Question about Virtual Analog Synths

[AnotherScott]

Actually I used the term advisedly. From the Yamaha website (verbatim):

"VA Synthesis

Unique to Yamaha, computer-based "physical modeling" technology which accurately simulates the complex vibrations, resonances, reflections and other acoustic phenomena that occur in real wind or string instruments-to produce high quality monophonic voice...

"

In that context, Yamaha used VA as an abbreviation for Virtual Acoustic. The more common use, and what we've been talking about in this thread, is Virtual Analog.

[Synthoid]

there is a strong argument that Roland's D-50 five years earlier was actually the first

 

There is? Do tell...

 

Yeah, I heard that before as well.

[AnotherScott]

The implication of a manufacturer's marketing department (and not just Roland's) that the circuit components of the original analog circuitry are being modeled is misleading.

I don't think they are saying that they are modeling the circuit components, rather they are modeling the behavior that results.

 

I would argue that a synth engine is VA if there is the clear INTENTION of emulating the traditional VCO -> VCF -> VCA chain of a typical analog synth.

Most (all?) workstations have allowed you to pick an oscillator waveform and send it through the equivalent of VCF and VCA, yet very few claim to be VA. From Korg, only the Kronos and OASYS (not Kross, Krome, M50, TR, M3 sans Radias, etc.) Nothing from Yamaha (they did offer it in things like the AN1X, but not in any workstations that I know of). Kurzweils have it, though.

[Bucktunes]

By definition, any digital synth (VA, ROMpler, clonewheel etc.) is a computer with proprietary software that tells it how to generate soundwaves. In a nutshell, it comes down to how many calculations per second are required to generate a particular sound, which is limited by the processor. As Governor Silver pointed out, VA waveforms are much more complex to generate than simply playing back ROM wave data.

 

What amazes me is synths like the Virus TI that can generate dozens of voices in multitimbral mode with the effects retained from each individual patch! There must be a whopper of a multicore processor under the hood. I guess that explains why it's a $3,500 synth. By comparison, my humble Blofeld desktop specs at 25 voices. However, once you get three oscillators kicking with FM, ring modulation and a half dozen other modulation routings going, the polyphony drops to 6 or 8. I guess that's why you can buy about 5 of them for the price of a new Virus.

[Synthoid]

Most (all?) workstations have allowed you to pick an oscillator waveform and send it through the equivalent of VCF and VCA, yet very few claim to be VA. From Korg, only the Kronos and OASYS (not Kross, Krome, M50, TR, M3 sans Radias, etc.) Nothing from Yamaha (they did offer it in things like the AN1X, but not in any workstations that I know of

 

http://www.deepsonic.ch/deep/pix/yamaha_ex5s.jpg

 

 

The EX5 features a unique Extended Synthesis system which incorporates a number of the most advanced tone generator technologies currently available...plus a full-featured sampling system. It offers a Four-Element Voice structure which allows for up to four independent waves from any of the tone generators. The tone generators implemented here include AWM2 (Advanced Wave Memory) Synthesis, AN (Analog Physical Modeling) Synthesis, FDSP (Formulated Digital Sound Processing) Synthesis and Virtual Acoustic (VL) Synthesis.

 

 

back in the day when Yamaha was innovative.

[RABid]

When you watch Roland videos they talk about circuit modeling. I don't really know if there is much, or any difference, in modeling a circuit or modeling the behavior of those circuits. What I do know is that years ago Roland claimed to be modeling drums, speaking about changing the depth of a snare, tightness of the head, etc... They lead us to believe it was all modeling technology. Then, after someone sampled the "modeled" drums Roland filed a lawsuit against the company selling samples of their "modeled" drums. The lawsuit reviled that they were not really modeled drums after all. They were modifying samples of drums. After that I'll always be skeptical of anything that Roland claims to model. Having said that, I still have a large V-Drum kit.

[Theo Verelst]

It's still an obscure subject isn't it ?

 

I sure had virtual analog synthesis in the first half of the 90s from a Tg500 (multiple samples as oscillators, digital filter simulations), and I recall there was a program from D. Smith to do analog simulation that predates a lot of what later was brought out.

 

It isn't necessarily rocket science to get an oscillator with frequency control, some sort of (resonating) low pass filter with frequency control, and a envelope generator with VCA going on some sort of computer of at least say a 486 level of power. However, it will suffer from all kinds of inaccuracies, that certainly are audible.

 

So in the course of history, cunning plots have been devised to make good use of the available processing power. This isn't easy, just like a lot of modern computers are hard to predict in how fast they'll respond exactly. So say you have a certain amount of (limited) computer power you can try to find ways to get better sounds, more parameters, more voices, etc. by giving you digital algorithms some sort of accuracy tradeoff, and trying make smart use of what hardware you have available with all the tricks you can A) come up with in the simple enough domain B) buy or research in case of more complicated (scientific) approaches.

