Oversampling?

[Cyberdyne387]

Already sent a support ticket about this, but in retrospect, this would be a better place for it.

I tried out the CMI V demo last night. It has a lot of potential, but sadly, what holds me back from using it as a sampler is that, sans the filter and the samplerate/bitrate adjustments, it’s just another piecewise interpolation sampler limited to the samplerate clock of the host, whereas the actual Fairlight (and the sampling engine of the Synclavier) were variable clock rate. What this means is that the amount of aliasing is fairly consistent no matter what key you press, a significant contrast to when you play, for instance, three whole steps down from the root note instead of an octave.

Are there any plans to implement oversampling like for the Synclavier V, that said? I realize it takes a bit of CPU as it is, so it makes perfect sense to me if you choose not to oversample by at least 8x. However, I’ve been working on a VST of my own that uses an interpolation method that sounds very faithful to variable-clock samplers without the need to oversample by more than 4x, an interpolation method I’m willing to give over if you express interest.

Regardless, thank you for your time, apologies for the technical content of this message, and definitely a great start!

[jackn2mpu]

This is just like the Synclavier - Arturia purports to have an accurate replica of the h/w and they miss ajor things that were part of the original synth/sampler. At least the CMI has the resynthesis that the Synclavier is missing and will probably never have.

Close only counts in horseshoes and hand grenades.

[baubry_arturia]

Already sent a support ticket about this, but in retrospect, this would be a better place for it.

I tried out the CMI V demo last night. It has a lot of potential, but sadly, what holds me back from using it as a sampler is that, sans the filter and the samplerate/bitrate adjustments, it’s just another piecewise interpolation sampler limited to the samplerate clock of the host, whereas the actual Fairlight (and the sampling engine of the Synclavier) were variable clock rate. What this means is that the amount of aliasing is fairly consistent no matter what key you press, a significant contrast to when you play, for instance, three whole steps down from the root note instead of an octave.

Are there any plans to implement oversampling like for the Synclavier V, that said? I realize it takes a bit of CPU as it is, so it makes perfect sense to me if you choose not to oversample by at least 8x. However, I’ve been working on a VST of my own that uses an interpolation method that sounds very faithful to variable-clock samplers without the need to oversample by more than 4x, an interpolation method I’m willing to give over if you express interest.

Regardless, thank you for your time, apologies for the technical content of this message, and definitely a great start!

Hello,

thank you for your feedback.
In order to be as faithful as possible to the original instrument, we decided not to implement oversampling which wasn't available then.

[jackn2mpu]

Already sent a support ticket about this, but in retrospect, this would be a better place for it.

I tried out the CMI V demo last night. It has a lot of potential, but sadly, what holds me back from using it as a sampler is that, sans the filter and the samplerate/bitrate adjustments, it’s just another piecewise interpolation sampler limited to the samplerate clock of the host, whereas the actual Fairlight (and the sampling engine of the Synclavier) were variable clock rate. What this means is that the amount of aliasing is fairly consistent no matter what key you press, a significant contrast to when you play, for instance, three whole steps down from the root note instead of an octave.

Are there any plans to implement oversampling like for the Synclavier V, that said? I realize it takes a bit of CPU as it is, so it makes perfect sense to me if you choose not to oversample by at least 8x. However, I’ve been working on a VST of my own that uses an interpolation method that sounds very faithful to variable-clock samplers without the need to oversample by more than 4x, an interpolation method I’m willing to give over if you express interest.

Regardless, thank you for your time, apologies for the technical content of this message, and definitely a great start!

Hello,

thank you for your feedback.
In order to be as faithful as possible to the original instrument, we decided not to implement oversampling which wasn't available then.

But it could be done though right? Would be a good thing to have. A mini Moog wasn't polyphonic nor was the Moog Modular but you gave them polyphony right? Given that I find your statement rather disingenuous.

