Spot diffusing technique and angle diversityperformance for high speed indoor diffuse infra-redwireless transmission

A.G. Al-Ghamdi and J.M.H. Elmirghani

Abstract: For future high-speed indoor wireless communication systems, spot diffusing techniquesin an indoor nondirected diffuse channel draw great attention. In the paper, a novel spot diffusingconfiguration based on a line strip multibeam transmitter (LSMT) in association with an anglediversity receiver is used to improve the performance of optical wireless (OW) systems. The impactof both ambient light noise and multipath dispersion is investigated. Original results are presentedevaluating and assessing the system performance when based on the proposed geometry, andcomparison is carried out with four other geometries under the same conditions. The authors alsocompare the results of using the diversity detection with the spot diffusing technique and the case ofno diversity with conventional diffuse systems where a single wide field-of-view (FOV) is used.The simulation results show that the performance improvement of the LSMT with a three directiondiversity receiver is an enhancement of ,15 dB (signal-to-noise ratio) over the other conventionalgeometries in the presence of very directive noise sources and multipath dispersion. The RMS delayspread performance as well as the signal-to-noise ratio for the proposed configurations, at differentpositions on the communication floor (CF), are evaluated and compared.

1 Introduction

Recently, indoor optical wireless (OW) systems haveattracted great attention due to their potential in high-speed, low cost applications [1–14]. Optical wirelesstransmission offers several potential advantages over radiofrequency (RF), including abundant unregulated spectrum, aradiation that does not pass through walls or opaque objects(hence the possibility of frequency reuse) and that does notinterfere with other wireless radio sources. Nondirected OWsystems can mainly be classified into two configurations:line-of-sight (LOS) and diffuse systems. LOS can only beestablished by having a direct path between transmitter andreceiver. Any moving objects can easily obstruct the directpath. Moreover, this class of system needs careful alignmentin order to set up the link. Diffuse systems are a veryattractive alternative – they offer robust links as well asalleviating the problem of shadowing, since diffuse systemsdo not rely on the transmitter and receiver alignment andonly count on reflections from walls, ceiling and otherreflectors. Both configurations are affected by multipathpropagation, where the impact of this effect is less for LOSsystems. Furthermore, in indoor OW systems, backgroundoptical noise sources such as sunlight or artificial light (forexample, incandescent lamps) induce significant shot noisein the photodetectors and can burn out the received data,especially when the receiver lies underneath the noisesource [2].

In this paper, several spot diffusing configurations using amultibeam transmitter are proposed in order to reduce theeffect of the above obstacles and to improve the systemperformance. The transmitter is placed on the communi-cation floor (CF) and pointed up. A holographic opticaldiffuser is assumed to be mounted on the emitter, resultingin multiple narrow beams, which illuminate multiple smallareas, forming a lattice of diffusing spots on the ceiling [1].Careful hologram design can even offer intensitydistribution within the spots. Multiple spots organised inuniform, diamond, and LSMT configurations are studiedand compared with the conventional diffuse system (CDS),which employs a wide transmitter beam and a wide field-of-view (FOV) receiver. Multibeam transmitters can beimplemented practically using computer-generatedholograms (CGH) for a particular spot intensity and/orintensity distribution [3].

Compared to the other geometries, the LSMT systemprovides a significant performance improvement when it isaccompanied by an angle diversity receiver that employsthree narrow FOV photodetectors tilted in differentdirections, thereby separating signals that arrive fromdifferent directions [4]. Furthermore, an angle diversityreceiver provides the possibility of using combiningtechniques to combine the output signals from differentreceivers. Using more than one receiver can ensureuninterrupted reception of the optical signal when there istransmission blockage. In addition, the use of LSMTcombined with angle diversity detection (three receivers)has demonstrated an SNR improvement of ,15 dB over thespot diffusing systems that employ wide FOV receiver and20 dB SNR improvements over the CDS system.

