and increased functionality. Key issues are miniaturization and mono-lithically integrated optoelectronic and photonic ICs, in particular for components basedon GaInAsP/InP.

Fig. 1: Add/drop response andphoto of a passive dou-ble ring resonator device

Fig. 2: Typical filter and comblaser response of an all-active ring resonator

Throughputport

Prop port

1.549 1.550 1.551Wavelength [µm]

- 40

- 30

- 20

- 30

- 20

- 10In

sert

ion

loss

[dB

]

ExperimentSimulation

Throughput port

Drop port

1.549 1.550 1.551Wavelength [µm]

- 25

- 20

- 15

- 10

- 5

Tra

nsm

issi

on [d

B] I = 50 mA

I = 90 mARing

Ring

IRing

Itr

66

Dry etched facets are very attractive forsemiconductor laser diode fabrication, sincethey have the potential for on-wafer devicefabrication and thus substantial cost reduc-tion. The chemically assisted ion beam etch-ing technique is commonly used for dryetching, as it offers both vertical sidewalletching and high etch rates. Work in HHI hasconcentrated on the fabrication of InGaAsP/InPridge waveguide (RW) lasers with dry-etchedfacets using a simple photoresist mask.

The mask definition for the facet etching isan important issue, as the masking materialmust be compatible with the overall fabrica-tion scheme and also cost-effective. In ourwork we used a 1.4 µm thick layer of pho-toresist AZ5214, which we patterned usinglithographic exposure. There was no addi-tional treatment. Etch selectivity proved to beabout 35:1.

The MOVPE grown laser structure com-prised a multiple quantum well (MQW) ac-tive region with strained InGaAsP QWs emit-ting at 1.55 µm. Prior to facet etching, 2 µmwide ridges were structured. The photoresistlayer efficiently masks the ridge region de-spite an overall etch depth of about 12 µm,as can be seen in Fig. 1. We could achievedepths of up to 20 µm without any apparentridge damage. We believe that the goodphotoresist mask efficiency is due to a bakingeffect taking place in the CAIBE reactor asthe sample is heated to process temperature.

Comparison of the devices with etched orcleaved facets showed practically identicaloutput characteristics (Ith = 10-15 mA,Pout@100mA = 12-15 mW). In order to ex-plore the possible influence of the CAIBE pro-cess, we systematically evaluated the far and

near field distributions of the emitted laserlight. A typical comparison for etched andcleaved facets is shown in Fig. 2. No de-tectable differences could be observed.

Another issue for the dry etched facets isthe possible accelerated aging of these de-vices. Aging at 85 °C and 150 mA for 660hours showed no substantial difference tocleaved facet lasers, all devices having an es-timated lifetime at 20 °C of more than100,000 hours (Fig. 3).

This work was supported by the LandBerlin and the European Fund for RegionalDevelopment.

Anagnostis Paraskevopoulos([email protected])


Fabrication of InGaAsP/InP Ridge Waveguide Lasers with Dry EtchedFacets

Dry etched laser facets for InGaAsP/InP ridge waveguide (RW) lasers were fabricated us-ing chemically assisted ion beam etching (CAIBE) and a cost-effective photoresist maskwithout additional treatment. The output characteristics, especially the optical field distri-bution, were identical to those of lasers with cleaved facets.

Fig. 2: Comparison of the

measured optical farfield distribution of the

light emitted from acleaved and a dry-etched

laser face.

Fig. 3: Aging results with dry

etched facet RW Lasers.The estimated lifetime is

105 hours at 25 °C.

Fig. 1: SEM photos of a RW

laser with CAIBE etchedlaser facets. The

photoresist mask hasbeen removed.

2 µm

- 1.0 - 0.5 0.0 + 0.5

up(air)

down(header)

+ 1.0Shift [µm]

0.4

0.6

0.8

1.0

Nor

m. f

ield

inte

nsity

nearfieldlat. cleavedvert. cleavedlat. etchedvert. etched

2000 400 600Ageing time [hours]

aging @ 85 °C, I = 100 mAbias

(aging criterium :

20 % output power reduction)

1.0

1.1

0.9

0.7

0.8

Nor

m. o

utpu

t pow

er @

20

°C[I

= 1

50 m

A]

bias


AWGs based on polymer materials havebeen gaining increasing attention becausepolymers have the potential to provideadded device functionality and to offer costbenefits. A serious problem with using AWGsin dense wavelength division multiplex(DWDM) systems is the temperature depen-dence of their channel wavelengths, whichmeans that they require a temperature con-trol unit for thermal stabilisation, an expen-sive overhead. There is therefore an urgentneed to create temperature-insensitive, orpassively athermal, AWGs.

Various solutions have been proposed toachieve temperature insensitivity for silica-based AWGs. However, these methods areeither somewhat susceptible to performancedegradation or somewhat intricate in theiruse of a moving input fibre. We have intro-duced an all-polymer approach [1] for realis-ing athermal AWGs that uses polymer mate-rials for both the waveguide structure andthe substrate. By properly matching the posi-tive coefficient of thermal expansion (CTE) ofthe polymer substrate to the negative ther-mo-optic coefficient of the waveguide mater-ial, athermal and polarisation-independentAWGs can be realised.

The fabrication of our AWG devices in-volves the deposition of the polymer wave-guide layers by spin coating, cross-linking byUV exposure, and structuring by means ofconventional photolithography and reactiveion etching (RIE). The measured optical lossof a polymer channel waveguide fabricatedin this way was less than 0.8 dB/cm in the1.55 µm wavelength region. The crosstalklevel is well below -30 dB, and the insertionloss was between 5.8 dB for the centre portand 7.5 dB for the edge ports. The polarisa-tion shift proved to be less than 0.02 nmwithout any extra compensation. These re-sults prove the all-polymer AWG has a cross-talk equivalent to that of standard silicaAWGs, but superior polarisation dependence(Fig. 1).

To achieve the desired athermal behaviour,a polymer substrate with a CTE of αsub =80 ppm/K was used. This value can be shownto compensate for the thermo-optic coeffi-

cient of the waveguide material, which was-1.16·10-4/K. Figure 2 shows the measuredtemperature dependence of our all-polymerAWG, compared with those of a standard sil-ica-on-silicon AWG and of a polymer-on-sili-con AWG. The all-polymer AWG has a tem-perature-dependent wavelength shift of lessthan ±0.05 nm over the entire temperaturerange between 25 – 65 °C.

[1] N. Keil et al., Electron. Lett., vol. 37,pp. 579-580, 2001.

Norbert Keil ([email protected])

Polymer-Based Athermal Arrayed-Waveguide Grating Multiplexer

An athermal arrayed-waveguide grating (AWG) multiplexer using an all-polymer ap-proach has been realised. By properly adjusting the coefficient of thermal expansion ofthe polymer substrate, athermal and polarisation-independent AWG devices featuring awavelength shift of less than ±0.05 nm in the 25 – 65°C temperature range could bedemonstrated.

Fig. 2: The shift in wavelength

with temperature forAWGs of various compositions

Fig. 1: Measured polarisationshift of an all-polymerAWG

1551 1552

∆λ = 0.02 nm

1553 1554Wavelength [nm]

TE

TM

0

- 10

- 40

- 30

- 20

Opt

ical

pow

er [d

Bm

]

3020 40 50 7060Temperature [°C]

0

- 1

+ 1

- 4polymer - on - silicon

silica - on - silicon

all -polymer

- 3

- 2

Wav

elen

gth

shift

[nm

]

68

Future mobile broadband communicationin the 60 GHz frequency range should offerbeam tracking between transmitter and re-ceiver to ensure high signal-to-noise ratio.This can be achieved by steerable beamform-ing with a phased-array antenna by means ofamplitude and phase control of the individ-ual millimetre-wave signals feeding the an-tenna elements. In our experimental systemthe radio frequency is generated by hetero-dyning two optical waves (of frequencies ν1,ν2) in the 1.5 µm band. This offers the possi-bility of obtaining the desired phase shifts inthe millimetre wave domain, at the beat fre-quency ν2-ν1, by shifting a phase in the opti-cal domain by the same amounts.

We have realised a silica-based PLC for op-tical beamforming that utilizes the thermo-optic effect for phase control. Planar wave-guides of 6 x 6 µm2 cross-section are embed-ded between 15 µm thick silica buffer andcladding layers. Each of the two optical inputwaves in this optical beamforming network(OBFN) is split up by a manifold by y-branch-es that depends on the number of antennaelements, as shown in Fig. 1a. The straightwaveguide sections that follow are coveredby metal heater strips on top of the claddinglayer. A phase shift of ∆ϕ = 2π/λ·(dn/dT·L·∆T)in the optical domain is caused by the ther-mal change of refractive index in silica, withdn/dT = 10-5 K-1. The maximum requiredphase shift for heaters of length L = 10 mm is2π, allowing for the modulo-2π excitation ofthe antenna steering. This is achieved by atemperature change of only ∆T = 10 K. Theheaters are implemented by 30 µm wide and200 nm thick platinum stripes with goldbond pads on both ends.

At the OBFN exit the optical waves arecombined in pairs using 3 dB couplers. Boththe beat frequency ν2-ν1 in the millimetre-wave domain and the mean optical frequen-cy appear at the coupler outputs.

The controllable attenuators consist ofMach-Zehnder Interferometers (MZIs) incor-porating two 3 dB couplers and thermo-op-tic (TO) phase shifters, as shown in Fig. 1b. Amaximum phase shift of π is required for fullsignal suppression. The chip size of the com-plete 4-channel phase shift PLC is about 2 x30 mm2 and that of a 4-channel attenuator

PLC about 2 x 25 mm2. Both componentswere housed and pigtailed and successfullytested in an experimental 60 GHz smart an-tenna system.

Figure 2 shows the obtained phase shiftand attenuation for 60 GHz signals as a func-tion of the heater power.

[1] B. Kuhlow, G. Przyrembel, E. Ehlers, R.Ziegler, G. Großkopf, R. Eggemann and D.Rohde, “Silica based optical beam former ina 60 GHz radio-over-fibre system”, Proc.Seminar on Broadband Communications,Zürich 2002.

Berndt Kuhlow ([email protected])


Silica-Based Beam-Forming Planar Lightwave Circuit (PLC)

A novel silica-based optical signal processor for beamforming in smart antenna systems inthe 60 GHz frequency range is presented. It is comprised of a phase shifter and an attenu-ator PLC network utilizing the thermo-optic effect in silica.

Fig. 1: Layout of 4-channel PLC,

each with 2x4 outputs(MMI: multimode inter-ference). (a) Phase con-

trol PLC, (b) amplitudecontrol PLC.

Fig. 2: Typical measured phase

shift (a) and attenuation(b) at 60 GHz in one

channel of a 4-channelOBFN

4

1ν1

ν2

Heater forphase control

3 dB - Coupler(MMI)

out 4

out1in 1

in 4

a)

b)

4000 800 1200Heater power [mW]a)

400

600

0

200

Pha

se s

hift

[deg

ees]

100 2000 300 400Heater power [mW]b)

- 10

- 20

0

Rel

. RF

- p

ower

[dB

]

69H H I R E P O R T 2 0 0 1 · S E L E C T E D C O N T R I B U T I O N S · M O B I L E B R O A D B A N D …

The main disadvantage of OFDM transmis-sion is its high peak-to-average power ratio(PAPR). If the peak power is limited by a reg-ulatory body, then the average power al-lowed under OFDM is reduced compared toconstant power modulation schemes. Also,the transmit power amplifier must operate ina power-inefficient region. On the otherhand limiting the output back-off causes clip-ping of the transmitted signals, leading tospectral re-growth and increased bit errorrates.

The clipping rate depends on the statisticaldistribution of the PAPR of the OFDM sig-nals. However, the distribution of the PAPR isdifficult to analyse, and so far an accurate es-timation is still not possible. Upper boundswere derived in [1] that can be applied in re-gions where simulations are time consuming.

Another key parameter is the highest pos-sible PAPR, referred to as the maximum PAPR.A method for computing the maximum PAPRfor coded OFDM systems was presented in[3]. It is clear that for standard subcarriermodulation methods, such as BPSK, thesquare root of the maximum PAPR grows lin-early with the number of subcarriers. How-ever, it was shown in [1] that the probabilitythat the maximum PAPR is attained is ex-tremely low. It was further shown that thePAPR of an OFDM signal is with very highprobability of order N log(N) for large sub-carrier numbers N. Therefore the PAPR of asystem should be compared with N log(N)rather than N. If the bounds in [1] are ap-plied, an effective PAPR can be found by as-suming that the probability of an OFDM sig-nal with this PAPR is negligible in practice.

A fundamental result about the estimationof the peak value of OFDM signals from thesamples of complex band-limited signals wasderived in this work [2]. The major result isthat, as long as oversampling is applied, thepeak values of a band-limited signal and itssamples differ by a factor that is upperbounded by a trigonometric expression (seethe figure). Moreover, this bound is sharp forall practical rates.

These results can be used to improve aninteresting approach for lowering the maxi-mum PAPR of codes [3]. Future work will fur-ther investigate the triple code rate, mini-mum distance and PAPR of codes.

This work was supported by the FederalMinistry of Education and Research undergrant FK 01 BU150.

[1] H. Boche and G. Wunder, “On the PA-PR problem in OFDM systems”, 39th AnnualAllerton Conference on Communication,Control, and Computing, October, 2001.

[2] H. Boche and G. Wunder, “Über eineVerallgemeinerung eines Resultats über trigo-nometrische Polynome auf allgemeine band-begrenzte Funktionen”, Zeitschrift für ange-wandte Mathematik und Mechanik, March2002.

[3] G. Wunder and H. Boche, “A basebandmodel for reducing the PAPR in OFDM sys-tems”, 4th ITG Conf. on Source and ChannelCoding, Berlin, January 2002.

Gerhard Wunder ([email protected])

Peak Value Estimation for Multi-Carrier Signals

Orthogonal frequency-division multiplexing (OFDM) communication systems enjoy awell-established history of 30 years of research and development. A major barrier for theirwidespread acceptance in low-cost applications is the high peak-to-average power ratio(PAPR) of the transmitted signals.

Upper bound and computed values of thedifference of the peakvalues of an OFDM sig-nal and its samples, as afunction of the oversam-pling factor L

1 3 75L

2.2

2.6

1.0

1.4

1.8

Pea

k er

ror

70

It has recently been shown that OFDM MI-MO (Multiple Input Multiple Output) sys-tems can increase the capacity of wirelesssystems to tremendous values. However, amajor obstacle to the practical achievementof this capacity is that the radio channelmust be perfectly known at the receiver,which is generally impossible. Moreover, tak-ing into account that future wireless systemsare expected to operate in highly mobile en-vironments, channel estimation can take asignificant part of the communication time,and the system performance will strongly de-pend on the quality of the estimationschemes.

For convenience, the quality of channel es-timation schemes is measured in terms of themean-squared error (MSE) between the realand the estimated channel coefficients. TheMSE measure also permits simple computa-tions to estimate the impact of imperfectchannel estimation on the capacity.

Clearly, the choice of the channel estima-tion scheme is crucial for the system perfor-mance, and various schemes have been pro-posed in the literature. In [1] a pilot signal es-timation scheme was proposed in whichknown OFDM symbols are periodically in-serted into the data stream from each trans-mit antenna, so that the multipath channelcoefficients from each transmit antenna toeach receive antenna can be estimated with-in one OFDM symbol interval. In this schemethe MSE depends on the number of subcarri-ers, representing the degrees of freedom inthe system, the number of transmit anten-nas, the channel delay profiles, and ofcourse, most importantly, the pilot signalsthemselves. It was shown in [1] that, if thenumber of subcarriers is larger than the delayspread (maximum number of taps over allchannel impulse responses) multiplied by thenumber of transmit antennas, then a very ef-ficient pilot signal design is possible. Addi-tionally, the pilot signals have a very lowcrest-factor.

In [1] a broadband scenario with low mo-bility was assumed. However, in scenarioswith high mobility, the number of subcarri-ers, i.e. the degrees of freedom, typicallycannot be made arbitrarily large, as thisnumber is limited by the coherence time of

the channel. Therefore, for future system de-sign the case where the number of antennastimes the delay spread exceeds the numberof subcarriers has also been considered. Alower bound on the MSE for channels withequal power delay profile was derived, and isshown in the figure. This upper bound servesas a fundamental limit with respect to thesystem performance and is particular suitedfor trading off MSE for throughput. More-over, for some special cases also the optimalpilot signals were designed.

Future work will include the design ofspace-frequency codes and the investigationof the performance of OFDM MIMO systemsin high mobility environments.

This work was supported by the FederalMinistry of Education and Research undergrant FK 01 BU150.

[1] G. Wunder and H. Boche, “Perform-ance bounds and optimal pilot signals inOFDM MIMO systems”, Aachen Symposiumon Signal Theory, September 2001.

Gerhard Wunder ([email protected])

H H I R E P O R T 2 0 0 1 · S E L E C T E D C O N T R I B U T I O N S · M O B I L E B R O A D B A N D …

Channel Estimation for OFDM MIMO Systems

The combination of orthogonal frequency division multiplexing (OFDM) and multiple an-tennas is a promising candidate for future wireless systems. The capacity, channel estima-tion schemes and space-frequency codes for such systems are investigated in this work.

- 20 0 + 20

M = 6T

N = 16, L = 4,

M = 5T

M = 4T

+ 40SNR [dB]

10-3

10-2

10-1

100

MS

E

Lower bound of MSE,with parameters numberof subcarriers N, channel

delay spread L and number of transmit

antennas M


Efficient use of bandwidth is needed forthe extremely fast growing wireless commu-nication market. The approach with the high-est spectral efficiency for wireless indoorcommunication is to use transmit (Tx) andreceive (Rx) antenna arrays to exploit the MI-MO channel.

The achievable BER is an important criteri-on for the design of experimental MIMO sys-tems. We investigate the BER performancefor MIMO systems and how it is influencedby antenna diversity, a LOS component, andchannel estimation errors.

How to achieve maximum capacity in theMIMO channel is well understood, but fur-ther investigation is needed for the morepractical aspect of implementation. The MIMOchannel capacity with no channel informa-tion at the Tx is given by

where Pk/σ2k is the signal-to-noise-ratio

(SNR) of the k-th sub-channel and λk is thek-th singular value of the MIMO channel ma-trix [1]. Each singular value (SV) determinesthe quality of one of the parallel sub-chan-nels, and therefore its capacity and BER.

Figure 1 shows the statistical distribution ofthe singular values for 8 Tx and 8 or 12 Rxantennas. Additional antennas at the Rx in-crease all singular values, which results in on-ly a small increase of the total capacity butgives a significant improvement of the BERperformance because the smallest SVs are in-creased, and these belong to the sub-chan-nels with the worst SNRs and cause most ofthe bit errors.

A LOS component (worst case = far fieldapproximation) simply increases the biggestSV, which gives little gain in capacity, and noBER improvement is observed.

It was shown that channel estimation er-rors with a correlation-based measurementin a separate time slot have an effect similarto additional noise power, and in the worstcase a power penalty of 3 dB applies.

From our investigations we conclude thatthe BER of a practical MIMO system benefitsstrongly from antenna diversity, althoughthe capacity is not significantly improved. ALOS component is of little importance forthe BER performance, but increases the ca-pacity, especially in the low SNR region.

This work was supported by the FederalMinistry of Education and Research undergrant 01 BU 150.

[1] G.J. Foschini, “Layered space-time ar-chitecture for wireless communication in afading environment when using multiple an-tennas”, Bell Lab. Tech. J., 1996, 1, (2), pp.41-59.

[2] T. Haustein et al., “Bit error rates for aMIMO system in Rayleigh and Rician chan-nels”, Proc. VTC-Fall ’01, Atlantic City,October 2001, pp. 1984-1987.

Thomas Haustein ([email protected])

BER for MIMO Systems in Rayleigh and Rician Channels

Multiple Input Multiple Output (MIMO) systems, which are under intensive research atpresent, are discussed, with particular attention to the bit error rates (BER) that can beachieved in indoor scenarios. We focus on the influences on the BER of antenna diversity,a line of sight (LOS) component, and channel estimation errors. Our simulation environ-ment allows us to determine the channel estimation accuracy required to achieve the de-sired BER performance under certain constraints.

Fig. 1: Distributions of the singular values for a MIMO system using 8 Txand 8 or 12 Rx antennas

0 2

2

2

1

1

3

3

4

4

5

5

6

6

7

7

8

88 x 12

8 x 8

4 106 8Singular values

Distribution of singular values

1000

Cou

nts

0 10 3020

16 dB

antenna diversity

MIMO - BER performance, 16 - QAM

LOS

avg. SNR per receive antenna [dB]in Rayleigh channel

0.01

1.0

0.1

1E-3

1E-4

1E-5

BE

R 8 x 8 Rayleigh8 x 10 Rayleigh8 x 12 Rayleigh8 x 12 K = 0 dB8 x 12 K = 10 dB Fig. 2:

BER with antenna diversity as a parameter,including the case of aLOS component

72

Multiple antennas can be used at thetransmitter and receiver to create a multiple-input multiple-output (MIMO) channel toenhance the performance and capacity of awireless communication system. It is neces-sary to know the statistical properties of theMIMO channels if the promised capacity andperformance gain are to be achieved. At leastthe first and second moments of the inverseeigenvalue distributions of the Rayleigh andRician MIMO channel matrices must beknown, especially for the analysis of the biterror performance and to estimate the re-quired transmit power in water-filling orchannel-inversion transmission schemes.

The Rayleigh MIMO channel models thecase of rich multipath propagation betweenthe receive and transmit antennas without aline-of-sight component, whereas the RicianMIMO channel contains a line-of-sight com-ponent. In Rayleigh MIMO channels thechannel matrix consists of independentlyand identically distributed zero-mean com-plex Gaussian random variables, whereas inRician MIMO channels the channel matrixentries have non-zero means. In indoor envi-ronments these channel models fit the mea-sured channel data well.

Tools for computing the empirical distribu-tion functions of the eigenvalues are re-viewed in [1], and new results for the inverseeigenvalues of Rayleigh and Rician MIMOchannel matrices using the Stieltjes transformare presented. A simple method for calculat-ing the moments of the inverse eigenvaluesin the transform domain yields analytical for-mulas for the means and variances of the in-verse eigenvalues.

The theory also yields closed form expres-sions for the empirical distribution functionsof the eigenvalues and inverse eigenvalues ofthese random matrices. In figure 1 the simu-lated and empirical eigenvalue distributionfunctions are shown. In figure 2 the simulat-ed data are compared with the theoreticalinverse eigenvalue distributions. The theoryfits the simulated data well.

It is worth mentioning that the means ofthe inverse eigenvalues of quadratic MIMOsystems do not exist. The average noise gainin MIMO zero-forcing receivers can easily be

reduced by increasing the number of receiveantennas, and system designers can decidein advance what minimum receive-to-trans-mit antenna number ratio is necessary toachieve a given error performance.

[1] E. Jorswieck, G. Wunder, V. Jungnickeland T. Haustein, “Eigenvalue and inverseeigenvalue statistics for Rayleigh and RicianMIMO channels”, IEE Workshop on MIMOSystems, London, December 2001.

Eduard Jorswieck ([email protected])


Eigenvalue and Inverse Eigenvalue Statistics for Rayleigh and Rician MIMO Channels

The probability density functions for the eigenvalues and inverse eigenvalues of Rayleighand Rician MIMO channels are derived, as well as useful formulas for the moments ofthese distributions. Examples of the use of these results for MIMO system design are given.

Fig. 2: Simulated and theoreti-

cal distributions of theinverse eigenvalues of

the 8 x 8 (solid) and 8 x 16 (dashed) Rayleigh

MIMO channel

Fig. 1: Simulated and theoreti-

cal distributions of theeigenvalues of the 8 x 8

(solid) and 8 x 16(dashed) Rayleigh MIMO

channel

0 21 3 4Value of eigenvalue x

0

0.5

1.0

1.5

2.0

Em

pir.

den

. fun

ctio

n -

edf [

x]

0 2 64Value of inverse eigenvalue y

0.0

0.2

0.4

0.6

0.8

1.0

Em

pir.

den

. fun

ctio

n -

edf [

y]


Multiple-Input Multiple-Output (MIMO)transmission is very attractive for an indoorwireless local area network (WLAN), since itexploits the enormous capacity hidden in therich scattering multi-path propagation that istypical of indoor environments. But relativelycomplex and powerful signal processing isneeded to implement the popular BLAST al-gorithms [1] at the high data rates typical inWLAN applications.

Signal processing before data transmissionis possible if channel state information (CSI)is available at the transmitter (Tx). Becausethe radio channel is reciprocal, CSI can bemade available at the Tx by taking it fromthe CSI used for detection of the receivedsignal. Linear Channel Inversion (LCI) is asimple processing scheme based on CSI atthe Tx. In the down-link at the Tx, data aremultiplied by the pseudo-inverse of thetransposed channel matrix HT, while slightlymore complex processing is needed in theup-link (see Fig. 1).

LCI results in a statistical distribution of theTx power which depends on H. This distribu-tion can be described in terms of the inverseeigenvalues of HHH, where HH is theHermitian transpose of H. For the Rayleighfading channel, an analytical expression forthe empirical density function of the inverseeigenvalues can be found, from which theaverage Tx power can be obtained (see thepreceding contribution). As a general result,at least one extra antenna is needed on oneside of the link. The noise gain is isotropicwith LCI, and the bit-error performance issimilar to that of the additive white Gaussiannoise (AWGN) channel. Compared to theAWGN channel the BER curve of LCI is shift-ed towards higher Tx/noise power values by

a factor n/(n-m), due to signal processing atthe Tx. The figures n and m are the numbersof antennas at the mobile unit and at thebase station respectively. For small bit errorrates, the performance in terms of Tx poweris significantly better than with Zero Forcing(ZF). In the case of many parallel datastreams VBLAST performs even better (seeFig. 2), but the complexity of the requiredsignal processing increases significantly. LCIis more sensitive than VBLAST to channel es-timation errors. Fortunately this is unprob-lematic at high data rates, where long pilotsequences can be used without losing toomuch time for channel estimation.

In conclusion, MIMO with LCI is a promis-ing candidate for WLAN applications. LikeVBLAST, LCI offers significantly better perfor-mance than ZF. In addition, LCI needs lesssignal processing than VBLAST when thenumber of data streams is large.

This work was supported by the FederalMinistry of Education and Research undergrant 01 BU 150.

[1] G. D. Golden, G. J. Foschini, R. A.Valenzuela and P. W. Wolniansky, ElectronicsLetters, 35, 14-15, 1999.

[2] V. Jungnickel, T. Haustein, E. Jorswieckand C. von Helmolt, Proc. IEE Technical Sem-inar on MIMO Communication Systems,London, 12.12.2001.

Clemens v. Helmolt ([email protected])

A MIMO WLAN based on Linear Channel Inversion

A Multiple-Input Multiple-Output (MIMO) system is proposed that uses radio channel rec-iprocity and linear channel inversion. This scheme is especially attractive for broadbandindoor wireless local area network (WLAN) applications, since the required signal process-ing schemes are inherently fast and of low complexity.

Fig. 1: Principle of linear channel inversion (LCI)for MIMO transmission.

The matrices U, D and Vrefer to the singular value decomposition of H = UDVH.

Fig. 2: Bit error rates as a func-tion of the transmitterpower with perfectchannel estimation. Dots are the simulationresults, while full anddashed lines are the theoretical curves for zero forcing (ZF) andLCI. (The dashed line forVBLAST is only there as aguide for the eye.)

