484 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. COM-31, NO. 4, APRIL 1983 A Self-organizing Random Array Communications Relay Abstract-A sparse random array with self-organizing propertiesis proposed as a communications relay. The relay consists of N randomly deployed, low-power transmitlreceive elements with capabilities for phasingtheirretransmissionstoarrivecoherentlyat the ultimate destination. Signal-to-noise ratio results for conventional modulation schemes corrupted with additive noise are given, together with possiblesystemblockdiagrams.Itisconcludedthat, in spite of the complexity, practical systems can be built with extremely low electromagnetic radiation densities with point-to-point characteristics at least as good as withconventionalsystemsemployinghighpower relays. The novel characteristics make the proposed system particularlysuitablefortacticalmilitarycommunication,andvariants thereof may find application in urban communication. \\ -.___.' a . I. INTRODUCTION I Fig. 1. Schematic model of the random array relay system. L ARGE, lowdensity, nonuniformlyspacedsensorarrays preciably. The scheme thus potentially offers space diversity, have long been applied to radio direction finding and, parreliability, and signal obscurity, the latter being a consequence ticularly in the field of astronomy, to high resolution imaging. of the low powers and the dispersion of the retransmitting reThe large size is associatedwith high resolutionornarrow lay elements. Potential applications may be found in military beamwidth, the limited number of elements with economy, communication,whereemittingsourcesmustbeconcealed, and the nonuniformity with ambiguity resolution in direction and the concept of a distributed emitter-receiver array may finders, or more generally, with the suppression of the lobing also find application in urban communication for its diversity phenomenon associated with lowdensity periodic arrays (low potential. densityimpliesmorethanhalf-wavelength spacing between Theschemedescribedhereinutilizesfrequencyinvariant elements) [ 11 . phase control only at the array elements. The constancy of the Herewedealwithanarrayof widely dispersed,sparse, phase shift with frequency imposes a bandwidth limitation on transmit-receive elements used as a two-way radio relay (see the signals to be handled by the array [8]. The use of phase Fig. 1). The elements may be viewed as randomly scattered or shift only also means that maximum signal-to-noise ratio will deliberatelydeployednonuniformly over somearea.Inuse, be obtained at the ultimate receiver only if the signal-to-noise thearrayelements will each receive a signal fromadistant ratios delivered by each element to the final receiver are equal transmitter and retransmit it so that the emissions from all ele- and if there is no incoming interference to the array. When ments converge coherently on a distant receiver. This is accom- these conditions are not satisfied, complex (phase and ampliplished by phase control of the signal transmitted by each of tude) element weighting, as opposed to simple phase shifting, the elements, the phase information being obtained with the is essential. Extensions to overcome these limitations are subcooperation,ofthedistantcommunicators.Thecommunijects of continuing study and are not pursued here. cators using the relay take advantage of the array gain to opIn the next section we present results on the system properate at low power. Furthermore, the relay acting as a distrib- erties of the array relay along with a numerical example using uted radiator will require low power from each of its elements.typical parameters to illustrate the advantages gained. This is If a small number of elements fail or are not mutually visible followed by details of a possible system implementation. The to the communicators, the channel quality will not suffer ap- essence of the method described here is also found in [2] and Paper approvedby,the Editor forCommunication Systems Disciplines of the IEEE Communications Society for publication after presentation at.the 22nd Midwest Symposium on Circuits and Systems, Philadelphia, PA, June1979. Manuscript received December 21,1981; revised August 3, 1982. J. B. Vespoli is with the Satellite Communications Division, Harris Corporation, Argentina. F. Haber and R. S. Berkowitz are with the Department of Systems Engineering,Moore School