Skip to Main Content
Turbo codes allow reliable data communication at signal-to-noise ratios very close to those predicted by Shannon. Further, breaking up the decoding task so that separate decoders work on the two constituent codes, passing extrinsic information between them, greatly simplifies the implementation of the decoder. Traditionally, the statistical model for noise used in the design and analysis of the decoder is the additive white Gaussian noise (AWGN) model. However many practical applications present situations where the AWGN model for noise is clearly sub-optimum. Optimum performance for a spread spectrum system when the noise is non-Gaussian in nature is possible by augmenting the traditional turbo decoder with a nonlinear preprocessor. A computationally efficient version of such a receiver for reliable decoding of turbo coded, direct-spread data in non-Gaussian noise is investigated here. The noise is represented using Middleton class-A model and a parametric receiver is designed. Performance of such a scheme based on the system bit error rate is studied.