Zero Delay Feedback Filter (How to test if your synth has a )- Xils-Lab White Paper -

[Lotuzia]

I just tried the hack method lotty posted, seems Steinbrrg Neon had 0DF

So te test is only 5 very simple steps : I'll put it here again as a c/p from the first page :

" The Test: You'll need a synthesizer (software of hardware) with a self-oscillating filter and a noise generator.

1. Open the noise generator level and mute the oscillators.

2. Set the filter frequency to a very noticeable value (500 Hz for instance)

3. Set the emphasis to a mid value (not to high, but enough to be noticeable)

4. Sweep the filter frequency toward the highest value.

In Zero Delay Feedback filters, the emphasis will remain the same, but in filters that do not manage this delay, the emphasis will grow until the filter self-oscillates. "

On the 2 first lines you'll see that you only need a synth equipped with a noise generator. Then you just have to mute the oscillators.

Does Neon has a noise oscillator ? No. Can you switch off the oscillators on neon ? No ( you can only choose one ). So what did you test ?

So a guy can't read the first two lines of a 5 lines test, and then declares erroneous things, showing in the same time that he doesn't understand a single thing about 0df problem and how it translates into audio, or a test.

What is more curious is that a developer, who has given his POV very loudly in this thread, and who is supposed to have more insights about all these problems cheers him for declaring so, without verifying the information himself.

To sum up : Try again, or eventually ....... just learn to read

[AdmiralQuality]

I just tried the hack method lotty posted, seems Steinbrrg Neon had 0DF

Hi,

i just thought this is a joke but i tried with the now free version of Neon and you seem to be right...
Of course this synth is so old that it should not include any 0df filter design.

I already recognized that for a few synths like e.g. Diva, Saurus and Xils Lab synths the result seems to be quite clear but also found it could be very unclear for some others.


Ingo

It's not a joke. Lotuzia's "test" criteria are complete bullshit. That's why all the developers who've come in here are unanimous that it's bullshit. (Really, how often do we all agree on anything? ) And if something looks like bullshit, smells like bullshit, and tastes like bullshit, you should be damn glad you didn't step in it!

[AdmiralQuality]

What is more curious is that a developer, who has given his POV very loudly in this thread, and who is supposed to have more insights about all these problems cheers him for declaring so, without verifying the information himself.

At least I'm not intentionally spreading lies for the sake of advertising my products in a thread disguised as helpful advice.

[Lotuzia]

I just tried the hack method lotty posted, seems Steinbrrg Neon had 0DF

Hi,

i just thought this is a joke but i tried with the now free version of Neon and you seem to be right...
Of course this synth is so old that it should not include any 0df filter design.

I already recognized that for a few synths like e.g. Diva, Saurus and Xils Lab synths the result seems to be quite clear but also found it could be very unclear for some others.


Ingo

It's not a joke. Lotuzia's "test" criteria are complete bullshit. That's why all the developers who've come in here are unanimous that it's bullshit. (Really, how often do we all agree on anything? ) And if something looks like bullshit, smells like bullshit, and tastes like bullshit, you should be damn glad you didn't step in it!

Really ? This is not what I read ....

[Lotuzia]

What is more curious is that a developer, who has given his POV very loudly in this thread, and who is supposed to have more insights about all these problems cheers him for declaring so, without verifying the information himself.

At least I'm not intentionally spreading lies for the sake of advertising my products in a thread disguised as helpful advice.

Really ? Thats curious I was under the impression that you were constantly promoting your very important self. But if you say so...

[ariston]

I just tried the hack method lotty posted, seems Steinbrrg Neon had 0DF

Hi,

i just thought this is a joke but i tried with the now free version of Neon and you seem to be right...
Of course this synth is so old that it should not include any 0df filter design.

TBH, Neon was the first synth I tried. I just didn't post the result because some people already thought I was weird for assuming ulterior motives behind this oh-so innocent test. I feel vindicated.

Not that that matters much, because it's still JUST about the sound, dammit!

Oops, broke my promise again.

[Howard]

Ltz is certainly an expert in the strategic placement of the shrug emoticon .
I must admit I prefer the French-typical shrug to the USA's "Just sayin'"!
(popcorn)

[AdmiralQuality]

What is more curious is that a developer, who has given his POV very loudly in this thread, and who is supposed to have more insights about all these problems cheers him for declaring so, without verifying the information himself.

At least I'm not intentionally spreading lies for the sake of advertising my products in a thread disguised as helpful advice.

Really ? Thats curious I was under the impression that you were constantly promoting your very important self. But if you say so...

