All pass filters?

[Magnus Danielson]

>>>>> "MC" == Martin Czech <martin.czech at intermetall.de> writes:

 >> This is something i just don't understand. I'm studying computer systems 
 >> engineering 2nd year, and i'm doing a lot of analogue electronics 
 >> subjects including filter design etc, but what the hell is an all pass 
 >> filter used for? If it passes everything then its not really filtering is it?
 >> 
 MC> When all poles of the transfer function have a mirrored zero,
 MC> then |H(s)| =1, thus allpass . But the phase of the signal will be frequency
 MC> dependend. Allpass structurs are used for phase shifting or time delay
 MC> therefore.

Being the nitpick that I am I must make a small objection here...

It is correct that you get the allpass by mirrowing the poles with
zeros (mirrowing occurs around the jw-axis). However, this does not by
far makes |H(s)| = 1 but |H(jw)| = 1 is more like it. The difference
will not show up on a sine-sweep but the difference is certainly ther
for nonstable signals. It is a small but important difference which
however does not directly govern the way we make allpass filters (but
very well could!).

BTW. I greatly enjoyed Eric Babours article in IEEE Spectrum!

Cheers,
Magnus