ofElectrical Engineering, University of Pennsylvania, Philadelphia, PA 19104. D. Yavuz is with the Communications Division, Shape Technical Centre, The Hague, The Netherlands. ingreaterdetail,including scheme, in [3]. an alternative implementation 11. SYSTEM ANALYSIS In this section, we will determine the signal-to-noise ratio propertiesofasystemforpoint-to-pointcommunication through a widely dispersed random relay. The scheme is depicted schematically in Fig. l . We will present results for the idealized, but reasonable, situation in which the noise sources are independentateachreceptorand wherepropagation 0090-6778/83/0400-0484$01.OO 0 1983 IEEE 485 VESPOLI er al.: SELF-ORGANIZING RANDOM ARRAY RELAY lossses are purely inverse square. We will also assume that the array, B in Fig. 1 , is somehow focused on the users. That is, each element of the arraybj contains a phase shifter whichean be set to make the signals emanating from the individual elements in phase on arrival at the destination receiver. We will elaborate on this importantfocusing problem in Section111. Referring to Fig. 1 , we denote by K b and K c the average free space propagation losses for transmission from communication center A to array B , and from array to mobile station C,respectively. For the purposes intended in this section we assume no significantdifferencesamong the individual distances between A and the array elements and between C and the array elements. Thus, d and D will be viewed as distances from A to each element of B and C to each element ofB, respectively. When station A is transmitting, denote by Pt, the equivalent isotropic power radiation from its antenna. Each element of the array bj isotropically receives powerPrbj which we will assume is equal for all elements. Hence, we have sumgd neghgible, the SNR at C would be equal to N times the SNR at an array element. Solving( 5 ) for Gb we obtain Since Gb must be positive and finite we have the following inequality : Using (l), (2), and ( 6 ) we may write an equality relationship between the SNR’s at B and C: The largest SNR improvement from B to C is N and occurs when G b m. At this point it is convenient to define an equivalent system gain G , as the ratio of the power that would be necessary to transmit from A to C withoutthe arrayfora where = number Of array and K b = x2/(4.rr4)2 ; given SNR at C to the power to attain the same SNR at C but similarly,K C = x2/(4nD)2 for the path between and Contransmission through the array B . The formeris given by sider first the case in whichA transmits through the array and C receives. Eachelement bj receives a signal from A with KaPm ‘ power Prb andretransmitsanoisyversion withtotal power (SNR),o = (9) Ptb and signal-to-noise ra,tio (SNR)b. is, That NC --f where Pra’ is the power radiated fromA , and K , = X’ /(4.rrDac)2, (2) D,, being the distance from A to C. Equating (9) to (5) and letting Pta’/Pta= G, we obtain where N b is the equivalent noise power at the input ofeach arrayelementand Gb is thetotal poweramplification of each array element (including antenna gain). Assuming that the N elements of the array are focused at C, the power received at C becomes Prc =N2K,Gd’rb -I- N K , G f l b . (3) G, = N2KbGbKc/Ka . NKcGb(NbINc) Our aim is to maximize G,. However,thisoccurswhen 00, as in Fig. 2 . Denoting by r,,0 < r < 1, the practical fraction of the maximum attainable value of G,, and calling this value G,, the acceptable gain, we have Gb -f Clearly, the first term in (3) is the power of the signal component at the input of C; it is proportional to fl as a consequence of coherent the addition of signals from the N eleG, =r N k ) (1 1 ) ments. The second term in (3) is the noise power transmitted . . from the array to C;it is proportional to N since the noise acG, max is given by (1 1) when I = 1 . Array element gain corcumulates noncoherently. The total noise at Cis that received responding to G,, may be expressed in terms of r as from B plus that generated in C and is given by (2). where N c is the equivalent noise power due to the receiver at C. The signal-to-noise ratio at C,denoted by (SNR),, may thus be written as (5) This result is obtained by setting (10) equal to (1 l), and solving for G b . In terms of the parameter r, we may rewrite the previous expressions relating SNR and transmitted powers. Putting(12) into (8) we get 1 486 