- 10 0 + 20+ 10

MIMOflat Rayleigh8 Tx, 12 Rx

QPSK

ZF

LCI

VBLAST

Total transmitter power /noise at one receiver [dB]

0.01

0.1

1E-6

1E-5

1E-4

1E-3

Unc

oded

bit

erro

r ra

te

mobileunit

basestation

x 1 1

22

3

(H)

(H )T

n < m

VD V x-1 H

mn

x´

x´ x1 y (H ) xT -1

V U yH

74

It is well known that multiple-element an-tenna arrays can improve the performance ofa wireless communication system in a fadingenvironment.

We consider the multiple-input single-out-put (MISO) single-user case with imperfectchannel state information at the transmit ar-ray. It has been shown that even partialchannel state information at the transmitter(CSIT) can increase the capacity of a MISOsystem. Recently transmission schemes foroptimizing the capacity of MISO mean-feed-back and covariance-feedback systems havebeen derived. It was shown that the optimalcapacity can be achieved with Gaussiantransmit signals with a particular covariancematrix. Further it has been proved that theoptimum transmit covariance matrix has thesame eigenvectors as the known channel co-variance matrix. The optimum eigenvaluescan be computed numerically.

The number of eigenvalues greater thanzero corresponds to the number of directionsin which the transmit signals are sent. If thenumber of directions is one, the system per-forms beamforming. If perfect channel stateinformation is available at the transmitter, orif the signal-to-noise ratio (SNR) is low,beamforming is known to be optimal. Forthe general case the required number ofeigenvalues to achieve maximum capacity isnot known a priori, and cannot be easilycomputed from the channel covariance ma-trix eigenvalues.

A necessary and sufficient condition for theoptimality of beamforming for MISO systemswith covariance feedback has been derivedin the literature. We extend these results in[1] to the general case with transmission in aspecified number of directions.

A necessary and sufficient condition hasbeen derived for the optimum eigenvalues ofthe transmit covariance matrix of an MISOsystem with many transmit antennas andone receive antenna, assuming a knownchannel covariance matrix. A necessary andsufficient condition for achieving maximumcapacity when transmitting in a fixed num-ber of directions was also obtained.

With these conditions it is possible to an-swer the following questions about the sys-tem design of MISO systems:

• For a fixed number of transmit antennas and a given channel covariance matrix, what is the required SNR to take advan-tage of full space diversity?

• Given the channel covariance matrix and the SNR, what is the minimum number of parallel data streams needed to achieve maximum capacity?

• What is the maximum spatial diversity for a fixed channel covariance matrix as a func-tion of the SNR?

These questions can be answered usingthe new results. For example, in the figurethe diversity function is shown as a functionof the SNR. The diversity function yields theminimum and maximum required number ofdirections to achieve capacity given a fixedchannel covariance matrix, and hence an-swers the third question.

Helpful tools for computing the optimumnumber of transmit antennas and the re-quired SNR to take advantage of full spacediversity were derived. These considerationsgave insight into the optimum single-userMISO system design. Open questions con-cern the multi-user case and the MIMO case.

[1] E. Jorswieck and H. Boche, “On trans-mit diversity with imperfect channel state in-formation”, accepted to ICASSP 2002.

Eduard Jorswieck ([email protected])


On Transmit Diversity with Imperfect Channel State Information

The optimal transmission strategy for a multiple-input single-output wireless communica-tion link is studied for the case in which the receiver has full channel state informationand the transmitter knows only the channel covariance matrix.

Diversity function as afunction of the SNR

- 20 - 10 0 + 10 + 20SNR

6

4

2

0

M [S

NR

]


Transmission of a pilot signal in wirelesssystems is a very useful way of obtaininggood amplitude and phase estimates of themultipath channel, thereby making possiblequasi-optimum coherent reception andweighted combining of multipath compo-nents. Pilot-based channel estimationschemes are widely used in the downlinks ofwireless communication systems. The muchmore difficult case of the many-to-one uplinkfrom multiple wireless users to a base stationis still the subject of intense research work.

The figure shows the underlying model ofa multiuser system under consideration.

Here, each user, say user k, transmits a dis-tinct N-periodic pilot signal Sk to a commonbase station, while autonomously operatingin an uncorrelated multipath radio environ-ment. The multipath channel is modelled bythe vector hk containing N independent iden-tically distributed Gaussian random variableswith variance σc. The received signal y is dis-turbed by the zero mean complex Gaussianrandom vector z with the variance σ. In eachperiod, the channel estimator of the k-th userwith impulse response Ck observes N samplesof the received signal and gives an estimateof the multipath vector hk. Apart from noiseand inter-symbol interference, the channelestimation system must also combat inter-user interference to obtain reliable channelestimates, which is in clear contrast to thedownlink scenario. In our analysis we as-sumed that neither σc nor σ are known tothe receiver. We used the mean squared esti-mation error (MSEE) as an optimization crite-rion.

For a given set of pilot signals, it may beeasily shown that a bank of inverse filters at

the receiver is optimal. However, it turns outthat the MSEE still depends on the choice ofpilot signals. Assuming a bank of inverse fil-ters, we proved that the MSEE for each useris bounded below by σc(K-1)+σ, where thelower bound is attained if and only if each pi-lot signal is a periodically self-invertible se-quence. Since the normalized discreteFourier transform of such sequences is equalto the complex conjugate of its inverse, asimple bank of matched filters is the opti-mum channel estimation system if the pilotsignals form a set of periodically self-invert-ible sequences. Thus, a careful choice of pilotsignals not only improves the average perfor-mance of a channel estimation system but al-so reduces its complexity. We also showedthat a bank of simple matched filters alsogives optimal performance if a complemen-tary set of periodic sequences is used.

Periodically self-invertible sequences arerare. In the bipolar case, there is most proba-bly only one such a sequence. In complexspaces, some construction methods areknown. In contrast, complementary sets ofbinary sequences are easy to construct, al-though these are not as robust as sets of pe-riodically self-invertible sequences.

These results show that obtaining reliableestimates of multipath components in multi-user systems can be a difficult task. This is es-pecially true for large systems, since the low-er performance bound increases linearly withthe number of users. Better performance canbe obtained if knowledge of the variances isavailable and utilized at the receiver.

[1] H. Boche and S. Stanczak, “Lowerbound on mean squared channel estimationerror for multiuser receiver”, Proceedings ofIEEE International Symposium on Circuitsand Systems, Sydney, Australia, 6-9 May,2001.

Slawomir Stanczak ([email protected])

Pilot Signal Design for Multiuser Channel Estimation Systems

We investigated the mean-squared error performance of a pilot-based channel estimationscheme in a code-division-multiple-access (CDMA) multipath environment. We derivedthe optimum linear estimator in the absence of any prior knowledge and obtained a low-er bound for the maximum mean-squared channel estimation error.

Model of a pilot-basedchannel estimation system with K users

Channelestimator

C1

Channelestimator

CK

S1 h1^z

y

SK hK^

Model of a pilot - based channel estimationsystem with K users.

Multipathchannel

h1

Multipathchannel

hK

76

We say that a code-division multiple-access(CDMA) system is quasi-synchronous (QS-CDMA) if the duration of the modulatedsymbols T is significantly longer than Ta + Tm,where Ta is the maximum time dispersion ofa multipath channel and Tm denotes themaximum relative time delay resulting fromnon-perfect synchronization of the spatiallyseparated mobile users. Since coarse syn-chronization of users can usually be estab-lished with a GPS-based common clock, oralternatively, with the aid of higher-layer net-work protocols, we can assume that Tm ≈ 0.The parameter Ta varies between about 2 µsand 20 µs, depending on the multipath radioenvironment. Figure 1 shows a typical con-figuration of a QS-CDMA transmitter.

In order to achieve high data rates in QS-CDMA systems, the duration of the encodedbinary symbols of length Ts needs to be ex-tended. To this end, we can employ a map-ping from a set of binary vectors of length qonto a finite set of complex numbers G of or-der 2q and/or a multicode technique that al-locates m sequences to each user. The para-meters q and m are chosen so that the quan-tity d = [NTa/(q m Ts)] + 1 is significantlysmaller than the length of the sequences N.A guard interval of length d-1 is also usuallyinserted between any two consecutive se-quences.

All this has tremendous implications onhow to design spreading sequences, sincewe only need to consider aperiodic correla-tions in the small window -d+1,…, d-1. Inparticular, to eliminate inter-symbol interfer-ence (ISI) and multiple access interference(MAI) completely in QS-CDMA systems, it is

sufficient to construct sequences so that allaperiodic autocorrelation sidelobes and cross-correlations are zero within this window.Aperiodic zero-correlation-window (AZCW)sequences have been shown to exist on thecomplex unit sphere. We proved that thenumber of AZCW sequences is boundedabove by N/d if d divides N, which meansthat sets of AZCW sequences may be toosmall for many applications.

To overcome this problem we can consid-er sets of sequences for which all aperiodiccorrelation magnitudes within the window(except for the autocorrelation at the origin)do not exceed a certain given small value.We have found an upper bound on the num-ber of such sequences on the complex unitsphere. We also constructed a class of mutu-ally orthogonal sequences for which themaximum aperiodic correlation magnitudein this window is bounded above by (d-1)/N,and showed that the use of such sequencesin QS-CDMA systems keeps the total interfer-ence (ISI+MAI) at a low level. We proved thatthe interference in a fully loaded system canbe expected to be significantly smaller than2d2/(3N). Figure 2 shows typical aperiodicautocorrelations of binary sequences ob-tained using our construction procedure.

[1] S. Stanczak and H. Boche, “Sequenceswith small aperiodic correlations in the vicini-ty of the zero shift”, to appear in 4th Inter-national Conference on Source and ChannelCoding, Berlin, 28-30 January, 2002.



Sequence Design for Quasi-Synchronous CDMA Systems

We present a unified theory for constructing sets of mutually orthogonal sequences withsmall aperiodic correlation magnitudes in the vicinity of the zero shift. We also provesome important bounds related to sets of such sequences.

Fig. 1: Typical configuration ofa QS-CDMA transmitter

Fig. 2: Aperiodic

autocorrelations - 60 - 30 0 + 30

N = 64d = 8

+ 60

0

- 0.5

+ 0.5

- 1.0

+ 1.0

Aperiodic autocorrelations

source+

channelcoding

S / P - conv.+

multicodespreading

chipsummer

+signal

forming

QPSK

Q

Q

I

1 m

q m S

P

I

2 - arymapping

0.1 G

q

q


Designers of future wireless communica-tions systems will be confronted with theproblem of how to use scarce resources suchas power and bandwidth in order to satisfystringent quality-of-service requirements.

The signal-to-interference+noise ratio(SINR) is a useful measure of the quality of amultiuser receiver, particularly when it isused in conjunction with error-control de-coders. SINR is a soft-decision variable thatreflects the reliability of decisions. Roughlyspeaking, the larger the SINR is, the more re-liable the decisions are, and hence there ishigher quality-of-service for the users. Thus,to guarantee a certain quality-of-service forthe users, we demand that Q ≤ SINRk foreach k, where SINRk is the SINR for the k-thuser and Q is a given real number, called theSINR requirement. We say that a set of usersis admissible in a CDMA system if we can as-sign sequences to the users and control theirpower so that each user meets the SINR re-quirement Q.

The problem of admissibility of the users ina synchronous CDMA (S-CDMA) system wascompletely solved in [1]. However, since thesimplistic setting of perfect symbol synchro-nization rarely holds in practice, there is astrong need to investigate asynchronous CD-MA (A-CDMA) systems. We thereforedropped the demand for perfect symbol syn-chronization and considered a chip-synchro-nous A-CDMA system embedded in zero-mean additive Gaussian noise with varianceσ. Such a CDMA system is obtained if therelative time offsets that model the lack ofsynchronization between the users at the re-ceiver are assumed to be discrete randomvariables uniformly distributed on the symbolinterval.

Let S be a set of K sequences assigned tothe users and P = (p1,…, pK) be a set of pow-ers at which users are received. To showwhen exactly K users are admissible in aCDMA system with processing gain N, it issufficient to identify a pair (P, S), called anoptimal allocation, for which the minimumSINR is as large as possible. We solved thismax-min optimization problem for N ≤ K as-suming that S forms a set of polyphase se-quences. Such sequences are widely used inpractical systems, since they can be trans-

formed easily into a carrier pulse train with amodulated phase. The optimum power allo-cation is p = p1 =…= pK, where p is subject toa given power constraint, and the optimumsequence allocation is a complementary setof sequences in which each member has theminimum merit factor for its aperiodic auto-correlations. We constructed such a set of se-quences for N ≤ K.

The figure depicts the minimum SINR as afunction of the signal-to-noise ratio(SNR = p/σ) for both optimal sequences andthe so-called cubic sequences, which areknown to have favourable periodic correla-tion properties. As seen in the figure, if opti-mal sequences are allocated, users meet theSINR requirement of 5 dB at a SNR of about8.2 dB. Pmin in the figure is the minimumpower for which this is still true. In the caseof cubic sequences, the SINR requirementcannot be satisfied for all users.

[1] P. Viswanath, V. Anantharam, D. Tse,“Optimal sequences, power control, and usercapacity of synchronous CDMA systems withlinear MMSE multiuser receivers”, IEEE Trans.on Inform. Theory, vol. 45, no. 6. September1999.

[2] S. Stanczak and H. Boche, “On re-source allocation in asynchronous CDMAchannels”, to appear in Proc. of the 2002IEEE International Symposium on InformationTheory (ISIT 2002), Lausanne, Switzerland,June 30-July 5, 2002.


Resource Allocation in Asynchronous CDMA Systems

We solve the problem of admissibility in an asynchronous code-division-multiple-access(A-CDMA) system, assuming that all users have the same quality-of-service requirements,expressed as the signal-to-interference+noise ratio (SINR).

Minimum SINR as afunction of SNR for optimal and cubic phasesequences

- 5 + 5

SIR

SIR

N = K = 31

PminQ

0 + 20+ 10 + 15SNR [dB]

- 4

+ 4

+ 8

+ 12

0

SIN

R [d

B]

optimalcubic

sequences

78

The system consists of an antenna arraywith RF receiver front end (Fig. 1), a digitalarray receiver (Fig. 2), a personal computer(PC) for controlling the measurements andperforming signal processing, and three testtransmitters with single antennas. The anten-na array represents the base station receiverantenna and the test transmitters play therole of mobile stations in the uplink mode.

The system operates in the UMTS radiofrequency range from 1.9 GHz to 2.2 GHzwith a signal bandwidth of 5.0 MHz. For testpurposes UTRA-TDD signals have been im-plemented at the test transmitters.

The receiver system consists of parallelRF/IF/ADC hardware. All output signals fromthe receiver antenna array are simultaneouslyconverted into complex digital baseband sig-nals. Signal combining is not performed inthe RF or IF domain, so that a wide range ofdigital signal processing algorithms can beused instead.

The experimental system uses offline signalprocessing with a PC instead of a real-timesignal processing unit. The digitalized re-ceived signals are stored in memory devicesbefore performing signal processing.

One of the main advantages of this con-cept for an experimental system is that wecan use algorithms with large computationaleffort because there is no strict time limita-tion for the processing. Furthermore, a vari-ety of signal processing algorithms can beapplied to the same measured data, which isanother advantage for evaluation purposes.

To perform the measurements, the testtransmitters and the receiver antenna arrayare placed in an anechoic measurement

chamber. Various measurement scenarioscan be arranged by changing the positions,the power levels and the test signals of thetest transmitters. Furthermore, the angularposition of the receiver antenna array can bevaried with a remote controlled positionerdevice.

To simulate multipath and fading effects,which are typical for wave propagation inmobile radio systems, reflecting and movingdevices can be placed in the measurementchamber. Scenarios with a larger numbers ofactive mobile stations can be simulated bysuperposition of the received signals in thedigital domain.

A software library containing device dri-vers, signal processing algorithms and con-trol procedures has been developed to oper-ate the system. To extend the system capa-bilities to downlink experiments, an arraysignal generator and an array transmitterfront end will be developed in future.

This work is supported by the Gesellschaftvon Freunden des Heinrich-Hertz-Instituts e.V.

[1] A. Kortke and G. Mönich, “Antenna ar-ray calibration algorithms”, MICRO.tec 2000,VDE World Microtechnologies Congress,Hannover, September 2000, vol. 2, pp. 193-198.

Andreas Kortke ([email protected])


Experimental Smart Antenna System for UMTS

Smart antenna techniques are proposed to enhance the system capacity of future 3G mo-bile communication systems. Smart antenna design involves various fields of electrical en-gineering. To perform experiments in the fields of antenna array design, RF hardware de-sign, array calibration techniques and array signal processing, an experimental smart an-tenna system has been developed.

Fig. 1: Experimental 2 GHz

linear microstrip antenna array

Fig. 2: Digital array receiver

and signal processing PC


The downlink of future multimedia-typewireless systems will exhibit very large peak-to-average capacity demands, which can on-ly be fulfilled by adaptive transmission strate-gies that are optimally matched to the actualchannel states. Significant gains in spectralefficiency can be achieved by exploiting thespatial diversity and processing gain offeredby the multipath channel. This requires theemployment of antenna arrays at the basestation and/or the hand-held device. In a typ-ical wireless system, however, the size andenergy storing capability of the hand-heldpreclude the use of arrays at that end. Recentresults show that in this case beamforming atthe base station is close to the optimal trans-mission strategy.

Classical results on beamforming focus onsingle-user systems. Typical wireless systems,however, are limited by severe co-channelinterference. Thus, high peak data rates canonly be achieved by jointly optimizing thebeamforming weights of all users, togetherwith dynamic resource management. In alarge system there is likely to be a set of userswith large separations in terms of azimuthangle at any time. By choosing the userswith the best channel characteristics and az-imuth distributions, large capacity gains canbe expected.

However, the actual link capacity of a par-ticular user depends strongly on the beam-forming weights. The optimum weights mustbe found by jointly optimizing the communi-cation links for all users.

The optimization goal is to guarantee eachuser a certain data rate, subject to strict delayconstraints. This leads to the problem ofjointly balancing the signal-to-interference-plus-noise-ratios (SINRs) – an individual SINRmust be achieved for each user. If a solutionto this non-linear optimization problem ex-ists, additional degrees of freedom can beused to minimize the total transmission pow-er, thereby reducing the amount of co-chan-nel interference received by neighbouringcells. This strategy not only increases theoverall system capacity, but also improvesthe power efficiency.

The HHI has developed a new iterative al-gorithm [1] that solves this problem withminimal computational burden. The pro-posed algorithm always converges to theglobal optimum within 3-4 iteration steps.Moreover, it is able to detect infeasible sce-narios, in which the required SINR levels can-not be achieved for the given power thresh-old. The algorithm provides a measure forthe achievable SINR margin, so that it can beused very efficiently for resource manage-ment.

Among a set of active users, the algorithmfinds those that are most favourable fortransmission. Then, it determines the beam-forming weights that minimize the totaltransmission power while satisfying individ-ual SINR requirements.

The trade-off between the transmissionpower and the achievable SINR margin fordifferent beamforming techniques is com-pared in the figure. Conventional techniquesare clearly outperformed by the new opti-mization scheme.

[1] M. Schubert and H. Boche, “Multi-an-tenna downlink transmission with individualSINR receiver constraints for cellular wirelesssystems”, Proc. ITG Conference on Sourceand Channel Coding, Berlin, Jan. 2002.

Martin Schubert ([email protected])

Optimal Multiuser Downlink Beamforming

We address the problem of joint downlink beamforming and power control in a cellularwireless system, assuming that an antenna array is employed at the base station and thateach user is equipped with a single antenna element. The goal is to minimize the trans-mitted power while satisfying individual link capacity requirements for all users. The HHIhas developed a computationally efficient optimization scheme for solving this nonlinearmulti-user problem.

Power minimization capability of the new algorithm

4 6 8 10 12

single antenna

conventionalbeamformer(independent)

new algorithm(joint beamforming+ power control)

SINR requirement [dB]

- 5

+ 5

+ 10

+ 15

0

Tra

nsm

issi

on p

ower

[dB

m]

80

Future cellular or personal communicationsystems must be able to achieve high peakdata rates with low-complexity mobile termi-nals. This challenging goal can only be a-chieved by employing cell-site antenna ar-rays that are able to exploit the spatial diver-sity and processing gain offered by thewireless channel. In this context, beamform-ing is a promising strategy for both transmis-sion and reception. By pointing a beam to-wards the desired user, co-channel interfer-ence can be significantly reduced. This resultsin improved signal-to-interference ratios(SIRs). The resulting gains can be traded forincreased capacity or reduced transmissionpower.

Most studies on beamforming so far con-sider the uplink multi-access channel, wherethe users are coupled by the transmissionpowers. Once the optimum power levels aredetermined, the optimum beamforming fil-ters can be found by solving a set of inde-pendent eigenvalue problems [1].

In the downlink, the data are transmittedover a broadcast channel and the spatial pro-cessing must be carried out prior to transmis-sion. As for the uplink, all users are coupledby crosstalk. This crosstalk, however, is notdetermined only by the transmission powers,but also by the choices of the individualbeamforming filters. This makes the down-link problem more complicated than the up-link problem. In particular, the downlinkbeamformers cannot be optimized by simpleeigenvalue decomposition, as was the casefor the uplink.

The HHI has shown [1] that there is an in-teresting duality between the uplink anddownlink beamforming problems. The beam-forming vectors that optimally balance thesignal-to-interference ratios (SIR) in the up-link are also optimal in the downlink. This,however, does not hold for the transmissionpowers. Whereas the optimum uplink pow-ers are given by the dominant right-handeigenvector of the system coupling matrix,the downlink powers are obtained from thedominant left-hand eigenvector. Conse-

quently, the complicated downlink SIR bal-ancing problem, which was first addressed in[2] and later by many authors, can be effi-ciently solved by considering the uplinkproblem instead.

This downlink/uplink duality has been usedto develop a new rapidly convergent opti-mization scheme [1]. By alternately optimiz-ing beamforming weights and transmissionpowers, we obtain an iteration sequence thatmonotonically and rapidly converges to theglobal optimum of the joint SIR balancingproblem [2], regardless of the chosen initial-ization.

Numerical simulations show that the newalgorithm typically converges within 3-4 iter-ation steps. Thus it is suited for real-time ap-plications in future wireless systems. The sim-ulation results in the figure show that largecapacity gains can be expected compared tosingle antenna transmission. These gains canbe traded for increased user density or de-creased transmission power.

[1] M. Schubert and H. Boche, “A unifyingtheory for uplink and downlink multi-userbeamforming”, Proc. International ZürichSeminar, 2002.

[2] D. Gerlach and A. Paulraj, “Base stationtransmitting antenna arrays for multipath en-vironments”, Signal Processing (Elsevier Sci-ence), vol. 54, pp. 59-73, 1996.

Martin Schubert ([email protected])


Joint Uplink/Downlink Beamforming Optimization

We address the problem of multi-user signal-to-interference ratio (SIR) balancing in a cel-lular wireless system in which an antenna array is employed at the base station and eachmobile is equipped with a single antenna. In this work it is shown that there is an inter-esting duality between the uplink and downlink beamforming problems. We use this du-ality to develop a rapidly converging algorithm that always converges to the optimummulti-user beamforming solution, for both uplink and downlink. A convergence analysisand proof of optimality are provided in [1].

Average SIR improve-ment for various num-

bers of co-channel usersuniformly distributed in

a 120° cell sector

5 7 9 11 13 15Number of co - channel users

0

1

3

2

Ave

rage

SIR

mar

gin

[dB

] 4 antenna elements(uniform linear array)

single antenna


The use of space division multiple access(SDMA) in the 60 GHz frequency range isbecoming more and more interesting formobile communications. An important build-ing block for a mobile system is the phasedarray antenna. By using SDMA, the mobileterminal considered here can be connectedto the base station via an optimum radiolink, at the same time suppressing the signalsfrom the other terminals and minimizingemission into the neighbouring environment.

Constrained beamforming is usually a-chieved by amplitude and phase control ofthe individual millimetre-wave signals feed-ing the array antenna elements, which is dif-ficult at 60 GHz. Thus beamforming by opti-cal means is an interesting alternative solu-tion. In an optical beamformer the phasesand amplitudes of a number of optical sig-nals are controlled individually, and hencethe millimetre-wave signals feeding the arrayantenna are also controlled.

Beamforming is achieved by a complexweighting of the individual antenna signals.The Maximum Directivity (MD) beamformeralgorithm [1] enables us to steer both a mainbeam and a number of nulls in the radiationcharacteristic, and achieves a beam with thehighest possible directivity. The number ofnulls can vary from 0 to M-1, where M is thenumber of antenna elements. In the case offour antenna elements we can specify onelook direction ϑLD and K null directions ϑOK,0 ≤ k ≤ K, with K between 0 and 3. Thechoice K = 3 gives a full rank 4 x 4 linearequation system:

where ϑ is the angular direction, wm is theweight and gm(ϑ) is the radiation character-istic of array element m. The four individualradiation characteristics gm are obtained fromsimulations of the single antenna elementand the complete array structure. These areinserted into the equation system to calcu-

late the antenna weights wm, after which thefield patterns for given look directions arecalculated.

An experimental 1 x 4 patch array antennawas developed with λ/2 spaced antenna ele-ments (size 1 x 1.2 mm). It was tested in a60 GHz system. The millimetre-wave signalswere generated by heterodyning the wavesλ1 and λ2 from an optical millimetre-wavetransmitter (OMT), whose optical signalswere distributed to four optic/millimetre-wave converters (OMCs) for opto-electronicconversion. Each OMC was connected toone of the antenna elements. Beamformingwas performed by tunable optical delay linesand attenuators (OFBN) preceding theOMCs. Figure 1 shows the calculated andmeasured field patterns at a look direction of–30° with zeros at 0°, 27°, and –70°.

[1] T. Kuhwald, H. Boche and M. Bronzel,“A new optimum constrained beamforming-algorithm for future mobile communicationsystems based on CDMA”, Proc. ACTS Mo-bile Communications Summit '99, Sorrento,Italy, pp. 963-968, June 1999

Gerd Grosskopf ([email protected])

60 GHz Optically Controlled Maximum Directivity Beamformer

A 60 GHz patch array antenna that is to be used as a smart antenna in a broadband mo-bile communication system is described. The antenna field pattern was synthesized usingthe maximum directivity beamforming algorithm. In the experiments the 60 GHz signalswere generated by optically heterodyning the waves from two laser diodes, and the lookdirection of the 1x4 antenna was steered by an optical beamformer.

Experimental setup, together with the calcu-lated (dashed) and mea-sured (full) field distribu-tions of the 1 x 4 patcharray antenna at 60 GHzin the –30° look direction

- 60 + 600Azimuth [degrees]

- 20

- 40

- 60

rel.

RF

- P

ower

[dB

]

Receiver

OFBN 1 x 4Antenna

4 xOMC

4 x ODL& ATT

OMT

λ1

λ2

82

The 60 GHz frequency range and space di-vision multiple access operation (SDMA) willbe important in future broadband mobilecommunication systems. Optical millimetre-wave techniques offer interesting solutions tothe problems of such systems. On the onehand these techniques allow the antennacharacteristics to be determined by a singleopto-electronic integrated circuit (OEIC),while on the other hand millimetre-wave sig-nals with excellent properties can be gener-ated and easily distributed via optical fibresto the base station antennas.

