© 2003 Mercury Computer Systems, Inc.
SDR – “Do You Care to Buy the Softest?”
SDR – “Do You Care to Buy the Softest?”
Jeff Smith, Jim Kulp, Murat Bicer, Tansu DemirbilekMercury Computer Systems, Inc.
199 Riverneck Road, Chelmsford, MA 01824
John Anton, Kestrel Technology
Military Communications Conference“Mobile Communications and Military Transformation”
Washington, D.C. - March 25, 2003
2
ProblemProblem Waveform portability vs. performance
Implementation specific architecture• Radios can be mostly ASIC implemented and be SCA compliant • Waveforms expect a lightweight component run-time env. that not yet
commercially available – Synchronize infrastructure requirements with commercial supply– Motivate commercial SDR to leverage defense work
Technology lifecycle Implementation-neutral components Fabric/infrastructure immunity to lifecycle “kinks” – PNP HW
Scope ELINT and COMINT superset of SR SCA compatible Coalition Waveform used as layer in proposed C4I
approach
Waveform development productivity Reuse and verification of waveforms vs. waveform parts Application to SIGINT
3
Elements of a “Softer” SDRElements of a “Softer” SDR
Reduction of SCA loopholes for performance Lack of scalable CCM/embedded
standards Convergence of SCA specs Levels of reprogrammability “Softer” waveform specification and
(re)generation technology SCA implementation viewpoints for
interoperability and portability Relation to larger apps e.g. SIGINT
and C4I
4
Portability IssuesPortability Issues
CORBA waivers Ports - components do not communicate through CORBA Communication paths can avoid CORBA messaging Efficiency Legacy
Too much of the waveform is not implemented in component software Proprietary code Logic in FPGA, DSPs and ASICs
Richness of Domain Profile CF must support all potential XML configurations Inhibits full spec implementation
5
Elements of a “Softer” SDRElements of a “Softer” SDR
Reduction of SCA loopholes for performance Lack of scalable CCM/embedded
standards Convergence of SCA specs Levels of reprogrammability “Softer” waveform specification and
(re)generation technology SCA implementation viewpoints for
interoperability and portability Relation to larger apps e.g. SIGINT
and C4I
6
Scalable CCM/Embedded Standards Issue Scalable CCM/Embedded Standards Issue
CCM and services wasn't ready for embedded, so SCA was done out of sync with CCM
Neither CCM nor SCA addresses scalable embedded multiprocessing where a component has a data-parallel implementation
Neither CCM nor SCA address non-GPPs or wideband dataflow in any reusable or interoperable way
CCM and services are being fixed (embedded profiles), so evolution of SCA can be pointers to CCM (lite)
Defining extensions based on experience and working them through standards: Reusable definitions for
DSP and FPGA codes Data parallel embedded
computer support (building on adopted spec)
Wideband interoperable dataflow (front end interoperability and RF/IF processing re-use)
Issue Solution
7
SCA extensions can address issuesSCA extensions can address issues
WDEs, Tool Kits
SCA
SCA extensions
CommonPrim. Ctrl.