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. COM-31, NO. 4, APRIL 1983 6% Knowing (SNR),,N,, and Kb, we obtain-thepower to be radiatedfrom A using (14) as Pt, = 6.30 mW. From (15), the r G, m o x power to be transmitted by each array element is P t b = 27.50 mW. From (1 1)) the system gain is G,, = 8000 (39 dB). Therefore, the necessary power transmitted by station A has been reduced by 39 dB. Furthermore, since broadcasting fromA is no longerneeded, wesee from Fig. 1 that adirectionalanGbo Gb tenna at A , focused toward the array area, could be used. An Fig. 2. Equivalentsystempower gainas a function of theamplification in relay elements. antenna gain of 15 dB is a reasonable value for this particular case. If this is done, the required transmitted power at A beUsing (1) and (2) with (1 3) we get comes P,, = 0.20 mW. To evaluate the power to be radiated by C, when C transmits through the array and A receives, we can make use ofthe sameequations by interchanging the values of K b and K c . We then obtainp,, = 0.63 w. Similarly, the power to be radiated from each array element to A becomes Ptb = 0.27 mw. Or, when using a 15 dB gain antenna at Using (l), (2), and (14) with Gb taken from (12)we get A , we have P t b = 0.01 mW. The necessary power transmitted by C has been reduced by a factor equal to rN = 80 (19 dB) and the power transmitted by A with the 15 dB gain antenna has been reduced by a factor equal to 250 000 (54 dB). A summary of the most relevant results of this example is shown in Table I. Expressions (14) and (15), together with the bound for N It is interesting to note thatwhen A transmits to C through given in (7), specify completely the array relay system we have the array, the totalpower of the main source of radiation (arhypothesized. ray) amounts to NPtb = 2.70 W, this power being distributed To this point we have assumed that A was transmitting to over a large area. C through the array B . Due to the obvious symmetry of the Originally, the system wasconceived to protect the comsystem, the results for the case when C transmits through the municationcenterormaininstallations ( A ) frombeingdearray to A couldbeobtained easily by interchanging sub- tected and located. As we saw in the example, the mobile stascripts A and C. tions (C) are also protected since their radiated power is decreased by a factor of 80 (19 dB). We therefore conclude that A Numerical Example by using the array relay system we obtain a number of significant advantages: a great reduction in the powers transmitted We will now give numerical results for a system with the by A and C, wide dispersion of the actual source of radiation, following parameters. anda conceivable improvementinthe transmissionquality 0 Number of active elements in the array: N = 100. under adverse propagation conditions since the radio link does 0 Signal-to-noise ratio required at A and C: 1000 (30 dB). not depend on a single path. 0 Distance from A to B : 5 km. Distance from A to C: 50 km. 111. SYSTEM IMPLEMENTATION AND OPERATION 0 Array area: I k m 2 . To this point the assumption was made that array elements Requiredpowertransmitteddirectlyfrom A to C to are, somehow, focused on the users. We will now elaborate on achieve the required SNR: 50 W. this matter. As a basis for discussion,ahalf-duplexrequireFrom the data we first estimate a value for the noise power ment will be assumed and a system for achieving focus to users that will be present at the front end of receivers A , B, and C at both ends will be proposed. In Fig. 1 we assume that the assuming that all three receivershave the samenoisefigure. mobile station at C will receive and transmit at frequency f Consistent with the signal power transmitted without the ar- and thatthecommunicationcenterat A will receive and ray, the distance. from A to C, and the SNR assumed at C, we transmit at frequency f' = f fro. The array elementswill upget' convert signals from C t o A by an increment in frequency fro G, m o x + and will downconvert signals from A to C by an equal increment.The arrayelements will be given the oscillatorreferencethroughaseparateorder-wirechannelfrom A operatFrom (13), taking r = 0.8 (note that this value of r gives a sys- ing at a frequency fo. That is, cos q O twill be generated at tem gain loss of only 1 dB with respect to maximum attainA , be broadcast to all elements at B , and hereceived at eacheleable G, and results in a reasonable GbO), we have ment bi as cos(wl0t ~ j ) j, = 1 , 2, .