[sdiy] SVF & phase

[Magnus Danielson]

From: Ingo Debus <debus at cityweb.de>
Subject: Re: [sdiy] SVF & phase
Date: Wed, 27 Nov 2002 15:38:33 +0100

> patchell wrote:
>  >     One thing for sure, you will have a difficult time making a 
> filter with
>  > zero change in phase.  You can make one that has a constant (or near
>  > constant) time delay vs frequency, but that is about it (time delay and
>  > phase shift are related)..  Your only hope may be in the digital 
> domain,
>  > using finite impulse response filters (a subject I am not at all 
> familiar
>  > with), but even with that, I am doubtful.
> 
> It is possible to build an analog FIR filter. Use S&Hs instead of the
> data latches in the digital version, and attenuators/amplifiers for
> the coefficient multipliers. It's still time-discrete, but analog.

Yes, but there is another variant, which is time-continous... if you have a
delay-line with multiple taps, you only need to mix the taps together according
to the coefficients. Now, that sounds theoretic and not very usefull, but thats
the basis for what is done in Surface Acoustic Wave (SAW) filters and in
dispersion compensation fibres, with the modifications that there you use up
reflections in the acoustic and optical case respectively.

> A special case of this is a BBD ;-)

Well... yes... maybe that's not the best example then... ;O)

>  >>what are some options for building a state-variable filter with a
>  >>better-than normal phase response ?
> 
> I'm wondering, what is "better" or "worse" regarding the phase
> response of a synthesizer filter?

The actual goal should really be formulated as having a minimal impact on the
sound, other than that modification you intended (amplitude modification).
Weither phase changes or close to no phase changes are better or not (or even
unrelated) is however something one has to discover, but it is not given as
being assumed in the original question. This is the difficult thing, separating
out the actual requested property and the underlying assumption that led to the
formulation used in the question.

Actually, when you come to think about it, sometimes you actually want a
particular modification in the phase AND amplitude response at the same time
since they should counteract that of some defect. This is how you ideally
should apply parametric equializers, to remove or at least move the defect such
that the end result is closer to some "ideal". Graphical EQs are quite horrible
when you come to think of it, but they are easy to use and understand.

A 30-band EQ has 60 poles and 60 zeros in total. It's a good exercise in
understanding the tool to figure out what parameters is actually changed by the
30 pots. What is done to these 60 poles and also the 60 zeros?

A full-parametric EQ of 3 bands have 6 poles and 6 zeros, but has 9 controls,
how does these controls change the 6 poles and 6 zeros?

It is quite interesting to learn these things, you tend to shift your way of
thinking when applying these things.

Cheers,
Magnus - fan of fullparametric EQs, so I've designed one!