Please, make a moog model 55 vst

[PoorFellow]

I went out of my way to post one simple reproducible example of the problem I was experiencing. You accuse me of being rude because of my very direct nature, but I think it's kind of rude to suggest I'm wrong without them trying it out for themselves. I came here to post this because I hope MRB/Cherry Audio is able to fix this glaring defect in an otherwise great product. I have been using soft and hardware synths for over 20 years and know when something isn't working right.

You act like people here are somehow against you personally if they do not agree with you and your 'direct' - 'cough' - 'nature' apparently consists of , when someone do not agree or if they go out of their way to try to help you investigate , then you start questioning even the most basic foundation for them to participate in this discussion , namely if they even have and use the software that the discussion is about ! (and best of all you don't think yourself that that might be just a tad rude !)
I try to assist , not because I have any stake in this whatsoever , but merely because you somehow seem to miss some of the point here when you think that the other participant is some noob which UTD Grant most certainly is not...
When participating in the debate then Grant even politely adds that " I'm not saying that VM900 doesn't have aliasing issues, but this experiment is inconclusive to me." , which apparently does not appease you either !
You might be angry with me right now because I dare respond to you or because I dare write. But perhaps you misunderstand the basics here ? This is a forum , normally a place for discussion , which you really are not interested in , you just want to have it your way , which is kind of fine enough , but just don't think that you appear as ready for any discussion or debate !
If I were you , which I am obviously not - and I do not have the experience of yours with respect to music - or Grant's for that matter , then I would contact Cherry Audio support and try to get them to fix it !
But do not expect anybody to wanting to keep going out of their way for your sake if you repel or shut down any discussion in advance because there is no room for any debate !

[SynthBaron]

OK.

[utdgrant]

Uhh, I have hardware CP3's [...] and they do nothing like this (in fact they are quite clean unless you deliberately overdrive them, and even if you do it doesn't sound like aliasing). I also set the levels way below anything that could be considered distortion going through any module in the chain.

Which revision(s) of the CP3 circuitry are your modules?

With the VM900 Revision switch set to REV 1, as shown in your patch, you certainly DO get significant distortion with those knob settings. When you switch to REV 2, the distortion is vastly reduced, along with the amplitude of the sum-and-difference sideband frequencies (introduced as a result of intermodulation).

Not going to post another example, if you don't have the software you can download a free demo version if you don't believe me...lol. This was just one quick obvious example, there are plenty more anyone could come up with.

You don't have to post another example. I was suggesting that you perform the single oscillator experiment for yourself, so that you can ascertain if aliasing is actually occuring.

As I said, it may well be the case that there is an aliasing issue with some of the VM900 modules. However any genuine aliasing effects will be swamped by intermodulation sidebands in your original example. This is because you have two high-bandwidth oscillators beating against one another and fed through a distorting amplifier. Using one oscillator only will eliminate any inharmonic intermodulation signals, and isolate any genuine aliasing sidebands. These alias frequencies will be clearly visible when fed into a spectrum analyser because you won't have any intermodulation sidebands confusing the image.

I bought VM900 in October 2021 and I'm extremely happy with it.

[SynthBaron]

I said I had CP3's ("Rev 1"), not CP3A's ("Rev 2"). I'm aware of the quirks of this discrete transistor module, I also know that it does not behave like this in the analog realm.

This is because you have two high-bandwidth oscillators beating against one another and fed through a distorting amplifier.

Once again, nothing in this signal chain as pictured and in the audio file is dialed in to deliberately cause distortion. I verified this long ago when I was trying to pinpoint the cause of the aliasing, and was trying to limit anything possibly dumb I was doing on my end to cause it.

It seems like you are trying to convince me that this is "normal behavior", when I know it's not. Other softsynths do not alias, not even the Arturia Modular V. And that uses "high-bandwidth oscillators" too.

In case it is not absolutely clear by my previous posts, which I have tried to be...the aliasing still occurs with just 1 oscillator. I provided the audio and visual example to make it so it's easily heard using a completely normal patch.

