Impact of SNR estimation error on turbo code with high-order modulation

The sensitivity of turbo-code performance to SNR estimation error is investigated in high-order modulated systems. The estimated SNR is applied in bit-metric calculation for a turbo decoder. We analyze the roles that the SNR value has played in binary transmission and multi-level transmission and identify the difference which finally leads to the different sensitivity behavior of the turbo decoder in different modulation schemes. Extensive simulation results are presented to demonstrate the performance sensitivity over AWGN and fading channels. We show that turbo code is more sensitive to SNR estimation accuracy in high-order modulation than in BPSK. Both under-estimation and overestimation lead to significant performance degradation. The BER performance versus SNR estimation offset exhibits an obvious "V"-like shape. Moreover, the turbo-code parameters such as interleaver size and constraint length are also important factors that affect the performance sensitivity to SNR mismatch. Codes with large interleaver size or long constraint length tend to be more sensitive to the estimation error.