The aliasing thread

[Sascha Franck]

Try replacing those old EXS tracks (ALL of them) with SFZ doing the playing. On second thought, maybe you shouldn't. You might be upset.

Fortunately that won't work anyways, as I had to convert all the patches first which might bring up a whole lot of problems.

[WilliamK]

We have just updated Wusikstation with a new anti-aliasing code. And we are working on a 3rd option too. So please, George, remove that Wusikstation graphic as it is not relevant anymore. (we already asked several times)

We are currently beta-testing V1.0.6 with Hermite-Interpolation and studying Sinc-Interpolation for future implementation.

"Wusik Dot Com would like to ask sorry for WilliamK's behavior. He was only trying to prove a personal point about aliasing. Somehow, a few users got the wrong impression of him and started to flame him without any consideration. We have removed all posts to avoid any extra confrontation. And we would like everybody to forgive us for any problem caused by this."

I'm sorry George and Rene for the whole confusion.

"We would also like to state that we never said SFZ had "poor" quality. We just don't know where Rene got this from us, as we did not say that. This post was about aliasing, and that was what we were talking about here, until things got off-hand."

I just want to throw a question. How come the largest companies on the market, like NI, Steinberg, ... produced programs that has so poor anti-aliasing codes? Not putting more flames to the topic, but since several developers joined this discussion, I also feel that I should put something so people can talk about.

Best Regards, WilliamK

[autloc]

Renoise is a bit curious. If the graphic is to be believed, it is the only sampler to reproduce a sample with (significant) distortion even when the pitch hasn't (??) been changed. Does this indicate a totally different algorithm (some parametric sound representation, granular/additive synthesis or a phase vocoder), I wonder?

[Big Tick]

Wusik Dot Com would like to ask sorry for WilliamK's behavior.

Hmmm. Isn't Wusik.com a one-man company owned by WilliamK ?

[george]

We have just updated Wusikstation with a new anti-aliasing code. And we are working on a 3rd option too. So please, George, remove that Wusikstation graphic as it is not relevant anymore. (we already asked several times)

http://www.kvr-vst.com/forum/viewtopic. ... 404#707404

"We would also like to state that we never said SFZ had "poor" quality. We just don't know where Rene got this from us, as we did not say that. This post was about aliasing, and that was what we were talking about here, until things got off-hand."

All your previous posts have been removed so...

How come the largest companies on the market, like NI, Steinberg, ... produced programs that has so poor anti-aliasing codes?

It has been discussed already. Interpolation quality is one of _many_ features of the product. Companies like Image Line, Synapse Audio or Propellerhead actually take care of this.

[René]

Well, in a sampler, just like in a synth, you can use a trick: If you algorithm has, say, 16 samples latency, you simply calculate 16 samples more at note start and discart the first 16 samples. You simply process with some look ahead

The only thing that screws it up is, the NoteOff will be processed later. (I guess that everybody here lowpasses other control signals anyway, so that 16 samples would be neglectable).

btw. I think I would prefer polyphase filters (IIRs) over FIRs when filtering. Works fine for me, with low latency.

That's correct Urs, I believe gol was describing it early in this thread. However, that only removes the note-on latency (unless you're deducting the plugin latency from the system latency). All modulation controls like bend, CCs, etc., and as you said, the note-off will remain delayed.

I mentioned the 16-sample value by random, real world values tend to be a bit higher: just as an example, a fir with 20kHz to 24kHz (allowing the Nyquist - 20kHz rolling back @ 44.1k), with 3dB of passband ripple and 96dB of stopband attenuation requires over 65 taps, about 0.7ms of delay. At higher samplerates (88.2k, 96k), things get worse.

On the other hand, I don't believe this is the worse drawback of oversampled systems. I believe CPU usage vs. quality ratio is. But they work very well.

-René

[WilliamK]

Wusik Dot Com would like to ask sorry for WilliamK's behavior.

Hmmm. Isn't Wusik.com a one-man company owned by WilliamK ?

Not really, please, check the About page of our site.