Device Def.D&C
Answer
8
SDR Middleware SDR Middleware
What CCM interface SCA compliant Revised GUI Streamlined to this app
Why FPGA connection Infrastructure for Data Reorg
and multiprocessors Visualization and test
interface
A. CORBA interfacesB. CCM ServicesC. Embeddable (Lite)D. ScaleableE. Data Streaming/FlowF. Data ReorgG. D&C
A,B,G
A,B,E,G
CCM
SCA
C,D,E,F,GSCE
9
Data Parallel Middleware Embedded Computer Support
Data Parallel Middleware Embedded Computer Support
Component implementation span multiple processors
Stripe/distribute/ scatter streams across multiple processors - for streams from I/O ports to processors and among processors
Support scalable implementations that can use varying amounts of different speed processors as required
Data parallel CORBA (partial objects and parallel servers)
Alternative assemblies based on QoS and HW
Support for data reorganization standard (DRI)
Components that scale
Need Solution
10
Scalable Heterogeneous Components
High-performance, scalable middleware
Heterogeneous hardware G4s, Pentiums
FPGA components Dynamically scale
application-performance based on available resources
Component authoring tool Data-reorg, data-flow
SCA implementation that operates with Conventional middlewares Unique CORBA-compliant middleware (SCE) that supports seamless FPGA, DSP, and SW component interoperability and low-level machine standards
11
Elements of a “Softer” SDRElements of a “Softer” SDR
Reduction of SCA loopholes for performance Lack of scalable CCM/embedded
standards Convergence of SCA specs Levels of reprogrammability “Softer” waveform specification and
(re)generation technology SCA implementation viewpoints for
interoperability and portability Relation to larger apps e.g. SIGINT
and C4I
12
Lt. Log, Lt. CCM, Lt. Services, Deployment & Config., SWRadio, Parts of UML 2.0,
CORBA Variants, …
13
Software Radio StandardsSoftware Radio Standards
OMG SWRadio DSIG is working on a PIM for software radios
The SWRadio spec provides an SCA PSM mapping for this PIM
OMG specifications cannot remain specs without implementations (of PSM)
OMG SCA-compliant implementations permit convergence between SDRF and OMG – viz. Harris, Mercury implementations
OMG suite of specs is more abstract – applicable to a greater number of systems
OMG adoption means commercial acceptance (800+ international members)
14
Roadmap of Primary SpecsRoadmap of Primary Specs
Activity
Deployment & Configuration
RFP SWRADIO
RFP
Lightweight Services
RFP
Lightweight Log Service
RFC
Lightweight CCM RFP
**
Preparation of RFP by TF 11/01 2-6/02 5-6/02 5-6/02 7-11/02
Review by TC,
Approval of RFP by Architecture Board * 1/02 6/02 6/02 6/02 11/02
TC votes to issue RFP * 1/02 6/02 6/02 6/02 11/02
Public Comments due 9/02
LOI to submit to RFP due 4/02 10/02 8/02 2/03
Initial submissions due * 8/02 1/03 10/02 3/03
Voter registration closes 9/02 1/03 11/02 3/03
Initial submission presentations 9/02 1/03 11/02 3/03
Preliminary evaluation by TF
Revised submissions due * 3/03 5/03 3/03 5/03
Revised submission presentations 3/03 6/03 3/03 6/03
Final evaluation and selection by TF
Recommendation to AB and TC
Review by TC,
Approval by Architecture Board
TC votes to recommend specifications 5/03 8/03 5/03 10/02 7/03
BOD votes to adopt specifications 6/03 9/03 6/03 11/02 8/03
Finalization Task Force 6/04 9/04 6/04 11/03 8/04
shaded areas indicate completed task
* “Three week rule”
** LW CCM RFP dates are estimated; RFP content dependent on Deployment RFP.which may affect timing
15
Co-chair
Collaboration
Standard Reference ModelsStandard Reference Models
Proposed Design Behavior (states, sequences/roles)
Responsibility Levels Contained InformationOMG, SDRF Architectural Interfaces, DTDs
(components, properties, relationships)
CRC, SDRF Implementation Platform-specific code
Required to validate OMG spec and behavior Early reference implementation already used for SCA compliance testing SCA Interoperability
16
Elements of a “Softer” SDRElements of a “Softer” SDR
Reduction of SCA loopholes for performance Lack of scalable CCM/embedded
standards Convergence of SCA specs Levels of reprogrammability “Softer” waveform specification and
(re)generation technology SCA implementation viewpoints for
interoperability and portability Relation to larger apps e.g. SIGINT
and C4I
17
18
SW Radio Definitions Have Increasing Levels of Reprogramability
SW Radio Definitions Have Increasing Levels of Reprogramability
DynamicDeployment,
Plan &Reconfig
Waveform Air prog.
Static Config &Plan
Up/downConversion
withinFabric
ReconfigWaveforms
WithinFabric
Waveform/Channel
Waveform/Modem
Waveform/Prog.
Modem
Cluster A
The Future…Today’s
Technology
Waveform/Prog.