2 Simulation set-up

In this Section, the characteristics of the channel formed bya multibeam transmitter (uniform, diamond, and LSMT

q IEE, 2004

IEE Proceedings online no. 20040140

doi: 10.1049/ip-opt:20040140

The authors are with the School of Engineering, University of WalesSwansea, Singleton Park, Swansea SA2 8PP, UK

Paper received 28th March 2003

IEE Proc.-Optoelectron., Vol. 151, No. 1, February 200446

illumination) are studied and compared with the CDS. Thetransmitted signal propagates to the receiver throughmultiple reflections from room surfaces. Propagationsimulations were conducted in an empty room with floordimensions of 8m� 4m (length � width) and a ceilingheight of 3 m. It is assumed that the room has neither doorsnor windows. Previous research work has shown that plasterwalls reflect a light ray in a form close to a Lambertianfunction [5]; therefore, the ceiling and the walls of the roomwere modelled as Lambertian reflectors. Up to second-orderreflections were taken into account and full walls reflectivityis assumed. High reflectivity is chosen as it results in thehighest multipath dispersion, and thus significant pulsespread.

The transmitter is placed in the middle of the CF, 1 mabove the floor, and is modelled as a generalised Lambertianemitter, with a radiant intensity (W=sr) given by

RðjÞ ¼ n þ l

2p� Ps � cosn j ð1Þ

where Ps is the total average transmitted optical powerradiated by the laser/LED source, j is the angle of incidencewith respect to the transmitter’s surface normal, and n is themode number describing the shape of the transmitted beam(the higher the mode n the narrower the light beam), whichis also related to the half-power semi-angle (hps) of thetransmitter pattern. Moreover, the transmitter is alwayslocated in the middle of the CF, 1 m above the floor, at theroom centre (4m� 2m), and emits 30 dBm total opticalpower with an ideal Lambertian radiation pattern whichcorresponds to a half-power semi-angle hps ¼ 60�: Themode number n can be given by n ¼ �0:693= lnðcosðhpsÞÞ:

A simulation tool similar to the one developed byBarry et al. [6] has been used to produce the impulseresponses and power distribution, and to calculate the delayspread. To model the reflections, the room reflectingsurfaces were divided into a number of equal size squareshaped reflection elements. The accuracy of the receivedpulse shape, and the received optical signal power arecontrolled by the size of the surface elements. For allgeometries (the multispot channels and the traditionaldiffuse link), the surface elements of 5 cm� 5 cm for thefirst order reflections, and 20 cm� 20 cm for the secondorder reflections, were used. These dimensions have beenselected in order to keep the computation within reasonabletime and measure. The reflecting elements have been treatedas small transmitters that diffuse the received signals fromtheir centres in the form of Lambertian pattern with aradiation lobe mode number n ¼ 1: In all the cases studied asingle photodiode has been located at different locations onthe CF, 1 m above the floor, with a photosensitive area ðArÞof 1 cm2; and with a wide angle of reception of 180�: Thesimulations were carried out at several receiving positionswithin the room.

In addition to the high reflectivity surfaces and in order toassess the system’s performance, in a realistic situation,eight halogen spotlights, which result in one of the moststringent optical spectral corruptions to the received datastream, have been chosen to illuminate the environment. Toevaluate the impact of ambient light, the background noise(BN) distribution pattern of an incandescent light wasinvestigated. The ‘Philips PAR 38 Economic’ (PAR38) wasinvestigated. PAR38 emits a power of �65W in a narrowbeamwidth in which it is modelled as having a generalisedLambertian radiant intensity with order n ¼ 33:1: The eightspotlights were placed 2 m above the CF and positionedequidistantly on the ceiling as shown in Fig. 1. These lamps

produced a well illuminated environment. Furthermore,simulation of the optical noise power along both axes of CFwas carried out in steps of 10 cm.

3 Channel model

In OW communication links, intensity modulation withdirect detection (IM/DD) is the preferred choice [7]. Anindoor OW channel using IM/DD can be fully characterisedby its impulse response h(t):

yðt;Az;ElÞ ¼XM

m¼1

R � SðtÞ hmðt;Az;ElÞ

þXM

m¼1

nmðt;Az;ElÞ ð2Þ

where yðt;Az;ElÞ is the received instantaneous current in thephotodetector at certain positions due to m reflectingelements, t is the absolute time, Az and El are the directionsof arrival in azimuth and elevation (angle), M is the totalnumber of reflecting elements, S(t) is the transmittedinstantaneous optical power of the transmitter, denotesconvolution, R represents the receiver responsivity, andn(t, Az, El) is the background noise modelled as Gaussiannoise independent of transmitted signal. Because of the OWtransmission nature, the indoor communication link issubjected to multipath dispersion, which can cause ISI.root-mean-square delay spread is a good measure of signalpulse spread due to temporal dispersion of the incomingsignal.