Thank you. WilliamK

[george]

Renoise is a bit curious. If the graphic is to be believed, it is the only sampler to reproduce a sample with (significant) distortion even when the pitch hasn't (??) been changed. Does this indicate a totally different algorithm (some parametric sound representation, granular/additive synthesis or a phase vocoder), I wonder?

Interesting question. I have noticed Reason samplers are also adding additional harmonics too (they shouldn't), although, I didn't made any of those. I should bother to download the demos and check it out by myself, though.

[Sicklecell666]

What's with this 'Dev kill Dev' shit?



It's really easy to elevate yourself above your competitors by chopping thier frichen legs off at the knees..

I love seing Human weakness & disparity at work..

[Big Tick]

Wusik Dot Com would like to ask sorry for WilliamK's behavior.

Hmmm. Isn't Wusik.com a one-man company owned by WilliamK ?

Not really, please, check the About page of our site.

"Wusik Dot Com is the creation of WilliamK." So having the company apologize for your words sounds, hmmm, kinda weird. Anyway.... I don't want to get in that thread anyway. I don't know anything about dsp. But I know enough, not to confront René when it comes to vsti coding.

[george]

What's with this 'Dev kill Dev' shit?

It's called product feature comparison, it's not shit and if you haven't read the "full" thread already I'd recommend to do so.

[Markleford]

If it sounds and feels good, it is good.

Again I fall into line with Sascha: I wonder how many people actually cared about this before threads like this popped up and frequency analysis web sites were touted? If it sounds and feels good, it *is* good! Well, until someone convinces you otherwise, so it seems... ("Does it sound like a Virus?")

Sure, if your song is crap then people may pay undue attention to the quality of your sample playback. I guess the trick is to make them more concerned about little things like rhythm and melody instead of spectrum graphics.

Then again, I must have a tin ear. I have sfz+ and don't think I've been able to hear a quality difference in the settings, so I generally leave it on the lowest to save CPU. I know I should put it on highest quality when rendering down, but I forget because I honestly don't notice when all is said and done. Maybe it's because I typically use SoundFonts with at least 4 samples per octave: who's to say?

Doesn't stop me from using sfz+ though: it's chock full of amazing features, even if the sound quality is what's touted the most. I love it.

I also have Wusikstation and it's definitely my favorite synth. I hadn't even considered aliasing problems with it, even when using my FFT resynthed K1 wavetables with it, even at one per octave. Again, though, I'm just not an audiophile: I'm a musician. (I realise there are lots of people that are both, though. More power to you.)

Or maybe it's just because my monitor speakers don't have the frequency range to make aliasing in my stuff apparent. You'll have to go through my songs in the link below to point out that stuff to me.

Oh, and as full disclosure, I also use Reason as my main synth rack through ReWire. NN-XT is my fave sampler, simply because of its surprising depth, ease of use, and friendly CPU consumption. However, people always complain about how "weak" Reason sounds, but I'm at a loss there, too. Oddly, nobody has complained about that in any of my songs, though. Hrrrmmm...

I suppose ignorance is bliss. If I'm dreaming, then please don't wake me up: I'm having fun!

Anyway, WilliamK is a nice guy, but he sure can get excited. Yeah, we like passionate developers, and sometimes they can be baited into saying things they regret (I'm also still a fan of Art@Bitshift: so sue me!). I know William will make things right, technically, and will make steps to try to repair the political mis-steps he made here if people give him the chance.

To William directly:

Don't worry so much about the pretty color spectrum graphics! People will buy your synth anyway. And when in doubt, blame cultural and language barriers on your mistakes!

- m

[george]

Sure, if your song is crap then people may pay undue attention to the quality of your sample playback. I guess the trick is to make them more concerned about little things like rhythm and melody instead of spectrum graphics.

I agree, but on the other hand, we can also have a superb song being craped out by not using some decent interpolation too.

Or maybe it's just because my monitor speakers don't have the frequency range to make aliasing in my stuff apparent. You'll have to go through my songs in the link below to point out that stuff to me.

