Filters and samplerates

[soapform]

Would it be possible to design more sample-rate-invariant filters?

For example, observe the drastic difference in the factory preset drums\Closed HH 1 at different sampling rates (44.1-192 khz).

These kinds of things bug me (OCD?) when I start paying attention to them. I've found this same issue in other synths too... but then again I've heard synths with filters that behave exactly same (regardless of rate), so maybe it's possible in Sonigen as well.

thanks,
soapform

[sonigen]

Would it be possible to design more sample-rate-invariant filters?

It's a bug, it's not always picking up the samplerate correctly, so any keytrack or pitch will be off when not at 44100. It'll be fixed in the next beta.

[soapform]

Would it be possible to design more sample-rate-invariant filters?

It's a bug, it's not always picking up the samplerate correctly, so any keytrack or pitch will be off when not at 44100. It'll be fixed in the next beta.

Nono, at least here the keytrack and pitch are correct. I'm talking about filter behavior when the cutoff is near nyquist.

[sonigen]

Nono, at least here the keytrack and pitch are correct. I'm talking about filter behavior when the cutoff is near nyquist.

Hmm, ok.

Having thought about it the filters do have higher cutoff limits with higher samplerate. And the patch you mentioned has the cutoff and envelope mod maxed out. So that would definitely sound different. The only way to make that consistent is to have a fixed maximum upper cutoff no matter what the samplerate.

Obviously this would need to be the lowest common denominator, say around 20khz. It could be available as an option I suppose.

Well it could be done with oversampling, but it's a big cpu cost for a small issue like that.

[soapform]

Hmm, ok.

Having thought about it the filters do have higher cutoff limits with higher samplerate. And the patch you mentioned has the cutoff and envelope mod maxed out. So that would definitely sound different. The only way to make that consistent is to have a fixed maximum upper cutoff no matter what the samplerate.

Obviously this would need to be the lowest common denominator, say around 20khz. It could be available as an option I suppose.

Well it could be done with oversampling, but it's a big cpu cost for a small issue like that.

Ah, it makes better sense to me now.

Then I just have to pay attention where the cutoff points are, if I want to make compatible patches. I don't like the idea of oversampling or hard-coded limits... (hmm, maybe an indicator light saying that a cutoff is over 20khz and therefore may sound different...?)

[nineofkings]

20khz is the highest frequency you can really hear anyway so I think it's a good limit, especially considering any frequencies above it are attenuated, not completely blocked out.

[SoulState]

I think you don't need to do anytning with it.
If you want higer sound quality - just make all projects and presets in 96k, who now works in 44100?
I definitely against hardcoded cutoff of 20k for all samplerates, with this attenuation sound lacks "air", it is the main thing why i working in 96k.

And synths which sounds all the same in 44 and 96k - sounds like crap, generally.

Most of vst synths sounds way better in 96k or higher, because of oversampling, and fx sounds better especially reverbs.
So this is normal i think.

[EvilDragon]

I see no wrong in having oversampling... Even if it's such a "small" issue.

Optional oversampling, and optional max cutoff frequency would be cool.

Of course, the patch made at 44.1 kHz should have the same cutoff frequency at 192 kHz. In the other direction, it should behave the same, except oversampled.


Oversampling is good, and today's CPUs are plenty powerful, so why not use it?

[sonigen]

Oversampling is good, and today's CPUs are plenty powerful, so why not use it?

Because in this case it'd be using a sledgehammer to crack a nut. It would more than double the CPU cost for something few people will ever notice and i think most people do still care about cpu usage.

[nineofkings]

I definitely do. I see no problem with a hard-coded limit of 20 khz, or maybe 22 khz if people decide to be picky. Anything above is inaudible, so unless you're playing music for cats I think that that would be a perfectly acceptable limit.

[soapform]

I definitely do. I see no problem with a hard-coded limit of 20 khz, or maybe 22 khz if people decide to be picky. Anything above is inaudible, so unless you're playing music for cats I think that that would be a perfectly acceptable limit.

I have to take back my earlier statements, I really don't know the right way to do things, whether to oversample, limit, or do nothing at all. It seems VST inherently has this problem (things sounding different on different sample rates) that is not present in hardware. Regarding limiting the cutoff point, it would be funny if somebody still worked at 22050hz or lower, so even that could become a problem. But the interesting thing about Sonigen is, every other part of the synth seem essentially sounds the same to me regardless of sampling rate, except filters.

It's an interesting problem. I would have thought perhaps there is a "standard solution", "industry recommendation", or something.