The millimetre-wave signals are obtainedby heterodyning two optical waves with afrequency spacing equal to the desired mil-limetre-wave frequency [1]. These waves areemitted by an optical millimetre-wave trans-mitter (OMT)(Fig. 1). The beamforming OEICis realised in silica technology (OBF). Toachieve beamforming, the amplitudes andphases of the millimetre-wave signals feedingthe array antenna must be individually con-trolled. Thus the two input waves are eachsplit into a number of paths equal to thenumber of array antenna elements. A ther-mo-optic effect is used to adjust the phases,which are controlled by heaters on top of theembedded optical waveguides. The ampli-tudes are controlled by Mach Zehnder (MZ)structures, also varied by heaters. The advan-tage of this method is that, for a desiredphase shift of the millimetre-wave signal, thephase of only one of the optical waves has tobe shifted by the same amount in the opticaldomain. Finally the optical waves are com-bined in pairs and coupled to the optic/mil-limetre-wave converters (OMC) for opto-electronic conversion.

The RF properties of the phase section ofthe 4-channel OBF were measured using a1 x 4 phased array antenna. The relativephases of the four 60 GHz signals were ad-

justed by setting the OBF heater powers togive the desired antenna look direction (Fig.2). The amplitudes were set to a constantvalue by fibre-based optical attenuators(these will later be replaced by a 4-channelsilica-based attenuator that was recently de-veloped). Constrained beamforming is possi-ble with the completed device [2], i.e. an op-timum wireless connection to a desired mo-bile receiver can be established with negligi-ble extraneous radiation to other terminalsand into the environment.

[1] G. Grosskopf, R. Eggemann, D. Rohdeand M.S. Choi, “155 Mbit/s data trans-mis-sion at 60 GHz using a 1x4 patch array an-tenna with variable optical delay lines”, IEEEIntern. Microwave Symp. 2001, Phoenix,USA, pp. 1821-1824.

[2] T. Kuhwald, H. Boche and M. Bronzel,“A new optimum constrained beam-forming-algorithm for future mobile communicationsystems based on CDMA”, Proc. ACTS Mo-bile Commun. Summit '99, Sorrento, Italy,pp. 963-968, June 1999.



Silica-Based Beamformer in a 60 GHz Radio-Over-Fibre System

First results with a novel silica-based optical beamformer intended for use in smart anten-na environments are reported. The device was tested in an experimental 60 GHz trans-mission system using optical millimetre-wave generation and a 1x4 phased array antenna.

30 dB

+ 40 °- 40 ° + 20 °- 20 °

Control

Arrayantenna

OBFPhase Amplitude

4

14 x

OMCOMTλ1

λ2

Fig. 1: Experimental setup for

optical beamformingand millimetre-wave

generation. OMT: optical millimetre-

wave transmitter, OBF: optical beam-

former, OMC: optic-mil-limetre-wave converter.

Fig. 2: Measured far-field

patterns in the H-planeof the 1x4 phased arrayantenna at 60.8 GHz in

different look directions


The convergence of the internet and mo-bile access creates demands for broadbandwireless interconnections. Hybrid-fibre-radio(HFR) systems at 60 GHz are of special inter-est for this application [1]. A complete pro-posed indoor system with picocellular struc-ture, shown in Fig. 1, consists of four build-ing blocks: a central master control station(CMCS), a number of control stations (CS),each of which supports several base stations(BS), and the mobile terminals (MT). TheCMCS, CS and BS are connected by opticalfibres.

Steered phased array antennas are an im-portant building block in the BS. They con-nect the selected MTs and minimise theemission of radiation into the environment.The second advantage of the demonstrator isthe cost-effective realisation of the BSs,which is achieved by centralised microwavesignal generation using an optical referencesource in the CMCS to control 60 GHz opti-cal sources in the different CSs. Their outputsignals are heterodyned at optic/microwaveconverters (OMCs) in the numerous BSs.

The 60 GHz source in the CS is a self-pul-sating laser (SPL) comprising two active dis-tributed feedback (DFB) sections and an inte-grated phase tuning section [2], controlledby three dc currents. The Bragg wavelengthsin the two DFB sections are detuned in orderto generate two DFB modes. The mutual in-jection of lasing modes from one DFB sectioninto the other, together with proper phaseadjustment of the light waves via the inte-grated phase tuning section, leads to a cou-pled dual-mode lasing system with a very

stable 60 GHz beating signal at 1540 nm.Further noise reduction (phase noise value of-100 dBc/Hz @ 100 kHz) is achieved by in-jecting subharmonic 10 GHz optical pulsesinto the SPL, which are generated by a gain-switched laser (GSL) in the CMCS.

The SPL output signal is externally modu-lated using an intensity modulator (MOD)for data transmission (Fig. 2). A 1.27 GHzsubcarrier modulated by 155 Mbit/s offsetquadrature phase-shift keying (OQPSK) is ap-plied to the external modulator. Followingtransmission over a single mode fibre to theBS, the optical waves are split and applied tothe OMCs. The RF signals feeding the 1 x 4phased array antenna are obtained by opticalheterodyning. The phases of the RF signals,and hence the antenna look direction, aresteered by a fibre-based beamformer preced-ing the OMCs. The received signal is down-converted at the mobile terminal (MT) andfed to an OQPSK receiver. Error-free opera-tion at 155 Mb/s (BER less than 10-10) wasdemonstrated, with no penalty for fibre dis-tribution of the optical microwave signal orfor remote stabilisation of the SPL.

Future broadband wireless connectionscan be made in a cost-effective way by usingan SPL as the optical microwave source.

[1] G. Grosskopf et al., Intern. Microw.Symp., IMS 2001, Digest, vol. 3, pp. 1821-1824, May 2001.

[2] C. Bornholdt et al., Electron. Lett., vol.36, no. 4, Febr. 2000, pp. 327-328.


60 GHz Microwave Broadband Wireless Access Demonstrator for theNext-Generation Mobile Internet

A novel demonstrator with error-free operation for 155 Mb/s wireless access to mobilestations is presented. Key elements are steered antennas with optical beamforming and aremotely controlled self-pulsating laser used as a microwave source at 60 GHz.

Fig. 1: Broadband access system with hybrid-fibre-radio technology. CMCS: central mastercontrol station, CS: control station, BS: base station, MT: mobile terminal.

Fig. 2: Setup for the transmission experiment

arrayantenna

OMCBS

SMF

SMF + DCF

10.13 GHzCMCS CS

SP Laser60.8 GHz

62.07 GHz

GSL

MT 2

MT 1

MOD

155 Mb/sOQPSK

@ 1.27 GHz

4 x

84

Retrieval of digital images based on imagefeatures rather than text descriptions hasbeen the subject of considerable research ef-fort in recent years.

Feature vectors, describing visual proper-ties such as colour, texture and edge statis-tics, are calculated for either the whole im-age or parts of the image. Images are consid-ered similar if the distance between theirfeature vectors is small.

The typical user searches for images inwhich semantic objects such as cars, housesor persons occur. These semantic objectscannot be described by feature vectors calcu-lated on the whole image. An adequate de-scription needs both a suitable representa-tion of the visual properties of image partsand their relative positions.

In our research a new distance measurethat can be used to compare either thewhole or parts of segmented images was de-veloped. It is based on the colour, shape andtexture properties of image segments andtheir relative positions. The user can searchfor semantic objects based on interactivelyselected segments of images.

There is still no system that is able to auto-matically segment still images into semanti-cally meaningful parts. Hence in our system,following a machine segmentation process,we offer the user the possibility of mergingregions that belong together semantically.Extraction of MPEG-7 descriptors for eachsegment then takes place. We use the follow-ing descriptors: Scalable Colour Descriptor,Homogenous Texture Descriptor, Edge His-togram Descriptor and Bounding Box De-scriptor.

In a second step the information-reducedimages are represented by labelled graphs.This allows an image representation that isinvariant to translation, rotation and scale.The set of all segment centroids is the set ofnodes. Every node is labelled with an MPEG-7 feature vector. Edges, labelled with the dis-tances between centroids, are introduced foreach pair of nodes. The distances are nor-malized with respect to the image diagonal.

To find similar images, we calculate a dis-tance measure for the labelled graphs. Wehave proved that this measure is a metric inthe space of isomorphic classes of labelled

graphs with the same lable sets. This can beimportant in using distance-based methodsin computer vision, for instance in semanticrecognition (person walks or sits), and in thearea of machine learning (cluster methods).

[1] S. Bischoff, T. Sikora and F. Wysotzki,“A new distance measure for segmented im-ages based on MPEG-7 descriptors”, submit-ted to European Conference on ComputerVision (ECCV ’02), May 2002, Copenhagen,Denmark.

Stefan Bischoff ([email protected])

H H I R E P O R T 2 0 0 1 · S E L E C T E D C O N T R I B U T I O N S · E L E C T R O N I C I M A G I N G …

Graph-Based Representation of Objects in Image Databases

Measures of similarity of images are important in computer vision. The goal of our re-search is to develop a representation of local image primitives that is invariant to transla-tion, rotation and scale. Its usability in image retrieval is shown.

Fig. 1: Graph-based description

of image segments

Fig. 2: Three sets of images.

Each set contains imageswhich were evaluated as

being similar on the basis of the developed

metric.

85H H I R E P O R T 2 0 0 1 · S E L E C T E D C O N T R I B U T I O N S · E L E C T R O N I C I M A G I N G …

Multimedia data may originate from ordi-nary cameras and microphones as well asfrom specialized devices integrated intonotebook PCs, mobile phones, wrist watch-es, household appliances, etc. A large num-ber of attractive applications become con-ceivable if computers, machines or the sen-sors themselves are provided with SensingPeople functionalities.

In June 2001, a new project called “SensingPeople – Intelligent Cameras and Sensors”started at the HHI. The project aims at theextraction of high level personal data fromoffice and other scenes observed by threelow level sensors: a microphone array, a CCDvideo camera and a CMOS optical devicethat gives a depth profile.

To accomplish this, four partners are coop-erating on this project. The Telecommunica-tions Department at the Technical Universityof Berlin is working on microphone arrays toderive speaker positions and genders fromacoustic data. The Fraunhofer Institute forMicroelectronic Circuits and Systems inDuisburg is building a CMOS camera fordepth profiling. The Fraunhofer Institute forIntegrated Circuits and Applied Electronics inErlangen concentrates on Sensor Fusion, andis responsible in this project for combiningand analysing the data from all the sources.Finally, the Department of Interactive Mediaand Human Factors at the HHI Berlin is work-ing on the analysis of the visual content ofvideo images. It also coordinates the overallproject.

The fusion of three sensor signals of differ-ent types (Fig. 2) makes the acquired infor-mation richer and more robust than wouldbe possible with a single device. For exam-ple, combining video images with depth

profiles gives more reliable estimates of peo-ples' positions, postures and activities. Thesedata also support the extraction of speaker-specific features and the visual description ofpeople by their appearance.

Further, simultaneous analysis of a set ofdifferent features helps to recognise who is inthe office and what he or she is doing or in-tends to do. This information may help ubiq-uitous computers adapt to different users,and it may also help the users to communi-cate in a natural way with the computers.

Sensing People technologies can be ex-ploited in smart houses, smart cars and smartmachines. In addition, many other useful ap-plications, including biometry, are conceiv-able with Sensing People technologies.

This work is supported by the FederalMinistry of Education and Research undergrant 01 AK 900 B.

Thomas Ellerbrock ([email protected])

Sensing People – a New Multimedia Project

Sensing People denotes functionalities that automatically yield high level informationabout people from lower level sensor data such as still images, videos or microphone sig-nals. This information includes a variety of aspects concerning people, e.g. their number,their activity, their state, their gender, their identity and their mutual interaction. Thiscontribution reports on a new project just started at the HHI.

Fig. 1: A general SensingPeople scenario

Fig. 2: a) Speaker signal locatedby a microphone arrayb) Video framec) Grey-value-codeddepth profile from theCMOS camera

86

Retrieval of digital images based on imagefeatures rather than text descriptions hasbeen the subject of considerable research ef-fort over the last few years. Within thisframework the MPEG-7 standard attempts tostandardise suitable descriptors to allowmany image search engines to access asmany distributed images as possible, usingunified and standardised descriptors. Variouslow level image descriptors are currently un-der consideration in the MPEG video group,such as colour and colour distribution de-scriptors and texture descriptors. While theMPEG-7 activity attempts to specify suitabledescriptors, neither the search engines northe methods for displaying the retrieval re-sults will be specified by the MPEG-7 stan-dard, which leaves a large degree of freedomfor product developers to tailor their searchand visualization technology to specific ap-plications.

Most approaches to image database man-agement have focussed on search-by-query.These methods require that users provide anexample image. The database is thensearched for images that are similar to thequery. However, the effectiveness of search-by-query can be questioned, since it is oftendifficult to find or produce good query im-ages, and, perhaps more importantly, re-peated queries often tend to becometrapped in a small group of undesirable im-ages.

To overcome these problems, the usershould to be provided with easy and intuitiveaccess to the information in image databas-es, particularly when confronted with verylarge volumes of images. Hence similarity-based retrieval should return images that aredisplayed not only in order of increasing dis-similarity from the query, but also accordingto their mutual dissimilarities, so that similarimages are grouped together rather than be-ing placed along the entire returned list ofimages. Furthermore, it is important to pro-vide the user with an intuitive browsing envi-ronment that iteratively guides the userthrough the image database – the humanshould be “in the loop”. Only this combina-tion of an intuitive visualization techniquewith a browsing environment based on an

appropriately structured image database pro-vides easy and intuitive access to informa-tion.

Therefore a new image retrieval andbrowsing system has been developed at theHHI [1]. Our approach uses the metaphor ofroad maps. Like road maps with differentscales, ranging from a world map to a citymap, the image space is represented by “im-age maps” with different scales and user-adapted similarity metrics. Beginning with aglobal view, containing only representativeimages (key images) of the entire imagedatabase, the user can enter any domain ofthe database by selecting appropriate keyframes.

Three different technologies in combina-tion guarantee an intuitive browsing environ-ment. These are:• 3D–visualization of the selected images,

the “image map”;• Use of a new relevance feedback tech-

nique, which computes, based on the user’s interaction, the appropriate image similarity metric for the retrieval and visual-ization process;

• Hierarchical organisation of the image database so it can be used with an iterative retrieval procedure which operates at scales ranging the global down to very fine.

The results obtained verify the attractive-ness of this approach for navigation and re-trieval applications.

This work was supported by the FederalMinistry of Economics and Technology of theFederal Republic of Germany under grant 01MD 917/3

[1] T. Meiers, I. Keller and T. Sikora, “3Dbrowsing environment for MPEG-7 imagedatabases”, submitted to S&T/SPIE Work-shop on Storage and Retrieval for MediaDatabases 2002, 20-25 Jan. 2002, San José,California, USA.

Thomas Meiers ([email protected])


Hierarchical 3D Navigation in Image Databases

A new image retrieval and navigation system was developed. This system combines athree dimensional visualization of the image space with an intuitive hierarchical browsingenvironment. The image space is represented by “image maps” with different scales anduser-adapted similarity metrics, analogous to road maps with different scales.


To be successful, user interfaces for imagedatabases must present objects in an appeal-ing manner, must allow easy and intuitivenavigation through the set of objects, andmust highlight their characteristics.

This can be done very successfully by utiliz-ing the very high recognition capability ofthe human visual system. Visual similarityand visual order are major factors in search-ing. If images are arranged according to theirsimilarity rather than the currently wide-spread list presentation, the search will notbe so exhausting. In addition, the user canmark intuitively visual areas matching his orher interest and can refine the search in theseareas. We prefer a 3D space for the visualiza-tion which is caused by a scene movement.

We use MPEG-7 features since this is likelyto become the standard image content de-scription language. However MPEG-7 is notrestricted to visual features, and can easily beexpanded with textural attributes, so thatthe features describing each image may cov-er a very high dimensional feature space, upto 300 or even more dimensions.

For visualization the feature space has tobe reduced to 2 or 3 dimensions. We usePrincipal Component Analysis (PCA) for di-mension reduction, which provides statisti-cally optimum uncorrelated features. How-ever, salient components in a pure statisticalsense might not be relevant to the user, whomay be interested in some particular aspectsthat are of minor statistical importance.Therefore a user-specific weighting of featurecomponents is needed to take into accountpersonal preferences.

This can be done in our system by an ap-propriate weighting of the covariance matrixused for PCA, followed by a new computa-tion of the entire PCA, all done in real time.The user only has to change a slider corre-sponding to his or her special interests, giv-ing an elegant and direct change of visualiza-tion. This enables the user to visualize thosesubspaces that appear relevant to him or her.

Though PCA is a linear technique, its per-formance is powerful and it is sufficient tocompress high dimensional feature spaces toonly three dimensions.

Additionally, we have implemented a truerelevance feedback technique that presentsonly those images from the whole databasethat appear statistically relevant to the user.This technique adapts automatically and in-conspicuously to the user’s hidden search in-tentions, and gives the user a feeling of com-municating with an “intelligent”, flexible en-gine.

This work was supported by the FederalMinistry of Economics and Technology of theFederal Republic of Germany under grant 01MD 917/3.

[1] I. Keller, T. Ellerbrock, T. Meiers and T.Sikora, “Fast user-adaptive weighting ofMPEG7 descriptors for a visual E-Commerceinterface”, 3rd European Workshop onImage Analysis for Multimedia InteractiveServices, Tampere, Finland, 16-17 May 2001.

[2] I. Keller, T. Meiers, T. Ellerbrock and T.Sikora, “Image browsing with PCA-assisteduser interaction”, IEEE Workshop on Con-tent-based Access of Image and Video Librar-ies (CBAIVL-2001), 14 Dec. 2001, Lihue,USA.

Ivo Keller ([email protected])

Fast User-Adaptive Image Search Engines

The visualization of image databases is a task with widespread applications. We have de-veloped a concise and intuitive user interface suitable for both E-Commerce and privateimage collections. It is possible to perform user-adaptive 3D visualization with very effec-tive relevance feedback in real time. Currently, our modules use MPEG-7 features, but canbe expanded to commercially relevant attributes.

Result of a query onshape orientation

88

Various studies have shown that informa-tion that supports group awareness is essen-tial for successful communication betweenmembers of distributed work groups. Thisapplies in particular to information about thecurrent availability of potential communica-tion partners and their willingness to com-municate [1].

That kind of information can be providedby displaying indicators of user states or ac-tivities relevant to their availability for com-munication, such as being absent from thecomputer, talking, etc. (cf. the next contri-bution).

Indicators in the form of availability iconscan be shown on the computer screen or ona mobile phone display. The figure showsicons for a WAP service developed at HHI.

Another possibility for displaying availabili-ty information to mobile users is to use syn-thetic speech that tells a user if the desiredpartner is currently not at his or her desk, isinvolved in a talk with another person, is in ameeting, etc.

Data from which availability informationcan be derived, for either fixed or mobilecomputers, can come from cameras (by ana-lysing motion as an indicator of presence/ab-

sence), microphones (by analysing the hu-man voice as an indicator for being involvedin an audio conversation), or keyboard andmouse activity (as an indicator for using thecomputer).

The KICK project has developed the soft-ware module BINDI (Busy Indicator), whichcan capture and process those data:

BINDI analyses the user’s presence by ana-lysing a video sequence captured by a web-cam located on the user’s monitor. The algo-rithm uses only small images from that ordi-nary webcam and has minimum processingrequirements.

In order to find out whether speech com-munication is taking place, BINDI analysesthe audio signal captured by a microphonenear the user’s desk and looks for frequencypatterns typical of the human voice.

However, such means can usually not beused with real mobile devices, such as mo-bile phones or PDAs. In this case, novelmeans for sensing and analysing availabilityinformation need to be developed. HHI iscurrently considering R&D work on that is-sue. Among other things, sensors for move-ment, temperature and pressure are underdiscussion.

This work is supported by the FederalMinistry of Economics and Technology of theFederal Republic of Germany (BMWi) undergrant 01 MT 901/7.

[1] B. Quante, R. Buss, L. Mühlbach, D.Przewozny and D. Runde, “An experimentalcomparison of two variants of an informalcommunication module for distributed work-groups”, Proc. of the 18th Int. Symp. onHuman Factors in Telecommunication,Bergen, Norway, 2001.

Lothar Mühlbach ([email protected])


Group Awareness for Mobile Applications

Information about the current availability of potential communication partners and theirwillingness to communicate is crucial for distributed groups (e.g. groups of teleworkers).In recent years HHI has developed ways of capturing and analysing data relevant to thisproblem for use in stationary computers. In future, mobile users will desire similar ser-vices.

Availability icons displayed on a WAP

mobile phone


A precondition for successful initiation ofinformal communication is to contact a per-son at a moment convenient to him or her.In face-to-face encounters one would usuallyrefrain from speaking to a person who seemsto be very busy or is involved in conversationwith another person. But when at anotherlocation, it is difficult to decide whether ornot a person is addressable due to the factthat communicative signals and awarenessinformation are missing.

Video communications, however, whichhave the potential to transmit such cues,have turned out to cause user resistance dueto fears of violating privacy.

In view of this we have developed a mod-ule BINDI (Busy Indicator) that automaticallyrecognises whether the other user is pre-sent/absent at his or her workplace, is cur-rently talking to someone else, or is using thecomputer. These cues serve as indicators of aperson’s availability for communication andare displayed in both a chat variant, with aconventional user interface similar to those ofcommon chat systems, and in a virtual envi-ronment (VE), in which each user is repre-sented by an avatar.

The various user states are presented eitheras availability icons or in the form of avataractions in the VE (see the figures).

Additionally, electronic calendar data areprovided to predict a person’s availability inthe near future.

Results of an experimental field trial, in

which three distributed project groups testedboth variants during their everyday work,showed that the availability information pro-vided by BINDI was useful and facilitated theassessment of a potential interlocutor’s avail-ability for informal communication, particu-larly when this information is provided byavailability icons.

Availability icons turned out to be easy tointerpret and to give an overview of ongoinggroup activities at a glance. The VE variantcould produce a considerable degree oftelepresence, and users who were used tovirtual environments experienced the tele-worker as being more present when usingthe VE variant, although subjects did nothave the feeling of sharing a common room.The graphical representation of the entirework group in a virtual team room integrat-ed the teleworkers visually, and this in turnincreased the feeling of telepresence.

The acceptance of the tested variants wassuch that all users could imagine continuingto use the Chat variant after the end of thefield trial.

Moreover, privacy concerns did not turnout to be severe with either variant.

This work is supported by the FederalMinistry of Economics and Technology of theFederal Republic of Germany (BMWi) undergrant 01 MT 901/7.

Birgit Quante ([email protected])

Supporting Group Awareness in Teleworking Applications – HumanFactors Aspects

Telework, distributed workgroups and virtual enterprises will be common in the near fu-ture. Novel communication systems with user-friendly communication management arerequired for these working schemes. At present, simultaneous informal communicationbetween distributed team members is rarely supported by conventional groupware sys-tems. Since informal communication is unplanned, users need to be made aware of thecurrent status of their co-workers.

Fig. 1: Scene taken from thevirtual environment

Fig. 2: Availability icons

talking& usingthe PC

talk-ing

pre-sent

ab-sent

statusnot

known

usingthePC

?

90

Most of the known 3D display conceptshave one serious disadvantage when it comesto supporting direct virtual object manipula-tion with the hands or other physical tools.This problem is arises because the stereo im-ages are displayed on a physical displayscreen, whereas the spatial position of thevirtual object may stereoscopically appear tobe at a closer distance (e.g. within reach).Users have to accommodate, or focus, theireyes at the depth of the screen surface, in or-der to see an unblurred, clear stereo image,whereas the lines of sight converge at theapparent position of the virtual object. Theresulting conflict between accommodationand vergence is a potential contributor toeye strain and, moreover, it is a major factorin perceptual errors with stereoscopic dis-plays (regarding the perceived depth andsize of the virtual object).

As well as the accommodation-vergenceconflict with current 3D displays, there willalmost always be an accommodation mis-match between real and virtual objects. Theaccommodation distance for virtual objects isalways the distance from the viewer to thedisplay screen, whereas the accommodationdistance for the real object corresponds tothe position of the object. When, for exam-ple, the user tries to align the finger or a toolwith the virtual object, he or she will see ei-ther the virtual object (accommodating onthe screen) or the real object (when lookingat it) in full focus, but not both in focus atonce. Natural interaction, in which both thevirtual and real objects are seen in full focus,is hence not possible.

We have found a solution to this problemin the Accommodation 3D Display [2],where the stereo image pair is focussed onan aerial image plane floating in the user'sgrasp area (Fig. 1). The prototype displayprovides a symmetrical stereo resolution of2 x (1024 x 768) pixels. The 3D display cre-ates two circular viewing windows of diame-ter 65 mm at a distance of 70 cm from thedisplay. Looking through the windows, the

left-eye and right-eye images are fused,forming a 3D image which floats 20 cm infront of the display. Hence the accommoda-tion and convergence distances are bothfixed at about 50 cm from the user's eyes, al-lowing comfortable and unhindered hand in-teraction with a virtual object (Fig. 2). Thetwo high-resolution images forming thestereo pair are optically channelled to theeyes without any visible crosstalk (perfectstereo separation).

This work was supported by the FederalMinistry of Education and Research undergrant 01 BK 802.

[1] The term Mixed Reality was coined byPaul Milgram, University of Toronto.

[2] S. Pastoor and J. Liu, “3D display andinteraction technologies for desktop comput-ing”, in Three-Dimensional Television, Videoand Display Technology, edited by B. Javidiand F. Okano, Springer Verlag (in print).

Siegmund Pastoor ([email protected])


Next Generation 3D Displays for Mixed-Reality Applications

Rapid progress in computer technologies will enable mixed-reality applications, in whichcomputer-generated views of virtual objects and environments are combined in someproportion with the user's real environment [1]. The goal is to seamlessly integrate thevirtual objects into the user's real workspace, thereby avoiding any unnatural perceptualartefacts and allowing direct interaction and precision manipulation. Ideally, there shouldbe no noticeable difference between interacting with real and computer-generated ob-jects. We have recently presented the first working prototype of a high-resolution mixed-reality 3D display to the public.

Fig. 1: Basic design of a 3D dis-play for desktop mixed-reality applications. The

aerial 3D image of thevirtual object floats inthe user's grasp area.

The two circular regionson the right indicate thestereo viewing windows.

Fig. 2: This original photo

demonstrates that both

the user's hand and thevirtual object appear at

the same focal distance.The 3D image is very

bright (more than 500cd/m2) because of the

optical principle used inthe display.


Computer vision systems for recognizingand tracking people and their behaviourshould be fast, accurate, robust and informa-tive. The common challenging problems areto deal with changing environment (illumi-nation, other objects of no interest movingin the background) and changing appear-ance of the tracked objects (size, non-rigidform, orientation). Also, to make vision sys-tems useful for human-computer interaction,the whole system delay must be kept to anacceptable range.

In the mUltimo3D project we use a set ofcameras to track people's head/eye position,gaze direction, hand position and simplegestures.

Head/eye tracker: In the initialisationphase, user-specific eye patterns are found invideo images by detecting the user's eyeblinking or by means of a pre-stored eyedatabase. After that, the eye patterns aretracked using the block matching technique.In order to cope with variation of the illumi-nation, two strategies are used: 1) both thepre-stored reference eye patterns and thecandidate patterns in the current image areluminance-adapted before being comparedwith each other; 2) the gain of the videograbber is actively controlled by the bright-ness of the actual image (or the region of in-

terest), to ensure that the captured imagesare not too dark or too bright due to thevarying lighting conditions.