To evaluate the effects of multipath dispersion on thenondirected optical channel, the rms delay is usuallyadopted to measure the dispersion of the collected signal.The delay spread of an impulse response is given by

D ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPðti � mÞ2P2

riPP2

ri

sand m ¼

PtiP

2riP

P2ri

ð3Þ

where ti is the delay time associated with the receivedoptical power Pri (Pri reflects the impulse response h(t)behaviour) and m is the mean delay. As shown in (3), therms delay spread is calculated from the simulated channelimpulse response. It has been evaluated for all presentedgeometries and at different points on the CF.

4 Transmitter structures

In this Section, four different configurations are presentedand analysed. This section compares the various proposedmultibeam transmitter geometries to identify the mostsuitable geometry for use in indoor OW. Furthermore,simulations were carried out to evaluate the improvement

Fig. 1 Eight-spotlight distribution in an OW configuration

IEE Proc.-Optoelectron., Vol. 151, No. 1, February 2004 47

achieved through the use of different mutlibeam transmitterconfigurations and the influence of the distribution of thediffusing spots on the ceiling. The three different multibeamstructures were simulated and compared with the CDS.

4.1 Conventional diffuse system (CDS)

This is the basic configuration and has been widelyinvestigated [8–10]. The conventional diffuse link uses asingle beam transmitter and a wide single element receiverðFOV ¼ 180�Þ: For comparison purposes, a conventionaldiffuse link has been simulated to generate channel impulseresponses, power distribution, and delay spread. Forimpulse response assessment, the receiver location waschosen at the room corner ðx ¼ 1m; y ¼ 1m; z ¼ 1mÞ toexamine the worst receiver position case. The impulseresponse of the CDS configuration is shown in Fig. 2.

In the case of the CDS, as the distance between thetransmitter and the receiver becomes large, the power of thecollected optical signal decreases rapidly and thereby thetotal coverage is reduced [8].

4.2 Uniform multibeam transmitter

The multi-spot diffusing link was first proposed by Yun andKavehrad [11]. It utilises narrow beams pointed in differentdirections aimed at the ceiling. This structure is evaluated toassess the potential gain to be made using our proposedstructures (line strip and diamond multibeam transmitters).A holographic optical element mounted on the transmitter isassumed to create multiple narrow beams and to form16 � 8 diffusing spots. These diffusing spots are evenlydistributed on the ceiling with equal intensities, in which thedistance between two adjacent spots is 50 cm. Such aconfiguration for the case of a wide FOV receiver isillustrated in Fig. 3.

Observing Fig. 2, it is clear that the impulse responsespread has significantly increased compared to the CDS.This is due to two major factors: the use of a single wideFOV detector and the contribution of the diffusing spots.Furthermore, due to the large distance between the diffusingspots and the detector (at room corner) and since thereceived signals, at any point on the CF, can come directlyfrom diffusing spots as well as other directions (walls andceiling), severe pulse spreading results, as shown in Fig. 2.

An improvement in the received optical power is clearlyvisible when the uniform multibeam transmitter isemployed instead of the CDS, as can be seen in Fig. 2.

This significant improvement is due to the full ceilingcoverage with uniform distribution of diffusing spots, whichleads to a reasonably uniform power distribution as shownin Fig. 3.

4.3 Diamond multibeam transmitter

In this section, another geometry of multibeam transmissionlink is simulated. The link produces four line strips forminga diamond shape of diffusing spots on the ceiling. Figure 4shows a diamond multibeam transmitter structure, when awide FOV receiver is used. Every line in the diamondconsists of 20 spots, where the separation between twoadjacent spots is �10 cm: The diffusing spots are traced oneby one to find their contribution to the received signal.