[Valentin Arturia]

Polyphony and oversampling are not the same things. Giving a synth a way to play several notes at a time is not the same as changing/giving extra to its sound. But that's not the point. Yes it could be done, everything could be done, but we didn't see oversampling as a major priority and we had to make choices . We wanted first to have a good recreation of the synth.
However you can adjust sample rate and bit rate per slot. Here are variable clock rates (but not oversampled ok

[Cyberdyne387]

Hello,

thank you for your feedback.
In order to be as faithful as possible to the original instrument, we decided not to implement oversampling which wasn't available then.

Thank you for getting back to me, first of all.

Valentin, I'm aware of the option to downsample a sample within CMI V: in fact, that's how I discovered how bad the aliasing is on certain notes, since the filter doesn't lower its cutoff value automatically with the samplerate (which I honestly can't complain about: I love those high end harmonics).

As for the original topic of oversampling, baubry, I know enough about old DAC's to know that lack of oversampling back then is only partially correct. Pretending the sample rate is higher than it is was indeed innovated later on, so what the Fairlight and Synclavier both did to control aliasing was set the master DAC clock speeds to have an unusually high sample rate, usually in mHz instead of kHz. Like I said, though, I realize emulating such a samplerate would require the fastest computers nowadays (even my own sample interpolation code takes a shortcut and assumes an infinite samplerate), so I understand completely that you felt it best to take shortcuts of your own, especially since, again, I am aware of how much CPU it's taking up for me as it is.

[jackn2mpu]

Polyphony and oversampling are not the same things. Giving a synth a way to play several notes at a time is not the same as changing/giving extra to its sound. But that's not the point. Yes it could be done, everything could be done, but we didn't see oversampling as a major priority and we had to make choices . We wanted first to have a good recreation of the synth.
However you can adjust sample rate and bit rate per slot. Here are variable clock rates (but not oversampled ok

I never said they were the same thing. What I said was if you can have a synth do something it wasn't designed for (like a polyphonic miniMoog) then why not the oversampling for the CMI? Adjust sample rate and bit depth (which is the correct term and not bit rate as you write) is okay but doesn't get away from what is being asked for.

It's sad really - I knew both these machines from the beginning here in the US and they were capable of some great things that Arturia has barely touched (or seems capable of doing).

Overall I have to say this - at the price Arturia is asking to pay for second rate development of new synths to add to their stable - my money will be best spent on other things. Bring the price down by half and maybe I and others will think about this again.

[Valentin Arturia]

Polyphony and oversampling are not the same things. Giving a synth a way to play several notes at a time is not the same as changing/giving extra to its sound.

What I said was if you can have a synth do something it wasn't designed for (like a polyphonic miniMoog) then why not the oversampling for the CMI?

As I said adding polyphony and adding oversampling are not the same thing. It is not the same kind of work. So yes we could extend possibilities, like we do with polyphony on mono synth, but it is not the same amount/kind of work.

Overall I have to say this - at the price Arturia is asking to pay for second rate development of new synths to add to their stable - my money will be best spent on other things. Bring the price down by half and maybe I and others will think about this again.

If you think that 20 synths for around 500$ or the update with 4 new synths plus a new Piano engine for 200$, are expensive this is your opinion. I don't find this is expensive regarding other brands but OK maybe there are more valuable solution out there.

[jackn2mpu]

Polyphony and oversampling are not the same things. Giving a synth a way to play several notes at a time is not the same as changing/giving extra to its sound.

What I said was if you can have a synth do something it wasn't designed for (like a polyphonic miniMoog) then why not the oversampling for the CMI?

As I said adding polyphony and adding oversampling are not the same thing. It is not the same kind of work. So yes we could extend possibilities, like we do with polyphony on mono synth, but it is not the same amount/kind of work.

Overall I have to say this - at the price Arturia is asking to pay for second rate development of new synths to add to their stable - my money will be best spent on other things. Bring the price down by half and maybe I and others will think about this again.

