An Efficient Realization of the Decision Feedback Equalizer using Block Floating Point Arithmetic

Abstract

An efficient scheme is presented for implementing the decision feedback equalizer (DFE) in block floating point (BFP) format, which permits processing of data over a wide dynamic range at a processor cost marginally higher than that of a fixed point processor. It uses block processing of the input which is block formatted by an appropriate algorithm, taking care so that there is no overflow at the feed forward filter output and also that a uniform BFP representation of the input data vector is available all throughout including the inter-block transition phase. Suitable steps are also taken for block formatting the data used in computation of the feedback filter output, again ensuring no overflow. As the computations involve mostly fixed point based operations, considerable saving in computational complexities is achieved, as against a conventional floating point based realization.