Just check this quote:

If someone likes the sound of non bandlimited samplers and wants to use the sound of them in a song/track as a solo instrument, thats no problem, it's a question of taste. But fitting an aliased sound in a mix lowers the dynamic range only. It strengths the amplitude of fundamentals you don't need at every part of the spectrum. So if the sound has to be the best in the mix, with the bigest dynamic range, then the best solution would be a sampler with perfect band limiting.

Oh, and as full disclosure, I also use Reason as my main synth rack through ReWire. NN-XT is my fave sampler, simply because of its surprising depth, ease of use, and friendly CPU consumption. However, people always complain about how "weak" Reason sounds, but I'm at a loss there, too. Oddly, nobody has complained about that in any of my songs, though. Hrrrmmm...

Because Reason sampler has got a decent interpolation.

Anyway, WilliamK is a nice guy, but he sure can get excited.

Yes, excited enough for a lawsuit and publicly talk about taking a site down It's called harassment, though. I'm sorry for William fans, but that's just wrong.

[pough]

If it sounds and feels good, it is good.

Again I fall into line with Sascha: I wonder how many people actually cared about this before threads like this popped up and frequency analysis web sites were touted? If it sounds and feels good, it *is* good! Well, until someone convinces you otherwise, so it seems...

Here's a test that's fun and easy to do. Load up that 15kHz wav file into WusikStation (or something else that has a linear interpolation mode) and play every note on the keyboard. Notice how some of the keys play notes that seem out of place. Way out of place. If I'm not mistaken, this is because of aliasing - frequencies too high to be represented properly can become much lower (and quite incorrect) frequencies.

Of course, there are (as far as I can tell) two very important things to remember. One is that this is likely to happen only in some circumstances (like not very many samples in a sampleset) - and probably only on the harmonic overtones of a given sound. The second is that, while aliasing can be in important issue (particularly for geeks), it is but one issue of many.

I think I can see the point of those who find the aliasing comparison site alarming. If it's not such a big deal, why is there a large area of a website dedicated solely to this issue? That makes it seems as though it's some kind of conspiracy ("The Big Guns of music are trying to make Your Music sound Bad!") being revealed. But like all conspiracy stuff, it ignores the more likely scenario, the simpler one: it's just some geek who finds some technical issue interesting.

Everyone just needs to calm down; some more than others, of course. Find something relaxing to do for a while. As for me, I'm about to fly into a hurricane!

[george]

If I'm not mistaken, this is because of aliasing - frequencies too high to be represented properly can become much lower (and quite incorrect) frequencies.

Not really, since a sample can suffer from aliasing no matter what range is played (except its root frequency). Why a 15Khz sine tone is used was previously explained.

The second is that, while aliasing can be in important issue (particularly for geeks), it is but one issue of many.

I disagree about this assumption, since there isn't any real proof about aliasing being a geek issue mostly. Take a (insert your non geek musician here), explain him the basics of aliasing phenomena (properly, of course) , then put him in front of a good pair of studio monitors and play something using linear interpolation, once finished switch to some of the high quality samplers listed.

I'd be curious to know many average musicians will notice and then take care about it, AFAIK there is no statement about this being a number-one-must-have-it-feature. As said, ignorance is bliss, but are you and your (good) music getting any benefits from it? At the moment we can only assume things instead having a solid proof.

If it's not such a big deal, why is there a large area of a website dedicated solely to this issue?

The reason is stated at the page... anyways, from a developer side, it is something to take care about. And by looking at the recent high quality products there, I'm not afraid to say it isn't a trivial thing. Best example is VSampler who managed to raise from "poor" to "high quality" within the time.

That makes it seems as though it's some kind of conspiracy ("The Big Guns of music are trying to make Your Music sound Bad!") being revealed. But like all conspiracy stuff, it ignores the more likely scenario, the simpler one: it's just some geek who finds some technical issue

That conspiracy thing sounds a bit paranoid to me And again, we are not aware if the issue concerns to "some geek" only.