Gaze tracker: In order to estimate thepoint of fixation of the user, the user's eye isilluminated by several low-intensity infraredlights. The middle points of both the pupiland the highlights are detected by thecornea-reflex method. A head-fixed transfor-mation is applied to compensate for the in-fluences of free head movement and posechanges of the gaze camera. A self-calibratingalgorithm is used, which significantly simpli-fies the troublesome calibration process.

Hand Tracker: Our hand tracker usestwo infrared-sensitive cameras and an in-frared light source arranged next to the cam-eras. Simple hand gestures (hand opened orclosed and the pointing direction of the in-dex finger) are recognized by analysing thedistribution of the hand vectors, which pointfrom the centre of the hand to the fingers.The determination of the 3D hand position isbased on a correspondence analysis.


Jin Liu ([email protected])

Tracking Technologies for Multimodal Human-Computer Interfaces

Sensing people and their behaviour is the most important part of perceptual user inter-faces. For this purpose, we have developed in the mUltimo3D project a number of real-time image segmentation and pattern recognition algorithms for detecting and trackingpeople's head/eye position, hand position and gestures, and gaze direction. Great effortshave been made to achieve robust tracking performance under various lighting condi-tions and varying sizes and orientations of the tracked objects.

Four cameras arearranged around the display and next to thekeyboard – one for measuring the head/eyeposition, one for tracking the gaze direction, and two fortracking the hand position and gesture

keyboard / mouse

speech

hand image

eye image

head image

head tracking3D eye position (display control)

(head -motionparallax)

3D eye position

point of fixation

hand positionand gesture

commandscontinuous speech

conventional input devices

gaze tracking

hand tracking

3D perceptualuser interface

autostereoscopicdisplayspeech recognition

92

Humans perceive information through mul-tiple input channels using the senses of sight,hearing, touch, smell, taste and balance.Perception via the visual and auditive modali-ties generally outperforms the haptic, olfac-tory and gustatory modalities in terms ofbandwidth and spatial and temporal resolu-tion of the information transmitted, which isone of the reasons why visual and auditorydisplays have dominated human-computerinteraction. Various output channels, such asspeech, gestures and written text, are usedby humans to communicate with others.

The interest in multimodal interfaces isbased on the assumptions that: (1) a muchhigher bandwidth is possible with these sys-tems; (2) different modalities excel at differ-ent tasks; (3) errors and disfluencies can bereduced dramatically, since the differentchannels can complement and disambiguateeach other; (4) multimodal interfaces are morenatural and hence more engaging.

Multimodal output leverages the humansenses and the perceptual system becausewe can perceive several things at once; mul-tiple input leverages the human motor capa-bilities and communication skills because wecan do several things at once [1]. The con-cept of multimodal interaction means notsimply the provision of the display and sen-sory techniques that enable such interactionwith a computer (which is of course not sim-ple from an engineering viewpoint). The realchallenge is to fuse the various modalities inan intelligent way, taking into account howthe integrating modalities relate in a naturalenvironment. To find guidelines for this is amajor task of our current research (see thefigure and reference [2]).

Existing results indicate that speech inputseems to be relevant in any multimodal sys-tem, probably because speech is the mostpowerful human output modality. Users gen-erally prefer combined inputs over systemsoffering only a single modality. Combiningseveral modalities, such as hand gesture, eyegaze and speech input, increases the overallsystem performance.

The various input modalities have quite

different characteristics. For example, posi-tioning a cursor by eye gaze interaction isquicker but less precise than positioning byhand movement; natural-language speechinput is perfectly suited to descriptive tasks,whereas hand gestures are the ideal solutionfor the direct manipulation of objects.Challenges in designing an efficient user in-terface lie in minimizing the time needed toswitch among the modalities available and insimplifying the learning process needed tounderstand how interactions with the com-puter relate metaphorically to correspondingprocesses in the real world.


[1] G. Robertson: “Leveraging human ca-pabilities in perceptual interfaces”, talk at thePUI Conference, 1998.

[2] S. Pastoor and J. Liu, “3D display andinteraction technologies for desktop comput-ing” in Three-Dimensional Television, Videoand Display Technology, edited by B. Javidiand F. Okano, Springer Verlag ( in print).

Siegmund Pastoor ([email protected])


Human Factors in Multimodal Human-Computer Interaction

Multimodal interaction is expected to enable end users with varied backgrounds to mas-ter with ease the increasing functionalities of computers, and to make the use of comput-ers in general an enjoyable experience. Humans have developed various modalities tocommunicate with others and to interact with their physical environment. The basic ideaof a multimodal interface is to adapt such familiar modalities to the particular require-ments of human-computer interaction.

Multimodal human computer interface usinga 3D display with speechinput and output as well

as conventional inputdevices (keyboard,

mouse). Additionally,various video trackers

sensing the user's headposition, eye movements

(gaze point) and handgestures are available for

interaction. This systemis currently being used in

user trials, in order tooptimise system parame-

ters and to developguidelines on how to

fuse various modalitiesin an intelligent way.


In 1998, the Video Coding Experts Groupof ITU-T Study Group 16, Q.6 (earlier knownas Q.15) started to develop a new videocompression algorithm in a standardizationproject named H.26L. The promotion of thefirst version (H.26L baseline) is scheduled forlate 2002. The main two targets of this pro-ject are: a) coding efficiency (it aims to re-duce the bit-rate for a given subjective quali-ty by about 50 %, compared to H.263+ orMPEG-4 Advanced Simple Profile); and b)network friendliness. Considerable improve-ments in coding efficiency of 30-50 % havealready been achieved with the current ver-sion of the test model (TML8). The main fea-tures of TML8 are as follows:• Spatial intra-block prediction• 4 x 4 block transform• Motion compensation with multiple block

sizes• Multiple-reference frame prediction• Quarter-pel motion accuracy• Rate-distortion optimised encoder

decisions

The basic entropy coding method in TML8uses a single, universal variable length code(UVLC) for all coding elements. It has beenshown that the UVLC is not optimal with re-spect to coding efficiency. The simple designof the UVLC assumes that the underlyingprobability distribution is static and ignorescorrelations between symbols. In addition,the need to use at least one bit per symbolcalls for a joint coding of symbols.

In order to improve on these restrictions ofthe UVLC design, we have developed a fun-damentally different approach based on con-text-based adaptive binary arithmetic coding(CABAC). The concepts behind CABAC areknown to lead to highly efficient entropycoding. Similar techniques are used in theemerging still image compression standardJPEG-2000, but so far they have rarely beenused for video compression.

The advantages of our approach are three-fold compared to entropy coding using afixed variable length code:

1. Context modelling provides estimates ofconditional probabilities of the coded sym-bols. Utilizing suitable context models, inter-symbol redundancy can be exploited byswitching between different probability mod-els, depending on previously coded symbols.

2. Arithmetic codes permit a non-integernumber of bits to be assigned to each sym-bol of the alphabet. Thus the symbols can becoded almost at their entropy rate. This is ex-tremely beneficial for symbol probabilitiesmuch greater than 0.5. In this case a variablelength code has to use at least one bit,whereas arithmetic codes may use a fractionof one bit.

3. Adaptive arithmetic codes permit theentropy coder to adapt to non-stationarysymbol statistics. For example, the statisticsof motion vector magnitudes vary over spaceand time as well as for different sequencesand bit-rates. Hence, an adaptive model tak-ing into account the cumulative probabilitiesof previously coded motion vectors leads to abetter fit of the arithmetic codes to the cur-rent symbol statistics.

By using our new CABAC scheme insteadof UVLC in the TML, large bit-rate savings ofup to 30 % can be achieved. As a remarkableoutcome we have observed high codinggains not only at high bit rates, but also atvery low rates.

Detlev Marpe ([email protected])

CABAC – Advanced Entropy Coding in the Draft ITU-T H.26L VideoCompression Standard

A new entropy coding scheme for video compression is presented. Context models areutilized for the efficient coding of symbols. A novel binary adaptive arithmetic codingtechnique is employed to match the conditional entropy of the coding symbols, given thecontext model estimates. An adaptation mechanism is used to keep track of non-station-ary symbol statistics. Our new approach has been adopted by ITU-T as an advanced en-tropy coding method in the current ITU-T H.26L test model (TML).

Sample reconstructionsat 16 kbit/sec. Left: H.26L using CABAC.Right: H.26L using UVLC

94

MPEG issued a Call for Proposals on NewTools For Video Compression Technology [1]at the beginning of 2001. For the subjectivetests, carried out in July 2001, various videosequences were used to test the coding effi-ciency, in comparison with MPEG-4 Visual, atdifferent bit rates.

The ITU-T Video Coding Expert Group(VCEG) is currently developing a new videocoding standard known as H.26L [2]. TheH.26L project, to which the HHI is activelycontributing, took part in the MPEG codingefficiency tests.

Common to the ISO/IEC standard MPEG-4Visual and the ITU-T project H.26L is thatthey both define so-called block-based hy-brid video coders. The basic source codingalgorithm is a hybrid of inter-picture predic-tion, to take advantage of temporal redun-dancy, and transform coding of the predic-tion error signal, to reduce spatial redundan-cy. Although the H.26L design basicallyfollows that of earlier video coding standardssuch as MPEG-4 Visual, it contains new fea-tures that enable it to achieve significant im-provements in compression efficiency com-pared to previous coding standards.

MPEG-4 Visual and H.26L specify only thebit-stream syntax and the decoding process.The encoding process is left out of the scopeto permit flexible implementations and a-chieve interoperability. A rate-distortion-opti-mized coder using Lagrangian techniqueswas developed at the HHI for both MPEG-4Visual and H.26L. In both cases the improvedencoding strategy provides visible perfor-mance gains compared to the previous en-coding strategies. Following a proposal bythe HHI, the Lagrangian encoding strategywas integrated into the current test modelTML-8 for the H.26L project. Both rate-dis-tortion-optimized coders have been chosenby the standardization bodies, MPEG andITU-T, as their standard encoders for thecoding efficiency tests in July 2001. ForMPEG-4 Visual, an alternative implementa-tion by UBVideo, which mainly follows the

approach of the HHI, was also provided, andthe MPEG-4 anchors were chosen from bothimplementations.

For the set of sequences and bit rates spec-ified in MPEG’s Call for Proposals on NewTools For Video Compression Efficiency, theH.26L-compliant encoder achieved perfor-mance gains of 1.1-3.6 dB over the MPEG-4encoder. Subjective tests carried out at theHHI showed that, for all cases in MPEG’s testset, sequences coded at 512 kbit/s withH.26L are subjectively equivalent to the samesequences coded at 1024 kbit/s with MPEG-4 Visual. This corresponds to a bit rate savingfor H.26L of about 50 % over MPEG-4 Visual.At lower bit-rates the tests seem to indicatebit rate savings of 30-40 %. The figure showsa PSNR vs bit-rate comparison of the twocodecs for the sequence Foreman.

[1] ISO/IEC JTC1/SC29/WG11, Call forProposals on New Tools For Video Compres-sion, Doc. N4065, 2001.

[2] G. Bjontegaard and T. Wiegand, H.26LTest Model Long Term Number 8 (TML-8)draft 0, ITU-T VCEG, June 2001.

Heiko Schwarz ([email protected])


MPEG Coding Efficiency Tests

The Moving Picture Experts Group (MPEG) recently issued a Call for Proposals on newtools to further improve video coding efficiency above that of their MPEG-4 Visual stan-dard. The current project of the ITU-T Video Coding Expert Group, known as H.26L, tookpart in MPEG’s coding efficiency tests in July 2001. The HHI has provided a rate-distor-tion-optimized encoding strategy for both MPEG-4 Visual and H.26L, and this was chosenby the standardization bodies as their standard encoding method for the coding efficien-cy tests.

Performance comparisonof the current H.26L test

model and the MPEG-4Advanced Simple Profile

for a typical video testsequence

0 200 600

H.26L(Sep. 2001)

42 %

MPEG - 4 ASP(1999)

AcceptableQualitity Level

800400Bitrate [kbit/s]

36

38

28

32

30

34

Qua

lity-

PS

NR

[dB

]


Streaming media distribution is a rapidlygrowing application, especially with the up-coming broadband mobile phone networksand the increasing number of internet sub-scribers with broadband access, which en-ables high quality multimedia applications.There are already several solutions availablefrom vendors such as Apple, Microsoft andRealnetworks. Unfortunately these solutionsuse different transport schemes and mediatypes, incompatible with each other andmostly not usable for mobile transmission.The recently released ISMA specification of-fers an open standard for media streaming,with the goal of reducing market fragmenta-tion. Especially for user convenience, it focus-es on a common media format. The experi-mental framework described here will be ex-tended to full compatibility with the ISMAstreaming standard.

The streaming framework is built aroundthe MPEG-4 standard and provides supportfor real-time encoding and decoding ofMPEG-4 media to satisfy the demands of livetransmission. Several aspects relating to widearea network distribution are considered,such as cascading and forwarding.

The signalling scheme is based on RTSP,which uses TCP/IP as the transport protocol.To fulfil the requirements of media stream-ing, a request-response scheme for server-client communication is implemented. Com-munication on the signalling channel is es-tablished using a small set of commands. Atypical streaming session can be divided intothree phases: negotiation, streaming andshutdown.

The negotiation phase incorporates twoobligatory steps. First of all, a DESCRIBE com-mand is sent to the server for the requestedmedia. The response contains the media andinitial object descriptions for decoder initiali-sation in SDP format. The client then choosesfrom the available streams and issues at leastone SETUP command, which contains thedesired method of stream transport. In thecase of RTP transport, the client's UDP portpair and unicast/multicast choice are passedas arguments to the setup command.

After establishing the streaming connec-

tion, the second phase starts with the send-ing of the PLAY command, which starts thetransmission over the RTP channels and al-lows positioning within the stream to a-chieve a “skip” or “jump” feature.

The user is able to interrupt the streamingtransmission by using the PAUSE command.The third phase, the shutdown, of a typicalmedia transmission is initiated by the TEAR-DOWN message, which ends the streamingsession. The media transmission is achievedwith RTP, which is based on the UDP data-gram protocol. For each media channel oneRTP channel is used. To avoid unnecessarydata packet overhead, the data are packedwithout using the MPEG-4 SL packetizationscheme. Synchronization is performed usingRTP timestamps, which are already definedin the RTP specification.

The currently implemented frameworkcomprises a server for data retrieval and real-time media encoding. The players are avail-able for different platforms, such as Win2kand Linux. To be able to test the applicationin real mobile environments, the player is al-so available for WinCe-based PDAs. The IP-based networks that were tested includedLAN, Wireless LAN, T-DSL and GRPS undervarious conditions. For more information, re-fer to [1].

[1] B. Stabernack and H. Richter, “A multi-media streaming framework for mobile appli-cations – A first approach”, Proc. 12th Packet-Video Workshop 2002, Pittsburgh PA, April2002.

Benno Stabernack ([email protected])

Multimedia Systems based on MPEG-4 for Internet and MobileTransmission

The transmission of multimedia data over currently available IP-based networks is de-scribed. The coding scheme used for transmission is compliant with both MPEG-4 and theproposed transmission scheme. The xperimental system described here is for the evalua-tion of IP-based multimedia data transmission over mobile networks.

96

Several HHI projects with industrial and re-search partners use MPEG-4 technologies.These projects differ in their application area,transmission technology and target platform.In order to support all the different require-ments defined by these projects, a projecthas been launched that implements a gener-ic concept with a flexible design and reusa-ble MPEG-4 software components. This con-tribution focuses on MPEG-4 DMIF, which isthe key to supporting various transmissiontechnologies through a common software li-brary.

It is a key feature of the DMIF standardthat it hides the transmission layer from theapplication by inserting the DMIF applicationinterface (DAI). Any application, whether inthe automotive, TV, mobile or PC industry,can access streamed content through thisAPI.

In a similar manner, a well-designed DMIFimplementation should separate issues thatare specific to the transmission media fromthose DMIF functionalities that are commonto all scenarios.

The diagram illustrates the architecture ofHHI’s DMIF implementation. The DAI allowsthe application to set up the correct instancefor the chosen transport media. The imple-mentation provides the specific features ofeach media access protocol in separate in-stances, and uses a common transport inter-

face, which allows the mapping of the DMIFdefault signalling to the transport-specificsignalling.

The software development of the DMIF li-brary was carried out using state-of-the-artsoftware design methodology and a varietyof tools. The architecture was designed usingthe standardised Unified Modelling Lan-guage (UML) with Rational Rose. The C++source code, its related design data and thedocumentation are all managed undersource control with the Concurrent VersionsSystem (CVS). Any project requiring MPEG-4streaming has access to the library via theCVS server. An automated process, using Ra-tional SoDA, generates a consistent and up-to-date version of the documentation. It ismade available, as both a hyperlinked Worddocument and an HTML web page, to allprojects over the Intranet of the image pro-cessing department. In order to support mul-tiple target platforms, the implementationmakes use of the HHI Multiple OperatingSystems Library (MSysLib), which provides acommon interface for objects that are specif-

ic to the operating system, such as threads,sockets and file I/O to name but a few.

The DMIF library is thus a powerful andeasy-to-use tool for MPEG-4 streaming solu-tions.

Jens Güther ([email protected])


MPEG-4 Streaming Solutions using the Delivery Multimedia IntegrationFramework

Next generation multimedia applications are based on the evolving MPEG-4 standard.The MPEG-4 Delivery Multimedia Integration Framework (DMIF) provides a standardisedand flexible technology to support both various transmission scenarios and local storage.This contribution describes a reusable and well-designed software implementation ofMPEG-4 streaming using DMIF.

Transport

Transport interface

DAI

DAI & control

Data flowControl flow

Defaultsignalling

Localinstance

DMIF filter (switch to the appropriate instance)

Application

Remote interactiveinstance

Broadcastinstance

Groupinstance

FlexMux

Generic design of theDelivery Multimedia

Integration Framework


The data rates of uncompressed high qual-ity audio signals (e.g. PCM 16 bit at 44.1kHz) are very large. There has therefore beena proliferation of audio compression tech-niques and formats over the last few years.MPEG-2 layers 2 and 3 are frequently usedalgorithms for different types of internet au-dio (layer 3 is well known as MP3). Otherstandards for multimedia applications areAAC and CELP.

AAC (Advanced Audio Coding) is more ef-ficient than previous compression techniquesbecause of its many improvements. Thisstandard supports sampling rates from 8 kHzto 96 kHz and provides audio of almost CDquality at a very low bit rate (64 kb per chan-nel). The AAC standard is also capable ofhandling up to 48 channels in one bitstream, so it is suitable for any multi-channelapplication. MPEG-4 AAC can be used in awide range of applications such as home cin-ema, internet audio and hand-held playersand recorders.

The high coding efficiency of AAC isachieved by a combination of several tech-niques, including high resolution filter banks,adaptive prediction and Huffman coding.Due to these computationally intensive algo-rithms, the encoding process requires a hugeamount of computational power. Varioushardware solutions for implementing audiocodecs have been evaluated.

Real-time implementation on a DSP

A floating point architecture is mandatoryfor the implementation of the encoder be-cause of the required precision, whereas theAAC decoder can be implemented on a 32bit integer processor. The TMS320 C6x DSPseries from Texas Instruments [1] has beenchosen for the implementation of the AACcodec to realize a hand-held stereo player/re-corder.

The core of this application is a DSP oper-ating with a 167 MHz clock and a 3.3 Vpower supply. The DSP is booted from flashand works with 1 MB of SBSRAM. A MultiMedia Card (MMC) and a standard RS232interface are available for data input and out-

put. Encoded audio data are stored on theMMC. The configuration of this hand-helddevice can be done by remote control, whilethe status of the application is shown on theintegrated device display.

Two inputs for analogue audio data areavailable: stereo LINE IN and MIC IN. Onestereo LINE OUT port is used for audio out-put. A high fidelity 16 Bit Audio ADC/DACconverts analogue data into digital data andvice versa.

Player implementation on a RISCprocessor

RISC processors such as the SA-1110(StrongARM) from Intel [2] are possible alter-natives to DSPs. The processor performanceand the low power consumption enable theimplementation of modules for hand-heldsystems, especially for audio applications.

The SA-1110 module used for the codecimplementation incorporates a 32 bitStrongARM with a CPU clock at 206 MHz.This platform supports up to 32 MB of fastflash memory and uses Embedded Linux asthe operating system.

The player supports a graphical user inter-face and includes decoders for AAC, MP3,CELP and MPEG-4 Video. The algorithmswere ported and optimised for theStrongARM processor architecture, resultingin high performance implementations. For allthree audio standards the decoding processcan easily be achieved, but real-time encod-ing with fixed-point arithmetic requiresmuch optimisation, as the ongoing workshows.

[1] Texas Instruments DSP Technology:http://dspvillage.ti.com/docs/dspproducthome.jhtml

[2] Intel StrongARM:http://www.unique.memec.com/uk/uniquesolutions/pdasolution/sa1110/sa1110.htm

Maati Talmi ([email protected])

Audio Player on DSP and RISC Processors

The availability of powerful DSP and RISC processors makes possible the implementationof complex audio codecs on dedicated modules in hand-held devices. MPEG-4 audio en-coding and decoding for real-time applications has been implemented using both a TexasInstruments DSP and the RISC processor “StrongARM”. This paper discusses the imple-mentation results.

98

Drastically increased processor perfor-mance, large varieties of PCI componentsand comprehensive software developmenttools for standard PCI architectures also givenew possibilities for complex video-process-ing architectures.

Video processing extensions for the VIRTUE(VIRtual Team User Environment) conferenceterminal are based on a newly developedVIRTUE Processor Station (VPS), and realizethe advantages of a PCI-based system archi-tecture. The VPS architecture aims to meetthe requirements for multiple camera inter-facing, support of flexible system topologiesusing a modular structure, and cost-orientedscalability.

The basic element of the VPS is a subsys-tem containing four Trimedia processor cellsembedded in the host’s PCI hierarchy usinga transparent PCI bridge. Multiple subsys-tems can be used to enhance the overall pro-cessing power. To bypass bandwidth limita-tions on the PCI buses, the system architec-ture offers three ways of communicatingbetween the processor cells and other sys-tem components (Fig. 1). Method A refers todata transfers using the transparent PCIbridge. This method is recommended onlyfor communication with non-VPS devicessuch as the graphic adapter or system mem-ory, and should be rarely used for communi-cation between VPS cells. VPS cells with highdata exchange rates should be placed in thesame package, so that the host PCI bus is notaccessed (Method B). If the data rate ex-ceeds the bandwidth limitation on the localPCI bus, unidirectional connections (MethodC) may be used to unload the bus. Unidirec-

tional connections are based on the compati-ble digital video interfaces of each processorcell. The combination of these three meth-ods gives the flexibility to balance bandwidthrequirements for various video processingtasks.

The VPS concept features an open archi-tecture based on a PCI processor card con-taining the Trimedia processor systems (Fig.2) and a separate daughter card containingthe video interfaces. The common video in-terfaces are the TriMedia video ports (low-voltage TTL 656 interfaces), which give low-cost designs for daughter boards containinginterface circuits for serial 656 ECL (SDI,SMPTE 259M) interfaces, Channel Link LVDSinterfaces or other digital video interfaces.

The mechanical specification results in apackage which does not exceed the maxi-mum component height of a PCI expansioncard. A low-cost Channel Link interface cardthat provides four video input ports, fourvideo output ports and a router FPGA hasbeen developed, and meets the require-ments of the VIRTUE system. A more flexiblesolution is a SDI/LVDS interface card thatcomes with dynamically configurable routerFPGAs (design downloaded from Trimedia).

VIRTUE is a project within the EuropeanUnion's Information Societies TechnologyProgramme (IST).

Michael Karl ([email protected])


VPS – a DSP-Based PCI Board for Real-Time 3D Video Processing

Based on low-cost multimedia processors, the VPS architecture combines state-of-the-artPC technology with standard video interfaces for studio applications. Originally designedfor an Immersive Videoconference System, the modular, scalable and cost efficient multi-processor structure can also be used in other demanding signal processing applications.

PCI Bridge 32 - bit / 33 MHz

A

C

B

TM - 1300Processor

cell

TM - 1300Processor

cell

TM - 1300Processor

cell

TM - 1300Processor

cell

Video 656interface

in outVideo 656interface



in out

Fig. 1: VPS subsystem

Fig. 2: VPS Trimedia board


Each year about 3.5 billion cinema ticketsare sold worldwide. Technology has notchanged over the past decades: film materi-al is exposed, processed, edited, multipliedin film labs, and shown using a 35 mm filmprojector. Digital processing in the film in-dustry is rarely used, except when special ef-fects are added. Copying the film material isan elaborate and expensive process, whichinduces a loss of quality, as does repeatedprojection. Today studios are responsible forsupplying prints to theatres. The studios man-ufacture them and pay their cost.

The absence of digital processing comesabout because of the very high technical re-quirements of a complete digital processingline. Even when films are produced digitally,the end of the chain is an analogue 35 mmfilm, because the main problem is the the-atre projector with an adequately high reso-lution. Therefore most of the key componentsare also analogue, which impacts on theconnections to other media and applications.

The objective of the E-Cinema project isthe realisation and standardisation of a com-plete digital cinema system chain, startingwith the capturing process, following withpostproduction and then distribution by dig-ital transmission (download and onlinemodes), and finishing with the reproductionand local media storage.

Key points being investigated in the E-Cinema project are:• Digital cameras with online compression

and mobile data storage• Real-time encoding and decoding for

distribution via cable or satellite• High resolution digital projection• High quality spatial audio recording and

playback• Electronic data storage systems, archives,

encryption and IP management

The HHI is working mainly on the distribu-tion and receiving part of the cinema systemchain. To solve the problem of high resolu-

tion reproduction, a modular configurationof the projection system is proposed. Thesystem consists of multiple projectors, eachworking at High Definition resolution. To geta real immersive impression, as required forcinema applications, the different views willbe merged together seamlessly by theUHDTV (Ultra High Definition Television)merger unit. The hardware architecture de-veloped for the receiving part of the E-Cine-ma prototype system is shown in Fig. 1. Itcomprises three Hyperboxes (MPEG-2 HDRecorder/Decoder/Synchronizer) and theHiMerger for combining the different viewsto an ultra-high resolution picture.

This work is supported by the FederalMinistry of Education and Research undergrant FKZ 01 AK 040D.

Ulrich Höfker ([email protected])

E-Cinema – an Approach to Digital Cinema

In contrast to video, audio and internet media, which are nowadays processed completelydigitally, the cinema industry seems to have missed the connection to up-to-date technol-ogy. One reason for this stagnating development is the unavailability of several key com-ponents because to their very high technological requirements. The E-Cinema project hasbeen started with the objective of finding an approach to digital technologies which canrealize its advantages for cinema, in order to close the gap between traditional cinemaand the new digital multimedia revolution. A completely new service possible with digitalcinema will be the ability to also show live events such as the Olympic games or greatperformances.

Fig. 1: Receiving part of thedigital cinema systemchain developed by HHI

Fig. 2: Seamless ultra-high resolution view by multiple HD projectors

Cascadable hyperboxes

DVB transmission

UHDTV merger unitRemote

controllingHD

projectors

HiMerger

HyperBox HyperBox HyperBox

High resolution panoramic view

MultipleHD projektion

100

Real-time depth estimation in naturalstereo image pairs plays an important role ina large number of applications. Depth esti-mation is performed by the matching ofpoints in two images such that the matchedpoints are the projections of the same pointin the 3D scene. The disparity map obtainedfrom this matching process can be used torender a virtual view of the scene by an im-age-based rendering technique using knowl-edge of the stereo geometry. This is per-formed in immersive media applications suchas immersive TV or immersive video confer-encing. Highly reliable dense disparity mapshave to be processed in real-time for theseapplications.

