KVR Audio

tony tony chopper · Post by **tony tony chopper** » Thu May 31, 2012 7:17 am

& how is a filter "zero delay" anyway, I mean, make an ideal filter in the freq domain, how wouldn't it have a delay considering that you need quite some samples to define a frequency, and that any change you make will obviously have to be ramped. I think it's a bad name, for something that's ramped, thus delayed by design.
(the other question being how does it sound, as I haven't heard anything special so far)

stratum · Post by **stratum** » Thu May 31, 2012 7:25 am

I guess some folks have identified the 1-3 samples of delay found in IIR's to be a source of a few other things that they dislike and that the delay itself is not a concern. At least that's what the xils-lab link above is about.

helium · Post by **helium** » Thu May 31, 2012 7:33 am

tony tony chopper wrote:& how is a filter "zero delay" anyway, I mean, make an ideal filter in the freq domain, how wouldn't it have a delay considering that you need quite some samples to define a frequency, and that any change you make will obviously have to be ramped. I think it's a bad name, for something that's ramped, thus delayed by design.
(the other question being how does it sound, as I haven't heard anything special so far)

This thread is about having no delay in the feedback path. In most digital implementations you just take the last output for the feedback (i.e. delayed by one sample). In analog implementations the output is directly connected to the input being (almost) immediate. Thus there have been several discussions on here about how you could do this in software either analytically or numerically.

stratum · Post by **stratum** » Thu May 31, 2012 7:42 am

So the thread is about the feedback delay of a recursive filter, and not about latency. Thanks for clarifying.

tony tony chopper · Post by **tony tony chopper** » Thu May 31, 2012 8:41 am

This thread is about having no delay in the feedback path.

Yes I remember now, from a past thread

however is there really a good demo of this in action?
(there was this thread that starts very wrong so I didn't read further
http://www.kvraudio.com/forum/viewtopic ... sc&start=0)

I also read somewhere about it being about the resonance being precisely on each harmonic & not in-between, as the filter's freq is changing. However this is something I considered a problem when filtering in the freq domain (& widened my fake resonance bumps on purpose to avoid it)

Richard_Synapse · Post by **Richard_Synapse** » Thu May 31, 2012 10:01 am

It's not just about resonance. Generally speaking, classic digital filter design is trying to mimic only the frequency response of an analog system. The topology of the system and it's nonlinear behavior is lost. Despite this flaw, most DSP literature is not even easy to understand with all this s-plane and z-plane stuff. Most authors just throw the final equations at you, without bothering to explain where they come from. Everything gets simplier when looking at the differential equations instead, which operate entirely in the time domain.

Richard

stratum · Post by **stratum** » Thu May 31, 2012 10:42 am

Most authors just throw the final equations at you, without bothering to explain where they come from.

I guess we need to invite Steven W. Smith to write a book about this VA filter problem:)

I've just seen Vadim Zavalishin's book, and it seems to start right at the place where Steven Smith's dspguide book ends: Bilinear transform. For me that seems to be the next cool book to read.

Ichad.c · Post by **Ichad.c** » Thu Jun 14, 2012 11:24 am

Final Count: 43% Yes. 56% No. 1% Rounding Error

Was actually expecting someting in the ballpark of 35% Yes/65% No. Obviously I'm in the company of smart people. Thanks to everybody that voted.

Now it's time for the 56% to catch up.

Regards
Andrew

KVR Audio

Poll - can you build a zero-delay filter?

Can you build a zero-delay filter?