Prediction of performance of the DVB-SH system relying on Mutual Information

DVB-SH (Digital Video Broadcasting- Satellite Handled) is a broadcasting standard dedicated to hybrid broadcasting systems combining a satellite and a terrestrial part. On the satellite part, dedicated interleaving and time slicing mechanisms are proposed to mitigate the effects of Land Mobile Satellite (LMS) channel, based on a convolutional interleaver. Depending on the parameters of this interleaver, this mechanism enables to split in time a codeword on duration from 100 ms to about 30s. This mechanism significantly improves the error recovery performance of the code but in literature, exact evaluation at system level of this improvement is missing. The objective of this paper is to propose a prediction method compatible with fast simulations, to quantitatively evaluate the system performance in terms of Packet Error Rate (PER). The main difficulty is to evaluate the decoding probability of a codeword submitted to several levels of attenuation. The method we propose consists in using as metric the Mutual Information (MI) between coded bit at the emitter side and the received symbol. It is shown that, by averaging the MI over the codeword and by using the decoding performance function g such that PER=g(MI) determined on the Gaussian channel, we can significantly improve the precision of the prediction compared to the two other methods based on SNR and Bit Error Rate (BER). We evaluated these methods on three artificial channels where each codeword is transmitted with three or four different levels of attenuations. The prediction error of the SNR-based (resp. the input BER-based) method varies from 0.5 to 1.7 dB (resp. from 0.7 to 1.2 dB) instead of the MI-based method achieves a precision in the order of 0.1 dB in the three cases. We then evaluate this method on real LMS channels with various DVB-SH interleavers and show that the instantaneous PER can also be predicted with high accuracy.