Whenever I was choosing values for filters I always ignored the phase angle plots, but should I be taking it into consideration when for example designing a high-pass filter for the input of an amplifier or a low-pass filter in the feedback network? What are the situations where it's important to know the phase angle and take it into consideration in the design?
-
\$\begingroup\$ What does the signal represent? If it's audio then there are sonic implications - i.e. you will be changing how the audio sounds. \$\endgroup\$– CharlieHansonDec 23, 2015 at 13:05
-
\$\begingroup\$ Stability is a concern and in audio you might get a hearable lag between frequency components which where at the same time in the unfiltered signal. \$\endgroup\$– ArsenalDec 23, 2015 at 13:06
-
1\$\begingroup\$ Your question is soliciting opinion - read the rules. \$\endgroup\$– Andy akaDec 23, 2015 at 14:53
-
\$\begingroup\$ This isn't as much opinion-based as rather broad. The answer is: it depends on the application [of the filter]. LvW gave you some decent examples when it does matter. \$\endgroup\$– the gods from engineeringDec 23, 2015 at 17:21
-
\$\begingroup\$ For contrast, here's a well-known example when it isn't: the Linkwitz–Riley (4th order) filter widely used in audio. \$\endgroup\$– the gods from engineeringDec 23, 2015 at 17:33
1 Answer
From system theory we know that the negative slope of the phase response gives the so called "group delay" of the filter. For some applications - e.g. filtering of squarewave signals - we want to have a group delay as constant as possible (identical to a linear phase response) in order to retain the original waveform (in spite of filtering).
This requirement was the reason to define the Thomson/Bessel response as one of the standard filter functions. As an another example: We have allpass filters (constant magnitude over frequency) with the only purpose to shape/correct the phase response (resp. group delay) of other filter ciruits (Delay equalizer).