EECS20N: Signals and Systems

Filter Design

Good filter design demands quite a bit of expertise. The following general observations may be helpful.

  • FIR filters are a costly way to implement narrowband filters, but are often reasonable for filters with wider bandwidth. It is also easier to control the phase response and to design filters with an arbitrarily shaped frequency response (as opposed to just lowpass, bandpass, or highpass). Use Matlab's remez or firls functions.
  • IIR filters typically require much less computation per output sample than FIR filters for similar frequency responses. They excel at narrowband filters. However, it is very difficult to control the phase response. Use Matlab's butter, cheby1, cheby2, or ellip functions.