[sdiy] group delay (of filters) and listening

[Magnus Danielson]

From: "Czech Martin" <Martin.Czech at Micronas.com>
Subject: RE: [sdiy] group delay (of filters) and listening
Date: Tue, 5 Oct 2004 13:40:05 +0200
Message-ID: <D9D56E8FA1A73542BE9A5EC7E35D37FFF39142 at EXCHANGE2.Micronas.com>

Martin,

> Yeah, it´s better than TV!
> Had my vacation in the Swiss Alps, computer in the backpack, lot´s of time
> to program at night. Stiff walks at daytime make you think better!

Indeed. I on the other hand has to stay home until my fever goes down. Wish I
was doing those walks with you! Just guess what interesting discussions we
could have had!

> All comments known & true.

More or less, yes...

> I had also an experiment with a second order allpass running, audible
> artefacts ocured only at insane Q values. At that time I had no
> group delay calculator, I have to repeat that.

Yeah. Plot the group-delay and dispersion.

> The first order allpass was now created, I piled up 30 of them
> in order to get a nice large group delay that is really somehow
> audible. The nice thing in digital world is that this could be done
> in minutes.

I agree. You can do these things quickly if you prepared yourself properly.
Also notice that the "3 dB corner" moves as you stack up many of these bastards
ontop of each other. You may or you may not want to compensate for it.

> After all, it is not important how the group delay was achieved, but
> only it´s shape.

Yes, but having a "low pass" figure compared to a "band pass" figure IS
different. Also, if the dispersion plot shows a number of bumps instread of
"simple slopes" then you are different anyway.

> I will continue this little investigation, like you said:
> -different corner frequnecies 

I was thinking more of center frequencies, from two-pole/two-zero APs, possibly
ganged together.

> -possibly in the main speach formant area

Right.

> Now: a clever audiophile could counter: your headphone creates allready so
> much phase distortion that you can not hear these little artefacts (which I
> can with my golden ears and my ***** brand speakers).
> 
> How to counter that?

All the phase distorsions add up. Of course I always come to a limit of a
particular system, where larger phase distorsions in the system hides the ones
in my measurement. Infact, I've done this exercise already and know VERY well
that larger phase-distorsions hide smaller ones, but that's also my point since
I've seen that many systems have very large phase distorsions to start with,
and people are fiddeling around with all the wrong tools to do anything about
it. When you confront people with a system being very phase-linear, they will
learn just how much their normal graphical EQ corrections actually destroys the
sound and they will back off. Back in the good old days when I did PA stuff, we
cleaned the pipe on that PA. Whenever a house technician came with a band he
put up his standard CD and put up his standard EQ correction. A fun trick we
had was to ask him, just for a brief moment, to hit the bypass switch on the
main EQs. The difference was so large that when they snapped back into reality
they really flattened their EQ out, since now each correction DID MORE.
Actually, most systems have gross phase-distorsions built into them and it is
really annoying. A good set of headphones has much less of that going on, even
if they too have phase distorsions happening. However, in THIS test, given a
good set of headphones, the point is to see how much additional phase
distorsion can bring on, that is, is it real. Now, once you have established
that phase distorsions is audiable, we must start to work with better and
better tools to establish the ground-floor of sensitivity for the human ear,
and only then we really need the absolute numbers, not the relative numbers
that we are experimenting with here. The trick is that we need to have a system
with truely linear phase in the correct measurement situation (that is, as we
hear it). Measurement and correction of phase on a system is possible, but I
don't think we need those advanced tricks just to show that we are moving in
the right direction. What I will say is that a true A/B test should make the
test-subject (you) totally unaware of which version is which and what is played
when (preferably). The best thing is to run a random sequence of A and B and
ask for judgement each time, but another way is to let the subject themselfs
know when they hear A and when they hear B, maybe even choose A and B at their
own time, but they must listen X times to either of them.

Audiatorial sensation is a mess, bringing it to crystal clear science is hard.
I gave up on the audiophiles AND "HiFi-world" to be franc.

So, I hope I have contributed somewhat to your series of experiments. I don't
have the propper tools to do any good tests myself (mainly too bad DACs and
supporting system). I can measure amplitude and phase responses quite
accuratly on a real-time system.

Cheers,
Magnus