Sensitivity to SNR mismatch in interference cancelling based turbo equalizers

We study the sensitivity of a special iterative detection scheme, called turbo-equalizer, to an SNR (signal-to-noise ratio) mismatch at the receiver. For turbo equalization, the same algorithm as for SISO (soft-input soft-output) decoding is used for equalization at the first iteration. This turbo equalizer uses a MAP (maximum a posteriori) or Viterbi based SISO equalizer at the first iteration, and employs a simple interference canceller in the next iterations. It is shown that the sensitivity to an SNR mismatch is more for larger encoder memory length, as well as for difficult-to-equalize (highly delay-dispersive) channels. For the case that no information is available on SNR, a simple online SNR estimation technique is proposed, which gives satisfying results.