Receiver
19
Elements of a “Softer” SDRElements of a “Softer” SDR
Reduction of SCA loopholes for performance Lack of scalable CCM/embedded
standards Convergence of SCA specs Levels of reprogrammability “Softer” waveform specification and
(re)generation technology SCA implementation viewpoints for
interoperability and portability Relation to larger apps e.g. SIGINT
and C4I
20
Motivation for “Softer” Waveform Generation Technology & Reusable Components
Motivation for “Softer” Waveform Generation Technology & Reusable Components
Waveform developer
JTeL
Government
SIGINT
Spec Dev
Spec Implementer
Forward engineering waveform from parts, separation of concerns
Repository and verification of waveform parts
Common understanding, non-redundancy, less money
Working platform to apply constrained search to waveform classification
Check on services, use case identification and reduction and concise uniform description of waveform requirements
Reusable waveform components and capture of waveform design options, parameters and trades
If you are a: You get:
21
Characteristic ConceptualizationCharacteristic Conceptualization
Characterize waveforms into waveform attributes such that one could describe a waveform by combinations of types and parameters:• Unfolded table dimension = x1* x2 * x3*…* xN
• Waveforms can be thought of as unique paths in this 2-space• Maximum size is cross-product of columns
Air interface (x1=5)
Packet content (x2=2)
Connection type (x3=3)
Access mode (x4=3)
Signal type (x5=2) …
FDMA voice networking hopping analog
CDMA data point-to-point non-hopping digital
TDMA broadcasting direct sequence
DAMA
Push-to-talk
Trunked
…
…
A
B
22
x 5 = sig
nal type
analog
digital
x2 = packetcontent
CDMAFDMA TDMA
data
voice
DAMA x1 = air interface
Waveform configuration in N dimensionsWaveform configuration in N dimensions
consistent attributes
inconsistent attributes
x2 = packetcontent
CDMAFDMA TDMA
data
voice
DAMA x1 = air interface
analog
digital
with analog data, QPSKmodulation cannot be used
23
Next 3 dimensionsNext 3 dimensions
x6 = source coding
x4 = spectral maskx 7 = ch
annel coding
x1 = TDMAx2 = digitalxN = data
24
Larger View of Waveform Characteristics Ontology Used at OMG
Larger View of Waveform Characteristics Ontology Used at OMG
Signal: SWRI classification (see http://www.ece.neu.edu/~mbicer/Mobilecom/doc.html)Packet content: voice (analog & digital), data (digital), voice & dataConnection type: networking, point-to-point, broadcast, multicastSwitching type: circuit, packetLOS: yes, beyondAccess mode: push to talk, trunked, DAMA,TDMA,CDMA (direct sequence, multi-carrier), CSMA, FDMASecurity: transmission, communication, noneSpectral mask: center frequency, instantaneous frequency, bandwidth, powerChannel coding: convolutional, turbo, Reed Solomon, CRC, noneRouting: RF packets, IO packetsSource coding: yes, noChannel estimation: adaptive, trained, blind, nonePower control: open loop, closed loop, none
25
FM3TR ExampleFM3TR Example
Packet Content
Transmission Type
Connection Type
LOS Access Mode
Security Freq Hopping
Channel Coding
Source Coding
Channel Estimation
Power Control
Transmit Diversity
Voice
Analog Networking Exist PTT Transmission Yes Convolutional CVSD Adaptive Open Loop
STTD
Data Digital Point to Point
Beyond FDMA Communication No Turbo MPEG Trained Closed Loop
None
Voice & Data
Broadcast TDMA None Reed Solomon
MP3 Blind None
Multicast CDMA None None None
Packet Content
Transmission Type
Connection Type
LOS Access Mode
Security Freq Hopping
Channel Coding
Source Coding
Channel Estimation
Power Control
Transmit Diversity
Voice
Analog Networking Exist PTT Transmission Yes Convolutional CVSD Adaptive Open Loop
STTD
Data Digital Point to Point
Beyond FDMA Communication No Turbo MPEG Trained Closed Loop
None
Voice & Data
Broadcast TDMA None Reed Solomon
MP3 Blind None
Multicast CDMA None None None
FM3TR Data Transmission Mode