To simulate the proposed geometry, the receiver wasplaced at various locations on the CF. The receivedmultipath profiles were stored for each spot at each location.The resultant power profile is the sum of the powers due to

Fig. 2 Impulse responses of diffuse OW link

Simulations were performed near room corner at (1 m, 1 m, 1 m) for the fourconfigurations

Fig. 4 Diamond multibeam transmitter

a Physical configurationb Power distribution

Fig. 3 System model of uniform multibeam transmitter

a Physical structureb Power distribution

IEE Proc.-Optoelectron., Vol. 151, No. 1, February 200448

the 80 impulse responses. The diffusing spot is assumed tooperate as a secondary transmitter with a Lambertiandistribution.

The impulse response, at the room corner, for thediamond configuration is shown in Fig. 2. The powerdistribution over the CF for the diamond multibeamtransmitter is illustrated in Fig. 4b, The results show a lessuniform power distribution since most of the collectedpower lies on the area close to the line strips, where thedistance between transmitter and receiver is the minimum.From Fig. 4b, it can be clearly seen that, at x ¼ 2m andalong the y-axis, the signal power level increases slightly, inparticular at room sides and corners due to the extra numberof diffusing spot contributions at these locations. In contrast,near the room centre, the signal power is small due to thelarge distance between the diffusing spots and the receiver,which makes the direct path link between spots and receiverweak. The collected power level near the side walls,however, increases due to the diamond spots construction,wherein the maximum collected power is found in theregions where two adjacent line strips can concurrentlyilluminate the receiver directly.

4.4 Line strip multibeam transmitter

A novel structure of diffusing spots that employs amultibeam transmitter is proposed and examined. Figure 5shows a diffuse optical wireless communication system thatemploys the proposed line strip multibeam transmitter inconjunction with a wide FOV receiver. The same multibeamtransmitter is assumed to produce 1 � 80 beams aimed atthe ceiling with equal intensities, and to form a line ofdiffusing spots in the middle of the ceiling at x ¼ 2m andalong the y-axis The difference in distance between eachtwo adjacent spots is 10 cm. These spots become secondarydistributed emitters, which emit Lambertian radiation. Theceiling and the walls of the room were modelled asLambertian reflectors with first order reflection [12].

The impulse responses of the four different configurationsat the room corner are depicted in Fig. 2. It is clearly seenthat the power received by the multibeam transmitterstructures is significantly better than that of the CDS. This isdue to the fact that the impulse response of theseconfigurations contains many peaks corresponding to thedifferent direct path components between the diffusing spots

and the receiver. In addition, our impulse response resultshave further confirmed the findings in [11] that most of thecollected signal is in the first order reflection, concentratedwithin a very short time period due to the contribution of themany direct path components. On the other hand, impulseresponses for these configurations (that use single wide FOVreceiver) suffer from pulse spread due to multipathpropagation. To reduce the effect of multipath dispersion,different techniques can be implemented.

In this paper, LSMT with angle diversity is studied(Section 5), where it reduces the delay spread and provides aremarkable system improvement over the otherconfigurations.

5 Performance comparisons

In this Section, we investigate the performance in terms ofdelay spread for the above four configurations and computethe resulting SNRs. Additionally, the systems are assessedand evaluated in the presence and in absence of backgroundnoise. To evaluate the system performance in a BNenvironment and under the most stringent conditions,neither an optical filter nor an optical concentrator havebeen used. Furthermore, LSMT, in conjunction with anglediversity, is also described and compared to otherconfigurations.

5.1 Delay spread performance

Figure 6 shows the delay spread performance of the fournondirected multibeam transmitter configurations as well asthe CDS, considering a receiver positioned on CF;x ¼ 1m and x ¼ 2 and along the y-axis. For the multibeamtransmitter case, where a single wide FOV receiver is used,the delay spread is clearly larger than that of the CDS (singlebeam transmitter and wide receiver FOV) over the entirecommunication floor. This is due to the fact that themultibeam transmitter features many signal propagationpaths between transmitter and receiver. Figure 6 also shows,for the CDS case, that there is a direct relation between thedelay spread and the distance from the transmitter.

