Aliasing on Arturia Plugins?

[bmanic]

I’m not sure if it’s aliasing, but the new Prophet 5 V’s audio rate mod via osc 2 sounds terrible.

Sounds like a Prophet 5 to me. Sounds better than Repro-5's audio rate moduation out-of-the-box. You can match them by adjusting tuning on Repro so they're nearly indistinguishable. In other words, the same thing you would have to do on two physical P5s.

Here's an A/B from the two of them back and forth. Saw wave modulated by an OSC2 square wave.

https://f001.backblazeb2.com/file/teilo ... pro-5.flac

Not sure which one is which, but one of them is "buzzing" a lot more. I would wager that's the Arturia one but I don't know for sure as I don't have nor have I tried Repro-5 (as I don't really like the sound of the real Prophet either so I never bothered).

[jamcat]

I don't think that's aliasing. It looks more like harmonics overlaid with "analogue modeled noise" to me. Aliasing has a telltale look to it where it doubles back on itself, as if you took a translucent picture and folded it in half.




The tall harmonic pattern would continue where it left off at 20kHz, but from right to left, so you would see distinct tall spikes increasing in volume interspersed between the decreasing ones that you see, hitting a peak of probably around -30dB near 20kHz. I'm not seeing that second layer of harmonics there.

It would look more like this:



Of course, it is possible they could just be hiding aliasing in the noise, which is essentially what dithering does with quantization error.

[xbow]

I don't think that's aliasing. It looks more like harmonics overlaid with "analogue modeled noise" to me. Aliasing has a telltale look to it where it doubles back on itself, as if you took a translucent picture and folded it in half.

The tall harmonic pattern would continue where it left off at 20kHz, but from right to left, so you would see distinct tall spikes increasing in volume interspersed between the decreasing ones that you see, hitting a peak of probably around -30dB near 20kHz.

But then how do you explain the harmonics (not noise) below the note played? I'm not an expert at all but it kinda looks like harmonics folding back from right to left.

[crickey13]

Definitely looks like aliasing on a harmonically rich waveform.

[Funkybot's Evil Twin]

Has anyone done the most important test: can you hear the aliasing?

[Ari-S]

Has anyone done the most important test: can you hear the aliasing?

Yes and yes. Compared to other softsynths, it was faintly audible. When I applied some distortion (with a plugin that oversamples, so no significant aliasing from the distortion itself) it became worse.

[Tj Shredder]

The best way to hear it is with pitch bend. If parts of the sound move in the opposite direction its aliasing… If not, its something else. If you incorporate any distortion its much more likely the distortion is aliasing - worst possible test and its completely irrelevant…

[bmanic]

Has anyone done the most important test: can you hear the aliasing?

Yes. Very easily (especially on all the synths where you can do audio rate modulation).

ALL the Arturia synthesizers benefit a lot from running much higher samplerates. Unfortunately, only a hand full work past 192kHz. The older ones only work at 96kHz but all the new ones from a few years ago can go to at least 192kHz.

SQ-80
Prophet-VS
CS-80 v4
Jupiter-8 v4
MS-20

These all can go up to 384kHz (but all of the crap out at 768kHz, aka x16 oversampling). So 8x oversampling is the maximum for these select few.

ALL of them benefit greatly. I highly recommend running these at x2 oversampling at the very minimum and using the maximum available before rendering.

[bmanic]

When I applied some distortion (with a plugin that oversamples, so no significant aliasing from the distortion itself) it became worse.

.. this too is a great way to quickly have the aliasing be a real problem. Thus the quite common setup of a real analogue synth + stomp box distortion, is usually a no-go in the digital domain, unless you heavily oversample everything.

It only gets worse if you have a long chain of non-linear processes (which is not at all uncommon in the analogue domain). This is what most people just don't understand when it comes to aliasing. It all adds up, very quickly. This is why you can have a huge chain of analogue goodness in the analogue domain, without it all becoming fuzzy/fizzy/thin.

It's something guitar players need to fight with almost all the time. This super common chain: gate + comp + pre-distortion -> pre-amp -> amp -> mixer or mojo input like Neve 1073/1084 -> AD converter, has a lot of places where a ton of harmonics are generated. Trying to duplicate the sound within the digital domain requires some pretty hefty oversampling to not have the aliasing come up into the clearly audible range.


In short: Just because the aliasing isn't a problem for a single audio source, it WILL very easily end up in the audible amplitude levels at the end of the process.