FM3TR Voice Transmission Mode
26
Each Point in 2-Space Has OSI Level Meanings
Each Point in 2-Space Has OSI Level Meanings
Characteristic type
Characteristicparam
Protocol stack Protocol stack
App
licat
ion
App
licat
ion
App
licat
ion
Symbol streams Symbol streams
SDR equipment
Sof
twar
eH
ardw
are
Sof
twar
eH
ardw
are
Sof
twar
eH
ardw
are
Sof
twar
eH
ardw
are
Sof
twar
eH
ardw
are
Sof
twar
eH
ardw
are
Sof
twar
eH
ardw
are
Sof
twar
eH
ardw
are
Sof
twar
eH
ardw
are
Sof
twar
eH
ardw
are
Sof
twar
eH
ardw
are
Sof
twar
eH
ardw
are
Software
Hardware
Phy
sica
l
Phy
sica
l
Link
Net
wor
k
Tra
nspo
rt
Ses
sion
Pre
sent
atio
n
App
licat
ion
Link
Net
wor
k
Tra
nspo
rt
Ses
sion
Pre
sent
atio
n
Antenna
Waveform channels
Info procchannels
IO channels
From École de technologie supérieure, Jean Belzile
• Characteristics can be realized in 7 OSI layers• Layers of characteristic realized as SCA components• Layer parts gathered from multiple waveforms & reused to compose a waveform protocol stack• Layering of components transparent to the SCA CF• Deal with only the Waveform channel but extendable to info proc and IO channels
27
Characteristic constraints can partitioned for parallel optimization
Characteristic constraints can partitioned for parallel optimization
A waveform space of N dimensions can be represented as, x1,x2, … ,xN}, where the set of N vectors need to be determined according to the constraint set A waveform may be represented as p(x1(3), x2 (5), x3 (2), …, xN (6)) = true, s.t. constraints e.g. dead-end, aggregation, etc.
The optimal waveform is found by choosing the parameters that minimize the
cost function J( .) s.t. such that:
opt = {arg min J( ) |, arg min J( ) |
}, P1 P2 Pm
s.t. Ki are independent, mutually exclusive & cover all constraints, partitions Pn are therefore mutually exclusive where,
J( ) |J(Pj) |jj=1
m
28
Reformulation of Problem…Reformulation of Problem…
A/D
Access mode
All All
A D
A AD D
TDMA FDMA CDMA
TDMA FDMA
InfoSecno
InfoSecyes
1. The relationship between every possible building block can be seen as a mesh network, where a connected line means an “allowed path” between two nodes
2. By choosing different root nodes, the mesh network can be viewed as a different tree
D
x1
x2
x3
x4
xn
P1
P2
29
Spec and Implementation GenerationSpec and Implementation Generation
Requirements
Waveform Specification
Waveform Implementation
Other Alternatives
Other Alternatives
Formally Validatable
Automated Process
Automated Process
3. Application dependent constraint analysis
Check feasibility of each application constraint
Create search tree given application priorities
4. Automated optimization
Use high-performance search technique to find optimal waveform specification that meets application constraints
30
Waveform Generation Technology Summary Waveform Generation Technology Summary
Refine and publish N-space, testing against JTRS legacy waveforms
Apply search and constraints against refined waveform characteristics Build constraints as predicates over characteristics
composing the points in waveform configuration space• Output power, baud rate, bit error rate, frame error rate
Use high-performance search techniques to identify viable candidates
31
Elements of a “Softer” SDRElements of a “Softer” SDR
Reduction of SCA loopholes for performance Lack of scalable CCM/embedded
standards Convergence of SCA specs Levels of reprogrammability “Softer” waveform specification and
(re)generation technology SCA implementation viewpoints for
interoperability and portability Relation to larger apps e.g. SIGINT
and C4I
32
SCA Implementation Viewpoints for Interoperability and Portability
SCA Implementation Viewpoints for Interoperability and Portability
Clients — software that requests SCA applications to be run and interacts with them
Applications — software that assembles and configures SCA components into a useful structure, in SDR-speak, called a “waveform”.
