Effects of finite coefficient precision on FIR filter spectra

Finite Impulse Response filters (FIR) and other digital signal processing algorithms are implemented either with special purpose hardware or as programs on special purpose computers or signal processors. In either case the state variables and the coefficients are stored in binary format in registers or memory elements with finite lengths. Register length is an economic factor in hardware implementations and in software implementations it is generally forced upon the designer by the architecture of the available processor or computer.