To reach this goal an algorithmic frame-work has to be developed which is a trade-off between quality of the result and the real-time constraint. The idea of this framework isto estimate disparities on a sparse grid (4 x 4or 8 x 8), to post-process them, and then tointerpolate them to form dense disparitymaps.

Disparities are estimated with a hybrid re-cursive matching (HRM) approach, whichunites the advantages of block-recursivematching and pixel-recursive optical flow es-timation in one common scheme. Its compu-tational requirements are met by the efficientselection of a small number of candidate vec-tors, guaranteeing both spatial and temporalconsistency of the disparities. This algorithmhas already been used successfully for fastmotion estimation in MPEG coding and stan-dards conversion.

However, disparity analysis requires morethan just a straightforward application ofHRM. The crucial problem is to produce cor-rect disparities in every area of the image. Inparticular, occlusions and homogeneous orless structured regions cause visible errors inthe disparity maps and lead to artefacts in asynthesized virtual view. To cope with theseproblematic areas, the sparse disparity mapsof the HRM have to be post-processed tocreate a map of highly reliable disparities. Ina first step the mismatches are detected andthen substituted using an intelligent interpo-lation and extrapolation scheme. A consis-tency check is used to detect mismatches.

The detected mismatches are rejected andreplaced using an interpolation/extrapolationscheme that guarantees that depth disconti-nuities at object borders are retained. Seg-mentation and colour information from theoriginal images is used to perform this task.

Another problem is that a simple interpola-tion scheme in the dense field calculationwould lead to incorrect disparities at depthdiscontinuities. To prevent these errors a sim-ilar intelligent interpolation/extrapolationscheme is used when enlarging the disparitymaps to full resolution.

The disparity estimation and the post-pro-cessing procedure run on a fast Pentium IIIsystem at full CCIR 601 resolution.

This work is supported by the DeutscheForschungsgemeinschaft (DFG) under grantDD 20 9 11.

Nicole Brandenburg ([email protected])


Real-Time Solution for Depth Estimation in Immersive Media Applications

The aim of immersive media applications being developed at HHI is to render arbitraryvirtual views of a scene. A key ingredient of immersive systems is disparity estimation,which provides depth information in real time. An algorithmic framework has to be devel-oped which is a trade-off between quality of the result and the real-time constraint.

segm. driveninterpolation

rectification rectification

leftdisparity map

leftsegmented image

rightsegmented image

rightdisparity map

segm. driveninterpolation

segm. drivenpost-processing

consistency check

segm

enta

tion

hybrid recursive matching

segm. driven segm

enta

tion

post-processing

Real-time disparity estimation framework


The goal of immersive video-conferencingsystems is to allow geographically separatedconference participants to have the full spec-trum of experiences they are used to in real-world meetings.

To achieve this goal, 3D images of the par-ticipants are synthesised and positionedaround a shared virtual table. In order togenerate the required realistic 3D video ob-jects, a multi-view camera setup captures theconferees, while disparities that represent thedepths of the video objects are estimatedfrom corresponding images. This virtual 3Dscene is then rendered onto a 2D display us-ing a virtual camera whose placement coin-cides with the current position of the confer-ee's head.

Regardless of the system configuration ofthe teleconferencing application, a multi-view representation of the 3D video objectsis required. The term ‘view combining’ refersto the way in which the original images fromthe cameras are initially combined to obtainthe required surface texture and disparity in-formation to synthesize virtual views of theconferee. Here I present two non-conflictingapproaches to view combining; the choice ofwhich technique to implement is dependenton whether compression or scalability ismore important to the overall system.

The first approach is Incomplete 3D, whichis a compact disparity-based multi-view datarepresentation that was developed in thecontext of the MPEG-4 multimedia standard[1]. Disparity analysis between reference im-ages is used to limit the number of pixelsthat must be encoded. Each area that is visi-ble in more than one reference image is en-coded only once, and with the highest possi-ble resolution.

This approach eliminates redundant imagecontent in multi-view images. This is donedirectly after using these redundancies in thesearch for point correspondences, but beforeencoding, thus giving higher compression.

The Intermediate View approach warps thetwo original images to an intermediate de-rived position that is system defined. Thiswarp is generated using trilinear warping

techniques. The intermediate view can bethought of as coming from a third, virtual,camera that has its own viewpoint, cameraparameters, and position in the world coor-dinate system. This view has a texture thatmay be used as a default view, and also acorresponding disparity map. The disparitymap provides depth information in systemswhere viewpoint adaptation is possible, andwhere it is not available the receiver is stillprovided with a realistic texture as an initialview.

The problems of occlusion errors and holesthat are associated with image warping ariseduring the generation of the intermediateview. These can be overcome by implement-ing a rendering technique that employs oc-clusion ordering, while holes can be filled us-ing a hidden layer technique.

This study has been supported by theFederal Ministry of Education and Researchunder grant AK 002.

[1] E. Cooke, O. Schreer, B. Pasewaldt andP. Kauff, “Extension of incomplete 3D for ar-bitrary multi-view-synthesis”, Proc. VMV’00,Saarbrücken, Germany, Nov. 2000.

Eddie Cooke ([email protected])

View Combining and Novel View Synthesis for Image-Based Rendering in3D Video-Conferencing

In order to obtain an image-based immersive presence in a virtual world, some form ofmulti-view representation is required. With 3D videoconferencing applications in mind,two non-conflicting view-combining approaches are introduced. The choice betweenthem is dependent on whether compression or scalability is important to the overall sys-tem.

View combining usingIncomplete 3D (left) and Intermediate View(right)

102

The Internet has provided us with the pos-sibility of connecting and communicatingwith computers anywhere in the world. Wecan access a huge amount of informationand can also provide our own information tothe global web community.

Starting from simple text transmission, therapid development of technology over thelast decade has enabled the transmission ofany kind of data, including images, graphicsand video. These technological advanceshave triggered the development of a hugevariety of new services, especially in the fieldof multimedia communication.

A particularly popular group of services,which we denote virtual tourism, is the abili-ty to access and retrieve visual informationabout other places in the world and to virtu-ally visit cities, sights, landscapes (even non-terrestrial ones), buildings, museums, etc.The simplest media for this purpose are im-ages, but they are restricted to a predefinedviewpoint and have no temporal dimension.Video (e.g. webcams, streaming) extends thetemporal dimension at the cost of an in-creased data rate.

A highly attractive additional feature ofsuch services is interactivity. The user shouldbe given the maximum possible freedom tonavigate within a scene (walk, rotate, zoom)and to choose his or her individually pre-ferred views when exploring it. This requiresnew forms of representation of such scenes,e.g. the use of VRML to construct the virtualenvironments. Photo-realistic viewing re-quires the use of original textures.

The very popular QuickTimeVR® system isan example of such an application. It allowsrotation and zoom within a 360° panoramicphoto-realistic view. All information is down-loaded completely before rendering in orderto assure an acceptable reaction time of thesystem in an Internet scenario, so that the vi-sual quality is quite bad (small images, lowresolution).

We have launched a new project on virtualtourism [1] that aims to provide increased vi-sual quality (by using high-resolution pano-ramic views) and to extend the navigationcapabilities. The key idea is to stream the da-ta (e.g. 1.8 Gbyte for a high-resolution pano-

rama), and to display the images while stilltransmitting.

For this purpose we have developed a newscene representation that uses MPEG-4 BIFS.The panorama is divided into typically 100-200 patches. Each patch is encoded usingJPEG2000. The actual view is displayed usingthe HHI 3D MPEG-4 player. Only thosepatches that are actually visible are dis-played, which ensures efficient, fast and flu-ent rendering. As shown in Fig. 1, the pano-rama is generated from the video imagescaptured by a rotating camera. The scenedescription is stored on an MPEG-4 serverand can be accessed via the Internet.

At the client, the user navigates throughthe scene. An intelligent streaming mecha-nism ensures that only those patches thatbecome visible over time are requested fortransmission from the server.

[1] C. Grünheit, A. Smolic and T. Wiegand,“Interactive streaming of 3-D environments”,project web page:http://bs.hhi.de/~smolic/3DStreaming.html.

Aljoscha Smolic ([email protected])


Virtual Tourism – Streaming of 360° Panoramas

A new system for interactive streaming of high-resolution 360° panoramic views over theInternet is presented. The scene is represented very efficiently using MPEG-4 BIFS and isdisplayed at the client using the HHI 3D MPEG-4 player. Streaming of the data is con-trolled by user interaction, and the visible parts are transmitted with priority.

MPEG-4server

Video

Internet

Fig. 1: Streaming of high-reso-lution 360° panoramas

over the Internet

Fig. 2: Rendered views from

360° panorama


The process of video mosaicing denotesthe transformation and blending of all im-ages of a video sequence into a common ref-erence coordinate system, as illustrated inFig. 1. The result is a very compact represen-tation of the visual content by a single staticimage. This is highly attractive for a variety ofvideo processing applications. Such videomosaics are used for image-based renderingand the generation of virtual scenes and en-vironments, and can be employed for videosegmentation, coding, editing (e.g. stable re-placement of parts of the background) andcontent description. Another application ofhigh-resolution mosaics is in the up-samplingand format conversion of video signals.

For some of these applications it is impor-tant that video mosaicing removes a signifi-cant amount of redundancy in the video sig-nal, since information that is visible in severalframes of the sequence is represented onlyonce.

Alternatively, the redundancy of the videosequence can be exploited to extract addi-tional visual information about the observedscene. For example, images and video with ahigher spatial resolution than the originalvideo can be produced by combining videomosaicing and super-resolution techniques[1].

Figure 2 illustrates the basic idea of ournew algorithm. Global warping parametersare estimated for each frame using a very ac-curate and robust algorithm. All frames aresuperimposed by warping towards a com-mon reference image with double spatial res-olution in both dimensions, avoiding anyblending, averaging or interpolation of sub-pixel values. This preserves the original

sharpness of the video and eliminates thespatial alias. The method has been furtherextended to generate super-resolution video.

Figure 3 shows details from an original im-age (top) and a corresponding detail from ahigh-resolution mosaic (bottom). The high-resolution mosaic clearly provides superior vi-sual quality – it is much sharper, has muchmore detail, appears less blocky, and most ofall, aliasing is greatly reduced. This results ina reduction of annoying flicker effects in avideo sequence of high-resolution mosaics.

This work was supported by the DeutscheForschungsgemeinschaft (DFG) under grantOH 50/8-1.

[1] A. Smolic and T. Wiegand, “High-reso-lution video mosaicing”, Proc. ICIP ’2001,Thessaloniki, Greece, 7-10 October 2001.

Aljoscha Smolic ([email protected])

High-Resolution Video Mosaicing

A new method for generating super-resolution mosaics from video is presented. The algo-rithm exploits the spatial alias of the video sequence to generate images and video with ahigher resolution than the source video. Global warping parameters are estimated forevery frame using a very accurate and robust algorithm. All frames are superimposed bywarping towards a common reference image with double spatial resolution in both di-mensions, avoiding any blending, averaging or interpolation of sub-pixel values. This pre-serves the original sharpness of the video and eliminates the spatial alias.

Fig. 1: The mosaicing process:warping and blending allframes of a video se-quence towards a com-mon reference image

Fig. 3: Detail from originalframe (top).Corresponding detailfrom high-resolutionmosaic (bottom).

Fig. 2: Transformation of videopixels into mosaic ofdouble resolution, without interpolation ofintermediate values

Warping

Actual frameMosaic

xy

Frame t0 Framehalf - pelmotion

t1

Mosaicdouble

resolution

t0 Mosaicdouble

resolution

t1

105H H I R E P O R T 2 0 0 1

Communications and Events

107H H I R E P O R T 2 0 0 1 · C O M M U N I C A T I O N S A N D E V E N T S

PUBLICATIONS*

Photonic Networks

H.-G. Bach, W. Schlaak, G. G. Mekonnen, R. Steingrüber, A. Seeger, W. Passenberg, W. Ebert, G. Jacumeit, Th. Eckhard, R. Ziegler, A. Beling, B. Schmauß1, A. Munk1, Th. Engel2, A. Umbach3, 50 GHzphotoreceiver moules for RZ and NRZmodulation format comprising InP-OEICs,27th Europ. Conf. on Optical Communica-tion (ECOC 2001), Amsterdam (NL),Th.M.2.5, pp. 560-561.

H.-G. Bach, 40 Gbit/s Photoempfänger fürdie hochratige Glasfaser-Telekommunika-tion (40 Gbit/s photoreceiver for a high-bit rate fiber optics communication), 10. Fachtagung VDE-GMM “Optische undelektronische Verbindungstechnik”, München(D), 2001, Fachbereich 34, pp. 65-70.

E.-J. Bachus, R. Freund, M. Malach, E. Schulze, Distributed Raman amplifica-tion in cascaded normilized sections:Modelling and 10 Gb/s transmission over900 km without EDFAs, Proc. 2. ITG-Fach-tagung Photonische Netze, Dresden (D),2001, pp. 133-134.

S. Bauer, O. Brox, C. Bornholdt, B. Sartorius,M. Moehrle, Laser based 3 R regenerationfor high-speed all-optical networks, Proc.2nd International Workshop on All-OpticalNetworks (WAON 2001), Zagreb (Croatia),pp. 59-61.

A. Beling, D. Schmidt, H.-G. Bach, G. G. Mekonnen, R. Ziegler, V. Eisner, M. Stollberg, G. Jacumeit, E. Gottwald, C.-J. Weiske, A. Umbach, High power1550 nm twin-photodetector moduleswith 45 GHz bandwidth based on InP,Conf. Optical Fiber Communication (OFC 2002), Anaheim (CA, USA) March 17-22, 2002, accepted.

C. Bornholdt, S. Bauer, M. Möhrle, H.-P. Nolting, B. Sartorius, All optical clockrecovery at 80 GHz and beyond, Proc. 27th

Europ. Conf. on Optical Communications(ECOC 2001), Amsterdam (NL), p. 502.

R. Brunner4, J. Bischoff4, K. Rudolf4, M. Ferstl,Diffractive solid immersion lenses: charac-terization and manufacturing, Proc. SPIEs

46th Annual Meeting on “Optical Scienceand Technology” San Diego, (USA), 2001,vol. 4449, pp. 235-243.

F. Buchali5, W. Baumert5, H. Bülow5, U. Feiste, R. Ludwig, H. G. Weber, Eye moni-toring in a 160 Gbit/s RZ field transmis-sion system, Proc. 27th Europ. Conf. onOptical Communications (ECOC 2001),Amsterdam (NL), postdeadline paper.

B. Cai6, T. Hattori6, H. H. Deng6, K. Komatsu6, C. Zawadzki, N. Keil, T. Kaino6,Refractive index control and grating fabri-cation of DAST crystal, J. Appl. Phys., (2001),vol. 40, part 2, no. 9A/B, pp. L964-L966.

C. Caspar, Dispersion management fortransparent WDM long distance networks,Proc. 2. ITG-Fachtagung Photonische Netze,Dresden, (D), 2001, pp. 179-183.

C. Caspar, Guidelines for a simplified de-sign of reliable transparent optical net-works, AEP-Sonderheft, 2001, vol. 55, no. 5,pp. 363-367.

F. Devaux7, O. Leclerc7, B. Lavigne7, H.-P. Nolting, B. Sartorius, Alcatel-HHI col-laboration on all-optical 3R regeneration,Alcatel Telecom Review 3rd Quarter, June2001, pp. 231-233.

D. Dias8, S. Stankovic8, H. Haidner9, L. L. Wang10, T. Tschudi8, M. Ferstl, R. Steingrüber, High frequency gratings forapplications to DVD pickup systems, J. Opt. A: pure Appl. Opt. vol. 3, no. 3, 2001,pp. 164-173.

C. Dreyer11, M. Bauer11, J. Bauer11, N. Keil,H. H. Yao, C. Zawadzki, Polycyanurate esterresins with low loss and low birefringencefor use in integrated optics, SPIE Conf. on“Linear and Non-Linear Optics of OrganicMaterials”, San Diego (USA), 2001, Conf.Dig. p. 4461.

C. Dreyer11, M. Bauer11, J. Bauer11, N. Keil,H. H. Yao, C. Zawadzki, Polycyanurate esterresins with low loss for use in integratedoptics, Polytronic 2001, Potsdam (D), Conf.Dig. pp. 276-283.

H. Ehlers, M. Schlak, U. Fischer, Packagingof integrated Mach-Zehnder interferome-ters for optical communication systems,Conf. Optical Fiber Communication (OFC

* If titles are given bilingually the communication is in German. The list of footnotes is on page 116.

2001), Anaheim (USA), poster session 2WDD66-1.

U. Feiste, R. Ludwig, C. Schubert, J. Berger,C. Schmidt, H. G. Weber, B. Schmauss1, A. Munk1, B. Buchold1, D. Briggmann12, F. Kueppers12, F. Rumpf12, 160 Gbit/s fieldtransmission over 116km standard singlemode fiber using 160 Gbit/s OTDM and40 Gbit/s ETDM demultiplexer, IEE Proc.Optoelectronics, vol. 148, no. 4, 2001.

U. Feiste, R. Ludwig, C. Schubert, J. Berger,C. Schmidt, H. G. Weber, B. Schmauss1, A. Munk1, B. Buchold1, D. Briggmann12, F. Kueppers12, F. Rumpf12, 160 Gbit/s trans-mission over 116 km field-installed fiberusing 160 Gbit/s OTDM and 40 Gbit/s ETDM-technique, Electron. Lett., vol. 37,no. 7, 2001, pp. 443-445.

M. Ferstl, High selective etching of deepsilica components using electron cyclotronresonance plasma, Intern. Conf. on Micro-and Nano-Engineering (MNE), Grenoble (F),2001, in press.

M. Gunkel12, F. Küppers12, J. Berger, U. Feiste, R. Ludwig, C. Schubert, C. Schmidt, H. G. Weber, Unrepeateredtransmission of 40 Gbit/s RZ over 252 kmSMF using Raman amplification, Electr.Lett., vol. 37, no. 10, 2001, pp. 646-648.

H. Heidrich, Analysis of optical crosstalkwithin opto-electronic integrated circuitsincluding lasers and photodetectors, Proc.10th European Conference on IntegratedOptics (ECIO 2001), Paderborn (D), pp. 415-418.

H. Heidrich, R. Kaiser, P. Albrecht, S. Fidorra,M. Hamacher, W. Rehbein, Analysis of opti-cal crosstalk within opto-electronic inte-grated circuits including lasers and pho-todetectors, Appl. Phys., vol. B73, 2001,pp. 581-583.

H. Heidrich, Monolithically integrated pho-tonic and optoelectronic circuits based onInP – technology, system applications,perspectives –, Springer special Issue onWorkshop “Microsystem Technologies”(Dortmund, June 2001).

E. Hilliger, J. Berger, H. G. Weber, K. Yvind13, P. M. W. Skovgaard13, J. Hanberg, The cascaded amplifier and

saturable absorber (CASA) all-optical switch,Proc. OSA-Konferenz Photonics in Switching2001, Monterey (CA, USA), June 10-14, 2001.

K. Janiak, U. Niggebrügge, Investigation ofmacroscopic uniformity during CH4/H2reactive ion etching of InP and its improvement by use of a guard ring, J. ofVac. Sci. Technol. B 20(1), 2002, in press.

C. Janke14, P. Haring14, H. Künzel, H. Kurz14,Increased THz emitter efficiency by coher-ent superposition in a high repetition rateresonator, Proc. Conf. on Lasers and Electro-Optics (CLEO/QELS), Baltimore (USA), May2001, Techn. Dig., in press.

L. A. Jiang15, E. P. Ippen15, U. Feiste, S. Diez,E. Hilliger, U. Feiste, J. Berger, R. Ludwig, H. G. Weber, Sampling pulses with semi-conductor optical amplifiers, IEEE Journal ofQuantum Electronics, vol. 37, no. 1, pp.118-126.

R. Kaiser, OEICs/PICs Based on InP be-tween technological faesibility and com-mercial success, 13th Intern. Conf. onIndium Phosphide and Related Materials(IPRM 2001), Nara (J), invited short course.

N. Keil, H. H. Yao, C. Zawadzki, J. Bauer11,M. Bauer11, C. Dreyer11, Athermal polarisa-tion-independent arrayed waveguide grat-ing multiplexer using an all-polymer ap-proach, Appl. Phys. B73, (2001), 5-6 (specialissue on Integrated Optics), pp. 619-622.

N. Keil, H. H. Yao, C. Zawadzki, J. Bauer11,M. Bauer11, C. Dreyer11, Athermal all-poly-mer arrayed waveguide grating multiplex-er, Electron. Lett. (2001), vol. 37, no. 9, pp.579-580.

N. Keil, H. H. Yao, C. Zawadzki, J. Bauer11,M. Bauer11, C. Dreyer11, Athermal polarisa-tion-independent all-polymer arrayedwaveguide grating (AWG) multi/demulti-plexer, Conf. Optical Fiber Communication(OFC 2001), Anaheim (USA), postdeadlinepaper, Conf. Dig., PD7.

N. Keil, H. H. Yao, C. Zawadzki, J. Bauer11,M. Bauer11, C. Dreyer11, J. Schneider11,Athermal polarisation-independent arrayed-waveguide grating (AWG) multi-plexer using an all-polymer approach, 10th

Europ. Conf. on Integrated Optics (ECIO2001), Paderborn (D), postdeadline-paper.

108 H H I R E P O R T 2 0 0 1 · C O M M U N I C A T I O N S A N D E V E N T S


N. Keil, H. H. Yao, C. Zawadzki, K. Lösch5, K. Satzke5, W. Wischmann5, J. v. Wirth5, J. Schneider11, J. Bauer11, M. Bauer11, Hybridpolymer/silica vertical coupler switches,Appl. Phys. B73, (special issue on IntegratedOptics), (2001), 5-6, pp. 469-473.

N. Keil, H. H. Yao, C. Zawadzki, K. Lösch5, K. Satzke5, W. Wischmann5, J. v. Wirth5, J. Schneider11, J. Bauer11, M. Bauer11, Hybridpolymer/silica vertical coupler switches,10th Europ. Conf. on Integrated Optics (ECIO2001), Paderborn (D), Conf. Dig., pp. 164-167.

J. Kreissl, U. Troppenz, W. Rehbein, B. Hüttl,E. Lenz, H. Venghaus, High single-mode-yield multiple-wavelength DFB-laser arraysin the 1.55 µm range, Proc. 27th Europ.Conf. on Optical Communications (ECOC2001), Amsterdam (NL), Tu.B.1.1, pp.118.

K.-D. Langer, KomNet – Transport- undNetztechniken für das Breitbandinternet(Optical transport and networking tech-nologies for the broadband-internet),Proc. Consulting Akademie “Forum OptischeNetze 2001”, Königswinter (D), November26-28, 2001.

K.-D. Langer, KomNet – optical transportand networking technologies for thebroadband-internet, Proc. 2. ITG-Fachtagung Photonische Netze, Dresden(D), 2001, pp. 45-54.

K.-D. Langer, KomNet – optical transportand networking technologies for thebroadband-internet, Proc. Intern. Conf. onTransparent Optical Networks (ICTON2001), Krakau (PL), pp. 183-190.

G. Levy-Yurista16, A. A. Friesem16, E. Pawlowski, L. Küller, R. Ludwig, H. G. Weber, A. Donval17, E. Toussaere17, J. Zyss17, Hybrid semiconductor polymerresonant grating waveguide structures,Optical Materials, 2001, vol. 17, pp. 149-154.

R. Ludwig, A. Sizmann18, U. Feiste, C. Schubert, M. Kroh, C. M. Weinert, H. G. Weber, Experimental verification ofnoise squeezing by an optical intensity filter in high-speed transmission, Proc. 27th

Europ. Conf. on Optical Communications(ECOC 2001), Amsterdam (NL), postdead-line paper.

K. Mantel19, J. Lamprecht19, R.Schreiner19,N. Lindlein19, J. Schwider19, M.Ferstl, Micro-cylinder-lens testing usingcomputer-generated -holograms, Fringe2001, W. Osten and W. Jüptner (Eds.),Elsevier, Paris, 2001, pp. 100-105.

C. Möller, N. A. Maleev20, W. Passenberg, J. Böttcher, H. Künzel, A. V. Skharov20, A. E. Zhukov20, A. F. Tsatsulnikov20, V. M. Ustinov20, Development of 1300 nmGaAs-based microcavity light emittingdiodes, Proc. Int. Symp. “Nanostructures:Physics and Technologies”, St. Petersburg(RUS) 2001, p.163.

G. Nachtwei21, A. Manolescu21, N. Nestle21,H. Künzel, Bistable resistance in a ferro-magnetic quantum hall system induced byexchange enhancement of the Zeemanenergy, Phys. Rev. B 63, 045306(8) (2000).

H.-P. Nolting, All-optical signal processingdevices for photonic networks: towardspacket switching, “New Trends in OpticalNetwork Design and Modeling”, KluwerAcademic Publishers, edited by AlexandrosA. Stavdas, 2001, pp. 49-76.

H.-P. Nolting, All-optical 3R-regenerationfor OTDM-systems (invited), Plenary Sessionat Contemporary Photonics Technology,Tokyo International Forum, Yurakucho,Tokyo (J) 2002, pp. 15-18.

H.-P. Nolting, Devices for 3R signal regen-eration at 160 Gbit/s and faster (invited),Proc. Workshop on Optical Signal Proces-sing, Organized by COST 267, at COM,Technical University of Denmark, Lyngby,2001, pp. 19-22.

H.-P. Nolting, H.-J. Wünsche22, Novel ultrahigh speed switching technique for all-op-tical signal processing, Integrated PhotonicsResearch (IPR 2001), Monterey (CA, USA),postdeadline paper PDP 1.

L. K. Oxenloewe13, E. Hilliger, A. Tersigni13,A. M. Nik13, S. Hoejfeldt13, F. Romstad13, K. Yvind13, P. M. W. Skovgaard13, K. Hoppe23, J. Hanberg23, All-optical demul-tiplexing and wavelength conversion in anelectroabsorption modulator, Proc. 27th

Europ. Conf. on Optical Communications(ECOC 2001), Amsterdam (NL).

L. K. Oxenlowe13, C. Schubert, C. Schmidt,

E. Hilliger, J. Berger, U. Feiste, R. Ludwig, H. G. Weber, Optical clock recovery em-ploying an optical PLL using cross-phasemodulation in a Sagnac-interferometer,Proc. Conf. on Lasers and Electro-Optics(CLEO 2001), Baltimore (USA), Tech. Dig.,paper CThU2.

A. Paraskevopoulos, H.-J. Hensel, W. Molzow,K. Janiak, E. Suryaputra, H. Roehle, P. Wolfram, W. Ebert, Fabrication ofInGaAsP/InP ridge waveguide lasers withdry etched facets using chemically assistedion beam etching and a simple photoresistmask, Proc. LEOS 2001 annual meeting, LaJolla (USA), paper, TuE 4.

K. Pfeiffer24, M. Fink24, G. Ahrens24, G. Gruetzner, F. Reuther24, J. Seekamp25, S. Zankovych25, C. M. Sotomayor Torres25,H. Schulz25, H.-C. Scheer25, R. Steingrüber,Polymer stamps for nanoimprinting, Intern.Conf. on Micro- and Nanoengineering (MNE2001), Grenoble (F), in press.

