1474 PROCEEDINGS OF THEIEEE, VOL. 67, NO.

1 1 , NOVEMBER 1979

Wireless In-House Data Communication via

Diffuse Infrared Radiation
FRITZ R. GFELLER AND URS BAPST

Abstmet-A novelwireless broadcast/multi-acwschannel is descn'bed

for flexibly interconnect@ a duster of data terminals located within
thesameroom.ThetMgniPsionmediumisdiffusivelyscatteredinfrored
radiation at 95Oam melength. Transniasion is low-twnedium speed
and the range up to 50 m. Theoretical analysis indicates that the time
dispersion limits the trmsmission bandwidth of the system to 260
"bit. m/s, but background noise produced by ambient. .daylightreduces
the trnnsmission speed bdow 1 "bit/& The trrnsrmsslon propertiesot
the diffiw optical channel are analyzed, and experimental digital links
for basebaud F"at 125 kbit/s and PSK 64 kbit/s are demonstmted.

I. INTRODUCTION

w E ARE at present facing an enormous growth of com-

puter-based terminal stations. These terminal stations
are interconnected t o form distributed or centralized
computer networks of increasingcomplexityanddensity
performavariety of functions.Intheofficeenvironment,
to
a
w
W
J

+
a

0
a 1
individual word processing, data processing, and information 2 ;
systemsmovetowardsintegratedofficesystems.Onthe
shop-floor level, process control,manufacturing,andtesting
depend increasingly on a multitude of interconnected terminal
stations such as sensors, transducers, microprocessors, or in-
dustrial robots. Similarly, in the merchandising environment,
cash-register terminals, stock taking, or order placements are
becomingincreasinglya part ofcommondata-processing
systems. This development has had an impact on the system
architecture and a number of pertinent architectures has.been
conceived [ 11-[ 3 I with an end to flexibility and adaptation
to growth.
Fig. 1. (a) Diffuse optical channels connecting terminalclustersinan
In the same room or shop floor there is often a multitude of in-houseenvironment. S = satellite, T = terminals. (b) Diffuse
terminals clustered together. There is a need to flexibly place optical-channel network.
terminals at different positions within the same room t o suit
particular working conditions, to reconfigurate existing terminal arecostlyandcausedowntime.Inaddition,theremay also
arrangements, or even t o -havemobilehand-heldterminals. be problems with ground loops and noise pickup.
Examples are cash-register terminals in supermarkets or data
terminals in a production plant where economic production A . Wireless Optical Channel
of a particular model requires frequent modifications in number An alternative to cabling is wireless communication. In this
and location of the terminals. paper, we propose diffuse optical radiationin the near-infrared
The conventional method of interconnecting terminals is by region as signal carrier to interconnect a cluster of terminals
wiring. There is often a multitude of different cable types to located in the same room to a common cluster controller [41.
suitparticulartransmissionspeeds. While mainspowering is In principle, each terminal interface is equipped with a light-
readily available and prewired with tapping points within easy emittingdiode LEDanda photodiodefor convertingan
reach, signal cables may have to be pulled individually. Actual electronic signal t o an optical signal, and vice versa, and the
situations are very diverse and difficult t o categorize, but, in corresponding driver and receiver circuits. Similarly, a central
general, the cabling problem ranges from lowcost solutionsof optical satellite station serves as an interface t o the common
loose cables on the floor to crowded ducts or even raised floors cluster controller which may be local or remotely connected
with hgh costs. Once the cables have been installed, changes via a conventional wire link (see Fig. l(a)). Optical radiation
fromthesatellite(downlink) is diffuselyscattered from the
surrounding walls, ceiling, and other objects in the room, thus
Manuscript received February 8, 1979; revised June 22, 1979.
The authors are withthe IBM ZurichResearch Laboratory,Riisch- filling the entire room with the optical signal carrier. Reception
likon, Switzerland. is nondirectional, i.e., the photodiodes receive the incoming

