KVR Audio

aciddose · Post by **aciddose** » Thu Jul 30, 2015 9:29 am

Even though as Urs, you say "Eliminating this delay and thus creating a "zero delay feedback filter" sounds better", you must admit that the words "eliminating this delay" are misleading.

What we're doing is more closely approximating where the value might actually be, assuming there was no such delay. Eliminating a delay requires time-travel which we have not yet invented.

Only in cases where the equation can be fully solved without feedback can we truly eliminate the delay. In those cases again, "zero delay feedback" is a misnomer. What we have instead is "feed forward" or "feedback eliminated"

To everyone else: this is a simple, self obvious truth. You can't compute something that depends upon the result of the computation. Each time you compute a new value, the value you would compute changes. We do have methods to cope with this but they are approximate.

Assuming anyone here knows math, an example of this problem: https://en.wikipedia.org/wiki/Navier%E2 ... _equations

See the discussion about the fact there is no proof that a solution exists at all. Many filters exhibit this same "chaotic" behavior and it is not possible to prove a solution exists to the "feedback problem".

Urs · Post by **Urs** » Thu Jul 30, 2015 9:37 am

aciddose wrote:Even though as Urs, you say "Eliminating this delay and thus creating a "zero delay feedback filter" sounds better", you must admit that the words "eliminating this delay" are misleading.

What we're doing is more closely approximating where the value might actually be, assuming there was no such delay. Eliminating a delay requires time-travel which we have not yet invented.

Only in cases where the equation can be fully solved without feedback can we truly eliminate the delay. In those cases again, "zero delay feedback" is a misnomer. What we have instead is "feed forward" or "feedback eliminated"

You're wrong and there is no need to deploy time travel. All you need is high school math, a bit of perspective and a standard root finding algorithm, such as the bisection method.

aciddose · Post by **aciddose** » Thu Jul 30, 2015 10:01 am

Yes, but the question here isn't whether you can use a root finding algorithm to make an approximation or even to solve an equation. It is whether it is accurate in any way to use the term "zero delay feedback".

If you want to claim I am wrong about this, that is fine. As far as I am aware though there is a major problem with the term "zero delay feedback".

Root finding algorithms can be used to approximate the value resulting from feedback assuming the delay were zero. None are direct solutions which will compute an exact value. To claim they "solve" an equation is much like claiming you have fully computed the value of PI.
In using such algorithms we haven't actually eliminated the delay resulting from discrete-time feedback. z[0] isn't known until afterward, so either z[-1] or an approximation of z[0] is used.

Due to that, to claim it is "zero delay feedback" is a contradiction. It is either discrete-time feedback (z[-1] == a delay) and therefore delayed, or it isn't discrete-time feedback. It can't be both.

The contradiction is the issue, it is extremely simple, simple enough for anyone to understand.

Urs · Post by **Urs** » Thu Jul 30, 2015 10:19 am

LOL. We don't need more than 7 digits of PI to get better signal to noise ratio than CD (or analogue circuits fwiw). Even the bisection method is fast enough to bring an estimate and a result of a State Variable Filter model to convergence. In realtime. And that is with 3 OpAmps using standard formulas from circuit modulation packets.

What are you trying to win here? - We've done this 4 years ago, it's tried and tested and proven and all that. It's quite lazy to come along and imply that I'm a liar when in fact you have not a single clue of what we've done. Wanna hear the difference? Try Diva's "Draft" mode versus "Fast". One has a unit delay in the feedback path, the other doesn't - and that is the *only* mathematical difference between the two. One uses a delayed input for resonance, one converges an estimated resonance input with the resulting output, down to 5 or so digits behind the comma.

Gawd, this is tiresome.

(note the "Uhbie" filter always uses zero delay feedback, even in draft mode - the delay has such a devastating effect on that one, the filter can explode if we leave it in)

H-man · Post by **H-man** » Thu Jul 30, 2015 10:30 am

I'm starting to lean toward the term "approximated value similar to zero delay feedback filter"

It's got a ring to it

aciddose · Post by **aciddose** » Thu Jul 30, 2015 10:30 am

Nice way to brush off the fact that it is a poor choice of words and technically inaccurate to call it "zero delay".

I'm happy though Urs, you've admitted it although in an awfully clumsy way.

There is nothing to be ashamed of! Yes, the method is great. It solves a lot of problems. It just isn't "zero delay". That would be a time-machine.

Full Bucket · Post by **Full Bucket** » Thu Jul 30, 2015 10:32 am

Correct, the right term is "Zero Delay Feedback Filter".

aciddose · Post by **aciddose** » Thu Jul 30, 2015 10:36 am

No, "zero delay" and "feedback" are contradictory. What it is, is a filter that approximates its state without using discrete-time feedback.

The term is a misnomer!

Anyway as far as the topic of this thread is concerned, no customer would actually understand what we're talking about here so as far as marketing blurbs go ahead, use a technically inaccurate term. Along with all the other terms in your marketing blurbs...

H-man · Post by **H-man** » Thu Jul 30, 2015 10:39 am

Back on topic, is anyone else out there concerned that there is now a drum machine with a soul!?

How the F$#k did they do that??? Maybe it's an approximation?

aciddose · Post by **aciddose** » Thu Jul 30, 2015 10:42 am

It steals the soul from the user with its bad interface.

Rcl · Post by **Rcl** » Thu Jul 30, 2015 11:00 am

aciddose wrote:It steals the soul from the user with its bad interface.

Blahhhhhghhhh!!!!!!!!!!!

aciddose · Post by **aciddose** » Thu Jul 30, 2015 11:04 am

Rcl wrote:Blahhhhhghhhh!!!!!!!!!!!

https://youtu.be/VJh9W3Gcpmo?t=103

H-man · Post by **H-man** » Thu Jul 30, 2015 11:10 am

Rcl wrote:
aciddose wrote:It steals the soul from the user with its bad interface.
Blahhhhhghhhh!!!!!!!!!!!

The drum machine with your soul?

aciddose · Post by **aciddose** » Thu Jul 30, 2015 11:13 am

Don't take my comment as honest critique, I have no idea which drum machine this is. I was merely speculating about possible methods by which it might extract a soul for itself.

Alternative:
http://www.balancedenergies.com/shamani ... ction.html

Symptoms of needing soul extraction:
The list of symptoms contains everything.

H-man · Post by **H-man** » Thu Jul 30, 2015 11:26 am

Actually I think it's more along the lines of:

http://www.imdb.com/title/tt0092563/

Daftest synth plugin press release blurb ever?