In contrast to the CDS configuration, the delay spreadvariation, for the case of the spot diffusing technique, issmall (towards the room corners), which is due to thepresence of spot diffusing transmission points near the roomcorners and the lower contribution of the far spot points.Furthermore, for the case of wide single receiver, Fig. 6shows that the delay spread for the diamond configuration islower than that associated with the uniform multibeamconfiguration, in particular at x ¼ 2m and along the y-axis.On the other hand, the delay spread values increased slightlyin the area where the diffusing spot illumination is low.Comparing the cases of spot diffusing techniques when asingle wide receiver is used (Fig. 6), it is to be noted that thelowest delay spread values are obtained by the proposedLSMT with angle diversity receiver. The smallest delayspread associated with the LSMT and three narrow branchesreceiver is 0.5 ns, which is 9.5 dB, 8.2 dB 7.6 dB and 6 dBlower than the smallest delay spread associated withuniform spot diffusing system, the diamond spot configur-ation, the LSMT when single wide FOV receiver is used,and the CDS geometry system, respectively. Also, the delayspread associated with LSMT and diversity detection issmaller than the one associated with uniform spot diffusingand angle diversity detection by �4 dB: The maximumdelay spread associated with the LSMT and angle diversityreceiver is also low and has a value of 2.2 ns. This is 4 dB,4 dB, 3.6 dB and 2 dB lower than the maximum delay spreadassociated with the three systems (in the same respective

Fig. 5 LSMT multibeam transmitter

a Physical configurationb Power distribution

IEE Proc.-Optoelectron., Vol. 151, No. 1, February 2004 49

order), while for a uniform multibeam transmitter, LSMTdemonstrates lower delay spread.

The LSMT with angle diversity structure has resulted in asignificant improvement over the cases of spot diffusing thathave single and diversity receiver configurations in thevarious metrics considered. For instance, impulse responseat the weakest reception point (room corner) on the CF iscompared. By comparing Figs. 2 and 7, the results showthat, in spite of the different number of beams employed bythe multibeam transmitter (128 for uniform, 80 for diamondand LSMT), redistribution of the diffusing spots on theceiling when it is accompanied with narrow directive FOVreceivers, can achieve improvement in delay spread as wellas power reception, as shown in Figs. 6 and 7. Also, Fig. 7shows that the power received by first order reflectioncontains most of the total collected power, whereas thehigher order reflections comprise the lowest power, whichmeans better-received power and less pulse spreading.

5.2 Performance assessment of the line stripmultibeam transmitter with angle diversityreceiver

In contrast to the single wide-FOV receiver, in this section,the receiver is a collection of narrow-FOV detectorsoriented in different directions, forming an angle diversityconfiguration. The optical signal power received in thevarious receivers can be treated separately, and can beprocessed using several techniques such as combining or

selection. Furthermore, to combat background noise as wellas multipath dispersion, diversity detection is an appropriatechoice, where a significant performance improvement canbe achieved [13]. The detectors are placed on squarepyramidal faces, which form a new geometry that isinvestigated in this work.

By using such configuration, and by optimising the FOV,directional interference can be minimised. The squarepyramidal detector diversity system considered consists ofthree photodetectors, mounted only on three-square pyr-amid faces. Each face bears a certain direction that can bedefined by two angles: azimuth (Az) and elevation (El)angles. While the El of two photodetector remains at 35�;the third one is facing up with El of 90�; and the Az for thethree faces of the detectors are fixed at 0�; 180� and 0�.In addition, their FOVs have been chosen to achieve the bestSNR; hence, two of them were restricted to 35�; whereas thedetector that faced up was set to 20�: The design of thereceiver structure was based on a line strip spot diffusingconfiguration. This structure is able to look at the spotdiffusing points from the entire CF. Moreover, the anglediversity receiver is designed so that at least five diffusingspots are always positioned within the receiver FOV,providing a robust link against diffusing spot blockage.Using such a configuration also leads to an additional degreeof freedom that can be used to eliminate a large amount ofthe background interference. Additionally, narrower FOVsare used to restrict the range of incident rays accepted andhence reduce the pulse spread, at the possible expense ofpower loss.

