Title :
Variable wordlengths in filters with least-squares error criterion
Author :
Dempster, A.G. ; Macleod, M.D.
Author_Institution :
Sch. of Electron. & Manuf. Syst. Eng., Westminster Univ., London, UK
fDate :
1/30/1997 12:00:00 AM
Abstract :
Variable coefficient wordlengths in combination with a multiplier reduces the computational complexity of fixed-point IIR digital filters with minimax error specification. When variable wordlengths are used for the least-squares error specification here, improved filters are achieved without resorting to discrete optimisation. Also. The wordlength allocation for the least squares criterion gives near-optimum results for filters with minimax specification
Keywords :
IIR filters; least squares approximations; minimax techniques; computational complexity; fixed-point IIR digital filter; least-squares error; minimax error; multiplier; variable wordlength;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:19970159
