On a periodic training sequence in DFE to reduce the steady-state error probability

A periodic transmission of a short resetting sequence used to “fill up” the decision-feedback equalizer feedback filter memory with correct symbols is proposed to reduce the propagation error. Two strategies are considered: full and partial reset, where the entire memory and some part of it is reset, respectively. For any given steady-state bit-error rate, the period of the resetting sequence is found by use of the Markov chain limit theorem. It is shown, via an illustrative example, that for low signal-to-noise ratio (SNR), the resetting strategy outperforms Bose-Chaudhuri-Hocquenghen (BCH) code with the same code rate. Moreover, a joint resetting and error correction coding strategy is considered, which is better than BCH code for low and medium SNR