You're under a lot of impressions. If I have news about one of our products, I'll make a thread about it clearly stating so. Not disguising itself as advice or general knowledge or a technique. If some factoid about a product of mine is relevant to a thread's topic or a question asked within, I'll mention it. And, being fair minded, I'll mention any other products I'm aware of that meet the same criteria. Like I did here, for just one example: http://www.kvraudio.com/forum/viewtopic ... 67#4891567 (Note my mention of ImpOscar along with Poly-Ana.)

I've explained, exactly the false assumptions you've made which make your test criteria incapable of determining whether a filter exhibits feedback delay. And I have the example of my own product, which I know doesn't, yet it still passes your "test". Therefore, your methodology is invalid.

So find me a thread I started that says other products are substandard because they don't do a particular thing that my product just happens to do. (What a coincidence!) And then suggesting some methodology to test for it that is a complete work of fiction. Fine me just one...

[Urs]

I don't understand why people insist on these zero-delay filters being expensive. You don't need to use iterative feedback solvers unless you want to. Just linearize the filter around it's known state once per sample and you can solve analytically. The overhead is pretty insignificant and it works quite fine in practice, at least until aliasing starts becoming a bigger problem anyway.

I'm not sure if this worked sufficiently in a filter loaded with non-linear elements, driven at high gain. You'll get quite some error, possibly noise. Some of our filters got fairly noisy with less than 14 bit precision.

The cost might even be similar to an iterative approach with a good prediction because you still have to build derivates for all non-linear elements, and you still have to calculate it once. If the average number of iterations is close to 1 then the iterative approach might be even faster.

[Lotuzia]

So find me a thread I started that says other products are substandard because they don't do a particular thing that my product just happens to do. (What a coincidence!) And then suggesting some methodology to test for it that is a complete work of fiction. Fine me just one...

We never said a product is substandard because they dont do a particular thing.

Actually we clearly stated the opposite at the end of the fist post. ( just reread it )

If your product passes the test ( provided the test is properly done ) its probably because you found a good method to emulate an analog filter behaviour.
And well its great

Properly emulating an analog filter IS taking care of the way the 1delay problem affect the sound in certain situations. ( and you said you were aware of this problem in the first post, long before Vadim talked about it, so it seems that the problem is real ) And finding a solution. Like we said there are probably many different solutions. No more, no less.

[Lotuzia]

Ltz is certainly an expert in the strategic placement of the shrug emoticon .
I must admit I prefer the French-typical shrug to the USA's "Just sayin'"!
(popcorn)

Thanks . Those who discuss about emoticons style are unvaluable contributors to this thread.

You're certainly an expert for commenting emoticons , and I'll honor you by taking some peanuts ( I'm not a fan of popcorn )

[Lotuzia]

Oops, broke my promise again.

Printed words last. Will can be more fragile.

[AdmiralQuality]

So find me a thread I started that says other products are substandard because they don't do a particular thing that my product just happens to do. (What a coincidence!) And then suggesting some methodology to test for it that is a complete work of fiction. Fine me just one...

We never said a product is substandard because they dont do a particular thing.

Actually we clearly stated the opposite at the end of the fist post. ( just reread it )

If your product passes the test ( provided the test is properly done ) its probably because you found a good method to emulate an analog filter behaviour.
And well its great

Like I said, not only does Poly-Ana pass the test, but at her lower quality settings she actually gets the inverse results (emphasis peak diminishes towards upper extremes of cutoff frequency). At higher quality (oversampling) settings, it's just about perfectly flat over the entire range.

So OF COURSE I found a good method, that's my job. As it is every other VA softsynth developer's job. If you happened to find some synth that does a bad job in this area, that is not indicative of any algorithm used in the filter. Only that the developer either wasn't aware it was a less than accurate emulation of classic synths' behavior, or because they didn't care, or maybe they lifted their filter code from some other resource and have no ability to manipulate it to their own taste. (Assuming they're lucky enough to possess taste.)

Properly emulating an analog filter IS taking care of the way the 1delay problem affect the sound in certain situations. ( and you said you were aware of this problem in the first post, long before Vadim talked about it, so it seems that the problem is real ) And finding a solution. Like we said there are probably many different solutions. No more, no less.

No it's not. You attest that it is. But as I've pointed out, using my own product as an example, it's an incorrect assumption to think that the sample delay in the feedback path is solely responsible for these artifacts, or that eliminating the delay through whatever technique is the only way to mitigate these artifacts. I don't NEED a 0 delay solution, as it wasn't a PROBLEM.

But there ARE other artifacts produced by the feedback delay that ARE determinable (albeit, it's VERY subtle). You've just failed to identify them. Your suggested methodology is meaningless garbage, and you're just confusing people. If myself or others seem a little perturbed, that's why.