0018-9219/79/1100-1474$00.75 0 1979 IEEE

GFELLER AND BAPST: DATA COMMUNICATION
DIFFUSE
VIA IR RADIATION 1475

radiation from a wide field of view (FOV). Hence, there is no an address and the numberof users is only bounded bychannel
direct line of sight required between transmitter and receiver, capacity. There is no limit to temporarily inactive users beyond
and the optical transmission path is very insensitive to inter- that of finite address field size. An in-house system based on
ruption. The same transmission method applies for the uplink diffuse optical channels may be expanded or contracted with-
(terminals to satellite). out major changes and offers great flexibility. A diffuse optical
A comparison between diffuse optical and radio-frequency channel could also serve as a subnet in a local area network.
transmission reveals the following aspects in favor of the optical
link: with radio-frequency transmission FCC regulations apply C. Overview of the Topics Addressed
within the range 10 kHz to 300 GHz. There are two classes, In this paper, we present a study on the potentialand limita-
namely, licensed transmitters and unprotected restricted radia- tions of the diffuse optical channel. Our effort has been con-
tion devices. Fortheapplications considered, licenses are centrated mainly onthe physicaltransmission properties,
obtainedwith increasing difficultydue to the overcrowded namely, bandwidth limitations
due to multipath signal
spectrum. dispersion, optical power demandfor satellite andterminal
The optical signal carrier considered does not fall under FCC interfaces, as well asrangeand distribution of the diffuse
regulations and thereis n o interference with the electromagnetic optical radiation field.
spectrum. In
contrast to radio-frequency antennae, the In Section 11, we describe an idealized diffuse optical channel.
dimensions of LED and photodiodes are small, and the optical It can be visualized as a transmitter and receiver sitting in an
radiation is essentially confined to the room where it is gen- optical cavity representinga room. If the cavity walls are
erated.Thereisthusnointerferencewith similar systems uniformly illuminatedby thetransmitterthentheoptical
operating next door and ahigher degree of data security is signal power incident on the receiver photodiode is constant,
affordedthan radio-frequencytransmissioncanoffer. An irrespective of thelocationandorientation of the receiver
interesting possibility in connectionwith diffuse optical within the cavity. This is a most desirable feature which, how-
transmission is the operation of wireless hand-held terminals. ever, can only be approximated in practice. We also investigate
Progress inlow powerconsumption microprocessors,liquid- theeffect of multipathpropagation within the cavity. The
crystal displays, andminiaturekeyboardshas broughta resulting time dispersion is analyzed assuming a signal which is
batterydriven personal terminal within the state of the art diffusely reflected from a plane surface(ceiling). A bandwidth-
[51, [61. Conceivable applications are message transmission, distance product is then derived indicating an upper bound for
paging, voice transmission, remotecontrol,stock taking, or the transmission speed for a given room size.
classroom teaching. In Section 111, abaseband PCM receiver is discussed both
Since the appearance of the first LED therehas been theoretically andexperimentally.Apartfromthe infrared
enormous progress in the technology of electrooptical com- signal, the receiver photodiode is also exposed toambient
ponentsforoptical communications. Of special importance light. A photon striking the photodiode produces, withagiven
to us are high radiant output, high efficiency LED’s, sensitive probability, ahole-electronpair which is manifested as an
large-area and lowcapacity pin photodiodes, as well as low- electrical currentintheexternal circuit.Since this is a
noise optical receiver circuits. Ontheconsumermarket, statistical process, the generation of the photocurrent due to
wireless infrared devices have already madetheir mark in ambient light is associated with noise which degrades the
remote controlof TV sets and earphones for audio transmission. transmission quality. Some typical ambient-light levels in
office environments have beenmeasuredand optical filters
B. System Aspects for blocking part of theambient light are discussed. The
The diffuse optical channel can be visualized as a completely optical receiver power required for a prescribed transmission
connected network (see Fig. l(b)). Optical radiation from the error rate is then calculated with the ambient-light level and
satellite (downlink) reaches every terminal within the room, transmission speed asparameters. Finally, an experimental
thus forming a broadcast channel. Since this channel is used PCM receiver operating at 125 kbit/s is described. Measure-
only for the satellite, no collisions of data packets can occur. ments of the error performance showed good agreement with
The terminals select only packets addressed to themselves and theory. However, with baseband receivers, transmission
discard the rest. Radiationfromtheterminals reaches the integrity may be impaired in the presence of fluorescent lamps
satellite as well as the other terminals. (However, there may emitting light with frequency components falling within the
be “hidden” terminals not in contact with one another.) The receiver passband.
uplink is thus a multi-access channel.Here, the problem of Section IV addresses the infrared field distribution in a com-
time sharing among all users arises. Access may either be plex environment. A brief survey of the infrared reflective
random or by polling. The polling process may be based on an properties of office interiors is given. A computer model is
SDLC-type protocol where the cluster controller serves as a then described simulating the distributionof the diffuse infrared
primary station. A performance analysis of various random field for a given room geometry and transmitter LED configura-
access methods suchas ALOHA or carrier-sense multiple-access tion. With the aid of this simulation model, suitable LED
CSMA has been given by Kleinrock et a2. [ 71. Data transmis- configurations can be found that provideasufficiently uni-
sionmay be either baseband PCM orbymodulated carrier form illumination of a given interior. A case study of three
methods such as FSK or PSK. The carrier method offers the different room geometries and satellite LED configuration is
possibility of having several separateoptical channels. The then presented. Finally, using the above three cases as an ex-
diffuse optical channel supports a large population of active ample, the opticalpower neededfor a 64 -kbit/slink is estimated
andinactive users. In wire communications,the maximum for both thesatellite interface and the terminal interfaces.
number of users is often bounded by some hardware limita- Section V deals with an experimental 64-kbit/s PSK trans-
tions. In wireless communications each user is represented by mitter-receiver interface using a 256-kHz carrier. The use of a
1476 PROCEEDINGS
IEEE,
OF THE VOL. 67, NO. 11, NOVEMBER 1979

carrier allows operation of several separate channels. Further-

more, thereceiver passband is shifted away from the baseband,
thus allowing effectiveelectricalfiltering of ambient-light
fluctuations.

II. IDEALIZEDDIFFUSEOPTICALCHANNEL
1
,q
-
/ REFLECTING
SURFACE

A. Optical Power Received from Diffusely

Reflecting Surface
The key features of the diffuse optical channel are readily
derived from a simple idealized model. An optical source
illuminates an arbitrarily shaped surface A such that the radiant
emittance w (power emitted per surface area into hemisphere)
is constant over theentire area A (see Fig. 2). We further
assume that the surface is an idealLambertianreflector.A
photodiodewith photosensitive area AR and field of view RECEIVER
FOV is then subject to the diffusely scattered radiation from
surface A . The system optical source (satellite), a cavity with Fig. 2. Optical power received from diffusely reflecting surface.
diffusely reflecting surface A (representing the physical bounds
of the room), and the photodetector (terminal) thus form a where p = 1 and FOV = 90'. This wouldrequire thatthe
diffuse optical channel. Theoptical power incidentonthe photodetectors be in close contact with the radiating surface
photosensitive area AR from the surface element dA is given A = A F . Note that P,, is also the power contained in direct
by line4-sight radiation from a distant source falling normally
upon the area AF.
Theproblem inareal environment will therefore be to
distributethesourceradiation to the reflectingsurfaces as
where homogeneously as
possible, bearing in mind their size,
geometry,and reflective properties, in orderto achieve an
1
dS1 =- AR COS 6, if AR << R Z approximately constant received signal power level over the
RZ entire working area of the room.
Thephotodiode represents a relatively high-cost high-
1
dA = -R Z sin 6 d6 de. capacitance device with area AR where the incident radiation
cos 7 is collected and converted to an electronic signal. Therefore,
All variables are functions of the coordinates 6 and E. Ifwe it appears attractive to collect optical radiation witha low-cost
consider only a homogeneously illuminated surface A with w a passive optical system (lenses, reflectors) having a large aperture
constant, then the total power incident on AR is obtained by A, andconcentratingtheincidentradiationon a small-area
integrating over the entire FOV lowcapacitance photodiode. However, from phase-space [ 81
or second-law thermodynamic considerations [9] it has been

PR = AR
R
iFov
i2n sin 6 COS 6 d6 de
shownthatthreedimensional
(2) relation
ideal concentratorsobeythe

AR = AL sin2 (FOV). (6)