*.,N. The -yj depend on the distancesfrom A to eachelement(andonpropagation (SNR)I, = 12.50 (1 1 dB). conditions) and are not a priori known. To focus the transmissions from, say, C to A requires that 1 The noise level obtained is perhaps two or three orders of magni- the up-conversion at each element be accompanied by a phase tude higher thanone would ordinarily expect. The example is arbitrary. shiftwhich will maketheseparateretransmissionsfromthe + 487 VESPOLI et al.: SELF-ORGANIZING RANDOM ARRAY RELAY Converting as shown in Fig.4(b), the received signal will be TABLE I POWER REQUIREMENTS FOR EXAMPLE SYSTEM USING A N ARRAY RELAY COMPARED WITH A CONVENTIONAL SYSTEM. 30 dB SNR ASSUMED REQUIRED AT TERMINALSA AND C. Array Relay System 15 dB antenna at A Conventional 0 dB antenna at A 50 W 54 W 6.30 mW 0.63 W system Parameter Pta ptc 0.20 mW 0.63 W Phose added of bl B Frequency f Focal Region OD, + w'd, again independent of the index j . Equations (16a) and (16b) require knowledge of the phases y i . If these are available, one can shft the local oscillators so that they are all cophased to, say, y1 in which case (16a) and (16b) become identica1:'In this case up and down conversion at each element would be carried out as shown in Fig.5. Estimation of the phases yi may be carried out as now described. Recall that the local oscillator waveform cos wlot is generated at A and broadcast to all elements at B . For the purpose of this measurement an instruction is sent from A t o B (throughtheorder wirechannel)requestingonly b , to respond. It shifts the received waveform at frequency o to frequency w ' with the phase of its local oscillator. Thisis now received back at A , is mixed at A with the originating waveform, and forms the difference frequency term Fig. 3. Diagram showing paths for point-to-point communication through the array relay with relevant focusing parameters inqicated. yi = oscillator phase at elementbp @i= focusing phase shift injected at element bi. bi arrive in phase at A . Such a process is indicated in Fig. 3 , showing two branches with phase shifts and @i introduced at the two elements for focusing. Branch 1 will be treated as the referencebranchand will be setequal to zero. For branch j to be coherent with.branch 1 atA requires that' @, @, c The phase shfts involved are shown in Fig. 6 . Next, in a similar manner, A transmits to bl at frequency a'= w + wIo and receives the shifted waveform at frequency w . Again, the returned waveform is compared with the orginating one and a difference frequency term (0 + o')d, C which is independentoftheindex transmitting fromA toC requires j . Coherentcombining Then, (18) and (19) are applied through a mixer retaining the low frequencyterm, which is (Al,'Al,~lf=cos 2Yl. 2 Quantities such as wDi/f represent the phase delay associated with the transmission path. Equatlons (16a) and (16b) are written assuming a single ray path. In multipath the phase delay would not depend on the line-of-sight distances Di and di alone. The approach to be described applies in the more general case, provlded the phase delays are the same regardless of the direction of transmission, A to C or C to A. In Fig. 4 the phasedelays are denoted 9.i. and are thereequated to -wDi/c and -w'dj/c, respectively, for he-of-sight nonmultipath conditions. - Yl] (20) Equations (1 8) and (1 9) are received waveforms resulting from nonsimultaneous transmissions. To compare them will require stretching the first one at the receiver 96 that the second will overlap it. This is readily accomplished:in a circuit of sufficiently narrow bandwidth. Although y1 obtained with (20) is not unique, the ambiguity can be resolved by testing each of the possible solutions using (18) and (19). The correct y1 used to phase shift (18) downward and (19) upward will produce cophased sinusoids. In a step-by-step manner, each of the local oscillator phases is similarly determined. Feeding the measured y j , j = 1 , 2 , .-, 488 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. COM-31, NO. 4, APRIL 1983 CosIW~t.vC,.via.Yj .VI) IUSEI Local oscillator (b) Fig. 4. Array element frequency andphasetranslation.Under single path propagation @a.