[xavier]

I just tried the hack method lotty posted, seems Steinbrrg Neon had 0DF

Hi,

i just thought this is a joke but i tried with the now free version of Neon and you seem to be right...
Of course this synth is so old that it should not include any 0df filter design.

I already recognized that for a few synths like e.g. Diva, Saurus and Xils Lab synths the result seems to be quite clear but also found it could be very unclear for some others.


Ingo

It's not a joke. Lotuzia's "test" criteria are complete bullshit. That's why all the developers who've come in here are unanimous that it's bullshit. (Really, how often do we all agree on anything? ) And if something looks like bullshit, smells like bullshit, and tastes like bullshit, you should be damn glad you didn't step in it!

Hi

Could please explain with some mathematical evidences (not related to the capacitor and so one in the analogue feedback path of course), related to digital signal processing, z-transform and so-on, why this test is bullshit ?
I would learn a lot then (and I'm always opened to learn)

Anyway, I did the test with Neon.

First it doesn't self oscillate.
Second, the frequency (but maybe I missed something) doesn't go beyond 1500 Hz
So How did you do the test ? You are kidding or didn't read more than two words of our explanation.

This thread is suppose to be scientific and when someone is claiming something, some proof of evidence (video, mathematical explanations, physical equations, whatever) could help all of us to learn.

This was one of the best thing some physicians did some months ago : They discovered a particle which ran faster than light. All the Einstein theory would have been drop-down. So they kindly gave all their data to all the other physicians: They didn't just claim their results. (The other physicians found a error in the measure, Einstein was safe)

So if the test is bullshit, please give us mathematical evidences we will learn !

We tried to explain this test with some maths. No one found an theoretical error there. maybe you found something? Please tell us.

But there ARE other artifacts produced by the feedback delay that ARE determinable (albeit, it's VERY subtle)

Which one ?

Best regards
Xavier

[AdmiralQuality]

Hi

Could please explain with some mathematical evidences (not related to the capacitor and so one in the analogue feedback path of course), related to digital signal processing, z-transform and so-on, why this test is bullshit ?
I would learn a lot then (and I'm always opened to learn)

No. I've already explained it with logic. (It's because you can't make assumptions that the resonance coefficient isn't being tweaked by who knows what formula.)

Anyway, I did the test with Neon.

First it doesn't self oscillate.

As far as I understand, the test is supposed to be done well below the level of self oscillation. Which makes sense, as you're likely to run into a big non-linearity (aka clipping) when filters self oscillate. But again, it's an entirely arbitrary decision of the filter designer as to what level this clipping happens. Other non-linearities in the feedback path are common too. This is of course where we also experience the worst effects of aliasing, though that has nothing to do with feedback delay.

Second, the frequency (but maybe I missed something) doesn't go beyond 1500 Hz
So How did you do the test ? You are kidding or didn't read more than two words of our explanation.

The frequency? Of what? Anyway, it was Kriminal who said he did the Neon test. Not me.

This thread is suppose to be scientific

That's exactly my problem with it. It's not.

and when someone is claiming something, some proof of evidence (video, mathematical explanations, physical equations, whatever) could help all of us to learn.

No evidence is more relevant than empirical. Lotuzia's method would fail to identify the 1 sample (or sub sample) delay in Poly-Ana's filters. And I'm sure a whole bunch of other softsynths out there as well. (I'm certainly not claiming to have done anything particularly impressive. Other than notice and care. I literally solved these issues to my satisfaction within the first two days I ever spent writing audio plug-ins, with oversampling as well as (a bit later on) application of some corrective curves.)

This was one of the best thing some physicians did some months ago : They discovered a particle which ran faster than light.

The particle (neutrino) was already discovered. And their finding was incorrect. Because they rushed to the press before doublechecking their work. Bad form! (And the lead investigator on that project has since resigned.)

All the Einstein theory would have been drop-down. So they kindly gave all their data to all the other physicians: They didn't just claim their results. (The other physicians found a error in the measure, Einstein was safe)

No, actually, they found their own error. It was a loose cable! (Seriously.)

So if the test is bullshit, please give us mathematical evidences we will learn !

Math isn't evidence. If it was, loose cables wouldn't cause incorrect results.

We tried to explain this test with some maths. No one found an theoretical error there. maybe you found something? Please tell us.

But there ARE other artifacts produced by the feedback delay that ARE determinable (albeit, it's VERY subtle)

Which one ?

Best regards
Xavier

The main artifact is the frequency of of the resonant peak relative to the position of the beginning of the cutoff slope changes very slightly from where it should be with zero delay. (I've already said this! )

And sorry, since its so valuable to you, where is Lotuzia's math in this thread?