The MMSE-DFE is a well established Intersymbol Interference (ISI) mitigating structure on linear, noisy, and dispersive channels. Time invariant communication channels exhibits ISI for severe frequency-selective channels. In this paper we investigate the iterative decoding with imperfect MMSE decision feedback equalizer using different modulation scheme. We assume that soft outputs from channel decoder are independent identically distributed Gaussian random variables with known mean and variance. The Imperfect MMSE DFE using different modulation outperforms other turbo equalization algorithms of similar computational complexity in terms of bit-error rate. The achieved improvement is up to 3dB for severe frequency-selective channels.