and
Hence, there is no gain possible over a single photodiode ex-
PR = w AR sin2(FOV), 0 < FOV G 90'. (3) posed to a homogeneously radiating surface A extending over
thefull hemisphericalFOV. Furthermore, practicalconcen-
Thus the received optical power PR is independent of position trating systems resemble thetheoreticalperformancefor
and angular orientation of the photodetector with respect to relatively small angles FOV < 30° only. This is already a con-
the surface A , provided the surface covers the entire FOV. A straint on the freedom of angular position of the photodiode
hypotheticalbut illustrativeexample is asphericalcavity with respect to the environment. For thesereasons, passive
illuminated by an isotropic source located in the center. An optical systems appear only to be useful for direct line-of-sight
estimate of the optical-power demand P, for the source illumi- transmission, for bounded diffusely reflecting areas appearing
nating the cavity wall A is given by below an angle smaller than 30°, or if a bright ambient-light
source is to be kept out of the FOV.
APmh
P, (4)
A R sin2
~ (FOV) B. Multipath Dispersion
where P m b is the minimum power requiredby the receiver We now consider the effect onsignal distortion due to multi-
circuit, and p is a reflection coefficient indicating the propor- pathpropagation.Inorder to gain a simple representative
tion of diffusively reflected incident radiation. result, thereflecting surface A of arbitrary shape is replaced by
The theoretical minimum power P,, to supply a population a plane surface A' (ceiling) with distance L from the photo-
of photodetectors arranged over a floor area AF is given by diode (see Fig. 2). At time t = 0 a Dirac pulse arriving from a
distant source is reflected by the surface A ' , producing an im-
AF
P,, = Pmh - pulsive uniform radiant emittance w over the entire surface.
AR The optical power incident on the photodiode at thepropaga-
GFELLER AND BAPST: DATA COMMUNICATION VIA DIFFUSE IR RADIATION
tion delay r is, using (1) and expressing R in terms of r,
where
r=- L
c cos 6
L RECTANGULAR PULSE RESPONSE q(tl
70 = -.
C
Here, c is the speed of light and T O the minimum delay. The
contribution dPR(r) within thetime element dT originated
from annular surface elements dA'. From (7), the normalized
impulse response F ( T )is calculated using (7) and (3)
Bearing in mind the impulse response, the resulting normalized
signal pulseshape g ( t ) arriving atthephotodiodemay be
obtained by convoluting F ( T ) with the original signal f(t) in-
cident on thereflecting surface
where
1: g(t) d t = 1.
In Fig. 3, the impulseresponse F ( r ) and the resulting pulse
shape g ( t ) for a rectangular pulse f(t) of duration To = L / c is
given with FOV as parameter. Multipath propagation causes a
spread of the transmitted pulse which may result in a loss of
pulse amplitude and intersymbol interference. There is, there-
fore, a maximum transmission speed ultimately depending on
room size, where these effects affect transmission if no signal
processing is used. In a given environment, the multipath con-
ditions are somewhat different for eachsource/receivercon-
figuration. However, an example shall serve as a roughestimate
for the maximum transmission speed with baseband transmis-
sion. Consider source and receiver located at the same position
at somedistance L fromthe reflectingsurface which may
representthe ceiling of a large room. The radiation charac-
teristics of the sourceare assumed to cause a homogeneous
radiant emittance w of the surface appearing withinthe circular
FOV of the receiver. Assuming rectangularpulses f(t) with
50-percent duty cycle and FOV = 60", the above-mentioned
criteria for zero amplitude loss and intersymbol interference
of the received pulses At) are just met (see Fig. 3). Here, we
1411
- DlRAC RESPONSE F(T)
8 6(t)
-
6 -
-
4 -
-
2 -
-
0 '
I
0 2 3 1 4
T4
"
0 1 2 3 1 4
Td
Fig. 3. Signal dispersion due t o multipath propagation. The top figure
showsthe impulse response of aplanediffuselyreflectingsurface.
The bottom figure shows the correspondingrectangular pulse response.
have to considera double propagation delay to account for
the individualsource-to-reflector and reflector-to-receiver
paths. We can approximately associate the circular area in the
reflectingsurface, as described by the FOV, with the room
size. For a diameter of 5 m (small room)the transmission
speed satisfying the above criteria is 52 Mbitls. Similarly, for a
diameter of 50 m (hall) the transmissionspeed is still 5.2
Mbitls. This result may be expressed in terms of a bandwidth-
distance productamounting to approximately260 Mbitmls.
These speedsaresufficient for local data distribution. How-
ever, if a modulated carrier is used these considerations apply
to the carrier frequency to limit phase distortions.
111. OPTICALRECEIVER
A. Ambient Light
In a diffuse optical channel, the photodiode is exposed to
the ambient light, introducing additional noise to the input
circuit of the receiver amplifier. This is in contrast to fiber-
optics receivers. Themostcommonlyencounteredambient-
lightsourcesare daylight,tungsten, andfluorescent lamps.
Their relative spectral power densities are shown in Fig. 4. In
general, there is a high level of stationary or slowly fluctuating
ambient light generating shot noise inthephotodiode.In
addition, artificial-light sources also emit rapidly fluctuating
components associated with higher harmonics of the mains
frequency. These have to be eliminated by electrical filtering.
Part of the incident ambientlight may be blocked byan optical
filter (see Fig. 4). Two types of filters may be used: optical
interference filters with a passband corresponding to the LED
source bandwidth(withdue allowance fornonnormal inci-
dence) and absorption edge filters blocking only the visible
part of thespectrum. An effective and inexpensive filter is
obtained by developing an unexposed color film. Due to the
different source spectra, the filter is able to block ambient
light to a varying degree of effectiveness [ 101. The transrnis-
sion is approximately 1 percent at 720 nm and 85 percent at
the LED wavelength of 950 nm. Even for the proposed indoor
use of the transmission system, the ambient-light level varies
over several orders of magnitude.A series of measurements
has indicated some typical ranges (see Fig. 5). The intensities
 1478 PROCEEDINGS OF THEIEEE, VOL. 67, NO. 11, NOVEMBER 1979

I R FILTER
1 BIAS
R‘ P

‘
I,L
t
AMPLI-
EOUALIZER FILTER
FIER

(4
W IDEAL
n
cn
0.4 0.6 0.8 1 1.2 1.4
ese .
AMPLIFIER

j R;1
< VISIBLE > X (pml

E
SPECTRAL
’ / \ TRANSMITTANCE
V
COLOR FILM (b)
0.5
Fig. 6. (a)
Block diagram of PCM baseband receivers. (b) Noise
I equivalent circuit diagram.
FERENCE
/ FILTER
0
0.4 0.6 0.8 1 1.2 1.4 may be used. Intheliterature, extensiveworkhasbeen
WAVELENGTH h publishedbyGoell [ 11 ] and Personick [ 121 on baseband
Fig. 4. Spectral power density of three common ambient-light sources.
PCM fiberdptics receivers, and we shall follow their approach.
The bottom figure shows the effect of two different opticalfdteI-5. The noise equivalent receiver circuit diagram is shown in Fig.
6 . The minimum average optical power Pmh is given by

-
1 1 1 1 I [ 1 1 1 1 I

BACKGROUND ILLUMINATION

0 INTERFERENCE
FILTER A i = 5 0 n m
where Q is a function of the error probability pe
@ AGFA FILM

and the effective noise current is

A OFFICE WITH FLUORESCENTLAMPS

B MIXED DAYLlGHTlARTlFlClAL
ILLUMINATION
C SKY BACKGROUND
F?m
D SUN
PHOTO
CURRENT DENSITY A/
I I I I I I 1 1 1 1 I

lo7 rc? 1$ 1g4 f63

Here, h is Planck’s constant, 71 thephotodiodequantum
16’