= @iu = - w ' d j / c and GCj = @jic = - w D . / c , where w' = w + wlo,. (a) Gp-coqverting for transmission from to A . @) Down-converting for transmission from A to C. CC FROM Clf) TO C l f l 7 7 T O A lf'l 7r Y Y TRANSMITTER CONJUGATOR PHASE SHIFTER vi TRANSMITTER !. . cos l wt ot RECEIVER If1 If') Downconvertermixer Fig. 5 . Basic operation of an array element. (f): signal is centered on frequencyf: (f'):signal is centered on frequency f' = f + .,-lf b y C through the array will arrive at A with less than maxi- mum amplitude, The scheme involves comparison of thephase of the total arriving signal at A with the phase of the signal from each element in turn, and shifting the latter to conform with that of the total signal. The scheme to be described deals with the separation of the two phase angles for measurement and the correction obtained with each step. l - h cd , . y , l Consider Fig. 7 where the conditions are shown relating to Fig. 6. One step in scheme for determining the common phase referthe measurement of the present value of &. A signal from C is ence for the array elements. assumed t o be transmitted through B t o A ; we identify it on N to the appropriate elements the local oscillator outputs in arrival at A by its carrier only, cos w't. A low frequency waveeach are shifted to put them all in phase. If the propagation form cos wmt is generated at A and transmitted on the orderpaths betweep A and the elements of B are fixed, one N-ele- wirechannel to all arrayelements. By instructionfrom A ment scan will suffice to keep the local oscillators cophased transmitted on the order-wirechannel,eachelement will, in for the system lifetime. turn, amplitude modulate its signal from C with the low freWith all the yi = yl, (16a) and (16b), the phases to be in- quency waveform cos wmt. Since information may be present jected at element bi for focusing in both directions of trans- on the signal coming from C, om is chosen outside the signal mission, are identical depending only on the distances. While band so thatthemodulated waveformcomingfrom the A and B are apt to be fixed in position, the communicators at element will have a spectrum as suggested by Fig. 7(a). Fig. C are mobile and the distances Dj are not fixed. A scheme will 7(b) indicates the total signal environment when element bj is be described for dynamically determining the required phases the one carrying the additional modulation. The totalreceived allowing for communication even while C is in motion. A limit signal at A is to the velocity, which depends on the size of the focal region N and the speed with which the phases can be tracked, is to exrA ( t ) = cos ('d't e i ) cos ( a mt - aj) pected; these matters aresubjects of other research. i= 1 Assume the array is initially defocused with phase angles @ arbitrary (they may all be set initially to zero). A signal sent COS (oft+ ei) C O S I W( 0 + + - 489 JAL V-fm CHANNEL 1’ f’. 1, f (b) (a) Fig. 8. Determination and correction of thejth element phase. compared in phase correction aand Aej = 0,- B j is applied to bj through the order-wire channel. This increment of phase has the effect of illcreasing the magnitude of the resultant given by (22) as suggested by the phasor diagram, Fig. 8, and as will be shown subsequently. Element-by-element correction of phase in this way, going through the array several times, will result (b) in convergence to coherence. Fig. 7. Method of determiningfocusing phase angles. (a) Phase relaTo show the increasein the magnitude o f the resultant with tionshipsamong signals involved. @) Spectrum of transmitted signal fromjth array element. the correction, we write the phasors representing the separate arrivals exp(jOi), i = 1 , 2 , -, N . The resultant phqor at A is N ELEMENTS COS w,t x where, to avoid unnecessary complication, we assume all comN ponents arriving to beofequalmagnitude. In (21) the phases O i , = exp (ioi) = exp (io,). i = 1, 2, -, N account for the phase delays in the path CBA i= 1 plus the added phase shifts @iat each element, and aj represents the phase delay intheroundtrippath A b / ( q = Aftercorrection is applied to bi, theresultant is 2wrndj/c). The two components of (21) being nonoverlapping in freN quency are separated appropriate in filters. The first term is 4 = exp ( j e i ) - exp ( i q ) exp (io,) i= 1 denoted N rA l(t) = 2 cos (w’t + ei) = R cos (w‘t + e,). i= 1 (24) (25) where we have assumed that the original phasor expOej) has simply been shifted in phase t o form exp(j0,). Denoting by (22) R and R ’ the magnitude of r and r ’ , we find