PIMRC’96, TAIWAN, OCT. 19961Effect ofcapture models on the performance oftheCDPA cellular architectureFlaminio Borgonovo , Luigi Fratta and Michele ZorziDipartimento di Elettronica, Politecnico di Milano, Piazza L. Da Vinci, 20133 Milano, Italy – lastname@elet. polimi. itCenter for Wireless Communications, UCSD, 9500 Gilman Dr., La Jolla, CA 92093-0407, USA – zorzi@ece. ucsd. edu— In this paper we present the performance of theCDPA cellular architecture obtained by using a more accuratepacket capture model thanthe one used in previously appearedanalyses. In the new model, the effect of co-channel interfer-ence is evaluated oneach bit ofthe packet. Theresults obtainedwith this new model are directly related to the propagation pa-rameters that are normally used in the literature on cellularsystems, such as the exponent of the near-far power attenua-tion law. In practice, we find that CDPA can achieve betterperformance than what predicted by the “threshold” modelused before. With this new model we also evaluate the effect offorward error correction techniques, which are found to pro-vide only a small improvement in the maximum throughput.I. INTRODUCTIONThe Capture DivisionPacket Access (CDPA) is a cellulararchitec-ture that has been recently introduced [1, 2, 3]. It is based on thepacket switching technique, so that it can easily integrate differentkinds of traffics, and can be an attractive alternative to the twoclassic cellular approaches, that are based on CDMA [4, 5] andFDMA/TDMA [6]. CDPA implements a cellular access architec-ture in which multiple access and coding against the co-channelinterference are designed in an integrated way to achieve the bestefficiency while using a single frequency in all cells. CDPA usesa hybrid reservation/polling mechanism (HRP) at the base sta-tion (BS) to solicit packet transmission from the mobile terminals(MT). Parallel transmissions from different cells exploit capture.Ifcapture fails, the MT is solicited again to retransmit the collidedpacket. This mechanism has been shown to be particularly ef-fective because it can easily adapt to rapidly changing co-channelinterference and propagation conditions. Finally, despite the delayintroduced by packet retransmissions, it has been shown that thistechnique can cope with delay sensitive traffic such as voice [3, 7].Packet capture, i.e., the capability to correctly receive a packeteven in the presence of noise and interference, is a complex phe-nomenon that depends on several parameters, such as the lengthofthe data string, the modulation and coding techniques, the numberofinterfering signals and their characteristics, and the propagationconditions, that affect the signal shape and level. In a packet trans-mission environment, a possible characterization of the receivercapture can be represented by the probability that the informa-tion conveyed by a packet is correctly detected at the receiver asfunction of the short-term signal to interference ratio (SIR) at thereceiver itself. Examples of such results are given in Figure 1. Inthisexample, we show the packet capture probabilitythe probability that all bits in the packet are correctly detected inSIR , i.e.,This work has been supported by MURST 40% 1995-1996 and by the Centerfor Wireless Communications, UC San Diego.the environment that is specified later indetail. The shortterm SIRis defined asSIR01whereco-channel interference component, i.e., the power received at theintendedBS dueto aconcurrentpacket transmissionincell . Inthefigure we also plotted the corresponding SIR probability-densityfunctionSIR . The average capture probability, which is directyrelated to the system capacity, can then be derived as0is the power ofthe received packet andrepresents a2The equation above explains the mechanism on which CDPA isbased but, unfortunately, can not be directly used in the analysisbecause the SIR density depends, in turn, onmines the retransmission policy. Moreover, also the curvedepends on the retransmission policy, because the capture prob-ability changes if, given SIRpattern change, as shown in Figure2, where the two capture curvesshown correspond to two different values of the transmission ratein the adjacent cells. In other words, this means that even theshort term SIR does not capture all the system variables.In order to obtain an analytical approach, in the past we haveadopted a simplified capture model, the capture-threshold model,exemplified by the dotted curve in Figure 2. In this model a packetis correctly detected if the SIR exceeds a given threshold , thecapture threshold. By this model, the capture phenomenon is de-coupled from the environment, and analytical evaluations becomepossible, as described in[3, 8, 9, 10], where theCDPAperformancehas been evaluated under a wide range ofhypotheses and environ-ment models, such as different cell layouts and reuse factors, fastand slow shadowing, site diversity, etc.In this paper we present a more accurate CDPA throughputeval-uation, by adopting a more realistic capture model that simulatesthe capture directly in the cell environment, without depending onthe reduced description given by the SIR. The packets, composedof511 bits, are BPSK modulated. The receiver demodulatesbit by bit the received signal, which is the sum of the intendedsignal and interfering signals, perfectly overlapped with randomphases.This model describes accurately the receiver behavior,thus providing a better understanding of the interference-capturemechanism. It also allows to investigate the effects ofpacket cap-ture enhancing methods, such as forward error correction (FEC)codes. In this case, a packet is considered captured by the receiverifno errors are present after decoding.The results are obtained by simulation, a technique that alsoallows to take into account realistic propagation models., which deter-the interference traffic and/or
