A novel combined approach of constant correction and variable correction method to minimize the mean square error and compensation hardware for fixed width multiplier design

This paper is based on a novel combined approach of constant correction and variable correction scheme for fixed width multiplier (FWM) with detail mathematical analysis of error correction function. The paper summarizes the various milestone topologies of fixed width multipliers and their implementation strategies. The mathematical analysis of a compensation function is based on the probability theory and mathematical expectation applied to least significant part of the partial product matrix, which is ultimately responsible for occurrence of error. The newly designed topology of an error compensation function is compared with the state of art FWM architectures in terms of hardware complexity and mean square error. The proposed combined approach gives the better error compensation without any cost of extra hardware. This design is tested as a part of DSP based application in 1024 and 8192 point FFT.