1 Channel Coding for Cooperative Broadcasting M. A. Imran, M. K. Gurcan and S. Kahn On-demand delivery of live media streams over down- link of 3G networks require minimum start-up delay and latency. To meet the increased demand for such services, existing capacity needs to be extended. Goldsmith [1] identifies the capacity region for the downlink multiuser channel. This channel is characterised by the co-channel interference as a result of multipath as identified by Mehta et al [2]. Bergmans and Cover [3] show that co- operative broadcasting can increase capacity in compar- ison to time-multiplex and frequency-multiplex schemes. This scheme involves interference cancellation requiring subtraction of the estimated interfering signals from the received signal, followed by decoding of the intended signal. In order to decode the desired signal satisfactorily, it is important that the estimation of interfering signal is error free. Automatic repeat request (ARQ) based techniques, like TCP, can ensure error free delivery [4] of the packets. However retransmission requests for failed ‘in- terfering packets’ are undesirable. We must, therefore, ensure that the failed interfering packets can be recovered without retransmissions. Digital Fountain schemes, surveyed in [5], can be used to extract the failed interfering packets from a set of successful packets from the interferer’s stream. An example scheme is LT-coding presented in [6]. For a given number of packets, k, LT codes require a larger number of encoded packets, K, to ensure successful decoding with high probability. Hence, effective rate of LT codes, k/K, is less than one. This has an associated E b /N 0 loss in packet error rate performance. In order to avoid this performance loss, the number of source packets must be large as shown in [6]. As a result the start-up delay and latency associated with the Fountain codes can be prohibitively large for on-demand live streaming services under consideration. Reducing the start up delay requires that k must be kept small and at the same time failed packets have some alternate source available for detection. One solution would be to repeat each packet n times, termed as repeat- n code. Repetition degrades the packet error performance against E b /N 0 for a fixed bandwidth and signal to noise ratio. Using the coding gain (exploiting the multiple The authors are with Communications and Signal Processing Group, Department of Electrical and Electronic Engineering, Imperial College London, SW7 2AZ, UK (email: m.gurcan@imperial.ac.uk) sources of packet if available) this loss in performance can be recovered for a range of E b /N 0 . This scheme however exhibits a packet error floor at high E b /N 0 values which depends on erasure probability and the number of repeats, n. In order to lower the error floor, we must provide many multiple sources for each message packet without sacri- ficing the performance against E b /N 0 . This is achieved by forming each encoded packet by bitwise exclusive OR operation on multiple message packets. In addition to lowering the error floor, the multiple sources for each message packet are used to recover the E b /N 0 loss due to extra packets transmitted. In this paper we demonstrate how a channel coding scheme can be used to improve the performance of cooperative broadcast systems on downlink. We present a simple code containing packets formed by XOR oper- ation on encoded packets. We show how the individual packet decoding performance can be improved by itera- tively using the information from the multiple sources. After the iterative decoding is complete, the message packets are regenerated from the decoded packets using an algorithm similar to LT-decoding process [6]. Packet error rate performance for the interferer’s stream is compared for turbo coding with repeat-n, LT- coding and proposed coding scheme. Simulation and analysis show that in comparison to repeat codes, the proposed scheme provides around 2dB improvement in performance at low E b /N 0 and also brings down the error floor at high E b /N 0 . In contrast to fountain codes, these codes have a slightly raised error floor but with much smaller overhead. REFERENCES [1] A. J. Goldsmith, “The capacity of downlink fading channels with variable rate and power,” IEEE Transactions on Vehicular Technology, vol. 46, pp. 569–580, 1997. [2] N. B. Mehta, L. J. Greenstein, T. M. W. III, and Z. Kostic, “Analysis and results for the orthogonality factor in WCDMA downlinks,” IEEE Trans. Wireless Commun., vol. 2, pp. 1138– 1149, 2003. [3] P. P. Bergmans and T. M. Cover, “Cooperative broadcasting,” IEEE Transactions on Information Theory, vol. IT-20, pp. 317– 324, 1974. [4] S.Lin and D. Jr., Error Control Coding: Fundamentals and Applications. Prentice-Hall, 1983. [5] M. Mitzenmacher, “Digital fountains: a survey and look for- ward,” 2004, pp. 271–276. [6] M. Luby, “LT codes,” in Proceedings 43rd Annual IEEE Sympo- sium on Foundations of Computer Science, 2002, pp. 271–280.