For performance analysis, the signals received by thethree receivers are calculated. Let R1 be the distancebetween the transmitter and the surface element dA, and R2

be the distance between the surface element and one of thedetectors. An arbitrary surface at ðxp; yp; zpÞ on the x-z wallis selected to be the reflector surface element. R1 and R2 aregiven by:

R1 ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðxt � xpÞ2 þ ðyt � ypÞ2 þ ðzt � zpÞ2

qð4Þ

R2 ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðxr � xpÞ2 þ ðyr � ypÞ2 þ ðzr � zpÞ2

qð5Þ

where ðxt; yt; ztÞ and ðxr; yr; zrÞ and are the transmitter andreceiver locations on the CF, respectively. The total powerreceived by the transmitter through one wall is given by

Fig. 6 Delay spread performance

(i) Diamond multibeam transmitter; (ii) uniform multibeam transmitter;(iii) single wide FOV receiver; (iv) line strip multibeam transmitter;(v) uniform multibeam transmitter (angle diversity); (vi) conventionaldiffuse system; (vii) line strip multibeam transmitter (angle diversity)a x ¼ 1mb x ¼ 2m

Fig. 7 Impulse response of nondirected diffuse OW link

Simulations were performed near room corner at (1 m, 1 m, 1 m) for LSMTuniform spot configuration and angle diversity receiver

IEE Proc.-Optoelectron., Vol. 151, No. 1, February 200450

Pr ¼Z x�z wall

xp;yp;zp

�½Psrðyt � ypÞðyr � ypÞ

� ðzp � ztÞð1 þ ð1= tanðElÞ2Þþ R2

2 � ðxp � ðxr þ ðcosAz= tanElÞÞÞ2

� ðyp � ðyr þ ðsinAz= tanElÞÞ2

� ðzp � zr � 1Þ2ÞðArÞrectðd=FOVÞ�=

½2p2R41R4

2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 þ ð1= tanElÞ2

q��

dA ð6Þ

The function rectðd=FOVÞ describes the FOV of thephotodetector. By controlling FOV, the receiver can havecontrol over the range of incident rays that are detected. Therectðd=FOVÞ is given by

rectðd=FOVÞ ¼ 1 for ðd=FOVÞ � 1

0 for ðd=FOVÞ > 1

ð7Þ

where d is the reception angle with respect to the receiver’ssurface normal. Function rectðd=FOVÞ describes the FOVof the photodetector. By controlling FOV, the receiver canhave control over the range of incident rays that aredetected, while the total collected light power in a receiveris the sum of contributions from all the room surfaces.Consequently, the total power in a one receives branch isgiven by

Prtotal ¼ 2Prx�z wall þ 2Pry�z wall þ Prx�y wall ð8Þ

Using these equations, we evaluate the SNR for eachconfiguration. For the diversity detection case, we considerone way of processing the resulting electrical signal fromthe different photodetectors, namely, selection of thephotodetector with the best SNR. Furthermore, the receivedpulse shapes for the four configurations have beenconsidered in calculating Ps1 and Ps0; the power associatedwith logic 0 and logic 1, respectively. The probability oferror ðPeÞ of the indoor OW communication system can bewritten as

Pe ¼ QðffiffiffiffiffiffiffiffiffiSNR

pÞ ð9Þ

where Q(x) is the Gaussian function, which assumes a valueof 6 at probability of error Pe ¼ 10�9; and the SNR takingPs1 and Ps0 into account (hence ISI) is given by

SNRs ¼ MAXi

R � ðPs1 � Ps0Þi

�t;i

� �2

; 1 � i � I ð10Þ

where I is number of photodetectors, R ¼ 0:5A=W is thephotodetector responsivity, and �t is the total noisevariance, which can classified into three categories:

(i) Background light-induced shot noise ð�bnÞ; which canbe evaluated by computing the corresponding shot noisecurrent. It can be calculated from its respective associatedpower level ðPbnÞ using

�bn ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 � q � Pbn � R � BW

pð11Þ

where q, Pbn and BW are the electron charge, receivedbackground optical power and receiver bandwidth,respectively(ii) noise induced by the received signal power, whichconsists of two components: shot noise current ð�s1 when a‘1’ is received and ð�s0 when a ‘0’ is received). This signaldependent noise is very small in this case and can beneglected

(iii) receiver noise normally generated in the preamplifiercomponents. The preamplifier used in this study is the PIN-BJT design proposed by Elmirghani et al. [14]. Thispreamplifier has a noise current density of 2.7 pA=

ffiffiffiffiffiffiHz

pand

a bandwidth of 70 MHz; therefore the preamplifier shotnoise is given by

�pr ¼ 2:7 � 10�12 �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi70 � 106

p¼ 0:023 mA ð12Þ

hence, �t is �0 þ �l; which represent the noises associated:

