[music-dsp] Linear and Nonlinear phase filters in practice

[Sampo Syreeni]

On 2003-08-12, Toto Takehara uttered:

>As far as I understand IIR filters do not ensure linear phase like FIR,
>is that correct?

Not really. No linear phase IIR's exist, but there are plenty of
non-linear phase FIR's. (There are also cascades of non-linear phase IIR's
which constitute linear phase FIR's, so beware. ;)

>But it seems to me that more real time filters are IIR but not FIR.

Both IIR's and FIR's can always be implemented in real time, if they're
causal. All it takes is processing power. The phase shift is a separate
concern entirely.

>I thought linear phase was more preferable.

It is when you want transparency. It isn't when you want a specific
effect, involving a phase shift. Both phasing and flanging are based on
that. So are many of the amp emulations out there. Phase shift is only bad
if you want what comes out to sound like what goes in. In DSP, that
assumption is for the most part just false.

>Especially in EQs.

True. Most of the time we wan't EQ's to affect just amplitude, not phase.
But EQ's are just a small part of audio processing as a whole. Not to
mention synthesis...

>But some really expensive EQs like the Sony Oxford's is advertised as non
>linear. Could anybody give me some clues on this please?

Well, EQ's can be used as an effect as well. They have their sound, and
sometimes we want to utilize that sound. Some of it comes from phase
nonlinearity, a part from complicated amplitude response, and a bit from
actual nonlinearity. All of those can contribute to a pleasurable sound.