Fig. 5. Photocurrentdensity of t pica1 ambient-lightenvironments. efficiency, u the optical frequency,e the charge of an electron,
2,
Measurements taken with a 1-cm pin Si-photodiode and 50‘ FOV. T the temperature; Zo = 0.3 and Zz = 0.13 are shape factors for
50-percent dutycycle signal pulses. and f is the bit rate. For
are given in terms of photocurrent density iB rather than radio- the Si-FET input stage thetransconductance is ,g = 0.005
metric units to avoid referring to .the different spectra. The mho and the noise multiplication factor is = 0.66 [ 131. CD r
approximateblockingeffectsofthefiltersare also shown. and R B arethephotodiodecapacitanceand biasresistance,
The ambient-light level determines, t o a significant degree, the respectively. They both dominate the other input capacitances
optical signal powerrequiredfor reliable transmission. For and resistances for the application considered. The first term
reference purposes, we have considered two particular ambient- of (12) describes the shot noise generated by the stationary
light environments. photocurrent 18 due t o ambient light. The remaining terms
In a predominantly daylight environment we chose iB = 1.5 representthermalandamplifiernoise.Inprinciple,thebias
X A/cm2, usingcolor
a film as filter. This
valuecor- resistor R B should be made large to reduce thermal noise, but
responds to approximately 380 Lux and was measurednear this requires an equalizer to compensate for the signal distor-
windows but with the photodiode not exposed t o direct sun- tion introduced by the input circuit with time constant TD =
light. The same photocurrent density is also obtained with the RBCD. Proper equalization with TD >> l/f, where f is the bit
photodiode placed under a 75-W tungsten desk lamp at 65-cm rate, requires careful adjustment. However, the diode capaci-
distance.Alternatively,ina well-lit windowless roomwith tance does not remain constant because of fluctuations of the
fluorescent lighting, the resulting photocurrent density is ig = biasvoltagecausedby the ambient-light photocurrent. We
3 X lo6 A/cm2corresponding to 200 Luxinphotometric thereforeconsidertwo cases, onewithmodestequalization
units. TD = l/f, andwith R B = 50 where thelatterdoesnot a,
requireequalizationforthebitratesconsidered.Thesignal
B. Required Optical Receiver Power photocurrent ZR increases proportionally with the photodiode
We shall now calculate the input optical power required by area A R , whereas the shot-noise current ZB increaseswith
the receiver in the presence of stationaryambient light. In 6.
An overall gain in signal-to-noise ratio is therefore ex-
principle,baseband PCM orcarrier-modulatedtransmission pected for large-area photodiodes. Combining (3) with (IO)
GFELLER AND BAPST: DATA COMMUNICATION VIA DIFFUSE IR RADIATION 1479

CLOCK
E(TRAcliDI
AMBIENT-LIGHT -
PHOTOCURRENT

RECEIVERClRCUlT
BASEBANDPC#

Fig. 8. PCM baseband receiver circuit.

-
I t I I I
0.4 I cm2 40
PHOTODIODE AREA AR

Fig. 7. Requiredradiant emittance of the diffusely reflecting environ-

ment.

we obtain an expression forthe required radiantemittance

w of the reflecting surface

W =
VeAR sn
hv
i' (FOV) Q [&I ". (13)
5 05 25 45 65 86 06 2 48 46 44 - d B m
The receiver photodiode thus collects sufficient optical power 30-8 WATT
if it sees a surface reflecting the infrared signal with a power AVERAGEOPTICALSIGNALPOWER
density given by w. With aknowledge of theradiant emit-
Fig. 9. Theoretical and experimental transmissionerrorperformance.
tance w , the optical transmitter power can be estimated by Baseband PCM,f= 125 kbit/s.
multiplying w with the total surface A of the cavity, bearing
in mind the reflectioncoefficient p. Note that this method
only applies for the case where the transmitter LED array is allows the introduction of a high-pass filter to attenuate low-
able t o produce an approximately uniform illumination of frequency components fromambient-light fluctuations.
the surrounding walls and ceiling. For numericalresults, we For the measurements, the receiver is equipped with a large-
assume the following parameters: the photodiode capacitance area photodiode A R = 1 cm' (type UDT PIN-IOD). The mea-
is ,C = (YAR, where (Y = 200 pF/cmZ, and the ambient-light sured diode capacitance is CD = 270 pF, with R g = 3 l .6 k!il to
photocurrent is ZB = iBAR , where ig = 1.5 x A/cm'. form the time constant TD = l/f, and the quantum efficiency
is 7) = 0.57 at A = 900 nm. The stationary ambient-light photo-
The projected error probability is lo-' (Q= 61, and FOV =
90'. In Fig. 7, therequiredradiantemittance is calculated current is ZB = 1.5 X Acorresponding to a photometric
as a function of thediode area A R . The total photodiode illumination of approximately 380 Lux.
areamay be obtained bysmaller photodiodesconnected in The high-pass filter is formed by a cascade of 5-RC coupling
parallel. circuits between amplifier stages and provides an asymptotic
The equalized receiver circuit with TD = RBCD is dominated lOO-dB/decade rolloff. The 3-dB point is chosen at 1.25 kHz
by ambient-light shot noise, whereas the low-impedance circuit which does not affect thequality of the eye pattern.The
is dominated by thermal noise from the bias resistor R g . Note resulting error performance is shown in good agreement with
that the numerical value of w for the diffusely reflecting sur- the theory (see Fig. 9). However, in the vicinity of fluorescent
face also corresponds to the power density of direct radiation lamps, additional transmission errors occur due to thepresence
withnormal incidence. Incomparison tothefiberaptics of high-frequency components. Baseband PCM transmission is
receiver a m e r optical inputpower is required to compensate therefore vulnerable to rapidambient-light fluctuations. A
for the ambient-light shot noise. much higher rejection of ambient-light fluctuations can, how-
ever, be achieved with modulated carrier transmission.
C. Experimental Baseband PCM Receiver
IV. DIFFUSEOPTICALCHANNEL
We have measured the transmission error probability with an
experimental baseband PCM receiver at f = 125 kbit/s. The A . SimulationModel
diagram of the receiver circuit is shown in Fig. 8. The circuit A real office room or shop-floor environment represents a
is provided with automatic gain control AGC, internal clock complex cavity with a variety of features such as geometry,
extraction, and code conversion from RZ to NRZ. The trans- size, and reflective properties. In order to establish guide lines
mission is coded diphase which facilitates clock extration and for the design of hardware components we describe acomputer
allows fixedthreshold detection.Furthermore,the signal model simulating the basic features. Of particular interest will
power spectrum [ 141 of coded diphase is zero at f = 0 which be the optimization of the LED arrays to produce an economic
1480 PROCEEDINGS
VOL. IEEE,
OF THE 67, NO. 1 1 , NOVEMBER 1979

0
' 10' 20' 30'