�0 ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�2

pr þ �2bn þ �2

s0

qand �l ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi�2

pr þ �2bn þ �2

sl

qð13Þ

The assumption of Gaussian noise statistics holds in ourcase, since thermal and shot noise can be modelledaccurately as Gaussian processes. To investigate andexamine the effect of the background noise (BN), twocases are presented. The first case is implemented when thecommunication environment is free of BN. The directiveBN is the second case. Figure 8 shows the detected SNR, forthe previous two cases, when the system operates underthe constraints of background noise (eight directed

Fig. 8 Signal-to-noise ratio distribution at: x ¼ 1 m and 2 m andalong y-axis

(i) diamond multibeam transmitter (single detector); (ii) line stripmultibeam transmitter (single detector); (iii) conventional diffuse system,(iv) uniform multibeam transmitter (single detector); (v) LSMT inconjunction with angle diversity effects; (vi) uniform multibeam transmitterand angle diversity effectsa x ¼ 1mb x ¼ 2m

IEE Proc.-Optoelectron., Vol. 151, No. 1, February 2004 51

spotlights n ¼ 33:1) and fully multipath dispersion, at the xco-ordinate that contains the weakest and the strongestreceived optical signal power along the y-axis as well aswhen there is no BN. Note also that neither opticalconcentrator nor optical filter was used. Under theseconditions, all the four configurations (single wideFOV ¼ 180�) are compared with the angle diversityreceiver when LSMT and uniform multibeam transmitterare used. Observing Fig. 8, a high level of SNR can easily beachieved, in particular when there is no BN. While under theconstraint of BN, it results in a very deteriorated systemperformance. In addition, Fig. 8 shows that SNR ismaximum at points close to the diffusing spots and faraway from the BN sources. This phenomenon comes fromtwo facts: firstly the distance between the transmitter and thereceiver is minimum, which results in a strong receivedsignal. Secondly, the noise distribution has a very low value,as the receiver is not underneath the spotlights. Furthermore,it shows that background noise has a significant impact onthe system performance when a single wide FOV (180�)receiver is used.

Furthermore, the impact is more evident when the wideFOV receiver is placed directly under a light source (ortowards the room corners where the distance between thetransmitter and the receiver is large, and hence the signal isweak). This is due to the fact that the increase in the FOVyields an increase in the amount of BN detected by thereceiver. This can be easily seen at y ¼ 1m; 3m; 5m and7m where the SNR has its minimum values. Comparing theresults shown in Fig. 8, it can be seen that, in spite ofemploying different types of transmitters, the signaldegradation is clearly visible at locations near room sidesand corners as the difference in distance between thetransmitter and the receiver increases and the diffusing spotilluminations decrease. In contrast, a remarkable improve-ment in the SNR is seen, in particular when an LSMT withan angle diversity receiver is used. Fig. 8 shows SNRimprovement, in particular at room corners and along they-axis, where the directional interference peaks exist. This isattributed to the fact that the noise levels at these locationsare significantly reduced due to diversity and due toreduction in the FOVs. Furthermore, due to the receiverstructure, Fig. 8 shows that the adapted LSMT as well asuniform spot diffusing configuration have not been affectedby the BN, where the SNR are almost the same in all pointson the CF. The improvement obtained by using the proposedstructure can be seen; a significant SNR improvement overboth CDS and the conventional multibeam structures isobtained as shown in Fig. 8. Compared to CDS, uniformspot diffusing and LSMT configurations, when they arecombined with angle diversity detection, yield an SNRimprovement of >20 dB and �15 dB; respectively. Since auniform multibeam transmitter with diversity detectionresults in a high level of SNR as well as high pulsespreading, LSMT, using a diversity detection technique, canoffer a very low probability of error ð�10�9Þ and low pulsespreading at the cost of a substantial decrease in complexityand in an economic fashion, from which it gains moreadvantage than using the uniform configuration. Conse-quently, an angle diversity detection technique would alsoseem to be an appropriate choice for reducing the BN as

well as the multipath dispersion effects as it selectivelyconfines the range of reception angles.