[bmanic]

The best way to hear it is with pitch bend. If parts of the sound move in the opposite direction its aliasing… If not, its something else. If you incorporate any distortion its much more likely the distortion is aliasing - worst possible test and its completely irrelevant…

Not at all true. There's plenty of distortion plugins out there that use 16x oversampling, bringing even quite heavily distorted signals aliasing levels down below -100dBFS.

.. and his point of distortion bringing the aliasing from the synth up, is 100% valid and a very good and easy test (at least at high notes). Those dissonant harmonics, in the synthesizers aliased oscillators will be super evident. Even if the distortion is aliasing too, you'll be able to hear the synthesizers aliasing first as the harmonic energy of non-aliased harmonics is simply way louder than the actual aliasing provided by the distortion.. thus it'll still modulate the already aliased things very clearly. This is why aliasing is so nasty in a longer chain of non-linear processes. The loud harmonic distortion will amplify the already existing aliased stuff and cause it to sound dissonant. etc. etc, further compounding the problem.

.. and if you are REALLY paranoid, just route it to an analogue distortion pedal, then it's equally obvious and 100% certain that all the aliasing that you hear is from the synth.

[zerocrossing]

Has anyone done the most important test: can you hear the aliasing?

Yes and yes. Compared to other softsynths, it was faintly audible. When I applied some distortion (with a plugin that oversamples, so no significant aliasing from the distortion itself) it became worse.

If you have to add distortion to be significant, is it really a problem?

If you really are concerned about aliasing, buy a good hardware synth. No aliases on my Prophet 6, or any other of my analogs… well, except the second ring mod on my Dominion 1, which is digital and on purpose. Monkeying around with ultra high sampling rates is for chumps.

[xbow]

If you have to add distortion to be significant, is it really a problem?

If you want to add distortion but it sounds crappy because of aliasing, that's a significant problem.

[OllieBoi]

People get so worked up about aliasing. Four of my favorite synths in my collection, the MicroWave I, Poly Evolver, Nord Lead 2, and Virus TI2 exhibit pronounced aliasing. And you know what, it makes these synths sound better. The 8 bit waveforms in the MicroWave in particular are grungy as hell and the 12 bit Prophet VS waveforms in the Poly Evolver can set your teeth on edge.

Plugin developers (especially Arturia) are responsible for this nonsense. Arturia has touted their "TAE" technology since day one where their plugins were supposed to be gloriously free of aliasing. But it was always bullshit. Their early plugins sounded too clean and lifeless because of their TAE tech.

[Tj Shredder]

The best way to hear it is with pitch bend. If parts of the sound move in the opposite direction its aliasing… If not, its something else. If you incorporate any distortion its much more likely the distortion is aliasing - worst possible test and its completely irrelevant…

Not at all true.

Not at all true…

[bmanic]

People get so worked up about aliasing. Four of my favorite synths in my collection, the MicroWave I, Poly Evolver, Nord Lead 2, and Virus TI2 exhibit pronounced aliasing. And you know what, it makes these synths sound better. The 8 bit waveforms in the MicroWave in particular are grungy as hell and the 12 bit Prophet VS waveforms in the Poly Evolver can set your teeth on edge.

Plugin developers (especially Arturia) are responsible for this nonsense. Arturia has touted their "TAE" technology since day one where their plugins were supposed to be gloriously free of aliasing. But it was always bullshit. Their early plugins sounded too clean and lifeless because of their TAE tech.

You are missing the point (or creating a straw man argument on purpose).

Aliasing is indeed very interesting, when it belongs in the emulation and is in the "right place" in the chain. But take for instance the Prophet VS. The real deal has very lo-fi oscillators and other interesting lo-fi parts of the sound processing.. that's all fine and dandy.

.. however, the thing also has an actual analogue filter which involves all kinds of non-linearities, that of course do not have any aliasing in the real world. Thus if you make an emulation of it, you need to also emulate the filter behavior as it is in the real world. Having a bunch of aliasing there causes the synth to sound different to the real deal in the way the harmonics from the oscillators interact with the filter resonance when cranked (or a distorted filter on some other synth). Thus minimizing aliasing of this part is necessary if you want the most authentic sound.

It's also worth noting that in a lot of different old school synths with digital guts and an analogue LP filter at the end, you do not end up with a ton of aliasing going past that filter cutoff.. unlike badly done emulations where all the aliasing of the guts is further compounded with aliasing from the filter model.