Components — software that performs some processing function.
System bootstrap/management — software that brings up a pre-configured SCA infrastructure, with various hardware devices
Devices (Drivers/Proxies) software that owns and uses hardware devices
Knowledge required to play this role (write this type of code), to use the SCA system in this way
Basic use case or scenario for playing the role
Core Framework interfaces used
Metadata formats use Testing issues
Types of software written to use an SCA-compliant environment: For each we describe:
No viewpointNo viewpointBut UIBut UI
CF or PlatformCF or PlatformDeveloperDeveloper
PlatformPlatformDeveloperDeveloper
WaveformWaveformDeveloperDeveloper
SCA Dev.GuideViewpoints
33
Roadmap to Implement SCA Partitioned by Viewpoint
Roadmap to Implement SCA Partitioned by Viewpoint
Application name, port names and interfaces, metadata file locations
Available components, their interfaces and descriptor file locations
SCA APIs and API Building blocks.
POSIX & SCA services & event channels
Name of DMD & DMD files to describe how to run bootstrap
Component writer’s info + CF interfaces for devices to be uniformly managed/ monitored
Find SCA server (via naming service), use CF to run application and use its external ports.
SAD is parsed and acted on by the Domain Manager and/or Application Factory
App.Factory instantiates, configures & connects components. Apps route client requests for retrieving refs to external ports to components.
Parse DMD & start described software that starts a DomainMgr. Partially parse DCD to start DeviceMgrs
Devices are used/created according to the DCD processed by the DeviceMgr, creating device at startup time
DomainManager,
ApplicationFactory,
Application
None Port, LifeCycle, TestableObject, Port Supplier, PropertySet, Resource, Resource Factory
DomainMgr., DeviceMgr.
DeviceMgr (used); Device, ExecutableDevice & LoadableDevice (implemented)
None Create SAD, reference SPD
IDL, SCD & SPD files SPD, SCD & DMD files.
DCD refs SPD for the device & DPD for the associated HW device
Test of skeleton apps that export appropriate interfaces & external ports
Apps can be built with non-compliant components
Test components independent of actual SCA applications, accessing SCA-limited OE.
Use log and event service to test exposure to key interfaces. Test SW launch.
A standalone device test or full apps consisting of components deployed on the device
Client Waveform Component System boot/mgt Device
KnowledgeRequired
Use-case
CFInterfacesUsed
Meta-dataFormatsUsed
TestingIssues
Viewpoint:
34
Elements of a “Softer” SDRElements of a “Softer” SDR
Reduction of SCA loopholes for performance
Lack of scalable CCM/embedded standards
Convergence of SCA specs Levels of reprogramability “Softer” waveform specification and
(re)generation technology SCA implementation viewpoints for
interoperability and portability Relation to larger apps e.g. SIGINT
and C4I
35
SDR Forms Lower Levels of One Projected C4I Protocol Stack
SDR Forms Lower Levels of One Projected C4I Protocol Stack
6. Content
2. Discovery (linking) waveform - ICS
5. Communications waveform-SCA/FM3TR
4. Networking protocols – TCP/IP
3. Identification process - ALE
1. C4I information (registry database)
Coalition waveform architecture leverages standards, communication, networking and modeling investments.
36
COMINT ReceiverCOMINT Receiver
37
RWR/ESM/ELINT ReceiverRWR/ESM/ELINT Receiver
38
Summary Summary
SCA letter for “efficiency” vs. spirit for waveform portability
Lack of CCM and embedded standards
Evolution of OMG and SDRF SCA versions
Waveform granularity/level of verification & portability
Levels of SDR reprogramability
Existence proof (remove excuses) + future standards, HW, SW
Component-centric middleware that supports D&C, Lwt. CCM and embedded standards
1) Ref design model that supports versions & 2) reduce SCA by using OMG portions
1) Waveform generation tech. & 2) reference design model
Identify/rate levels and phase above per level
Issues Suggestion