D. Rabus, Resonance frequency tuning of adouble ring resonator in GaInAsP/InP:Experiment and simulation, Special Issueon Indium Phosphide and Related Materialsin JJAP, Part 1, vol. 41, no. 2, 2002.

D. Rabus, M. Hamacher, MMI-coupled ringresonators on GaInAsP-InP, IEEE PhotonicsTechnology Letters, vol. 13, no. 8, 2001, pp.812-814.

D. Rabus, M. Hamacher, Optical filter andlaser applications using micro ring res-onators, Second Online Symposium forElectronic Engineers (OSEE), sequence num-ber B-93.

D. Rabus, M. Hamacher, H. Heidrich, Activeand passive microring resonator filter applications on GaInAsP/InP, Proc. 13th

Intern. Conf. on Indium Phosphide andRelated Materials (IPRM 2001), Nara (J) pp.477-480.

M. Rohde, HF-control modulation for per-formance monitoring of WDM-networks,Proc. 2. ITG-Fachtagung Photonische Netze,Dresden (D), 2001, pp. 165-168.

H. Rosenfeldt26, Ch. Knothe26, E. Brinkmeyer26, R. Ulrich26, U. Feiste, C. Schubert, J. Berger, R. Ludwig, H. G. Weber, A. Ehrhardt, Automatic PMD

compensation at 40 Gbit/s and 80 Gbit/susing a DOP-based feedback signal, Proc.Optical Fiber Communication (OFC 2001),Anaheim (CA, USA), postdeadline paperPD27.

H. Rosenfeldt26, R. Ulrich26, E. Brinkmeyer26,U. Feiste, C. Schubert, J. Berger, R. Ludwig,H. G. Weber, A. Ehrhardt27, Feed-forwardapproach for automatic PMD-compensa-tion at 80 Gbit/s over 45 km. installed sin-gle mode fiber, Proc. 27th Europ. Conf. onOptical Communications (ECOC 2001),Amsterdam (NL), postdeadline paper.

J. Saniter, V. Kravcenko, H. Feng, E. Patzak,TransiNet – Innovative transport networksfor the broadband internet, Proc. 2. ITG-Fachtagung Photonische Netze, (jointly withTU-München, T-Systems Nova, UniversitätStuttgart and TU-Berlin) Dresden (D), 2001,pp. 23-31.

B. Sartorius, 3R all-optical signal regenera-tion, Proc. 27th Europ. Conf. on Optical Com-munications (ECOC 2001), Amsterdam (NL),September 30 - October 4, 2001, tutorial.

B. Sartorius, C. Bornholdt, S. Bauer, M. Möhrle, 40 GHz optical clock recoveryfor application in asynchronous networks,Proc. 27th Europ. Conf. on Optical Commu-nications (ECOC 2001), Amsterdam (NL), p. 442.

B. Schauwecker28, M. Arnold28, C. Radehaus28, G. Przyrembel, B. Kuhlow,Opical waveguide components with highrefractive index difference in silicon-oxyni-tride for application in integrated opto-electronics, Opt. Engineering vol. 41 (1),2002, pp. 237-243.

C. Schubert, S. Diez. J. Berger, R. Ludwig, U. Feiste, H. G. Weber, G. Toptchiyski29, K. Petermann29, V. Krajinovic30, 160 Gb/sall-optical demultiplexing using a gain-transparent ultrafast-nonlinear interfer-ometer (GT-UNI), IEEE Photon. Techn. Lett.,vol. 13, no. 5, 2001, pp. 475-477.

C. Schubert, J. Berger, U. Feiste, R. Ludwig,C. Schmidt, H. G. Weber, 160 Gb/s polar-ization insensitive all-optical demultiplex-ing using a gain-transparent ultrafast-non-linear interferometer (GT-UNI), Techn.Dig. of Optical Amplifiers and their Applica-tions (OAA 2001), Stresa (I), paper OWA4.



C. Schubert, J. Berger, U. Feiste, R. Ludwig,C. Schmidt, H. G. Weber, 160 Gb/s polar-ization insensitive all-optical demultiplex-ing using a gain-transparent ultrafast-non-linear interferometer (GT-UNI), IEEEPhoton. Techn. Lett., vol. 13, no. 11, 2001,pp. 1200-1202.

C. Schubert, J. Berger, E. Hilliger, H. G. Weber, G. Toptchyski29, S. Randel29, K. Petermann29, All-optical switching usinggain-transparent semiconductor opticalamplifiers, In: Nonlinear Guided Waves andTheir Applications, OSA Technical Digest,Washington, USA), 2001, pp. 23-25.

E. Schulze, R. Freund, M. Malach, F. Raub,10 Gb/s NRZ transmission over 1800 kmmultiple pumed distributed Raman ampli-fied transmission link without lumped am-plifiers, Proc. 27th Europ. Conf. on OpticalCommunications (ECOC 2001), Amsterdam(NL), vol. 2, pp. 160-161.

S. Stankovic8, D. Dias8, M. Hain8, S. Bhattacharya8, R. Glöckner8, C. Karaboue8, L. Benmohammadi8, F. Guessous8, T. Tschudi8, L. Wang10, M. Ferstl, E. Pawlowski, R. Steingrüber,Integrated optical pickup system for axialdual focus, Appl. Opt., vol. 40, no. 5,(2001), pp. 614-621.

R. Steingrüber, M. Ferstl, W. Pilz31, Micro-optical elements fabricated by electron-beam lithography and dry etching tech-nique using top conductive coatings,Microelectronic Engineering 57-58, Elsevier(2001), pp. 285-289.

R. Steingrüber, M. Möhrle, A. Sigmund, W. Fürst, Continuously chirped gratings forDFB-lasers fabricated by direct write elec-tron-beam lithography, Intern. Conf. onMicro- and Nanoengineering (MNE),Grenoble (F), September 16-19, 2001, inpress.T. Tekin, All-optical demultiplexing perfor-mance of monolithically integrated GS-MZI module, Intern. Workshop on OpticalMEMS and Integrated Optics 2001,Dortmund (D).

T. Tekin, M. Schlak, W. Brinker, J. Berger, C. Schubert, B. Maul, R. Molt, All-optical de-multiplexing using monolithically integrat-ed Mach-Zehnder interferometer basedon band gap shifted semiconductor opti-

cal amplifier, Proc. 10th Europ. Conf. onIntegrated Optics (ECIO 2001), Paderborn(D), postdeadline paper.

T. Tekin, M. Schlak, W. Brinker, J. Berger, C. Schubert, B. Maul, R. Molt, Ultrafast all-optical demuliplexing performance ofmonolithically integrated band gap shift-ed Mach-Zehnder interferometer, Proc.27th Europ. Conf. on Optical Communica-tions (ECOC 2001), Amsterdam (NL), vol. 4,pp. 504-505.

A. Umbach32, µm-wave photoreceivers andoptoelectronic integrated circuits (OEICs),invited, International Microwave Sym. (IMS2001) Workshop on Microwave PhotonicComponents, Phoenix (USA), in press.

H. Venghaus, Optoelektronische und optis-che Bauelemente und Schaltungen(Photonic integrated devices and circuits),Handbuch für die Telekommunikation, Jung,Warnecke (Hrsg.), Springer, 2. Auflage.

B. V. Volovik20, A. R. Kovsh20, W. Passenberg,H. Künzel, N. Grote, N. A. Cherkashin20,Yu.G. Musikhin20, N. N. Ledentsov29, D. Bimberg29, V. M. Ustinov20, Optical andstructural properties of self-organizedInGaAsN/GaAs nanostructures, Sem.Sci.Techn. 16, 186 (2001).

G. Walf, KomNet – Scope and recent ac-complishments, Proc. 27th Europ. Conf. onOptical Communications (ECOC 2001),Amsterdam (NL), p. 36.

H. G. Weber, R. Ludwig, U. Feiste, C. Schmidt, E. Hilliger, C. Schubert, J. Berger,M. Kroh, T. Yamamoto, Optical signal pro-cessing for high speed communicationsystems (invited), Proc. 10th EuropeanConference on Integrated Optics (ECIO2001), Paderborn (D), 2001.

C. M. Weinert, J. Berger, U. Feiste, E. Hilliger,B. Konrad29, M. Kroh, R. Ludwig, K. Petermann29, C. Schmidt, C. Schubert, H.G. Weber, T. Yamamoto33, High bit rateOTDM-transmission on standard fibre,Archiv für Elektronik und Übertragungstech-nik (AEÜ), dedicated to Professor Voges onthe occasion of his 60th birthday, 1997,Gustav Fischer Verlag, Jena.

C. M. Weinert, A. Sizmann18, R. Ludwig, C. Schubert, U. Feiste, H. G. Weber,

Nonlinear optical intensity filters: Exper-iment and design rules, Proc. 27th Europ.Conf. on Optical Communications (ECOC2001), Amsterdam (NL), pp. 274-275.

H.-J. Wünsche22, M. Radzunias22, H.-P. Nolting, Modeling of devices for all-optical 3R-regeneration (invited), Proc.Integrated Photonics Research (IPR 2001),Monterey (CA, USA), pp. IWA1-3.

T. Yamamoto33, U. Feiste, J. Berger, C. Schubert, C. Schmidt, R. Ludwig, H. G. Weber, 160 Gbit/s demultiplexerwith clock recovery using SOA-based in-terferometric switches and ist applicationto 120 km fiber transmission, Proc. 27th

Europ. Conf. on Optical Communications(ECOC 2001), Amsterdam (NL), pp. 192-193.

T. Yamamoto33, C. Schubert, C. Schmidt, L. K. Oxenlowe13, U. Feiste, H. G. Weber, A semiconductor laser amplifier in a loopmirror as a phase comparator in PLL-based clock recovery for 160 Gb/s datasignals, Techn. Dig. Of Optical Amplifiersand their Applications (OAA 2001), Stresa (I),paper OTuC2, 2001.

T. Yamamoto33, L. K. Oxenlowe13, C. Schmidt, C. Schubert, E. Hilliger, U. Feiste,J. Berger, Clock recovery from 160 Gb/s da-ta signals using a phase-locked loop withan interferometric optical switch based ona semiconductor optical amplifier, Elec-tron. Lett., vol. 37, no. 8, 2001, pp. 509-510.

H. H. Yao, N. Keil, C. Zawadzki, K. Lösch5, K. Satzke5, W. Wischmann5, J. v. Wirth5, J. Schneider11, J. Bauer11, M. Bauer11,Polymeric planar waveguide devices forphotonic network applications (invited),SPIE Conf. on Design, Manufacturing, andTesting of Planar Optical Waveguide Devices,San Diego (USA) 2001, Conf. Dig., p. 4439.

H. H. Yao ,N. Keil, C. Zawadzki, M. Bauer11,J. Bauer11, C. Dreyer11, J. Schneider11,Polymeric waveguide devices for DWDMapplications, 10th Intern. Plastic Optic FibresConf. (POF 2001), Amsterdam (NL), Conf.Dig., pp. 139-146.

H. H. Yao, N. Keil, C. Zawadzki, K. Lösch5, K. Satzke5, W. Wischmann5, J. v. Wirth5, J. Schneider11, J. Bauer11, M. Bauer11,Integrated thermo-optic switches in sili-ca/polymer waveguide technology (invit-

ed), SPIE Conf. on Materials and Devices forPhotonics Circuits, San Diego (USA), 2001,Conf. Dig., p. 4453.


H. Boche, E. Jugl1, Review of Viterbi's mod-el estimating the capacity of CDMA basedmobile systems, IEICE Trans. on Communi-cations, vol. E84-B, no. 12, 2001, pp. 3212-2117.

H. Boche, M. Protzmann34, Oversamplingand boundation of signals, IEEE Trans. onCircuits and Systems-I: Fundamental Theoryan Applications, vol. 48, no. 3, 2001, pp.364-365.

H. Boche, G. Wunder, Analytical methodsfor estimating the statistical distributionof the crest-factor in OFDM systems, Proc.10th Aachen Symp. on Signal Theory, 2001,Aachen (D), pp. 223-228.

H. Boche, M. Schubert, A new approach topower adjustment for spatial covariancebased downlink beamforming, ICASSP 2001,Salt Lake City, Utah (USA), vol. V SAM-P3.

H. Boche, M. Schubert, Theoretical and experimental comparison of optimisationcriteria for downlink beamforming, Euro-pean Transactions on Telecommunications,Special Issue on Smart Antennas, 2001, vol.12, no. 5, pp. 417-426.

H. Boche, S. Stanczak, Lower bound onmean squared channel estimation errorfor multiuser receiver, ISCAS 2001, Sydney(AUS), II-61-II-64.

H. Boche, S. Stanczak, Optimal sequencesfor asynchronous CDMA channels with dif-ferent SIR requirements, Proc. 35th AsilomarConf. on Signals, Systems, and Computers,2001, Pacific Grove, CA (USA), pp. 415-419.

H. Boche, G. Wunder, Analytical methodsfor estimating the statistical distributionof the crest-factor in OFDM systems, Proc.10th Aachen Symp. on Signal Theory, 2001,Aachen (D), pp. 223-228.

H. Boche, G. Wunder, On the PAPR prob-lem in OFDM systems (invited paper), 39th

Annual Allerton Conference on Communica-



tion, Allerton (IL, USA), Control, and Com-puting, 2001, in press.

G. Grosskopf, R. Eggemann, D. Rohde, M. S. Choi, 155 Mbits data transmission at60 GHz using a 1 x 4 patch array antennawith variable optical delay lines, Proc. IEEEMTT-S Intern. Microwave Symp., Phoenix,Arizona, USA, 2001, Digest vol. 3, paperTH3D-5, pp. 1821-1824.

G. Grosskopf, B. Kuhlow, G. Przyrembel, R. Eggemann, J. Knueppel, D. Rohde, Silicabased beamformer for 60 GHz smart an-tennas, 27th European Conf. on OpticalCommunication (ECOC 2001), Amsterdam(NL), vol. 1, paper Mo. F2.6, pp. 12-13.

G. Grosskopf, D. Rohde, A. Kortke, R. Eggemann, R. Ziegler, M. S. Choi, 155 Mbit/s Datenübertragung im 60 GHz-Bereich mit einer optisch gesteuertenArray Antenne (155 Mbit/s data transmis-sion at 60 GHz using an optically steeredarray antenna), Proc. ITG-Diskussionssit-zung “Systeme mit intelligenten Antennen”2001, TU Ilmenau (D).

T. Haustein, E. Jorswieck, V. Jungnickel, U. Krueger, V. Pohl, C. v. Helmolt, Bit errorrates for a mimo system in Rayleigh andRician channels, IEEE 54rd Vehicular Tech-nology Conference, 2001, Atlantic City (NJ,USA), vol. 4, pp. 1984-1987.

M. Haardt35, H. Boche, G. Wunder, The fu-ture of mobile communications beyondthe third generation, 30. Annual IEEEComm.-Theory Workshop, San Diego (USA),April 2001.

A. Ihlow36, H. Boche, Kapazitätsgewinndurch SDMA mit linearen Antennenarraysin realer Umgebung (Capacity gain via SDMA with linear antenna arrays in realenvironments), Proc. ITG-Diskussionssitzung“Systeme mit intelligenten Antennen” 2001,Ilmenau (D).

E. Jorswieck, G. Wunder, V. Jungnickel, T. Haustein, Eigenvalue and inverse eigen-value statistics for Rayleigh and Rician MIMO channels, Proc. IEE Colloquium onMIMO Communication Systems, 2001,London (GB) pp. 3/1-3/6.

V. Jungnickel, A. Forck, T. Haustein, U. Krüger, V. Pohl, C. v. Helmolt, Wireless

infrared communications beyond100 Mbit/s: system design, transmissionexperiments and potential, (invited), Proc.5th World Multiconference on Systemics,Cybernetics and Informatics 2001, Orlando(FL, USA), vol. IV, pp. 413-420.

V. Jungnickel, T. Haustein, A. Forck, C. v. Helmolt, 155 Mbit/s wireless trans-mission with an imaging infrared receiver,IEE Electronics Letters, vol. 37, no. 5, (2001),pp. 314-315.

V. Jungnickel, T. Haustein, E. Jorswieck, C. v. Helmolt, A MIMO WLAN based on lin-ear channel inversion, Proc. IEE Colloquiumon MIMO Communication Systems: FromConcept to Implementation, London (GB),2001, pp. 20/1-20/6.

V. Jungnickel, T. Haustein, E. Jorswieck, V. Pohl, C. v. Helmolt, Performance of a MIMO system with overlay pilots, IEEEGlobeCom 2001, San Antonio (Texas), onCD-ROM.

V. Jungnickel, V. Pohl, U. Krüger, C. v. Helmolt, T. Haustein, S. Stanczak, A ra-dio system with multi-element antennas,Proc. IEEE 53rd Vehicular Technology Con-ference, 2001, Rhodos (GR), on CD-ROM.

V. Kravcenko, H. Boche, E. Jugl, Models forBER estimation of CDMA-systems and sequences design, Proc. 10th Aachen Symp.on Signal Theory, 2001, Aachen (D), pp.275-280.

U. Krüger, C. v. Helmolt, V. Jungnickel, T. Haustein, A power and bandwidth effi-cient modulation for diffuse wireless in-frared communication, Journal of OpticalCommunication, vol. 22 (2001), pp. 721-723.

V. Pohl, V. Jungnickel, U. Krüger, C. v. Helmolt, Integrating sphere diffuserfor wireless infrared communication, IEEPhotonics Professional Network, available online at www. iee.org.uk.

D. Rohde, A. Kortke, R. Ziegler, R. Eggemann, G. Grosskopf, M. S. Choi,Optically steered 1x4 patch array antennafor 60 GHz, Proc. Microwaves and Optronics,11th The German Wireless Week(MIOP’2001), Sindelfingen (D), pp. 128ff.

M. Schubert, H. Boche, Leistungsoptimier-

tes Downlink-Beamforming für zellulareMobilfunksysteme (Power optimizeddownlink beamforming for cellular wire-less systems), Proc. ITG-Diskussionssitzung“Antennen für Mobile Systeme 2001”,Ilmenau (D), pp. 113-122.

M. Schubert, H. Boche, CIR-based multiuserdownlink beamforming, Proc. 10th AachenSymp. on Signal Theory, Aachen (D), 2001,pp. 237-242.

M. Schubert, H. Boche, Solvability of cou-pled downlink beamforming problems,Proc. IEEE Globecom Conf., San Antonio,Texas, 2001, BWS06-2.

S. Stanczak, H. Boche, M. Haardt35, Are LAS-codes a miracle?, Proc. IEEE GlobecomConf., San Antonio, Texas, 2001, BWS05-4.

G. Wunder, H. Boche, Trigonometric sam-pling series and their application to mo-bile communications, Kleinheubacher Be-richte, T-Nova Technologiezentrum, vol. 44,pp. 183-190.

G. Wunder, H. Boche, Performance boundsand optimal pilot signals in OFDM-MIMOsystems, Proc. 10th Aachen Symp. on SignalTheory, 2001, Aachen (D), pp. 123-128.

Electronic Imaging Technology forMultimedia

C. Buchner, T. Stockhammer, D. Marpe, G. Blättermann, G. Heising, Efficient finegranular scalable video coding, Proc. IEEEInt. Conf. on Image Processing (ICIP 2001),Thessaloniki (GR), pp. 997-1000.

C. Buchner, T. Stockhammer, D. Marpe, G. Blättermann, G. Heising, Progressive tex-ture video coding, Proc. Int. Conf. onAcoustics, Speech, and Signal Processing(ICASSP 2001), Salt Lake City, Utah (USA), inpress.

R. Buss, U. Hambrock, The effect of the perceived work situation and backgroundknowledge on assessing people's willing-ness to communicate, Proc. 18th Int. Symp.on Human Factors in Telecommunication(HFT 2001), Bergen (N), pp. 81-88.

C. Fehn, E. Cooke, O. Schreer, P. Kauff, 3Danalysis and image-based rendering for

immersive TV applications, Proc. Int. Conf.on Augmented, Virtual Environments and 3DImaging (ICAV3D 2001), Mykonos (GR), pp.192-195.

C. Fehn, P. Kauff, R. Schäfer, InteractiveVirtual View Video – Eine Erweiterung desimmersiven Fernsehens (Interactive virtualview video – An extension of immersivetelevision), Proc.9. Dortmunder Fernsehse-minars, Dortmund (D), 2001, pp. 163-168.

C. Fehn, P. Kauff, R. Schäfer, O. Schreer,Interactive virtual view video for immer-sive TV applications, Proc. IBC 2001,Amsterdam (NL), vol. 2, pp. 53-61.

B. Feustel, T. C. Schmidt, D. Marpe, M. Palkow, H. L. Cycon, Compound mediastreaming in time, Proc. 9th Int. Conf. inCentral Europe on Computer Graphics,Visualization and Computer Vision (WSCG2001), Univ. of West Bohemia, Plzen (CzechRep.), pp. 161-167.

M. Hahn, C. Huck, Digital projectors forvideorepresentation, FKT, vol. 55, no. 1-2,Febr. 2001, pp. 36-43.

P. Kauff, C. Fehn, E. Cooke, O. Schreer,Advanced incomplete 3D representationof video objects using trilinear warping fornovel view synthesis, Proc. Picture CodingSymposium (PCS 2001), Seoul (Korea), pp.429-432.

P. Kauff, O. Schreer, An immersive 3Dvideoconferencing system based on ashared virtual environment, Proc. of Int.Conf. on Media Futures, Florence (I) 2001,pp. 49-52.

P. Kauff, O. Schreer, J.-R. Ohm14, A universalalgorithm for real-time estimation ofdense displacement vector fields, Proc. ofInt. Conf. on Media Futures, Florence, (I)2001, pp. 311-314.

I. Keller, T. Ellerbrock, T. Meiers, T. Sikora,Fast user-adaptive weighting of MPEG7descriptors for a visual E-commerce inter-face, Proc. 3th European Workshop on ImageAnalysis for Multimedia Services (WIAMIS2001), Tampere (SF), pp. 147-152.

I. Keller, T. Meiers, T. Ellerbrock, T. Sikora,Image browsing with PCA-assisted user-in-teraction, Proc. IEEE Workshop on Content-



based Access of Image and Video Libraries(CBAIVL 2001), (Kauai, HI, USA), pp. 102-108.

J. Liu, C. Conomis, Z. Zhu, S. Pastoor, Real-time head tracking in multimodal inter-face with changes in geometry and illumi-nation, Proc. IASTED InternationalConference “Visualization, Imaging andImage Processing (VIIP 2001)”, Marbella (E),pp. 608-612.

D. Marpe, G. Blättermann, G. Heising, T. Wiegand, Video compression using con-text-based adaptive arithmetic coding,Proc. IEEE Int. Conf. on Image Processing(ICIP 2001), Thessaloniki (GR), pp. 558-561.

D. Marpe, T. Wiegand, H. L. Cycon,Wavelet-based video compression usinglong-term memory motion-compensatedprediction and context-based adaptivearithmetic coding, Lecture Notes on onWavelet Analysis and Its Applications,Computer Science, Springer Verlag Berlin,vol. 2251, pp. 76-86.

T. Meiers, I. Keller, T. Sikora, Image visual-ization and navigation based on MPEG-7descriptors, Proc. Intern. Conf. onAugmented, Virtual Environments and 3DImaging (EUROIMAGE 2001), Mykonos (GR),pp. 165-168.

T. Meiers, I. Keller, T. Sikora, 3D browsingenvironment for MPEG-7 image data-bases, Proc. IS&T/SPIE Workshop “Storageand Retrieval for Media Databases 2002”,San José, CA (USA), pp. 324-335.

B. Quante, R. Buss, L. Mühlbach, D. Przewozny, D. Runde, An experimentalcomparison of two variants of an informalcommunication module for distributedworkgroups, Proc. 18th Int. Symp. onHuman Factors in Telecommunication (HFT2001), Bergen (N), pp. 89-98.

J. Restat, An inventory of terms for precisecognitive theories, 5th Meeting of the Ger-man Cognitive Science Society, in: LeipzigSeries in Cognitive Sciences 2, T. Arnold andCh. S. Herrmann (Eds.), Leipziger Universi-tätsverlag, Leipzig (D), 2001, p. 73.

J. Restat, Psychological relevant features ofcolor patterns, Proc. Workshop on ImageAnalysis for Multimedia Services (WIAMIS2001), Tampere (SF), pp. 107-112.

H. Richter37, A. Smolic, B. Stabernack, E. Müller, Real time global motion estima-tion for an MPEG-4 video encoder, Proc.Picture Coding Symposium (PCS 2001),Seoul (Korea), pp. 401-404 .

R. Schäfer et. al., Kooperative Postproduk-tion von Film und Video in High-SpeedNetzwerken (System for distributed post-production of video using a high-speednetwork), Proc. ITG/FKTG-Fachtagung, 9th

Dortmunder Fernsehseminar, Dortmund (D),2001, pp. 143-148.

R. Schäfer, T. Sikora, Performance-Messungam HHI – Konzept und Erfahrungen(Performance-measurement in the HHI –draft and experiences), Wissenschafts man-agement – Zeitschrift für Innovation, 7. Jahr-gang (2001), Heft 8, pp. 11-16.

O. Schreer, N. Brandenburg, S. Askar, A vir-tual 3D video-conference system provid-ing semi-immersive telepresence: a real-time solution in hardware and software,eBusiness 2001, Venice (I), vol. 1, pp. 184-190.

O. Schreer, N. Brandenburg, S. Askar, P. Kauff, Hybrid recursive matching andsegmentation-based postprocessing in real-time immersive video conferencing,Vision, Modeling and Visualization, Stuttgart(D), 2001, pp. 383-390.

O. Schreer, N. Brandenburg, Peter Kauff, A combination of census transform and ahybrid block and pixel-recursive disparityanalysis approach for real-time videocon-ferencing applications, Proc. Int. Conf. onAugmented, Virtual Environments and 3DImaging (ICAV3D 2001), Mykonos (GR), pp.29-32.

O. Schreer, N. Brandenburg, P. Kauff, Real-time disparity analysis for applications inimmersive tele-conference scenarios – acomparative study, ICIAP 2001, Palermo (I),pp. 346-353.

O. Schreer, C. Fehn, N. Brandenburg, M. Karl, P. Kauff, Fast disparity estimatorfor real-time video processing using a hy-brid block- and pixel-recursive matchingtechnique, Proc. Picture Coding Symposium(PCS 2001), Seoul (Korea), pp. 405-409.

K. Seifert29, J. Liu, J. Hurtienne, T. Baumgarten, S. Pastoor, Development ofan evaluation concept for a multimodalcomputer system, Proc. 8th IFAC Symposi-um on Analysis, Design, and Evaluation ofHuman-Machine Systems, Gesamthochschu-le Kassel, 2001, in press.

K. Seifert29, S. Pastoor, Multimodal human-computer interaction: When the computeris watching us, TU Berlin, Forschung Aktuell,1/2001, Mensch und Maschine, pp. 82-83.

A. Smolic, Global motion compensationand video mosaicing using different 2-Dmotion models, Proc. Picture Coding Sym-posium (PCS ‘2001), Seoul (Korea), pp. 331-334.

A. Smolic, Parametric motion models andvideo mosaics and their applications atcoding, post-production and informationmanagement, Proc. ITG/FKTG-Fachtagung,9th Dortmunder Fernsehseminar, Dortmund(D), 2001, pp. 175-180.