The second term is mixed with the original waveform cos w, t generated at A and the component at frequencyw ’ is retained; that is, form R 1 - exp [ @ j - e,)l R / >1. (26) This ratio is unity if andonly if B i = e,, and greater otherwise., Thus, convergence is assured. The focusingschemedescribed is onein w h c h element phases are adjusted in two steps: one to put all element local where NBF stands for narrow band filtering around frequency oscillators to the same phase and the second to correct for the w c r .We point out that the form(23) depends on having ident- different path lengths through the elements. If we were only ical sidebands on the double sideballd modulated signal; devi- concerned with transmission along the path CBA, the first step ations from the condition will cause an error in the phase of would not be needed. In this case the phase correction process the local the time-varying sinusoid. Provided laii < n/2, the phase of just described wouldalso account for deviations among (23) is 0, regardless of the value of ai. We can ensure a result oscillator phases. The firststep is required only because wewant the bidirective property whereby the same phase shifts f3i can such as (23) with phase independent of aj by rep.bcingcos wmt on the left of (23) with cos(o,t - 6) where & is a close be used in either direction of transmission through the array. estimate of ai. For 6 we may have 2wmd/c where d is the Should it be possible to add circuitry to the equipment at C to mean of the distances fromA to thearray dements. The latter assist in arrayorganization,thelocaloscillatorswouldnot is known, at least approximately. With this change made, the have to be cophased and two sets of values B j would be used right side of (23) has cos aj replaced by cos(aj - 6). Now if depending on the direction of thetransmission. Observe that the phase organization or focusing scheme deJ ( q - &)I = 2w, i(dj - a l / c < 77/2 the output phase of the use of the system;it narrow band is independent of a,. If, for instance,fm = 15 scribed does not interfere with the normal use continuously to compensatefor kHz, the maximum Idj - dl is 2.5 km. We envision an array of maythereforebein diameter approximately 1 km to operate at VHF so that the changes in the propagation characteristics and for movement of the mobile stations C. maximum condition is satisfied. The organizationscheme describedis robust in the sense With the phase in (23) not affected by aj,(22) and (23) are 490 PHASE DET. E- CARRIER rfm I I i ; : & E[ORDER tiCY WIRE I, CHANNEL1 OSC. Fig. 9. MOD RF AMP A possible functional block diagram of array organizer situated at A . that errors in the determination of the various phases will not switch controller SD to open the circuit in one direction when affect system performance catastrophically. Furthermore, transmission is in the other direction.Switches S3 and S4 consince the focusing system will be operating continuously, ertrol the local oscillator inputs to the mixers in each element, rors in successive scans of the array elements will tend to aver- admitting cos (ozot y j ) initially for measurement'of y j , then age out. Another important fact related to arrayfocusing is admitting cos (olot @ j ) for cohering the multiple arrivals at that the accuracy requirements of the organizational circuitry the destination. are not stringent. This is so because in an array with a large IV. DISCUSSION AND CONCLUSION number of elements, if the focusing phases have random independenterrorsof iess then1/2 rad rms, the resulting exThe intent in t h s paper was to show the feasibility and the pected loss in gain of the array 'with respect to perfect focus value of the array relay concept.Thescheme devisedwas is only about 1 dB [ l , p. 3071 . In obtaining this result, use is based on the assumed requirement of compatibility with current mobile transceivers. No hardware was therefore added to madeofthefactthatthe phase errorsarenormallydistributed. the mobile equipment, and the systemwas designed to operate Possible implementationsforthearrayorganizer(located in a half-duplex mode. Also, the scheme deals only with pointaccess communication.Extenat A ) and the array elements are given in the block diagrams, to-pointratherthanmultiple Figs. 9 and 10, respectively. In Fig. 9, the subcarrier oscillator sions to satisfy other requirements can be imagined and the deis a