[sdiy] group delay (of filters) and listening

[Magnus Danielson]

From: mirwin at qouest.net
Subject: Re: [sdiy] group delay (of filters) and listening
Date: Tue, 05 Oct 2004 09:32:09 -0400
Message-ID: <4162A259.4B6F at qouest.net>

> Martin (and list),

Mike,

> I have done some experiments with long allpass chains (including up to 
> 24 first-order analog allpass stages, identical sections). If you listen 
> to a narrow pulse (say one millisecond) the non-constant group 
> delay (dispersion) is quite obvious - the "click" turns into a short 
> "blip" with rapidly descending pitch. The effect on recorded music is 
> much less obvious, more so with fewer allpass stages. 

In the short term, it's the transient part (impulse) which gets modified. In
the long term, the all-pass characteristics take over so that the slow decay
behaves roughly as if it was just delayed.

Thus, the sharp attacks is where you can expect the differencies due to group
delay aspets to become apparent.

There is then another effect of poles and zeros in the music, zeroes in the
middle of where your music is will also cancel out certain attacks, so it is
not "punchy" enought. No wonder, the attack part gets swamped by the system.
Moving those zeros away will help. Similarly will resonances be quite annoying
since key keep stick things out improperly, when it happends in the mid the
system gets "piercy". Move those out of the way and it cools down.

The sharper change in amplitude you have in a sound, the further away you are
from the traditional constant sine amplitude responce curves. The direction of
the amplitude (rising or falling) is also of importance, since that gives you
right or left (respectively) in the s-plane diagram. Music traditionally have
much quicker rising than falling amplitude characteristics, so the switft
increase of amplitude is usually where one should listen, but for more
experimental music you can have sometimes almost equal amounts of same speed.

Cheers,
Magnus