The effects of channel characteristics on turbo code performance

Turbo codes provide greater coding gain than commonly used convolutional codes. The maximum a-posteriori turbo decoder, unlike the Viterbi decoder often used for convolutional codes, explicitly uses E/sub b//N/sub 0/ as a parameter in the decoding process. We have investigated the effects of errors in estimating the E/sub b//N/sub 0/ on turbo code bit error rate (BER) performance in additive white Gaussian noise (AWGN) channels and binary symmetric channels (BSCs). Both small (220 bit) and large (65535 bit) buffers have been tested. For a BSC without any E/sub b//N/sub 0/ estimation error, the small buffer turbo code yields a BER of 10/sup -3/ at an E/sub b//N/sub 0/ of 3.9 dB and the large buffer turbo code yields a BER of 10/sup -5/ at an E/sub b//N/sub 0/ of 3 dB. The small interleaver turbo code operating over a two link satellite demod/remod channel produces a net BER of 10/sup -3/ with an uplink E/sub b//N/sub 0/ of 5 dB and a downlink E/sub b//N/sub 0/ of 6.6 dB. When E/sub b//N/sub 0/ estimation errors are included in the AWGN or BSCs, the results indicate that errors of up to several dB can be tolerated before the onset of significant degradation in turbo code performance.