IEE Proceedings G (Electronic Circuits and Systems) · 1988 · 20 citations · 6 references
ComputingComputer EngineeringComputer ScienceAlgebraAlgorithmsNumber TheoryArithmetic FunctionsDigital Signal ProcessingNumber-theoretic AlgorithmsConvolutionFermat FactorizationMultipliersMultiply-accumulate Units
Work on using the Fermat number transform (FNT) for convolution and correlation has recently appeared in the literature, and it is known that by using the diminished-1 number scheme all the operations required for the FNT can be computed exactly. The paper presents a diminished-1 parallel multiplier that improves the overall speed of an FNT based convolver/ correlator. The multiplication algorithm used realises the diminished-1 product (A · B − 1) = (A · B) − 1. The work has also highlighted an error in the above algorithm when the data is equal to −1 and offers a method to overcome this problem.
6