low frequency oscillator used to form the sidebands at fre- tails of a system to fit different requirements will be different. R F sources as ameansfor quency 9, for determining the phase corrections O j and to Yet the ideaofdistributingthe formthe LO signal at frequency azo. Circuitsarehere in- reducing the spatial power density, and as a diversity measure, cluded for obtaining estimates of the LO phases yi [see (17)- is applicable generally. Among possible extensions we see no achieving full duplexoperation, (20) and accompanying discussion] and the correction phases fundamentalproblemsin 8, - bj [obtained by copparing (22) and (23)] . Note the in- simultaneous communication between communication center clusion of the estimated phase & as discussed below (23). The and multiple mobile users', and even communication 'between microprocessor control is programmed to instruct the elements mobile units through the array with the aid of the communicaon when to respond and which measuring waveforms are to tion center. The scheme described utilizes phase control only to coherbe transmitted from the array ongaqizer. The transmitter on the array organizer functions only when the yi are being deter- ently combine the arriving signals from the different array elemined. Instructions and subcarrier information are sent on the ments. When signals of equal magnitude arrive at the receiving end, and when thenoise sources are not directional, phase conorder-wirechannel at frequency fo, differentfromthefretrol alone is adequate to achieve maximum output signal-toquencies 7 and f' used for information transmission and for measuring the LQ phases,at the elements. The array elements noise ratio (SNR). More likely, thesignals relayed through difcontain circuits for separating the signals arriving on the order- ferent elements encounter different conditions of propagation wire channel into' the control phase correction, and subcarrier and arrive with different'magnitudes relative to noise. In such components. The' elements also contain the sideband adding cases, amplitude and phase (complex) weighting at the array circuit iqed in'the Oj - 8 , measuring circuit, the phase shifters elements is required to achieve maximum SNR. The weighting for equalizing the LO phases and for injecting the correct @ j , required is similar to thatfoundin maximalratiodiversity and the phase conjugation circuit which makes transmission combining [4],the differences arising fromthe two-legged in both directions possible. The switches S1 and S2 are con- journeytakenbythe signal whereitencounterstwo noise trolled bp the microprocessor in the array organizer via the sources in two receivers. With directional noise (interference) + + 49 1 VESPOLI e t al.: SELF-ORGANIZING RANDOM ARRAY RELAY T,O CA1FREQUENCY f TO A AT FREQUENCY 1.1’ RF AMP. COS % t,, , Note: Y i s treated araisl e r e n cien this d i a g r a m a n d i s set t o zero. TO S3 AND SL cos IwLet.yl I FREQUEYCY=frn TIMER DRIVER g STORAGE TO S3’S‘ S I N U S O I D AT FREQUENCY = 1, RF If,) CODE D ET. Fig. 10. A possible functional block diagram of an array element bf also present, optimization of SNR requires further adjustment in the complex weights at the array elements. Adaptive sidelobe cancelling schemes abound [SI -[7] and variants of some of these may lendthemselves to thearray relay. The frequency range and spatial focal region over which the array remains in focus is reported elsewhere [8] and is not dealt with here. Beamwidth and depth of field of focus are very nearly the usual functions of the aperture size measured i n the plane of the arriving wavefront. The bandwidth is close to the inverse of the travel timeofthe arriving wavefront through the array. The scheme as described requires estimation of two sets of phase differences, one set to equalize the initial phases of the local oscillators at the array elements, thesecond set to correct for path length differences through the elements. It was not our purpose here to deal with the parameters of the estimation process such as estimation time and power forgiven noise levels. We should point out, though, that the cophasing procedureinwhichthe angles @j areadjusted begins withlow SNR signals. Recall that the signal froma single element is initiallycompared in phase with a signal from all elements combined noncoherently. The SNR at the output of each element is.not high (in the