A novel algorithm of iterative channel estimation and turbo decoding for pilot symbol assisted modulation (PSAM) is proposed. The proposed channel estimation is based on least squared fitting (LSF) theorem. Compared with the linear optimum MMSE (minimum mean square error) channel estimation, it doesn't require any channel characteristic like as fading rate or noise variance. We discuss the effect of fading rate and noise variance estimation error in MMSE systems. The influence of noise variance in MMSE systems is slight, but fading rate mismatch effect is huge. The proposed system is less sensitive than MMSE system about fading rate variation.