Fast Prediction of Error Propagation in DFE Detectors

Abstract

to that of a DPLL (Digital Phase Locked Loop). The approach presented in this paper differs from this usual approach. The method we propose uses an auxiliary filter which attempts to predict the amplitude error due to sampling phase error. In the case of a PR4 (Partial Response class IV) channel this filter can be driven using the desired response samples which are from the set {O,? I} which removes the need to use any multiply operations thereby greatly reducing complexity. By using this auxiliary filter it is demonstrated that, for the PR4 case, the same performance can be achieved as with the usual interpolating approach, but with a less complex interpolator. This is attributed to the fact that in the usual approach the interpolator must provide interpolants to the detector based on noisy samples. In our approach the interpolants are instead used to drive the auxiliary filter from which the detector input is derived.

Techniques for minimizing error propagation in decision feedback detectors for recording channels

“…Further, the conditions and are equivalent to and , respectively, since in (12). Substituting the above and (13) in (12) and also simplifying, we get (14) Since is large compared to (see Fig. 2), the second term in (14) is negligibly small compared to the first term.…”

“…The values of the parameters , , , and were chosen by running a few trials. The BLD was estimated using the fast simulation procedure given in [14]. Since the details of this simulation procedure are not available in [14], we have given a brief summary of this algorithm in Appendix I.…”

“…The BLD was estimated using the fast simulation procedure given in [14]. Since the details of this simulation procedure are not available in [14], we have given a brief summary of this algorithm in Appendix I. In the detected bit sequence, an error event is said to begin at instant and end at instant if: 1) for and ; 2) there are no decision errors in bits prior to and after ; and 3) there is at least one decision error in any block of bits between and .…”

“…Hence, by applying error correction on the magnitude-constrained detector, we can have the good slope produced by the magnitude-constrained design and reduced error event rate due to error correction. As before, to save time, we want to use the fast simulation procedure [14] for estimating the BLD curve. However, we need to be cautious while using this method for a system with error correction.…”

“…Theoretical analysis has been carried out to quantify the false and correct detection probabilities associated with the threshold techniques. By combining the results of this analysis with an existing fast prediction algorithm [14], a fast approach is developed for estimating the overall BLD and/or bit error rate (BER) of the detector when the threshold techniques are applied. This paper is organized as follows.…”