Is there any equalizer with a zero delay feedback filters?

[AXP]

In my understanding, the non-zero delay feedback is when you explicitly introduce extra unit delay into the feedback path to simplify the calculations.

For example, consider a system:
output[n] = function(input[n], output[n])

if <function> is linear and invertible you can calculate the output algebraically:

output[n] = function2(input[n])

However if the function is nonlinear, there are 2 approaches:

1. introduce a unit delay into the feedback path:
output[n] = function(input[n], output[n-1])
Since the <function> remained the unchanged, the overall system behavior changes (different phase and frequency response)

2. Solve the system by numerical iterative methods.
This is the "0df" method. The drawback is the computational complexity.

Correct me if I'm wrong.

[docdued]

I found this link pretty informative: http://www.xils-lab.com/pages/Zero-Dela ... lters.html

[DaveGamble]

Are we talking about filters with nonlinearities in the feedback path? Like this article does: http://images-l3.native-instruments.com ... xNLZDF.pdf?

(snip)

Please correct me if I'm missing a point.

For nonlinear systems it's pretty obvious that it's tempting to introduce a unit delay into the feedback path to break it up into separate nonlinear and differential parts.

Absolutely as you describe.

Dave.

[DaveGamble]

I don't quite follow.. Wasn't the subject about zero feedback delay filters? Don't all IIR's by definition have a delay in their feedback structure?

Yes, I'm a bit puzzled too (sorry Dave )

Anyhow, it's of course possible to eliminate the unit delay in EQ feedback paths too. Might make a difference at high q setting, but maybe not as big a difference as fast modulated synth filters.

Well, it depends. If you're limiting yourself to conformal mappings then yes, you should worry about these things.

I haven't thought that way about EQ in a decade, so it never really bothers me.

The "eliminate the unit delay in EQ" strategy basically leads to some form of oversampling.
It's unnecessary - you can do just fine, and in fact achieve greater precision without oversampling.

Everything I just said applies STRICTLY to the linear case.

Dave.

[DaveGamble]

So when did zero delay filters become about correcting some kind of time delay? All I know about it is the feedback is solved before anyone cares.

again-can you show me this delay on dirac?
or...can you explain the details?
if not...I believe in Dave

Exactly. In a linear system, you can solve the feedback equations algebraically. Always.

This is the way it's always been done. No-one has been introducing unnecessary sample delays into EQ designs. The EQ shape would literally fall apart if you did.

On the other hand, if you're moving AWAY from the NI definition of 0df, and instead want to treat z like it's s, then:
1. Good luck. You'll need a lot of oversampling.
2. I can do better without any oversampling just being a lot more judicious about my coefficients.
Time and time again, people measure my stuff and this point is proven.

Cheers!

Dave.

[Urs]

I don't quite follow.. Wasn't the subject about zero feedback delay filters? Don't all IIR's by definition have a delay in their feedback structure?

Yes, I'm a bit puzzled too (sorry Dave )

Anyhow, it's of course possible to eliminate the unit delay in EQ feedback paths too. Might make a difference at high q setting, but maybe not as big a difference as fast modulated synth filters.

Well, it depends. If you're limiting yourself to conformal mappings then yes, you should worry about these things.

I haven't thought that way about EQ in a decade, so it never really bothers me.

The "eliminate the unit delay in EQ" strategy basically leads to some form of oversampling.
It's unnecessary - you can do just fine, and in fact achieve greater precision without oversampling.

Everything I just said applies STRICTLY to the linear case.

Dave.

Well, if you take IIR topologies commonly described in DSP literature such as the direct form biquad filters, the Chamberlin SVF or the Stilson Smith Ladder, then clearly those introduce unit delays aka z-1 blocks in the feedback path. The same accounts for all implementations of Sallen Key I have found in description or source code on the net.

Of course it's fairly easy to eliminate this unit delay in the linear case without oversampling for any analogue topology, but not so much for biquad filters. There are papers about this (e.G. Fontana IIRC), but I don't think these methods are commonly applied.

Recently I would say considerable efforts have been made to eliminate the unit delay even in the non-linear case. These have nothing in common with oversampling, they just rquire different methods of integration. Nevertheless, as nonlinearities are introduced, oversampling is advisable.

[33tetragammon]

http://blog.dmgaudio.com/

It will be called Equilibrium and it's feature set is looking stunning.

I believe the beta is imminent.

There will be circuit models of many great analog EQs (API, SSL, Neve, Harrison, Focusrite, Pultec etc)

Multi channel support, flexible phase, low latency analog phase modes, parallel modes, every conceivable filter type and loads of awsome mouth wateringly cool mastering features. Dave hasn't even had enough space or time to let us testers know the full feature set yet!

Can't wait for this mutha!

STOP IT!!!!!!!!!!!
I'm getting severe GAS symptoms just by reading your post! Circuit models as well?!?!?!

Credit card ready........bring it on, Dave...... I'm in need of a 'Compassionate' EQ.....

In Dave we can trust.......

Sorry for the derail all, got too excited!