Smolic, T. Wiegand, High-resolution videomosaicing, IEEE Intern. Conf. on ImageProcessing (ICIP ‘2001), Thessaloniki (GR),on CD-ROM.

E. Trucco, C. Plakas, N. Brandenburg, P. Kauff, M. Karl, O. Schreer, Real-time disparity analysis for Immersive 3-DTeleconferencing by Hybrid RecursiveMatching and Census Transform, ICCV2001, Vancouver (CDN), in press.


1 Lucent Technologies, Nürnberg, D2 Siemens AG, Berlin, D3 u2t Photonics AG, Berlin, D4 Carl Zeiss, Jena, D5 Alcatel, Stuttgart, D6 Tohuku University, Sendai, J7 Alcatel, Paris, F8 Technical University Darmstadt, D9 Carl Zeiss, Oberkochen, D

10 Philips Eindhoven, NL11 Fraunhofer Institut (IZM), Berlin-Teltow, D12 T-Nova – Deutsche Telekom,

Darmstadt, D13 Denmark Technical University,

Kopenhagen, DK14 RWTH Aachen, D15 Institute of Technology, Massachusetts,

USA16 Weizmann Institute of Science,

Rehovot, IL

17 Laboratoire de Photonique Quantique et Moleculaire, F

18 Innovance Networks Inc., CDN19 University of Erlangen, D20 Ioffe Institut, St. Petersburg, RUS21 Technical University Braunschweig, D22 Humboldt-University, Berlin, D23 GiGA-Intel, DK24 micro resist technology GmbH, Berlin, D25 University of Wuppertal, D26 Technical University Hamburg-Harburg, D27 T-Nova, Berlin, D28 Technical University Chemnitz, D29 Technical University Berlin, D30 Technical University Wien, A31 ISIT, Itzehoe, D32 u2t Photonics AG, Berlin, D33 Fujitsu Laboratories Ldt., Wakamiya, J34 Technical University Dresden, D35 Siemens AG München, D36 Technical University Ilmenau, D37 University of Rostock, D

REPORTS

E.-J. Bachus, C. Caspar, M. Rohde, K. Habel,System aspects and test of componentsfor the cityring network, Final Report oncontract with T-Nova Deutsche Telekom AG,HHI, Berlin, January 2001.

T. Ellerbrock, Bildinhaltserkennung undSegmentierung für Multimedia-Anwen-dungen (Recognition of image contentand segmentation for multimedia applica-tions), Final Report on DFG Research ProjectSE 456/5-1, HHI jointly with TU Berlin,August 2001.

G. Heising, D. Marpe, Open issues relatingto H.26L picture header, ITU-T Study Group16/Question 6 Document VCEG-M-62,Austin (USA), April 2001.

D. Marpe, G. Blättermann, G. Heising, T. Wiegand, TML5 software includingCABAC, ITU-T Study Group 16/Question 6Doc. VCEG-M-60, Austin (USA), April 2001.

D. Marpe, G. Blättermann, G. Heising, T. Wiegand, Further results on CABAC en-tropy coding scheme, ITU-T Study Group16/Question 6 Doc. VCEG-M-59, Austin(USA), April 2001.

D. Marpe, G. Blättermann, T. Wiegand,Improved CABAC, ITU-T Study Group


meterbereich (Millimeter-wave patch an-tenna module for V-band applications),101 02 468.1-35

O. Schreer, P. Kauff, N. Brandenburg, M. Karl, Verfahren zur Korrespondenzana-lyse von Bildmerkmalen in korrespondie-renden Videobildern in Echtzeit (Methodfor real-time correspondence analysis ofimage features in corresponding video images), 101 05 423.8-53

R. Stenzel, H. Roehle, H. Stolpe, Probenhal-ter (Sample holder), 101 10 297.6-52

D. Brand, Gasbündelstation mit Druckmin-dererstation und Sicherheitsabschaltung(Pressure reducing valve and safety shut-down in a bundle of gas cylinders), 201 03 682.7

N. Keil, H. Yao, C. Zawadzki, Einstellbareroptischer Pegelangleicher (Active gainequalizer), 101 12 021.4-51

M. Rohde, E.-J. Bachus, K.-D. Langer, Ver-fahren zur Qualitätsüberwachung undIdentitätskontrolle von Wellenlängenkanä-len sowie zur Übertragung von Netzma-nagement-Informationen in transparentenoptischen Übertragungsnetzen (Methodfor quality and identity monitoring ofwavelength channels and for transport ofnetwork management information intransparent optical networks), 101 12 511.9-35

V. Jungnickel, V. Pohl, U. Krüger, C. v. Helmolt, Th. Haustein, S. Stanczak,Funkübertragungsverfahren im Innen-raumbereich zur parallelen Funkübertra-gung von digitalen Datenteilströmen undmobiles Funkübertragungssystem (Parallelradio transmission of cochannel signals inindoor environments and system for wire-less data transmission), 101 14 052.5-35

W. Liebsch, D. Zier, Elektronisches Verfah-ren zur zuschauerindividuellen Suche undAuswahl von Fernsehprogrammangebo-ten (Electronic procedure for viewer-indi-vidual search and selection of televisionprogram supplies), 101 18 966.4-31

H.-P. Nolting, M. Gravert, Abstimmbarer,selbstpulsierender Zweisektions-Laser zuroptischen Taktfrequenzerzeugung unddessen Verwendung (Tunable selfpulsat-

16/Question 6 Doc. VCEG-O-18, Pattaya(Thailand), December 2001.

K. Mertens, Investigations on Erbium-doped fibre amplifiers, Report on a practi-cal semester, HHI and TFH-Berlin, January2001.

M. Rohde, Controlling of WDM channels inphotonic networks by means of controlmodulation, Report on DFG Research Pro-ject Ba 1790/1-1 (WDMCONT), July 2001.

M. Schlak, Final report DEMUX’ Alcatel-HHI collaboration – All-optical signal pro-cessing, May 2001.

M. Schubert, A. Kortke, H. Boche, Space-time array processing for 3G WCDMA -channel modeling and directivity con-trolled beamforming, Final Report, 2nd YearCollaborative Research ETRI/HHI, August2001.

E. Schulze, Influence of distributed Raman-amplification on high data rate WDM net-works, Final Report on DFG Research ProjectBa 1790/2-1 (RAMAN), HHI, Berlin, July2001.

A. Smolic, Videocodierung unter Verwen-dung von 2D und 3D Objekt- und Bewe-gungsmodellen (Application of 2D and 3Dobject and motion models for video cod-ing), Technical Report on DFG ResearchProject OH 50/8-1, September 2001.

A. Smolic, Y. Guo, J. Guether, T. Selinger,Demonstration of streaming of MPEG-4 3-D scenes with live video, ISO/IECJTC1/SC29/WG11, M7811, December 2001.

T. Stockhammer, D. Marpe, Test modeldocument changes for data partitioningand NAL, ITU-T Study Group 16/Question 6Doc. VCEG-M-52, Austin (USA), April 2001.

T. Wiegand, K. Müller, H. Schwarz, H.26LTML 8 for Digital Cinema Applications,ISO/IEC JTC1/SC29/WG11 MPEG01, m7848,Pattaya, Thailand, December 2001.

PATENT APPLICATIONS

D. Rohde, G. Großkopf, A. Kortke, Planar-antennenmodul für Wellenlängen im Milli-

ing multi-section-laser for clock recovery),101 18 965.6-33

B. Sartorius, St. Bauer, C. Bornholdt, O. Brox,Verfahren zur Ansteuerung eines Laser-basierenden 3R Regenerators (Scheme forthe operation of a laser based 3 R regen-erator), 101 18 959.1-35

B. Sartorius, 3R Regenerator mit Wellenlän-genumsetzung (3R regenerator withwavelength conversion), 101 18 958.3-35

H. Heidrich, C. M. Weinert, Integriert-opti-scher Feldweitentransformator zur adiaba-tischen, monomodigen Feldanpassung(Integrated optical beam spread trans-former for adiabatic, single mode fieldadaptation), 101 23 137.7-51

D. Rohde, G. Großkopf, A. Kortke, Planaran-tennenmodul für Wellenlängen im Milli-meterbereich (Millimeter-wave patch an-tenna module for V-band applications),101 23 571.2-35

M. Hamacher, H. Heidrich, M. Hentschel, D. G. Rabus, K. Richter, Monolithisch inte-grierter Mikrolaser mit einem nur eineSpiegelebene aufweisenden Zirkularreso-nator (Monolithically integrated micro-laser based on a circular resonator withonly one mirror plane), 101 32 479.0-33

V. Jungnickel, Th. Haustein, E. Jorswieck, G. Wunder, U. Krüger, V. Pohl, C. v. Helmolt,Adaptives Signalverarbeitungsverfahrenzur bidirektionalen Funkübertragung ineinem MIMO-Kanal und MIMO-Systemzur Verfahrensdurchführung (Adaptive bi-directional transmission in a MIMO radiochannel and MIMO system for wireless da-ta transmission), 101 32 492.8-35

H. Heidrich, M. Hamacher, D. G. Rabus, U. Troppenz, Verfahren zum kraftschlüssi-gen Ganzwaferlöten (Process for force-fitfull-wafer soldering), 101 53 054.4

R. Schmidt, Schweißbacken für das Schutz-gasschweißen von Rohrverbindungen mitSchweißfittingen aus Edelstahl (Weldingjaw for orbital welding of stainless steeltube fittings), 201 19 625.5

S. Bauer, O. Brox, Optische Mikrowellen-quelle (Optical microwave source), 101 60 502.1

D. Brand, Werkzeug zum Übertragen posi-tiver oder negativer Drehmomente in en-gen Montageräumen (Tool to transformpositive or negative torques in narry as-sembly sites), 101 62 112.4

J. Holzäpfel, Entsorgungssystem für gefähr-liche Abfallflüssigkeiten an einem Labor-arbeitsplatz (Disposal system for haz-ardous liquid waste at a laboratory work-bench), 101 64 645.3

H.-P. Nolting, M. Gravert, Verfahren zurKontrolle und Steuerung der Datensyn-chronisation in einer Pulsquelle zur rein-optischen Taktrückgewinnung (Methodfor control of data synchronizationprocess of an all-optical clockrecoverybased on multi-section lasers), 101 64 644.5

H.-P. Nolting, Verfahren zur direktenQualitätsermittlung von optischen Über-tragungskanälen und Anordnung zudessen Durchführung (All-optical methodfor direct control of optical signal quality),101 64 643.7

AWARDS

C. Fehn, P. Kauff, O. Schreer and R. Schaefer,President's Award for Best Technical Paperat IBC '01, International Broadcast Conven-tion, Amsterdam, Sept. 2001.

DOCTORATE THESES

D. Runde, Verzerrungsfreie Reproduktionräumlicher Szenen bei Bewegungen desBetrachters (A natural image reproductionusing stereoscopic displays with motionparallax), TU Berlin, 2000, FB (12)Elektrotechnik, Prof. Boerger, Prof. Filbert.

A. Smolic, Globale Bewegungsbeschrei-bung und Video Mosaiking unter Verwen-dung parametrischer 2-D Modelle, Schätz-verfahren und Anwendungen (Descriptionof global motion and video mosaicing ap-plying parametric 2-D motion models, es-timation and applications), Rheinisch-Westfälischen Technischen HochschuleAachen (RWTH), Fakultät für Elektrotechnikund Informationstechnik, Prof. Ohm.



DIPLOMA THESES

U. Grabner, Einbettung eines RISC-Kerns ineinen Echtzeit-Video-FormatkonverterASIC (Embedding of a RISC core into a real time video format conversion ASIC),TU Berlin, FB13 (Informatik). Supervisor atHHI: T. Weber.

I. Grotelüschen, Analyse und Vergleich vonOptimierungskriterien für die Downlink-Strahlformung in UMTS (TD/CDMA)(Analysis and comparison of optimizationcriteria for downlink beamforming inUMTS (TD/CDMA) systems), TU Berlin, FB 12 (Elektrotechnik). Supervisor at HHI: H. Boche.

K. Habel, Performance and transmissionproperties of regional and long-haulWDM ring networks, TU Berlin, FB 12(Elektrotechnik). Supervisor at HHI: Dr. Bachus.

M. Heidenreich, 2D Beschreibung von 3DObjekten (2D-description of 3D-objects),Hochschule für Technik und WirtschaftDresden, Fachbereich Informatik/Mathema-tik. Supervisor at HHI: K. Müller.

S. Kalyoncu, Modellierung eines echtzeitre-levanten Steuerungskonzeptes für kom-plexen ASIC zur Video-Formatkonvertie-rung (Modelling of a real time controlmechanism for a complex ASIC for videoformat conversion), TU Berlin, FB 13(Informatik). Supervisor at HHI: T. Weber, K. Rümmler.

D. Karadoulamas, Modellierung und Imple-mentierung eines räumlichen Kanalmo-dells für den zellularen Mobilfunk(Modelling and implementation of a spa-tially resolved channel model for cellularwireless systems), TU Berlin, FB 12(Elektrotechnik). Supervisor at HHI: H. Boche.

M. Krahn, Entwurf und Entwicklung einesLinux-Treibers für eine High Level MPEG-2Decoderkarte mit PCI-Schnittstelle sowieeiner verteilten Applikationssoftware zumFunktionstest des Treibers (Design anddevelopment of a LINUX driver for a HighLevel MPEG-2 decoder card including aPCI interface and of an application soft-ware for functional checking),Fachhochschule für Technik und Wirtschaft(FHTW) Berlin. Supervisor at HHI: U. Höfker.

O. Krips, Konstruktion einer Justiereinrich-tung zur Ausrichtung von Lichtwellenlei-tern in einem Gerät zur Chip-Faser-Kopp-lung, (Construction of an adjusting devicefor adjustment of fiber-optic cables in adevice for chip fiber coupling), TechnischeFachhochschule (TFH) Berlin. Supervisor atHHI: U. Fischer.

S. Lange, 6-D Automatisierung derFaserarray-Chip-Kopplung (6D automationof fiber chip coupling), Fachhochschule fürTechnik und Wirtschaft (FHTW) Berlin.Supervisor at HHI: Th. Rosin.

A. Lerch, Qualitätsbeurteilung von codier-ten Audiosignalen mittels eines objektivenVerfahrens (Quality assessment of codedaudio signals by means of an objectiveprocedure), TU Berlin, FB 12 (Elektrotechnik).Supervisor at HHI: M. Talmi.

Y. Li, Ähnlichkeitsanalyse von 2D-Ansich-ten und 3D-Objekten in Szenen (Similarityanalysis of 2D-views and 3D-objects inscenes), TU-Berlin, Fakultät IV (Elektrotech-nik und Informatik). Supervisor at HHI: K. Müller.

Ch. Schmidt, Analyses and Modelling of integrated optical Mach-Zehnder interfer-ometer, TU Berlin, Fakultät IV (Elektrotechnikund Informatik). Supervisor at HHI: M. Schlak.

A. Schweinzer, Charakterisierung planarerthermo-optischer Schalter auf Polymer-basis (Characterization of polymer basedplanar thermo optic switches), Fachhoch-schule Nürnberg, FB Nachrichten- und Fein-werktechnik. Supervisor at HHI: C. Zawadzki.

Y. Song, Untersuchung von Verfahren zurfeingranularen qualitätsskalierbarenVideocodierung (Investigation of finegranular scalable video coding methods),TU Berlin, Lehrstuhl für Fernmeldetechnik.Supervisor at HHI: D. Marpe.

J. Tembridis, Modellierung einer Testumge-bung eines komplexen ASIC zur Format-konvertierung (Modelling of a test envi-ronment of a complex ASIC for formatconversion), TU Berlin, FB 13 (Informatik).Supervisor at HHI: T. Weber.

J. Wilhelmy, Festkommaimplementierungeines psychoakustischen Modells auf dem

Signalprozessor TMS 320C6211 für dieAudiocodierung nach AAC-Standard(Fixed point implementation of a psychoacoustic model on the signal processorTMS 320C6211 for audio coding accord-ing to AAC standard), TU Berlin, FB 12(Elektrotechnik). Supervisor at HHI: M. Talmi.

G. Zander, Untersuchung von Verfahrenzur Wavelet-basierten Rauschunter-drückung in der Bilddatenverarbeitung(Investigation of wavelet-based denoisingmethods for image processing),Technische Fachhochschule (TFH) Berlin.Supervisor at HHI: D. Marpe.

S. Zech, 3D-Automatisierung der Faser-Chip-Kopplung (3D-automation of fiberchip coupling), Fachhochschule für Technikund Wirtschaft (FHTW) Berlin. Supervisor atHHI: U. Fischer.

GRADUATE THESES

S. Brahma, Broadening the free spectralrange of double-ring resonators includingtwo different radii by resonance tuning,TU Berlin, FB 12, Fachgebiet Hochfrequenz-technik. Supervisors at HHI: H. Heidrich, D. Rabus.

Y. Effendie, Portierung eines AAC (Ad-vanced Audio Coding) Decoders auf dendigitalen Signalprozessor TMS320C6701(Implementation of an AAC (AdvancedAudio Coding) decoder on digital signalprocessor TMS320C6701), TU Berlin, FB 12(Elektrotechnik). Supervisor at HHI: M. Talmi.

X. Fan, Morphologische Nachverarbeitungvon Segmentierungsergebnissen unterEchtzeitbedingungen (Morphological postprocessing of segmentation results underreal-time constraints), TU Berlin, FB 12(Elektrotechnik), Supervisor at HHI: R. Tanger.

J. Hurtienne, mUltimo3D – Psychologicalstudies on the design and evaluation of amultimodal computer system, HU Berlin(Institut für Psychologie). Supervisor at HHI:S. Pastoor.

D. Kossack, Modellierung und Implemen-tierung eines CDMA basierten zellularenMobilfunksystems (Modelling and imple-mentation of a CDMA based cellular mo-

bile communication system), TU Berlin, FB 12 (Elektrotechnik). Supervisor at HHI: H. Boche.

W. T. Marc, Anwendungen des Mehrträ-gerverfahrens OFDM (The applications ofOFDM), TU Berlin, FB 12 (Elektrotechnik).Supervisor at HHI: H. Boche.

Z. Zhu, Tracking of the eyes in a video sequence under changing illumination, TUBerlin (Institut für Elektronik und Lichttech-nik). Supervisor at HHI: J. Liu.

ORAL PRESENTATIONS

T. Sikora, Entwicklungstendenzen objekt-und modellbasierter Videocodierung (in-vited lecture), Media Seminar of TechnischeUniversität Ilmenau, 30.1.2001.

R. Steingrüber, Elektronenstrahllithogra-phie-Methode und Anwendung (invited),Universität der Bundeswehr München,30.1.2001.

M. Rohde, DPSK as an alternative modula-tion format for WDM networks, WorkshopITG-Fachgruppe 5.3.1, Darmstadt, January2001.

R. Schäfer, Verteilte Postproduktion inGbit-Netzen: Konzepte, Implementierun-gen, Erfahrungen, DFN Betriebstagung,Berlin, 6.-7.2.2001.

W. Passenberg, Investigation of MBEgrowth parameters on the optical qualityof GaInNAs quantum wells, poster, Cost2001, Würzburg, 27.2. 2001.

C. Caspar, Dispersion management in highspeed transmission systems, Workshop T-Nova, HHI, VPI, COM, Lyngby (DK),February 2001.

M. Rohde, Control modulation techniquefor the performance monitoring of WDM-Channels, Workshop T-Nova, HHI, VPI,COM, Lyngby (DK), February 2001.

M. Rohde, 10 Gb/s DPSK transmission inWDM-Links, Workshop T-Nova, HHI, VPI,COM, Lyngby (DK), February 2001.

C. Baack, Das Breitband-Internet der Zu-kunft, Vortrag beim ETV (anlässlich der Mit-



gliederversammlung des ETV), VDE Berlin,17.3.2001.

H.-P. Nolting, M. Gravert, Modeling of ultrafast optical clock based on tandem laser,Workshop on Simulation of Optical Systems,OFC 2001, Anaheim (CA, USA), 19.-22.3.2001.

H. Künzel, C. Janke, P. Haring Bolivar, A. Bartls, H. Kurz, Increased THz emitter efficiency by coherent superposition in ahigh repetition rate resonator, Frühjahrs-tagung der DPG, Hamburg, 26.-30.3.2001.

M. Hahn, C. Huck, Digitale Projektoren, 5. Potsdamer Filmkolleg, Potsdam, 30.3.2001.

H.-P. Nolting, All-optical 3R signal regener-ation, invited talk at Rank Prize Fund Work-shop on “Terabit Networks”, Windmere(UK), 4.4.2001.

M. Haardt, H. Boche, G. Wunder, The futureof mobile communications beyond thethird generation, 30. Annual IEEE Comm.-Theory Workshop, San Diego, April 2001.

P. Kauff, C. Fehn, E. Cooke and O. Schreer,Advanced incomplete 3D representationof video objects using trilinear warping fornovel view synthesis, PCS '01, Seoul, Korea,April 2001.

D. Marpe, ITU-T H.26L, current status ofdevelopment of the generic next-genera-tion video coding standard, FachbereichInformatik, Universität Rostock, 15.5.2001.

R. Schäfer, Bildkompression für die Video-übertragung in Mobilfunknetzen (invited),Kolloquium Digitale Mobilkommunikation,TU Berlin, 16.5.2001.

T. Sikora, Next generation user interfaces(invited plenary talk), Workshop ImageAnalysis for Multimedia Interactive Services,WIAMIS, Tampere (SF), 16.5.2001.

R. Schäfer, Visuelle Maskierung und Kunst-kopf – Meilensteine der Bildcodierung (in-vited), IRT-Festkolloquium zum 60. Geburts-tag von Dr. H. Wilkens, München, 30.5.2001.

H. Boche, Multi-user receiver for CDMA-systems, Elektrotechnisches Kolloquium derTH Karlsruhe, May 2001.

K. Müller, Der MPEG-7-Standard: Inhalts-beschreibung multimedialer Daten (invit-ed), 1. Thüringer Medianseminar der FKTG,May 2001.

C. Fehn, E. Cooke, O. Schreer and P. Kauff,3D analysis and image-based renderingfor immersive TV applications, ICAV 3D'01, Mykonos (GR), May-June, 2001.

J. Bischoff, R. Brunner, S. Gliech, A. Duparre,J. Bauer, M. Ferstl, Charakterisierung vonlateralen Nanometerstrukturen mit opti-schen und nichtoptischen Messverfahren,Jahrestagung der Deutschen Gesellschaft fürangewandte Optik, (DGaO), Göttingen, 6.-9.6.2001.

H. Heidrich, Monolithically integrated pho-tonic and optoelectronic circuits based onInP – technology, system applications,perspectives – (invited), Workshop OpticalMEMS and Integrated Optics, Dortmund, 11.-12.6.2001.

T. Tekin, H. Ehlers, M. Schlak, J. Berger, C. Schubert, B. Maul, R. Ziegler, All-opticaldemultiplexing performance of monolithi-cally integrated GS-MZI module, 1. Int.Workshop on Optical MEMS and IntegratedOptics, Universität Dortmund, 11.-12.6.2001.

R. Schäfer, Immersive media, BMBF-Work-shop, Internet und Medien – Anforderungenfür die zukünftige Entwicklung in der Inter-net-Technologie, Bonn, 13.6.2001.

S. Bauer, Laser based 3R regeneration forhigh-speed all-optical networks,WAON2001, Zagreb (Croatia), 14.6.2001.

V. Jungnickel, Breitbandige MIMO-Funk-übertragung, TU Dresden, Institut für Nach-richtentechnik, 18.6.2001.

R. Schäfer, Immersive media for communi-cation and infotainment, DLR-Workshop,Successful Technology Transfer, Berlin,21.6.2001.

A. Smolic, Globale Bewegungsmodelle undVideo-Mosaiken und deren Anwendungenin der Videoverarbeitung, ITG-Fachgrup-pentreffen 3.1.2, Digitale BildcodierungRWTH-Aachen, 22.6.2001.

G. Walf, Impact of Optical technologies onfuture networks (invited), Optical Networks

der EUROFORUM Deutschland GmbH;Freising, 27.-28.6.2001.

H.-G. Bach, OEIC-Entwicklung für 40 Gbit/sPhotoempfänger bis hin zur Modultechnik(invited), Infineon Fiber Optics, Seminarreihe“Fiber Optics Produkte”, Berlin, 13.7.2001.

N. Keil, Komponenten der optischen Nach-richtentechnik mit Polymeren (invited),Instituts-Kolloquium FhG IZM, Berlin,17.7.2001.

V. Jungnickel, A. Forck, T. Haustein, U. Krüger, V. Pohl, C. v. Helmolt, Wirelessinfrared communications beyond100 Mbit/s: system design, transmissionexperiments and potential, (invited paper),5th World Multiconference on Systemics,Cybernetics and Informatics (SCI), Orlando(FL, USA), 22.-25.7.2001.

H. Boche, Multi-user downlink beamform-ing, Elektrotechnisches Kolloquium derRWTH Aachen, July 2001.

H. H. Yao, Polymer waveguide devices forphotonic networks, Sorrento Networks Inc.,San Diego (USA), 3.8.2001.

H. H. Yao, Polymer waveguide devices: design and simulation (invited), SeminarUniversity of California, Los Angeles (USA),6.8.2001.

H. H. Yao, Research and development ofpolymer waveguide devices at HHI (invit-ed), Seminar Photon-X Inc., Malvern (USA),8.8.2001.

R. Schäfer, 3D-Fernsehen und 3D-Dienste –die Technologie ist reif, IFA 2001, Talk imTWF – Zukunft in 3D, Berlin, 29.8.2001.

C. Baack, Internetzugangsnetze mit hoherBandbreite und geringer Elektrosmogbe-lastung, Workshop (Stand und Trends derKommunikationstechnik) im Berliner Verbin-dungsbüro der Infineon Technologies AG,13.9.2001.

H.-P. Nolting, Devices for ultra-high speedall optical signal processing, Workshop DYNAMICS OF SEMICONDUCTOR LASERS(Modeling, Analysis, Experiments), organizedby WIAS, HHI, HU, Berlin, 13.-15.9.2001.

W. Passenberg, GaInNAs/GaAs quantum

well structures investigated by transientcapacitance spectroscopy, photoreflec-tance and photoconductivity, COST 268MC/WG Fall Meeting, Lecce (I), 16.-19.9.2001.

H. Künzel, J. Böttcher, C. Möller, MBEWachstum bei reduzierter Wachstums-temperataur für optimierte GaAs Tunnel-übergänge, Nationaler MBE Workshop,Zeuthen, 17.-18.9.2001.

G. Walf, Trends of optical Networks (invit-ed), Optical Internetworking der MarcusEvans Conferences, Berlin, 17.-19.9.2001.

C. Baack, Spitzenforschung in Berlin, Eröff-nungsveranstaltung von u2t InnovativeOptoelectronic Components GmbH, Berlin,24.9.2001.

C. Fehn, P. Kauff, O. Schreer and R. Schaefer,Interactive virtual view video for immer-sive TV applications, IBC '01, Amsterdam(NL), September 2001.

C. Fehn, P. Kauff and R. Schaefer, InteractiveVirtual View Video – Eine Erweiterung desimmersiven Fernsehens, DFS '01,Dortmund, September 2001.

