Part IV MULTIPLE ACCESS AND ADVANCED TRANSCEIVER SCHEMES 363
18.5 Time Hopping Impulse Radio
18.5.3 Impulse Radio in Delay-Dispersive Channels
Up to now, we have discussed impulse radio in AWGN channels. We have found that the transmitter as well as the receiver can be made very simple in these cases. However, TH-IR almost never works in AWGN channels. The purpose of such a system is the use of a very large bandwidth (typically 500 MHz or more), which in turn implies that the channel will certainly show variations over that bandwidth. It is then necessary to build a receiver that can work well in a dispersive channel.
Let us first consider coherent reception of the incoming signal. In this case, we need a Rake receiver, just like the one discussed in Section 18.2.3. Essentially, the Rake receiver consists of multiple correlators or fingers. In each of the fingers, correlation of the incoming signal is done with a delayed version of the time-hopping sequence, where the delay is equal to the delay of the MPC that we want to “rake up.” The output of the Rake fingers is then weighted (according to the principles of maximum-ratio combining or optimum combining) and summed up. Because Impulse Radio (IR) systems usually use very large bandwidth, they always use structures whose number of fingers is smaller than the number of available MPCs (SRake, PRake, see Section 18.2.3). In order to get reasonable performance, it might be required to have 20 or more Rake fingers even in indoor environments. Differentially coherent (transmitted reference) or noncoherent schemes thus become an attractive alternative.
In order to understand the principle ofTransmitted Reference(TR) schemes, remember that the ideal matched filter in a delay-dispersive channel is matched to the convolution of the time-hopping sequence with the impulse response of the channel. For coherent reception, we first have a filter matched to the time-hopping sequence, followed by the Rake receiver, which is matched to the impulse response of the channel. A so-called TR scheme creates this composite matched filter in a different way. A TR transmitter sends out two pulses in each frame: one unmodulated (reference) pulse, and, a fixed timeTpdlater, the modulated (data) pulse. The sequence of reference pulses is convolved with the channel impulse response when it is transmitted through the wireless channel;
this signal thus constitutes a noisy version of the system impulse response (convolution of transmit basis waveform and channel impulse response). In order to perform a matched filtering on the data- carrying part of the received signal, the receiver just multiplies the received signal by the received reference signal.
Spread Spectrum Systems 415
Let us now have a look at the mathematical expression of the transmit signal for one symbol:
p(t )= 1
2
Nf
j=0
dj[g(t−j Tf−cjTC)+bãg(t−j Tf−cjTC−Tpd)] (18.44) Inspection shows that correlation with the received reference signal can be done by just multiplying the total received signal by a delayed (by time Tpd) version of itself and integration. To be more precise: the first step at the receiver involves filtering by a receive filterhR(t ); this filter should be wide enough not to introduce any signal distortions, and just limit the available noise. The filtered received signalr(t )is multiplied by a delayed version of itself, and integrated over a finite interval Tint, which should be long enough to collect most multipath energy, but short enough not to collect too much noise energy.
TR schemes have the advantage of being exceedingly simple (though implementation of the delay in the receiver may be nontrivial). On the downside, they show poorer performance than coherent schemes for two reasons: (i) reference pulses waste energy, in the sense that they do not carry any information (this results in a 3-dB penalty); (ii) the reference part of the signal is noisy, as is the data-carrying part of the signal. Multiplication of the two signals in the receiver gives rise to noise–noise cross-terms that worsen the SNR. Now remember that impulse radio is a spread spectrum system, so that the SNR of the received signal is negative. Noise–noise cross-terms can thus become large.
Finally, we note that noncoherent detection can be an attractive alternative as well. This is espe- cially true if only a single user is to be served. However, noncoherent receivers have problems with multiaccess interference. Due to multipath propagation, energy is dispersed over several adjacent chips. Thus, energy detection will see much more interference (as exemplified in Figure 18.15).
Desired user – integration times Energy from interferer
– three chips shift
Figure 18.15 Interference from a delay-dispersed interferer signal (lower row) to the desired signal, whose integration time (two chip durations per frame) is sketched in the upper row.
Further Reading
An overview of spread spectrum communications in general can be found in the classic monographs of Dixon [1994] and Simon et al. [1994], which provide good coverage of FH and DS systems.
Books that are more tuned to cellular systems, especially CDMA, are Glisic and Vucetic [1997], Goiser [1998], Li and Miller [1998], Viterbi [1995], and Ziemer et al. [1995]; the overview paper [Kohno et al. 1995] gives a shorter description. Scholtz [1982] gives an overview of the history of the spread spectrum (though more from a military perspective, since cellular applications were not yet considered at the time of that article). Milstein [1988] discusses the interference rejection capabilities of various spread spectrum techniques.
Frequency-hopping codes are designed in Maric and Titlebaum [1992]. Estimates for the capacity of CDMA systems were first published in the widely cited paper of Gilhousen et al. [1991]. The effect of multipath on CDMA – namely, the use of Rake receivers – is reviewed in Goiser et al.
[2000], Swarts et al. [1998] particularly, the effect of a finite number of Rake fingers [Win and Chrisikos 2000]. Other important papers on Rake reception include Holtzman and Jalloul [1994] and Bottomley et al. [2000], among many others. For synchronization aspects, the papers by Polydoros and Weber [1984] are still interesting to read. Spreading codes are reviewed in Dinan and Jabbari [1998]. An authoritative description of sequence design (for radar as well as communications) is Golomb and Gong [2005]. While the Gaussian approximation is widely used for intercell inter- ference, it can give significant deviations from reality when shadowing is present; a more refined model is presented in Singh et al. [2010]. A complete TH-IR radio system is described in Molisch et al. [2004].
For multiuser detection, the book by Verdu [1998] is a must-read. Also, the overview papers (Duel-Hallen et al. [1995], Moshavi [1996], and Poor [2001]) give useful insights. Advanced con- cepts like blind multiuser detection are described extensively in Wang and Poor [2003] and Honig [2009]; the interaction between multiuser detection and decoding is explored in Poor [2004].
Time-hopping impulse radio was pioneered by Win and Scholtz in the 1990s. For a description of the basic principles, see Win and Scholtz [1998], and more detailed results in Win and Scholtz [2000]. An overview of the different aspects of ultrawideband system design can be found in the monographs Shen et al. [2006], Ghavami et al. [2006], Reed [2005], diBenedetto et al. [2005], Roy et al. [2004]; TR receivers and other simplified receiver structures are discussed in the overview paper of Witrisal et al. [2009].
For updates and errata for this chapter, see wides.usc.edu/teaching/textbook
19
Orthogonal Frequency Division Multiplexing (OFDM)