Abstract
Digital data communication systems are degraded by statistical noise and intersymbol interference caused by bandwidth constraints. This paper presents a new class of nonlinear equalizers with simplified structure implemented by digital filter circuits involving only decision slicers, which can approximately give the maximum likelihood (ML) bit estimates. The structure of the proposed generalized decision feedback equalizer (DFE) is obtained by inducing the concept of the decision feedback into the ML bit estimation. This paper also derives the computational methods for lower and upper bounds of the bit error probability and demonstrates that the proposed technique can give a significant improvement in error rate performance compared with the conventional DFE.