AR
Fig. 10. Diffuse reflection polardiagram of aplaster wal. The letters
indicate the angle of the incident-plane wave radiation. A = 950 nm. Fig. 1 1 . Simulation model of the diffuse optical channel.
The dotted circle represents an ideal Lambertian diffuser.
illustrates the gist of the diffuseoptical-channelmodel. The
and relatively homogeneous diffuse radiation field over the power dP, radiated by the LED into the solid angle dS1 is
whole floor area with respect to a given environment. incident on the surface element dA from where a part dPA is
The transmission relies on efficient scattering of the incident diffusely reflected with a Lambert distribution. The resulting
LED radiationfromthesurrounding surfaces formingthe radiant emittance w of the surface element is then given by
structuralenvironment of theroom. We have made rough
measurements of the reflection coefficient p at x = 950 nm for
a variety of commonly used materials: painted surfaces, wood
panes, textiles, and plaster walls. Here, the reflection coefficient where
p is defined as the ratio of total power reflected into thehemi-
sphere to plane wave incident radiation. We found that most R 2 >> dA.
of these materials reflect infrared radiation to 40-90 percent.
The reflection characteristics are generally composed of a dif- Hence, the power dPA received from dA by the photosensitive
fuse and a specular component, the latterbecoming significant areaAR is
for very shallow angles of the incident radiation. Fig. 10 shows
the polar scattering diagrams of a white plaster wall for dif-
ferent angles of incidence. Typical values of p for plaster walls
have been found to vary between 0.7-0.85, depending on sur-
face texture and angle of incidence, and the radiation charac- where
teristics are a close fit to a Lambertian distribution. Conclu-
sively, most environments exhibitsufficiently good to excellent
reflective properties for ourpurpose.
rect ( S i ) =
I1, if lSil< FOV
0, if IS,( > FOV.
We now formulate the simulation model. The basic source The function rect ( S i ) describes the FOV of the photosensitive
element fortheopticalradiation consists of an LED. The
area which may be limited by a lens or an aperture to shield
angular distribution of radiant output power (radiation char- excessive ambient light. The total power received bydiffuse
acteristics) is modeled with a generalized Lambert law reflection is then obtained by integrating over the entirereflect-
n+l ing area A and by summing over all LED sources. In symbolic
dP, = -cos" aPs dS1 notation
2n
where
P, =IHemisphere
dP,.

Here, P, represents the total power radiated by the LED, dPn

is the power radiated into the solid angle dSZ, n is a number
As a first approximation, we have only considered single reflec-
describing the shape of the radiation characteristics, and a is tions. Similarly, the radiation incident onAR via direct h e of
the polar angle. With an epoxy lens attached to the LED hous- sight is
ing,various radiation characteristics may be realized. The
angular spread of the beam is indicated by the half-power angle
a,, where the intensity dropsby a factor of 2
(0.5)".
= C0s-l (15)
Numericalresultsare obtainedwith a computer program
In practice, half-power angles down to aH 4', with n = 300, calculating the optical signal power received as a function of
may be achieved with simple molded epoxy lenses. Fig. 11 position and angular orientation of the photodiode. The pro-
GFELLER AND BAPST: DATA COMMUNICATION
DIFFUSE
VIA
Fig. 12. Distribution of diffuse optical radiation in an office room en-
vironment. The plot in the upper-leftcorner shows the radiant
emittance of the ceiling as produced by the satellite Sat. with a source
array of nine LED’s. The upperright plot shows the radiant emittance
produced by aterminal T with aLarnbertianLED.The lowerplot
indicates the resultingreceived optical power (downlink) as a func-
tion of the terminal position. S = 10 m , H = 3 m, FOV = 90°, non-
directional case.
gram operateswiththe followingparameters: the reflecting
enclosure is defined as a rectangular box where each face may
have a different reflection coefficient. The optical source pa-
rameters are represented by the positional coordinates(xs, y s ,
zs), the orientation of the radiation axis, the emitted power
P,, and the shape factor n of the radiation lobe. Any number
of different LED’s may be simulated to form complex arrays.
The receiver parameters are the positional coordinates ( x r , y r ,
z r ) , theorientation of thephotodiode axis (normal tothe
photosensitive area A R ) , and the FOV. Fig. 12 shows example
plots of the radiant emittances generated by the satellite and
terminal LED arrays, and of the resulting optical powerreceived
as a function of theterminal position. Fig. 13 shows the
photograph of a scaled-down version of a spotlight pattern at
the ceiling as produced by an array of LED’s. Measurements
of the diffuse radiation distribution produced with LED arrays
ina scaled modelenvironment have shown good agreement
with the theoreticalmodel.
B. Case Study: Three Types of Environment
We now discuss three examples of the downlinkradiation
distribution. The first two are simulated office room environ-
ments with dimension S X S X H , where S is the room size
(length) ranging from 5 to 30 m, and H = 3m is the height.
The reflectioncoefficient of one side wall is set to zero to
study the effect of a gross inhomogeneity of the environmental
reflective propertiessuch as awindowfacade. Glass panes
reflect only specular images of the source where the reflection
coefficient is determined by the Fresnel equations. With
normal incidence the reflection is only about 5 percent and
represents a direct line-of-sight link which may easily be
blocked. The reflection coefficient of all other surfaces is as-
sumed p = 0.7 (plaster walls and ceiling). All terminal photo-
diodes are located on desk level with zs = 1 m. The coordinates
xs and ys are variables, and FOV= 90’. We consider two angular
orientations of thephotodiodes,Thefirstone is withthe
photosensitive surfaces parallel to the ceiling (nondirectional
case). Inthis case the hemispherical FOV receives scattered
radiationfrom all walls and the ceiling. In the second case,
thephotodiodes are tiltedtowardsthecenter of the room
(semidirectional case). Here, thenormal axes to the photo-
IR RADIATION 1481
Fig. 13. Infrared spotlight pattern on diffusely reflectingsurface. Image
convertor photograph.The sourceconsists of nine LEDarrays.The
irregularities of the individualLEDradiation lobes are causedby
manufacturing tolerances.
sensitive surface meet at a common f i x point with coordinates
(S/2,S / 2 , 6 m) and most radiationis received from the ceiling.
The position of the satellite LED arrays is assumed to be in the
center with coordinates(S/2, S / 2 , 1.5 m).
We firstconsiderasatellite with a simple Lambert source.
The axis of the radiationcharacteristics is normal to theceiling.
Fig. 14 shows the received signal power in function of the
positional coordinates of the photodiodes for room sizes of
S = 5 m to S = 15 m. The calculationswere performed for
boththenondirectionaland semidirectional cases. The dB
values indicate the variation encountered between maximum
(center) and minimum (corner) signal power received.
The plots show that for rooms with an aspect ratio S/H close
to unity, walls and ceiling are irradiated uniformly resulting in
a smooth distribution of the signal power received. Position
and orientation of the source have little influence. With in-
creasing room size S most of the radiant emittance is concen-
trated at the centerof the ceiling. Also, the effective projected
area A R of the receiving photodiodes becomes smaller with
increased distance from the center. Considering a Lambertian
source and the given room geometry with a low room height
(S= 3 m), the maximum room size is limited to S = 10 m to
15 m. With larger sizes the variation of the signal power
received with position becomes excessive. Our criteria for the
variation allowed is the economic use of optical-source power
and simple AGC in the receiver circuit.
For large office rooms an LED array with nine sources has
been investigated. Here, the ceiling represents the main reflect-
ing surface. Narrow beams from the LED array spotlight the
ceiling in a regular pattern, thus transporting the signal power
with little divergence. The spotlights act as secondary sources
with Lambertian radiation characteristics, which in turn scatter
the optical radiation at the predetermined location at some
distance from the main source. Low partition walls or shelves
separating individual work stations do not interfere with the
 1482 PROCEEDINGS O F THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