numerical example of SectionI1 it was 11 dB, but lower valuesof SNR are not unreasonable). This will impact the convergence time to achieve focus since both components being compared are noisy, and it will also affect the tracking rate since one of the components is still noisy in this mode. We have here described theconceptofthe arrayrelay, showing it to behavelikea lens systemcontrolled to focus emissions in one region toward a receiver in another region. Power requirements at each element are indeed low and may be decreased at the expense of increased numbers of elements. Individual elements in the shade ofsome obstacle, or not functioning, will not significantly alter the behavior of the system. The gain provided by the array also lowers the power requirements of the remote users and, because of the path diversity, increases reliability of communication. Finally,weremarkthatthe systemimplementationdescribed here is intended mainly as a demonstration of possibility. The functions to be carried out are not extraordinary, although circuit realization, with all the imperfections which attend practical devices, may require effort. REFERENCES [I] [2] [3] [4] [5] [6] [7] [8] B . Steinberg, Principles of ApertureandArraySystemDesign. New York: Wiley, 1976. J. B . Vespoli, F. Haber, R . S . Berkowitz, and D. Yavuz, “A self-organizing randomarraycommunications relay,” in Proc. 22nd Midwest Symp. Circuits Syst., Philadelphia, PA, June 1979, pp. 452-460. J . B. Vespoli, “The widely dispersedrandomarrayas a radio relay,” M.Sc. thesis, Moore School Elec. Eng. Sci., Univ. Pennsylvania, Philadelphia, June 1979; also in Valley Forge Res. Cen. Quart. Progress Rep.. no. 28, pp.46-63, Feb. 1979, and no. 29, pp. 35-48, May 1979. F. Haber, “Element gain setting formaximum receiver SNR,” ValleyForgeRes.Cen.Quart.ProgressRep., no.31,Univ. Pennsylvania, Philadelphia, pp. 21-24, Nov. 1979. S . P. Applebaum, “Adaptive arrays,” IEEE Trans. Antennas Propagat.. vol. AP-24, pp. 585-598, Sept. 1976. R. T. Compton, Jr.,“An adaptive array inspread-spectrum communication,” Proc. IEEE, vol. 66, pp. 289-298, Mar. 1978. W. F. Gabriel, “Spectral analysis and adaptive array superresolution techniques,” Proc.IEEE, vol. 68, pp. 654-666, June 1980. D. Yavuz, “Frequency response and focal region characteristics of random arrays,” Valley Forge Res. Cen. Quart. Progress Rep., no. 31, Univ. Pennsylvania, Philadelphia, pp. 15-20, Nov: 1979; also no. 29, pp. 49-60, and no. 30, pp. 27-33. 492 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. COM-31, NO. 4, APRIL 1983 Jorge B. Vespoli (A’78)wasborn in Mar del Plata, Argentina, in 1951.He received the Engineer Diploma from the University of Mar del Plata in 1975, andthe M.S.E.E. and M.S.S.E. degrees from the University of Pennsylvania, Philadelphia, in 1978 and 1979,respectively. From 1972 to 1982, he was with the National Telephone Administration of Argentina, ENTEL, where hewas engaged in the area of satellite communications. From 1975 to 1982,he worked as part-time Professor in the Department of Electrical Engineering at the University ofMar del Plata, instructing in electromagnetics, communication theory, and electronic measurements. From 1977 to 1979,he pursued graduate studies at the Moore School of ElectricalEngineering, University of Pennsylvania, under a Fulbright Scholarship. In October 1982, he joined the Satellite Communications Division of Harris Corporation, where he currently works as a Systems Engineer in the implementation of the National Satellite Communication Network of Argentina. * Fred Haber (S’44-A’49-M’55-SM’68-F’78) was born in New York, NY, on July 1, 1921.He received the B.S.E.E. degree from Pennsylvania State University, University Park, in 1948, and the M.S.E.E. and Ph.D. degrees from the University of Pennsylvania, Philadelphia, in 1953 and 1960,respectively. He was employed byRCA Corporation from 1948 to 1951 and has been associated for short intervals with the Anna Corporation, General Precision Inc., the U.S. Navy Underwater SystemCenter, andthe Aerospace Corporation. During 1968-1969, he was a Visiting Professor at Pahlavi University, Shiraz, Iran. Since 195 1, he has been at the University of Pennsylvania, where he is now Professor of Systems Engineering. He teaches courses in communication theory and relatedareas, and his current research deals with adaptive arrays for communication and signal location.