6 Conclusions

This paper has presented a new spot diffusing configurationbased on a line strip multibeam transmitter. The LSMT,in conjunction with the angle diversity receiver (havingnarrow directive FOVs), has improved the performance ofnondirected (diffuse) OW systems. The proposed systemhas demonstrated significant optical power as well as adecrease in the delay spread towards the walls and corners.Such a system combines the advantages of both direct pathlink and diffuse transmission in an economic attractivefashion. Additionally, it employs a smaller number ofdetectors (only three), in contrast to the existing diversitydetection methods, which results in lower complexity andcost. The improvement in performance achieved is due tothe significant reduction in background noise as well as areduction in ISI effects. The LSMT configuration has alsodemonstrated a remarkable improvement, including anextensive drop in the noise power level, a strong receivedsignal due to the decrease in transmitter–receiver separation(as the diffusing spots are now large in number), reduceddelay spread, and improved SNR (�20 dB).

7 References

1 Eardley, P.L., Wisely, D.R., Wood, D., and McKee, P.: ‘Holograms foroptical wireless LANs’, IEE Proc., Optoelectron., 1996, 143, (6),pp. 365–369

2 Boucouvalas, A.C.: ‘Indoor ambient light noise and its effect onwireless optical links’, IEE Proc., Optoelectron., 1996, 143, (6),pp. 334–338

3 Simova, E., Tai, M., and Kavehrad, M.: ‘Indoor wireless infrared linkwith a holographic multiple-spot diffuser’. Proc. Applications ofphotonic technology 2, Montreal, Canada, 1996, 48, pp. 223–228

4 Al-Ghamdi, A.G., and Elmirghani, J.M.H.: ‘Optimization of apyramidal fly-eye diversity receiver for optical wireless systemsunder the influence of multipath dispersion and background noise’,Microw. Opt. Technol. Lett., 2003, 36, (5), pp. 401–406

5 Gfeller, F.R., and Bapbst, U.H.: ‘Wireless in-house data communicationvia diffuse infrared radiation’, Proc. IEEE, 1979, 67, (11),pp. 1474–1486

6 Barry, J.R., Kahn, J.M., Krause, W.J., Lee, E.A., and Messercgmitt,D.G.: ‘Simulation of multipath impulse response for indoor wirelessoptical channels’, IEEE J. Sel. Areas Commun., 1993, 11, (3),pp. 367–379

7 Kahn, J., and Barry, J.: ‘Wireless infrared communications’, Proc.IEEE, 1997, 85, (2), pp. 265–298

8 Al-Ghamdi, A.G., and Elmirghani, J.M.H.: ‘Performance and field ofview optimisation of an optical wireless pyramidal fly-eye diversityreceiver’, J. Opt. Commun., 2002, 23, (6), pp. 215–222

9 Tavares, A.M.R., Valadas, R.J.M.T., and De Oliveira Duarte, A.M.:‘Performance of an optical sectored receiver for indoor wirelesscommunication systems in presence of artificial and natural noisesources’, Proc. SPIE – Int. Soc. Opt. Eng., 1995, 2601, pp. 264–273

10 Al-Ghamdi, A.G., and Elmirghani, J.M.H.: ‘Performance of an opticalpyramidal fly-eye diversity receiver for indoor wireless communicationsystems in the presence of background noise and multipath dispersion’.Proc. London Communications Symp., September 2002, pp. 221–224

11 Yun G., and Kavehrad M.: ‘Spot diffusing and fly-eye receivers forindoor infrared wireless communications’. Proc. 1992 IEEE Conf. onSelected topics in wireless communication, Vancouver, BC, Canada,1992, pp. 286–292

12 Jivkova, S.T., and Kavehrad, M.: ‘Multispot diffusing configuration forwireless infrared access’, IEEE Trans. Commun., 2000, 48, (6),pp. 970–978

13 Al-Ghamdi, A.G., and Elmirghani, J.M.H.: ‘Analysis of optical wirelesssystems based on line strip multibeam transmitters and diversitydetection’, J. Opt. Commun., in press

14 Elmirghani, J.M.H., Chan, H.H., and Cryan, R.A.: ‘Sensitivityevaluation of optical wireless PPM systems utilising PIN-BJTreceivers’, IEE Proc., Optoelectron., 1996, 143, (6), pp. 355–359

IEE Proc.-Optoelectron., Vol. 151, No. 1, February 200452