[Steve W]

I am not jumping in on any disagreement here. just trying to understand. I have heard people talk about aliasing (and antialiasing). I have been aware of the "third tone" created by beats. I understand that digital is different than analog.

I recreated the patch and hear the two tones, the beating, and the "third tone." I have seen mathematical-based explanations, but I am too old to understand complex math. So, is aliasing the result of phase differences* that cause in essence "a fourth tone?" If so, is antialiasing the use of mathematics to compensate for this "fourth tone" so to speak?

If I do not understand it correctly, please keep your explanation/correction understandable. Thanks.

*I just looked at the patch in a spectrum analysis program and saw that the rate of change of the beats is not very smooth (as I expect from analog audio). So is it the non-smoothness of the change as the two tones converge and diverge the digital "fourth tone?"

[utdgrant]

Here is a Lissajou plot of the REV 2 transfer function - essentially 100% linear:

CP3Rev2Plot.png (495.84 KiB) Viewed 3089 times

Here is a plot of the REV 1 transfer function - decidedly NON-linear (and even appears to have an element of hysteresis):

CP3Rev1Plot.png (506.01 KiB) Viewed 3089 times

I created a video (YouTube link) showing the spectrum analysis of a single 901-B sawtooth oscillator in three states:

1. Direct output (CP3 bypassed)
2. CP3 REV 2 preamp model
3. CP3 REV 1 preamp model

Only in the REV 1 case, with high-pitched oscillator, can aliasing clearly be seen. So there's definitely a case for internal oversampling in the CP3 module when switched to REV 1 mode. Whether x2 oversampling would fix it, or it needs to go higher still, I don't know.

The direct output of the 901-B shows that band-limiting is performed internally, so no alias frequencies appear at its output. This is also the case for the stock CA Oscillator module - clever!

[SynthBaron]

Only in the REV 1 case, with high-pitched oscillator, can aliasing clearly be seen. So there's definitely a case for internal oversampling in the CP3 module when switched to REV 1 mode. Whether x2 oversampling would fix it, or it needs to go higher still, I don't know.

The direct output of the 901-B shows that band-limiting is performed internally, so no alias frequencies appear at its output. This is also the case for the stock CA Oscillator module - clever!

Very interesting. Thank you for taking the time to go in depth on this more. I hope something as easy as oversampling the CP3 will fix it!

I noticed the stock CA oscillator looks like its using additive synthesis with sine waves when you look at the waveforms on an oscilloscope, and is described as an "alias-free oscillator". That would make it truly alias-free, I think.

[ColinP]

If I do not understand it correctly, please keep your explanation/correction understandable. Thanks.

Hi Steve,

Aliasing is caused by sampling a signal that contains frequencies higher than half the sample rate.

The following gives a reasonable, non-mathematical explanation of the effect...

https://xiengineering.com/sampling-freq ... -aliasing/

[PoorFellow]

Thank you @ ColinP
I wanted to ask same question as Steve did but I figured that I might not understand the answer anyway,
but you putting it into simple terms made it very easy for me to understand ! (not suggesting that I understand the mathematics of it though !)

[Steve W]

Thanks for the "so-Steve-and-others-can understand" explanations and details. I now know why its called aliasing and a few other things that make sense. It seems to square somewhat with what I observed with my eyes and ears.

*I just looked at the patch in a spectrum analysis program and saw that the rate of change of the beats is not very smooth (as I expect from analog audio). So is it the non-smoothness of the change as the two tones converge and diverge the digital "fourth tone?"

The lack of smoothness of the beats is the auditory result of sample-rate induced "rounding errors" created by the digitized nature of the process.

Using the example of the similarity between a 1kHz tone and a 7kHz tone when sampled at 8kHz, the "rounding errors" [or should we call them "squaring errors"?] distort what the analog sound should actually be because there are sudden shifts. So to my ear it sounds like the change rates of the beats are not smooth like we would have with analog sound.