T. Sikora and T. Meiers, Hierarchical imagesearch engines based on MPEG-7 descrip-tors (invited lecture), Intern. Conf. On ImageProcessing, ICIP, Thessaloniki (GR), 10.10.2001.

H. R. Schäfer, Image analysis and metadataextraction: Prerequisites for efficient im-age representation and coding, VLBV2001, Athen (GR), 11.-12.10.2001.

J. Faber, D. Ruschin, Subjektive Bildqualität– eine Begriffsbestimmung zur Untersu-chung von analoger und digitaler Kamera-technik im visuellen Vergleich (invited), ARRI-Technologieforum, München,19.10.2001.

J. Saniter, TransiNet: Network elementsand knode architectures in future opticalnetworks, Workshop on “How many dy-namic will be needed in future transport net-works?”, HHI Berlin, 23.10.2001.

H. Boche, G. Wunder, Channel estimationfor MIMO-OFDM, 39th Annual AllertonConf. on Communication, Control andComputing, Urbana (IL, USA), October2001.



S. Bauer, Phasenkontrollierte 2-Moden-Pulsationen: Theorie und Experiment,DFG-Kolloquium, Darmstadt, 5.11.2001.

W. Schlaak, Phontonische ICs und Netze(invited), Kolloquium des VDE, (Düsseldorf,Germany), 6.11.2001.

T. Tekin, 160 Gbit/s demultiplexer', semi-nar in optical fibre communication, HighFrequency Division (Photonics), TU Berlin,16.11.2001.

G. Walf, New evolutions of optical net-works (invited), DFN-Symposium des DFN-Vereins, Berlin, Germany, 19.-20.11.2001.

T. Sikora, MPEG-7 content description in-terface – eine Übersicht (invited lecture),Electrical Engineering Seminar of UniversitätRostock, 23.11.2001.

U. Busolt, T. Tekin, C. Schubert, J. Berger, M. Schlak, B. Maul, W. Brinker, R. Molt, Ch. Schmidt, H. Ehlers, Ultrafast all-opticalsignal processing by a monolithically inte-grated Mach-Zehnder-Interferometer, Int. Workshop on Optical Signal Processing,Copenhagen (DK), 29.-30.11.2001.

H. Boche, Optimale Algorithmen für dasJoint Multiuse Downlink Beamforming,Kolloquium über Elektronik und Nachrich-tentechnik der ETH Zürich, November 2001.H. Boche, Optimal SIR balancing with mul-ti-user downlink beamforming, Prof.Paulraj, Stanford University, November 2001.

H. Boche, G. Wunder, Peak to averagepower ratio for OFDM – new results,Elektrotechnisches Kolloquium der ETHZürich, November 2001.

M. Rohde, Control modulation techniquesfor performance monitoring in transpar-ent optical networks, DFG Workshop,Darmstadt, November 2001.

E. Schulze, R. Freund, M. Malach, F. Raub, 8 x10 Gb/s NRZ transmission over 1826 kmand 10 Gb/s over 4000 km with distrib-uted high-gain Raman amplification, DFG-Colloquium, Darmstadt, November 2001.

R. Schäfer, Immersive tele-conferencing – akey application for the next generationInternet (invited), IST Conference 2001,Düsseldorf, 3.-5.12.2001.

K. Biermann, H. Künzel, D. Nickel, K. Reimann, M. Woerner, T. Elsaesser, Ultra-schnelle optische Nichtlinearität von LTGaInAs/AlInAs Quantentöpfen bei1.55 µm, DGKK-Workshop “Epitaxie vonIII/V-Halbleitern”, Berlin, 6.-7.12.2001.

W. Passenberg, S. Mikhrin, H. Künzel, MBEWachstum und Materialeigenschaften vonGaInAsN und GaAsN, DGKK-Workshop“Epitaxie von III/V-Halbleitern”, Berlin, 6.-7.12.2001.

F. W. Reier, Optimierung von MOVPE-AlGaInAs/InP SL-MQW-Schichten für1,3 µm-Laser, DGKK-Workshop “Epitaxievon III/V-Halbleitern”, Berlin, 6.-7.12.2001.

P. Wolfram, E. Steimetz, W. Ebert, B. Henninger, M. Rakel, J.-T. Zettler, Opti-sche in-situ Messungen an InP-basieren-den Halbleiter-Strukturen, DGKK-Workshop“Epitaxie von III/V-Halbleitern”, Berlin, 6.-7.12.2001.

H. H. Yao, Polymer components for DWDMnetworks (invited Seminar) Shanghai Jiao-Tong University (China), 17.12.2001.

H. H. Yao, Polymer waveguide devices forphotonic network applications (invitedSeminar), Shanghai Jiao-Tong University,(China), 20.12.2001.

LECTURES

H. Boche, Digitale Mobilkommunikation I,TU Berlin

H. Boche, Digitale Mobilkommunikation II,TU Berlin

H. Boche, Mehrnutzerempfänger in derMobilkommunikation, TU Berlin

H. Boche, Space-Time-Signalprocessing fürdie Mobilkommunikation, TU Berlin

H. Boche, Mehrnutzer-Informationstheorie,TU Berlin

H. Boche, TU-Kolloquium “DigitaleMobilkommunikation” jointly with Prof. P. Noll and Prof. A. Wolisz, TU Berlin

H.-G. Bach, Grundlagen derHalbleiterbauelemente, TU Berlin

H. G. Bach, Messverfahren fürHalbleiterbauelemente, TU Berlin

E.-J. Bachus, PhotonischeKommunikationsnetze, TU Berlin

L. Ihlenburg, Entwicklungstendenzen derMultimediakommunikation, TU Berlin

A. Kortke, H. Boche, Praktikum Space-Time-Signalprocessing, TU Berlin

B. Kuhlow, Einführung in die Photonik, TU Berlin

A. Paraskevopoulos, Halbleitertechnologie für die Integration in der Optoelektronik, TU Berlin

M. Rohde, “Optische Netze”, as part of “Op-tische Nachrichtentechnik” by U. Fischer,TFH Berlin

O. Schreer, Multi-View-Geometrie in 3DVision, TU Berlin

H. G. Weber, Grundlagen und Anwendun-gen der linearen und nichtlinearen Faserop-tik, TU Berlin

G. Wunder, H. Boche, Statistische Signalver-arbeitung, Detection and Estimation Theory,TU Berlin

WORKSHOPS ORGANISED

Workshop on Switching and Routing inAdvanced Photonic Networks, organised bythe IST OPTIMIST project in conjunction withONDM 2001, Vienna University of Techno-logy, February

ITG-Workshop Optische Aufbau- undVerbindungstechnik, HHI Berlin, May

Dynamics of Semiconductor Lasers, (in coop-eration with WIAS, HU, FBH), Berlin,September

Annual review of national R&D programmesKomNet and OptoSys, HHI Berlin,September

Workshop on How many dynamic will beneeded in future transport networks?, HHIBerlin, October

Press conference on the occasion of the annual review of R&D programmes KomNetand OptoSys, HHI Berlin, November

16. Workshop des DGKK-Arbeitskreises“Epitaxie von III/V-Halbleitern”, LudwigErhard Haus, Berlin, December

CONTRIBUTIONS TO EXHIBITIONS

OFC 2001, Anaheim (CA, USA), March:Demonstration of 40 GHz all-optical clock recoveryThe German R&D Initiative KomNet

CeBit 2001, Hanover, February/March:Mixed Reality: Interaction with virtual objectsBINDI – A busy indicator for telework andtelecooperation

ECIO 2001, Paderborn, April:Optical high speed components

Industrie-Messe-Hannover 2001, April:Usability engineering at HHI

IFA 2001, Berlin, August/September:3D PC – Towards novel forms of human-computer-interactionKICK – Communication management in theInternetInteractive streaming MPEG-4 playerNavigation in 360° PanoramenImmersive TV- a new broadcast serviceStereo television demonstrator system (3D-TV)PC- and DSP-based AAC codecPC-, DSP- and RISC-processor-basedaudio/video decoderBroadband mobile communication

IBC 2001, Amsterdam (NL), September:PC- and DSP-based AAC codec

Poster Exhibition, Berlin, September:The national R & D programmes KomNetand OptoSys

ECOC 2001, Amsterdam (NL),September/October:KomNet – German R & D initiative on opticaltransport and networking technologies forthe emerging broadband InternetDemonstration of an all-polymer athermalarrayed waveguide grating multi-/demulti-plexer



MMC Workshop, Berlin, HHI, November:Interactive Streaming MPEG-4 PlayerEffiziente Repräsentation und Darstellunghochaufgelöster 360°-Panoramen

Poster Exhibition, Berlin, November: KomNet – On the way to the next genera-tion Internet

Elektronische Automatisierung, Nürnberg,November:CabrioScreen (autostereoscopic 3D displaywith infrared head tracker, jointly with ACTKern)

IST Event 2001, Düsseldorf, December:Demonstration of the next generation video-conferencing system by the IST projectVIRTUE

WDR Computerclub: Lange Nacht derComputer III, Köln, December: Interactivestreaming MPEG-4 player

COMMITTEE ACTIVITIES

Standardisation Committees

DVB Technical Module: MemberISO/MPEG

ITU-T, SG 16, Associated Rapporteur

Research Program Committees

COST 211: Member

COST 266, Progress of PhotonicInfrastructure towards the IT-Age: Member

COST 267, Semiconductor devices for signalprocessing WG2: Chairman

COST 268, Management Committee:Member (deputy)

DFG Keyprogram “Optical Signalprocessing”:Evaluator

IST (Inform. Society Technologies): EvaluatorStrategische Plattform Informationstechnik:Member

Technology Foundation STW: Evaluator

Conference and Workshop ProgramCommittees

Consultations Meeting PCM ProgramCommittee: Member

DGKK Workshop: Epitaxie von III-IV-Halb-leitern: Chairman

9th Dortmunder Fernsehseminar: ProgramCommittee

ECOC Technical Program Committee:Member

10th European Conference on IntegratedOptics (ECIO '01), 2001, Paderborn,Germany: Program Committee Member

Human Factors in Telecommunications:Permanent Steering Committee

Integrated Photonic Research 2001,Monterey, CA, USA

10th Intern. Plastic Optical Fibres Conf. (POF 2001): Program Committee Member

14th International Conference on InP &Related Compounds (IPRM '02), Stockholm:Program Chair

International Picture Coding Symposium:Steering and Program Committee Member

International Zurich Seminar (IZS 2002):Program Committee

Media Futures Conference: ProgramCommittee

Münchner Kreis, Congress “eCompaniesfounding, growing, harvesting”: ProgramCommittee Member

Nonlinear Guided Waves and theirApplications Topical Meeting, Cost 211:Member

Packet Video Workshop: Program Committee

SPIE Design, Manufacturing and Testing ofPlanar Optical Waveguide Devices: ProgramCommittee Member

Technisch-Wissenschaftliches Forum IFA '01Very Low Bitrate Video Workshop: ProgramCommittee

Editorial Boards

EURASIP Signal Processing

IEEE Signal Processing Magazine: AssociateEditor

IEEE Transactions on Circuits and Systems forVideo Technology: Editor-In-Chief

Image Communication: Associate Editor

Image Communication: Guest Editor

Institute of Physics (GB)/SemiconductorScience and Technology: Referee

Springer Verlag Berlin, Heidelberg, Series“Photonics”: Co-Editor

Advisory Boards

Photonic Network Communications

Other Committees

Arbeitskreis Integrierte Optik (AKIO):Member

Aspen Institut Berlin: Member

Competence Center for the Application ofNanostructures in Optoelectronics (NanOp):Member of Executive Board

FKTG, Urtel-Preis-Komitee: Curatorship

ITG, FA 3.1 Fernsehtechnik und elektronischeMedien: Chairman

ITG, FG 3.1.2 Digitale Bildcodierung:Chairman

ITG FG 5.2.5 Access- und Inhouse-Netze:Member

ITG, FA 5.3 Optische Nachrichtentechnik:Member

ITG, FG 5.3.1 Modellierung photonischerKomponenten und Systeme: CommitteeMember

ITG, FG 5.3.2 Photonische Integrations- undAufbautechnik: Chairman

ITG, FG 5.3.3 Photonische Netze: Member

ITG, FG 5.4.1 Optische Polymerfasern:Member

LMTB - Laser und Medizin-TechnologiegGmbH, Berlin: Advisory Committee

Münchner Kreis, Supranational Associationfor Communications Research: ResearchCommittee

OPTEC BB, Berlin: Member

TSB-Technologiestiftung InnovationszentrumBerlin: Board of Curators

TWF (Technisch-wissenschaftliches Forum)IFA '01

VDI optische Technologien: Advisory Board

EXCHANGE PROGRAM

Scientists Visiting HHI

K. Biermann, Max-Born Institut, Berlin, financed by Max-Born Institut, for four years

H. Bölcskei, University of Illinois at Urbana-Champaign (USA), financed by by Universityof Illinois and HHI, for one month

R. Freund, Fa. VPI, Berlin, financed by VPI, fortwo years

F. Futami, Fujitsu Laboratories Ltd.,Nakahara-ku, Kawasaki (J), financed byFujitsu, for one month

M. Gionannini, Institute Politecnico di Torino(I), financed by COST 267, for two weeks

V. Haisler, Institute of Semiconductors,Novosibirsk (RUS), financed by TU Berlin, forten months

K. Iiyama, Kanazawa University (J), financedby Kanazawa Univ., for eight months

I. Koichi, Kanazawa University (J), financedby Kanazawa University, for eight months

V. Kravcenko, TU Berlin, financed by TUBerlin, for one year

V. Mamutin, Abraham Ioffe Institute, St.



Petersburg (RUS), financed by HHI, for oneyear

S. Mikhrine, Abraham Ioffe Institute, St.Petersburg (RUS), financed by HHI, for sevenmonths

C. Plakas, Heriot-Watt-University, Edinburgh(UK), financed by Heriot-Watt-University, forthree weeks

G. Reißig, Massachussetts Institute ofTechnology (MIT), financed by MIT and HHI,for 2 weeks

N. Sabelfeld, MergeOptics GmbH, Berlin, fi-nanced by MergeOptics GmbH, for one year

S. Watanabe, Fujitsu Laboratories Ltd.,Nakahara-ku, Kawasaki (J), financed byFujitsu, for one month

T. Yamamoto, Fujitsu Laboratories Ltd.,Wakamiya (J), financed by Fujitsu, for oneyear

HHI Scientists Visiting Foreign Institutes

U. Kowalik, Tokyo University, Tokyo (J), financed by HHI, for one year

B. Rathke, TU Berlin, financed by HHI, forone year

COOPERATIONS

Industry

Acterna, Eningen

A.C.T. Kern, Donaueschingen

Agilent Technology, Ipswich (GB)

Aifotec GmbH, München, Berlin

Aixtron, Aachen

Alcatel SEL, Stuttgart

Alcatel, Paris

ASM Lithography, Veldhoven (NL)

ATIP GmbH, Frankfurt/M.

Atomika Instruments GmbH, München

AXON Photonics, Livermore (USA)

Bayerischer Rundfunk, München

BBC, London

Bertelsmann, Gütersloh

Bioshape AG, Berlin

Blaupunkt-Werke GmbH, Hildesheim

British Telecom, Martelsham Heath, Corning(GB)

Canto, Berlin

Carl Zeiss, Jena, Oberkochen

com center, Kopenhagen (DK)

Cybertron, Berlin

DaimlerChrysler, Ulm

das werk, München

Deutsche Bank, Berlin

Dr. Detlef Rose Fotomasken Mikroelektronik,Bergisch Gladbach

D-Research Digital Media Systems GmbH,Berlin

DSPecialist, Berlin

EPIGAP Optoelektronik GmbH, Berlin

Ericsson Eurolab, Aachen

Fresnel Optics GmbH, Apolda

Fujitsu Laboratories Ltd., Wakamiya (J),London (UK)

Fujitsu Mikroelektronik GmbH, Dreieich-Buchschlag

Hitachi Central Research Laboratory, Tokyo,Cambridge (UK)

HoloEye GmbH, Berlin

Infineon Technologies, Berlin, Regensburg,München

INM – Institut für neue Medien, Frankfurt/M

Ionas, Lyngby (DK)

ISIS Optronics, Mannheim

JDS Uniphase, Eindhoven (NL)

Kerpenwerke, Stolberg

Knowbotic Systems, Frankfurt/M

Laser Components GmbH, Olching

Laytec GmbH, Berlin

Loewe-Opta GmbH, Kronach

Lucent Technologies, Nürnberg, Holmdel(USA)

Lynx Photonic Networks (USA and IL)Medav Digitale Signalverarbeitung GmbH,Uttenreuth

MergeOptics, Berlin

Mikrom, Berlin

moove, Leverkusen

Nawotec GmbH, Rossdorf

neomagic (IL)

NTT Advanced Technology Corporation,Tokyo

Nokia, Espoo (SF)

Optibase, Herzliya (IL)

Opto+, Paris

OptoSpeed SA, Darmstadt, Zürich

Optovation Inc., Kanata (CDN)

Philips BV, Eindhoven (NL)

Quantum Devices Inc. (QDI), Yorba Linda,CA (USA), Berlin

Robert Bosch GmbH, Hildesheim, Stuttgart

SCHOTT GLAS, Mainz

Siemens AG, Berlin, München, Regensburg

Sony UK, Sony Stuttgart

Tecsi, Paris

Telenor, Oslo

Teracom AB, Stockholm

Tesat-Spacecom GmbH & Co. KG, Backnang

Thomson CSF Optonique, Paris

TOPTICA Photonics AG, München

T-Systems Nova, Berlin, Darmstadt

Tyco Electronics GmbH, Berlin

u2t Innovative Optoelectronic ComponentsGmbH, Berlin

Vcon Telecommunications Ltd. (IL)

Virtual Photonics Inc., Berlin, Melbourne,Freehold (USA)

vision pearls, Berlin

Vodafone Ltd., Newbury (GB)

2SK Media Technologies, Berlin

Universities and Institutes

Brunel University, Uxbridge (GB)

Central Electronic Engineering ResearchInstitute, Pilani, Indien

Denmark Technical University, Kopenhagen

DFN-Verein, Berlin

DLR Stuttgart

ETH Zürich

ETRI, Taejon (Korea)

European Broadcasting Union, Genf

Fachhochschule für Wirtschaft und Technik(FHTW) Berlin

Ferdinand-Braun-Institut, Berlin



FhG Institut für Biomedizinische Technik

FhG Institut für Integrierte Schaltungen,Erlangen

FhG Institut für Medienkommunikation (FhG-FOKUS), Berlin

FhG Institut für MikroelektronischeSchaltungen und Systeme, Duisburg

FhG Institut für Zuverlässigkeit undMikrointegration, Berlin und Teltow

FhG / IGD, Darmstadt

FhG IOF Jena

FhG Institut Siliziumtechnologie, Itzehoe

Forschungszentrum Kossendorf

Freie Universität Berlin, Fachbereich PhysikHahn-Meitner-Institut, Berlin

Heriot-Watt University, Edinburgh (GB)

Hochschule Harz, Wernigerode

Humboldt-Universität zu Berlin

IAF Freiburg

INRIA, Paris

Institut für Physikalische Hochtechnologie –IPHT, Jena

Institut für Rundfunktechnik, München

Ioffe-Institut, St. Petersburg (RUS)

IRISA, Rennes (F)

Konrad-Zuse-Institut, Berlin

Max-Born-Institut, Berlin

MPI Halle

Paul-Drude-Institut, Berlin

Poznan University of Technology (PL)

Queen Mary Westminster College, London

Senatsverwaltung für Finanzen, Berlin

Senatsverwaltung für Wirtschaft u.Technologie, Berlin

Stanford University, CA (USA)

Technische Fachhochschule Berlin

Tohoku University, Sendai (J)

Tokyo University

TU Berlin

TU Braunschweig

TU Chemnitz/Zwickau

TU Darmstadt

TU Delft (NL)

TU Denmark, Lyngby

TU Dresden

TU Hamburg-Harburg

TU Ilmenau

TU München

TU Wien (A)

Universität Aachen

Universität Bremen

Universität Dortmund

Universität Erlangen-Nürnberg

Universität Jena

Universität Rostock

Universität Stuttgart

University of Torino (I)

Universität Würzburg

University of Illinois at Chicago (USA)

University of Linz (A)

University of North Carolina, Chappel Hill(USA)

University of South Carolina, Columbia,(USA)

University of Patras (GR)

University of Torun (PL)

Weierstraß-Institut für Angewandte Analysisund Stochastik (WIAS), Berlin

Zentral- und Landesbibliothek, Berlin

Technology Collaboration Networks

NanOp, Berlin

Optec BB, Berlin

START UP COMPANIES

DSPecialists GmbH, BerlinDSPecialists develops systems and tools fordigital signal processing using digital signalprocessors (DSP). Its focus is to provide plat-forms in software and hardware for cus-tomers in various branches, such as audio,video, telecom and measurement/control.DSPecialists sells its products and providesdifferent kinds of services, such as technicaltraining and application development.

Virtual Photonics Incorporated (VPI)VPI Virtual Photonics makes design and plan-ning tools for access providers, network op-erators, system integrators and manufactur-ers of WDM transmission systems and opticalcomponents. Lead customers include Alcatel,Hitachi, Deutsche Telekom, Lucent, Pirelli,MCI and Sprint. Siemens, Telcordia andOpen Telecommunication are worldwidebusiness partners. VPI employs 150+ technical experts, consul-tants, developers and sales persons. Officesare located in Berlin, Munich (BroadbandAccess), Holmdel, NJ (IP/Optical Network-ing), Australia (Optical Systems) and Minsk(Software Production). Investors includeWeiss, Peck & Greer, Siemens, Techno Ven-ture Management and Deutsche Telekom.

2SK Media Technologies GbR, Berlin2SK Media Technologies develops and mar-kets software for compression and decom-pression of audio and video signals accord-ing to the MPEG standards. Its main productis the MPEG SoftEngine, which currently sup-

ports MPEG-1 and MPEG-2. Future products,which will support MPEG-4 and MPEG-7, areunder development.

MikroM, BerlinMikroM develops and designs VLSI compo-nents for image and sound processing andcompression. Its main product is HiPEG+, asingle chip HDTV decoder according to theMain Profile@High Level of MPEG-2. Thischip, which is based on the HiPEG chip pre-viously developed at HHI, contains the videodecoder and the systems demultiplex.

u2t Photonics AG, Berlinu2t Photonics AG was founded in 1998 bythree scientists of HHI and is now recognizedworldwide as a leading-edge supplier for theoptical telecommunications market. U2t de-velops and manufactures innovative opto-electronic components for the highest speedapplications. Flagship is an ultrafast photode-tector with more than 50 GHz bandwidth,capable of handling very high input power.Other products such as photoreceivers andmodulators are especially designed for40 Gbit/s communication systems.In 2001, u2t merged with another HHI spin-off, LKF Advanced Optics GmbH to combinetheir complemetary product portfolios. Formore information please visit www.u2t.de ormail to [email protected].

Usability Lab am HHI, BerlinThe company supplies Human Factors andUsability support at all phases of the productlife cycle. For all information and communi-cation technology products (services, sys-tems, terminals, user interfaces, etc.) it offersuser requirements analyses (based on empiri-cal studies), design studies with detailed rec-ommendations, evaluation of prototypes andproducts already on the market (by means ofCognitive Walkthroughs, HeuristicEvaluation, and Usability tests with test sub-jects), and consultancy in terms of strategicdecisions on product lines.(www:http://ulab.hhi.de)

MicroShape, BerlinMicroSHAPE (Software and HardwareApplication Engineering) develops softwareand hardware components and tools focus-ing on digital image processing. FurthermoreMicroSHAPE develops complete prototypesystems and peripheral units. Starting atschematic architectures via structure simula-tions and layout designs up to high integrat-



ed FPGA developments, all steps are microaccurately done by this company. The firstproduct will be a digital to analogue inter-face module for HDTV projection systemscompliant to the DVB and ATSC standard.

Perspective Technologies GmbH (PT),BerlinPerspective Technologies GmbH Berlin is a3D display and interaction systems develop-ment and marketing company founded by ateam of scientists of the HHI and experts inthe fields of market research, business strate-gy and innovation management. PerspectiveTechnologies is pioneering the combinationof highest-quality 3D technologies with nov-el non-intrusive user-interaction systems. As aresult, PT’s displays produce stunning 3D ef-fects without the need of any encumberingviewing aids, such as stereo glasses or head-mounted devices. The proprietary interactiontechnologies are based on the latest resultsin digital image processing and computer vi-sion research. First 3D display products willapply the lenticular lens technology, adaptedto range of target applications, e.g. in med-ical imaging, molecular design and 3D CAD.

HHI AT A GLANCE

Government research institute (FederalRepublic of Germany and State of Berlin)Total staff at end of 2001: 272 employees

Areas of Research andDevelopment

Photonic Networks• Design, development and demonstration

of optical communication networks and subsystems (access and customer networks, core networks)

• Investigation and development of WDMand high-speed OTDM techniques forhigh capacity transmission and routing

• Exploration of high speed transmissionperformance of photonic networks

• Development of techniques for networkoperation and maintenance

• Development and fabrication of photonicdevices and integrated circuits (lasers,modulators, switches, optical amplifiers,filters, multiplexers and demultiplexers,signal regenerators, transceivers, receiverfrontends) based on InP, for passive components on SiO2/Si and polymers

• Development of new concepts for carrying IP traffic over WDM optical networks and wireless networks in an efficient manner (TransiNet)

• Switchable network design and testing

Mobile Broadband Systems• Development of space-time-receiver for

the uplink of mobile communication systems

• Design of downlink beamforming for CDMA-based mobile communication systems

• Development of calibration algorithms for smart antennas

• Teletraffic engineering for mobile communication systems

• Development of signal processing algorithms for OFDM systems

• Development of sequences for CDMA applications

• Development of optical microwave generation and transmission systems forcellular mobile communication systems

• Implementation of MIMO systems• Development of signal processing

algorithms for MIMO systems

ISSN 1435-1587


R E P O R T 2 0 0 1

HE

IN

RI

CH

-H

ER

TZ

-I

NS

TI

TU

T•

RE

PO

RT

20

01

Electronic Imaging Technology for Multimedia• Development of algorithms and hardware

architectures for video and audio compression

• Development of algorithms and hardwarearchitectures for image analysis and synthesis

• 3D signal processing for tele-immersion• Image processing for studio applications• Video-streaming over IP and mobile

networks• Design of integrated circuits for image

processing• Development of 3-D display technologies• Man-machine-interaction for future

3D-desktop applications• Virtual reality for telework and

telecommunication• Development of user interfaces for

multimedia applications• Analysis and optimization of

communication services• Image and video retrieval systems

HEINRICH-HERTZ-INSTITUT FÜR NACHRICHTENTECHNIK BERLINGMBH

Einsteinufer 37, 10587 BerlinGermanyPhone: +49 (0)30 310 02-0Fax: +49 (0)30 310 02-213Email: [email protected]://www.hhi.de

Scientific Managing Director:Prof. Dr. Clemens BaackAdministrative Managing Director:Dr. Wolfgang Grunow

indoor-multipath channel

data

input

T1

Tn

R1

Rm

transmitter receiver

data

output

Multiple Input • Multiple OutputM iM o

Radio System for Wireless Local Area Networks

Date post:	02-Aug-2020
Category:	Documents
Upload:	others
View:	3 times
Download:	0 times

HHI AT A GLANCE HEINRICH-HERTZ-INSTITUT Areas of Research … · HHI REPORT 2001 5 TABLE OF...

Documents