Fig. 14. Smallofficeroom.Receivedopticalpower as a function of

terminalposition.ThedBfiguresindicate theencountered signal
power variations forthe semi-andnondirectionalcases,respec-
tively. ThesymbolindicatestheLambertiausource.Plotsrefer to
photodiodw in semidirectional orientation. FOV = 90' ; room
geometry S X S X 3 m.

Fig. 15. Large officeroom.Receivedopticalpower as a function of

terminalposition.ThedBfiguresindicate theencountered signal
power variations for the semi- and nondirectional casea, respectively.
The satellite source consists of eight LED3 with n = 100, and one
LED with n = 1. Thelocation of the window front (S = 1) cor-
responds to the sidemarkedwith the room size S. Plotsrefer to
photodiodes in nondirectional orientation. FOV = 90'; room
geometry S X S X 3 m.

transmission as long as the FOV is sufficiently large. In Fig.
15, the plots of the signal power received as a function of the
room size S show a more uniform distribution. With the array
described a limit is reached at approximately S = 30 m, where
the signal-power variation of 12-1 7 dB is exaggerated due to
the central spotlight which is emphasized too strongly in com-
parison with the intensity of the sidelobes.
So far the angular orientations of the photodiodes have been
assumed fixed. We are now interested in knowing the angular
range within whichsufficientopticalpower is received.
Especially forhand-heldterminals,it is importantthatthe
operation is not subject to severe angular and positional con-
straints. Fig. 16 shows calculated polar diagrams of the optical
power received for three extreme locationsA , B, and C within
a room with S = 20 m. The polar diagrams indicate a rotation
of the photodiode axis (normal t o the photosensitive surface)
by 180' aboutthecoordinate axes x and y . Thereference
value 0 dB representstheminimumopticalpowerrequired
for a given transmission error rate. The total satellite power
is adjusted so that at locationC the power received in the non-
directional case (photodiode axis 90' elevation) reaches 0 dB.
Atlocation B theopticalpower received is amaximum,
whereas at location C it is a minimum. The polar diagrams
show that, apart from location along the window facade, the
photodiode orientation may cover thefullhemisphere.This
feature has been confinned by experimental tests.
GFELLER AND BAPST: DATA COMMUNICATION VIA DIFFUSE IR RADlATION 1483

-sGGLLITE OPTICAL POWER 4Hxx)

4 I I I I I I1
0 IO 20 30 40 50 60
ROOM SIZE S(m)
b
0 Fig. 18. Satellite optical power as a function ofroom size and geometry.
Transmission speed f = 6 4 kbit/s. Pso represents the theoretical
Fig. 16.Receivedopticalpower versusangular orientationforthree minimum power for plane wave radiation.
specificterminallocations. Thepolardiagrams indicate PR/P-
(dB). S = 20 m; FOV = 80’.

MB’T’S 1 .
MAXIMUMBIT RATE ACHIEVABLE WITH
1OPTICAL
W SATELLITE POWER
1

1 SAFETY
MBITIS FACTOR 1 5 X

I00 k B I T I S

SL RANGE 6 dB
-\..--, y/
IO k BITIS
Fig. 17. Hall. Receivedopticalpower as afunction of terminalposi- 0 10 20 30 40 50 60
tion. There is tittledifferencebetweensemi- and nondirectional ROOM SIZE Slm)
photodiode orientation. Satellite source consists of four LED’s with
n = 3. FOV = 90’; room geometry S X S X S/4. Fig. 19. Transmission speed versus room size.

Finally, we considerahall with dimensions S X S X S/4. satellite LED arrays. Theupper-bound applies tothe non-
Here, the aspect ratio S/H is fixed. Due to the favorable geom- directional case with all photosensitive surfaces parallel to the
etry a simple LED array with four lobes provides a relatively ceiling, whereas the lower boundapplies for thesemidirectional
uniform power distribution (see Fig. 17). Because of scaling, case. Thesmooth curve indicates thetheoretical minimum
the powervariationremains approximatelyconstantfor all power P,, if the optical radiation were available in the form
sizes S. of a plane wave with normal incidence and an aperture cor-
responding to the floorarea S X S, equation (5). The increased
C. Optical Satellite Power power demand P, with respect to P,, is explainedbya
We now estimate the total optical satellite power required reflecting surface (walls and ceiling) larger than the floor area,
forthe previous example environments. The transmission nonuniform radiant emittance, and absorption. However, the
speed assumed is 64 kbit/s and the receiver area A R = 1 cm2. satellite powerapproximately follows the theoretical square
From (10) the receiver power requiredfor atransmission law if the LED arraysare adequatelyadapted to theroom
error rate p e = lo-’ is then P m h = 9.6 nW for daylight, and geometry.In practice, the satellite power calculatedshould
Pmh = 1.5 nW for fluorescent ambient light, respectively, and be multiplied with a safety factor allowing ‘for temporary or
R s = TIC,. Assuming constant ambient light, we obtain the permanent partial obstruction of the individual FOV’s, aging
total satellite power by equating thereceived power PR, equa- of the LED’s, optical-filter losses, and receiver imperfections.
tion ( 1 8), at the position where the minimum occurs with the Alternatively, Fig. 19 shows thebitrates achievable as a
required power P m h . Hence, atthisposition,thepower function of the room size, given a total satellite power of 1 W
received from the satellite via diffuse reflections is just suf- includinga safety factor of 5. It is apparent that for large
ficient to establish the signal-to-noise ratio Q required for the rooms, an excessive power demand is more likely to be the
transmission errorrate prescribed. For all otherterminal limiting factorthan signal dispersion due to multiplepath
positions, the signal-to-noise ratio is greaterthan required. propagation.
This is illustrated in Fig. 16 where location C represents the
position where leastoptical poweris received with the particular D. Optical Terminal Power
satellite LED array. In Fig. 18, the total satellite power P, is We shall now discuss the terminal-powerrequirements. In
shown as a function of the room size S for the three different principle, to form a totally connected network, thesame power
1484 PROCEEDINGS OF THE IEEE, VOL. 67, NO. 1 1 , NOVEMBER 1979

