Digital signal processing

digital filter design

Digital filter design is the craft of choosing the coefficients of an FIR or IIR filter so that it lets some frequencies through and blocks others, matching a specification you draw as a target shape: how flat the passband, how deep the stopband, how steep the transition, how much phase distortion is allowed. It is goal-driven engineering — you sketch the wall you want sound or data to hit, then a method computes the numbers that build it.

Several well-worn recipes exist. The window method takes the ideal 'brick-wall' filter, which would need infinitely many taps, and truncates it with a smooth window for a workable FIR. The Parks–McClellan (Remez) algorithm produces the optimal equiripple FIR for a given length. For IIR, designers often start from a known analog prototype (Butterworth, Chebyshev) and warp it into the digital domain with the bilinear transform. Each choice trades sharpness, ripple, delay, and computational cost.

The bilinear transform maps the entire analog frequency axis onto the finite digital one, which non-linearly 'warps' frequencies; designers pre-warp the critical cutoff so it lands exactly where intended after the mapping.

Also called
filter synthesiswindowed-sinc methodParks–McClellanbilinear transform濾波器設計