If you think that 20 synths for around 500$ or the update with 4 new synths plus a new Piano engine for 200$, are expensive this is your opinion. I don't find this is expensive regarding other brands but OK maybe there are more valuable solution out there.

I'm not the only one that thinks the $200 upgrade is exhorbitant - there's plenty more out there. Every person over on the Avid Pro Tools forum (the DUC) that has an interest in this because they're Arturia users is of the same opinion. Should have been more like $100 given VCollection 5 has been available these last few months for $250.

[MarkBog]

When V-Collection 5 was rolled out, I believe that existing customers received 5 new synths, PLUS a MAJOR overhaul to the GUI on all prior products for $200.  My opinion is that customers got a lot more bang for our buck back then.  It's very disappointing that Arturia is not listening to their customer base on this issue.

[jackn2mpu]

When V-Collection 5 was rolled out, I believe that existing customers received 5 new synths, PLUS a MAJOR overhaul to the GUI on all prior products for $200.  My opinion is that customers got a lot more bang for our buck back then.  It's very disappointing that Arturia is not listening to their customer base on this issue.

True that. But if you waited like I and probably others did to upgrade that price dropped later in the year to IIRC $150 and even lower in a couple of vendors. So it's either wait for the price drop or hope that Arturia comes to their senses (which I doubt) and drops this buy-in right now. They definitely do NOT have the Christmas spirit.

[madtheory]

... what the Fairlight and Synclavier both did to control aliasing was set the master DAC clock speeds to have an unusually high sample rate, usually in mHz instead of kHz.

This is not correct for the Fairlight CMI I/II/IIx. There is no high sample rate trick to reduce aliasing. On the real machine there is a DAC for each voice each with its own clock. Pitch change is achieved by changing the clock. This gives the aliasing characteristic of the CMI. These clocks also have a lot of jitter. Oversampling would change that raw sound so it would not be like a real CMI any more.

PS I agree with your correction, it is called bit depth because that is to do with the dynamic range- very important to the sound. Bit rate is a different thing

PPS correct SI unit is MHz not mHz

[hedvre]

This is not correct for the Fairlight CMI I/II/IIx. There is no high sample rate trick to reduce aliasing. On the real machine there is a DAC for each voice each with its own clock. Pitch change is achieved by changing the clock. This gives the aliasing characteristic of the CMI. These clocks also have a lot of jitter. Oversampling would change that raw sound so it would not be like a real CMI any more.

I'm agree too. This is quite strange to want a perfect recreation of the instrument in a way and in another way want something to reduce aliasing... which is characteristic of CMI. If you want an Arturia sampler go for Spark.

[LBH]

CMI III did have a higher samplerate. 100000 Hz in mono and 50000 Hz in stereo according to this: https://en.wikipedia.org/wiki/Fairlight_CMI

Arturias version is not version III. But perhaps up to 48000 Hz would be appropiate. And in todays context perhaps up to 24 bit. That's only my view.
Better loop features could perhaps also be relevant in to todays perspective.  Did even CMI III have better looping features? I don't know.

But no doubt the poor sample quality and audioble loop clicks was part of the sound. I don't see CMI as a sampler (sample player) that shall compete with samplers of todays standard and possibilities. It's still about sound to me, and then being able to function in a modern production technically.

CMI for example also offers a workstadion to create 10 tracks sequences of 256 steps. No other Arturia application does that.
It has a re-synthesis feature too.
And  CMI V is relatively easy to use.

[madtheory]

Yes CMI 3 had high SR, and no oversampling. DAC was very early 16 bit. But AFAIK it is still clocking to transpose, not linear interpolation. Series 3 has full 8 bit mode as well, same voice card architecture as the II but with shared memory instead of memory for each voice.

But yes as you say this is a CMI II emulation, it has a big beautiful sound. If someone wants oversampling there are other samplers.