I TERMINALOPTICALPOWER

4000
- CLOCK
-
u)
PSK IRUSYITIERIRECIIYER
400 % IWTfRHff
J

N
a!
40

I I I I I
0 40 20 30 40 50 60
ROOM SIZE S(m) RECEIVER PULSE REGEWERATIOI,DECODl~G

be keptconstant. This calls for a modularconstruction of

forthe satellite is needed. However, theterminals are theoptical satellite andterminal interfaces. The basic trans-
randomly located W i t h i n the room and each terminal would mitter and receiver circuits of both types of interfaces remain
therefore require an individual optical-source array to evenly the same, but individual modules consisting of an LED driver
disperse theradiantpower to the environment. Intheinterestand flexible LED arraymayadded to adapttheoptical
of a simple LED array at the terminal interface, it is favorablesystem to the various range, geometry,and ambient-light
to establishadedicatedchannel from each terminal to theconditions.Inmodem buddingsinfrared-reflectingwindow
satelliteonly.This approach requires different carrier fre- panes are often installed to save energy for air conditioning
quencies for the up- and downlinks if full duplex operation is and to provide a pleasant working climate. Transmission of
desired. Fig. 20 showscalculatedresults of the maximum incident daylight in the near-infrared region is thusreduced
terminal powerrequiredasa function of the room size. The by up to 75 percent which represents a further gain in the
transmission speed is again 64 kbit/s,andcommunication is signal-tenoiseratio.
assumed between a terminal located in a comer and thesatellite The simulations were performed for a relatively well-defined
located centrally. The satellite photodetector hasahemi- reflecting environment.Productionplants are often housed in
spherical receiver characteristicwith an effective projected acomplex structuralenvironmentwherethemodel described
area AR = 1 cm'. With a cosine law characteristic (plane can only give the coarse features.Finer details of the signal-
photosensitive area) the projected area for radiation incident power distribution should be obtained from field tests.
fromdistant terminals is smaller,resulting in alower signal- A somewhat speculative thought is the use of laser diodes
to-noise ratio. Theisotropic receiver characteristics can beinstead of LED's. The narrow Source linewidthwould allow
approximated by photodiodes resting onthe faces of a the use of narrow-band opticalinterference filters offering a
tetrahedron. Fig. 20 shows both the nondirectional case with greatly reduced shot-noise level, and hence less signal power,
the LED axis normal to the ceiling and a semidirectional case. depending on the spectrum of the ambient light. Furthermore,
For larger values of n (narrower radiation lobe) and with the additional separate channels may be gained by using different
LED tilted tOwards the center of the ceiling, a further reduc- optical wavelengths. An optical system to disperse the laser
tion of the terminal power can be achieved at the expense radiation to a safe level would be needed.
of a higher degree of directionality.
V. EXPERIMENTAL
TRANSMITTER-RECEIVER INTERFACE
E. Discussion In this section,we describe an experimental PSK transmitter-
The previous examples have shown that efficientillumina- receiver interface operating at a channel transmission speed of
tion may be achieved and that the optical power received is 64 kbitls. The use of a modulated carrier allows the signal
quite insensitive to the angular orientation of the photodiodes. spectrum to be shifted away from the spectrum of ambient-
The dynamic range of the receiver circuitrequired to cope light fluctuations. Furthermore, separate channels for the u p
with the signal-power variation over the floor area is less than and downlinks,maybeprovided. Fluorescent lamps, for in-
that for fiber-optic receivers having to adapt to different fiber stance,exhibit a spectrum reaching as far as 50 kHz. The
lengths. So far, we have assumed that the entire floor area is carrier frequencyshould,therefore, be higher than a few
to beprovided with the signal carrier, but diffuse spotlights hundred kHz but less than afew MHz, depending on room
serving a local cluster in an otherwise large hall may also be size, to avoid phase distortions due to multipath distortion.
considered. To extend the range, for instance in a large in- The block diagram of the transmitter-receiver interface is
house plant, several distributed satellites operating at different shown in Fig. 21. At bothends of theinterface,dataare
carrier frequencies to avoid mutual interference, may be em- available in NRZ form.Atthetransmitter side, datafrom
ployed. In practice, a high level of ambient light is found near theterminal in NRZ form arefirstconverted to RZ form.
windows, whereas the interior of large rooms is comparatively We have chosendiphase coding whichprovides easy clock
darkor fluorescent lighting with alowerinfrared portion extraction at the receiver end. As an alternative,scrambling
prevails. can be used.
In the interest of an economical use of optical signal power A PLL circuit synchronizes the carrier frequency with the
in large rooms, the signal-to-ambient-light-noise ratio should clock frequency where the carrier is a multiple N of the clock
Fig. 22. Packaged 64-kbitls PSK terminal transmitter-receiver interface.

frequency. The carrier is then keyed with the RZ data, and

the resulting PSK signal is amplified to drive the LED array.
On the receiver side the incoming signal is equalized after
distortion by the photodiode capacitance and bias resistance.
The channelfilterselects theappropriate carrier-frequency
band,andanautomatic gain controlkeepsthe signal at a
constant level. The carrier is recovered with aPLL circuit,
and after subsequent multiplication with the PSK signal, the
baseband signal is obtained. Clock extraction is achieved with
a second PLL circuit, and finally, pulse regeneration and con-
version to NRZ form follow. Fig. 22 shows the packaged
transmitter-receiver interface. The LED array consists of ten
LD 27 1 diodes in series with an average optical radiant output
power of 100 mW. An investigationhasshown that emitted
radiation from this diode type poses n o safety hazard for the Fig. 23. (a) Packaged satelliteinterface. The plexiglasscover is trans-
parent fop infrared radiation. (b) Packaged satelliteinterface. Image
eyes [ I S ] . convertor
photograph. EachLED-array spotlightemits
100-mW
The carrier frequency is 256 kHz, and with N = 4, the cor- average optical power.
responding channel
transmissionspeed is 64kbit/s.The
photodiode array consists of 9 BPW 34 diodes connected in
medium speed transmission at low error rates andfree from
parallel forming a total area of 0.67 cm2. The transmission
range between these terminal interfaces is 10-20 m depending EMI.
on ambient-light conditions. With present low-cost state-of-the-art LED’s, the maximum
We have measured the stationary transmission error probabil- transmission speed is around100kbit/sdue to the limited
ity under an ambient-light environment of 380 Lux to values up modulation capability of the LED. Higher transmission speeds
to lo-’. No interference arising from fluorescent lamps was up to 1 Mbit/s appear feasible if increased optical power raises
detected. However, during switching-on of fluorescent lamps, no objection.
theerrorprobability may rise to approximately IO-’. This There is scope for improving the performance and reducing
period of increased error probability occurs only during the the cost of the satellite and terminal interfaces. Current trends
indicate a decrease of hardware costs for optical components
ignition phase of the fluorescent lamp and is typically shorter
and circuitry, whereas costs of copper cables and labor for
than 100 ms. The effect depends on intensity, proximity, and
pulling cables are not likely to be reduced. Expenditure for
type of fluorescent lamp. With higher carrier frequencies, this
fiber optics is still relatively high and the present state of
effect is expected to be reduced or even eliminated. But con-
development favors individual point-to-point links rather than
sidering the relative rarity of switching-on events per day, these
distributed systems with a large population of users. There is,
short periods of increased error probability may not pose a
therefore, a challenge in identifying new applications of local-
problem.
area networks where the wireless optical link can offer distinct
Fig. 23 shows a prototype of the satellite interface. Up to
advantages.
eight flexible spotlights of 100-mW optical power each may be
added to adapt the radiation pattern to the environment. ACKNOWLEDGMENT
VI. CONCLUSIONS We should like to express our gratitude to H. R. Miller, P.
Vettiger, and A. H. Frei for the help received and continuous
We have analyzed the physical aspects of the diffuse optical
support and encouragement during the course of the project.
channel and demonstrated technical feasibility. With the wire-
less transmission scheme great flexibility in terminal arrange- REFERENCES
ment may be achieved. An experimental PCM baseband link
[ 11 D. J. Farber and K. C. Larson, “Thesystemarchitecture of the
and a PSK link, operating at 125 kbit/s and 64 kbit/s, respec- distributed
computer
system-The
communicationsystem,”
tively,havebeenbuilt. Thechannelis suitable forlow to presentedat The PolytechnicInst.BrooklynSymp.Computer
 1486 PROCEEDINGS OF THE IEEE, VOL. 67, NO. 11, NOVEMBER 1979

N e w o r b , Brooklyn, NY, Apr. 1972.
R Metcalfe and D. R Bo-, “Ethernet:Distributedpacket
switching for local computer networks,” presented at Xerox Palo
Alto Research Center, Palo Alto, CA, Nov. 1975.
k G. Fraser, “Spider-An experimental data communications
system,”in 1974 Znt Commun Conf., pp. 2lF-1, 10.
F. R. Gfeller, H. R Miiller, and P. Vettiger, “Infrared communi-
cation for in-house applications,” presented at IEEE COMPCON
‘78 (Washington DC), Sept. 5 4 , 1 9 7 9 .
L.G. Roberts, “Extension of packet communication technology
t o a hand-held personal terminal,” in Proc. AFlPS Spring Join2
Comp. Conf., VOL 40, pp. 295-298, 1972.
R. Bryan, “Research in micro-terminal development andnet-
work end-to-end error recovery,” California University Advanced
Research Projects Agency Defense Supply Service, AD/A-006589,
Feb. 1975.
L. Kleinrock and S S Lam, “Packet switching in a multi-access
broadcast channel: Performance evaluation,” ZEEE Trans.Com-
mun., VOL COM-23, Apr. 1975.
[ 81 R Winston, “Light collection within the framework of geometrical
optics,”J. Opt. SOC.Amer., vol. 60, pp. 245-247, Feb. 1970.
[ 91 A. Rabl, “Comparison of solar concentrators,” Solar Energy, VOL
18, pp. 93-111,1976.
[ l o ] R Knauer, “Toniibertragung und Fernsteuerverfahrenmit In-
frarot,” Siemens A. G., Miinchen, Germany, Bereich Bauelemente
Vertrieb, Tech. Schrift.
[ 111 J. E. Goell, “Input amplifers ofoptical PCM receivers,” Bell
Sysr. Tech. J.,vol. 53, pp. 1771-1793, Nov. 1974.
[12] S D. Personick, “Receive design for digital fiber optic com-
munication systems,” Bell Syst. Tech. J., VOL 52, pp. 843-874,
1973.
[ 1 3J D. B. MuKunda, “FET noise sources and their effects on amplifier
performance at low frequencies,” IEEE Trans Electron Devices,
VOL ED-19, pp. 338-340, ~ a r 1972.
.
[ 141 P. Bocker,Dateniibemagung Band Z. New York: Springer, 1976,
PP. 129-133.
[ 151 Kuny, “Augenschadigung durch Infrarot-Strahler,” Siemens A. G.,
Bericht. Aktenvermerk 150 000 8.7 1 2181.

Transform Approach to Electromagnetic

Scattering

Abstmct-In this paper, we present a comprehensive review of the Ut Total field.

Fourier W o r m technique as applied to the problem ofhigh-fre- E
quency scattering and introduce the concepts of the spectral theory of Electric field.
diffraction (STD). In contrast to the more commonly employed ray- H Magnetic field.
opticalmethod for high-frequency scattering, viz. the geometrid J Induced electric current.
theory of diffraction (GTD), the SII)approach interprets the scattered F Fourier transform operator.
&Id as the s pecrtum, or the Fourier transform of the induced current F-’
on the scatterer. Such an interpretation offers several important Inverse Fourier transform operator.
Pdvantager uniform nature of representation, capacity to improve aud U Transformed function.
extend the ray+ptical formulas in a systematic manner, and convenient ? Transformed function.
accuracy tests for the results Methods for combining integral equation X Spectral diffraction coefficient.
methods with the Galerkin procedure and asymptotic techniques in the Hd Edge diffracted field.
transform domain are descnied, and representative examples illustrat-
ing the application of the spectral approach are included. HP Physical optics field.
Hg Geometrical optics field.
NOMENCLATURE HI Field of edge 1.
STD Spectral theory of diffraction. H2 Field of edge 2.
UTD Uniform theory of diffraction. :
F
Fresnel function.
Asymptotic term of Fresnel function.
UAT Uniform asymptotic theory.
ui Incident
field.
Jb Physical optics current.
J*‘ Truncated physical optics current.
us Scattered
field.
J’ Induced current on a semi-infinite half-plane.
HD Diffracted field in UTD formulation.
Manuscript received December 4, 1978;revised July 5 , 1979.
This work was supported in part by Joint Services Electronics Pro- e Step function in Section 111.
jects under Grant N00014-79C-0424, and in part by Office of Naval e Truncation operator in Section IV.
Research under Grant N00014-75C-0293.
R. Mitra and W. L. KO are with the Electromagnetics Laboratory, el Complimentary operator.
Department of Electrical Engineering, University of Illinois, Urbana, G Dyadic Green’s function.
IL 61801.
Y. Rahmat-Samii is with the Jet Propulsion Laboratory, California 5 Basis function.
Institute of Technology, Pasadena, CA 91 103. % Testing function.

0018-9219/79/1100-1486$00.75 0 1979 IEEE