 

Modern PCs are an example of having quite some processing power in terms of operations per second, memory and disk size and peak bandwidth, and mathematical operations per second. But this power needs to be effectively used for the synthesizer simulation you want, and the programmers need to have access to right advanced algorithms from a certain level of complexity onward. In the PC realm, major errors have been made with the translation of analog circuits into the digital domain. Some will not even admit it unless pressured, but essentially most programmers aren't by far scientists enough to *properly* simulate the analog synthesizer circuit, and the more accurace is required, the harder that gets.

 

And the PC has the disadvantage for Digital Signal Processing that it's sort of a super computer derived architecture that can compute on big streams of data with some sort of accuracy, but because of it's pipelines and all kinds of latencies (memory for processor instructions and data), it cannot respond accurately on a fine grained level without losing a lot of the speed that comes from caches and predictive branching, and isn't really a proper real-time processing architecture by design. Some improvement is possible, which probably is in libraries from well known audio software/hardware manufacturers (installations that change the OS a bit), and in Linux.

 

The real smart stuff like the well known hardware manufacturers have occupied themselves with isn't always easy to understand, so there is little known about the differences in accuracy and (smart and less good) tunings that are part of all kinds of digital synthesizer programs. The Physical Modeling was a term I think came from the the time of the Yamaha instrument modeling, a little after the SYs IIRC. Of course you could say that virtual analog is the simulation of a virtual electronics circuit, which is true, but I don't want to confuse the two. What is common difficulty about both the PM and the VA is the accuracy of the sampled signal.

 

Fully virtual analog synths have at their output a DA convertor, and contrary to specific digital hardware synths, not even one of a given type, but just some soundcard or an interface with certain chips in it. The point is that even good DA convertors will create certain errors in the signal, with the exception of A) real low frequencies (they can be ok), B) possible long convoluted signal components that can pass through the convertor moderately well, and C) specific (DA convertor dependent!) pre-processing that is hard, sounds horrible when done wrong, and isn't easy to do.

 

So just like in the time of the rise of the sampler machines, it can be interesting to make the filter and possibly the VCA analag, like some of the ancient Akai S series and the more recent Prophet 12.

 

To the point about the processor power for virtual synthesis: it's hard to explain all the issues that can go on in a pipe-lined/parallel implementation of a virtual synthesizer, which is undoubtedly why there are special chips in Yamaha's, Lexicons, Kurzweils, some Rolands and Casio (as far as I know). Standard DSPs (like in some Rolands, the Nords as far as I know, and recent and older effect equipment) aren't always as powerful as most new PCs, but can do certain essential things more efficient.

 

I'm sure more can be said about this all.

 

T.

 

[ManInTheBox]

VA used to mean modeled behavior. This allowed for audio rate frequency and pulse width modulation of the oscillators.

 

Now it's a meaningless marketing term bestowed everywhere they can get away with it.

 

This is the way I've come to understand it & the Jupiter80 is the main reason why. It's got 256 voice polyphony that varies depending on how you allocate your live sets, but I'd think it would cost three times as much if it were creating those voices the same way a Virus or Nord Lead does.

 

But it does sound good. I don't think Roland uses the term "Virtual Analog" but many people who use it/sell it do.

 

I don't think it matters, as long as it sounds like a Jupiter 8, or Juno 60, or Prophet 5, if that's what you're looking for & the Jupiter 80 gets pretty darn close. Close enough for me anyway. Not that I own one, just saying.

[WillNeverPost]

there is a strong argument that Roland's D-50 five years earlier was actually the first

 

There is? Do tell...

 

Here:

http://www.gearslutz.com/board/10090121-post49.html

 

And here:

http://www.vintagesynth.com/forum/viewtopic.php?f=1&t=56593&start=17

 

And many elements of the argument are contained in the seven pages of this thread:

http://www.vintagesynth.com/forum/viewtopic.php?f=1&t=56593&sid=d434cf0ab0a5a9b91418fe9e1a79ac39

 

It's an _argument_ that you can agree or disagree with. In the end, it makes no real difference what each of us believes. The only reality is that these machines were created and that they do what they do. What pigeon hole we stuff them in is irrelevant.

 

PS. As you can see, what is and what is not "worthy" of being called Virtual Analog has been flogged literally TO DEATH on other forums (and no doubt here as well in the past). There is no way to reach a consensus and no point in trying. Next topic.

 

[WillNeverPost]

The implication of a manufacturer's marketing department (and not just Roland's) that the circuit components of the original analog circuitry are being modeled is misleading.

I don't think they are saying that they are modeling the circuit components, rather they are modeling the behavior that results.

Most of the literature is worded (intentionally I believe) to avoid making that distinction one way or another. The unsurprising result is that most people think they are modeling the circuits component by component, as if they are implementing a real-time SPICE model (not that most musicians know what that is, I'm just trying to explain the impression they get).

 

I would argue that a synth engine is VA if there is the clear INTENTION of emulating the traditional VCO -> VCF -> VCA chain of a typical analog synth.

Most (all?) workstations have allowed you to pick an oscillator waveform and send it through the equivalent of VCF and VCA, yet very few claim to be VA. From Korg, only the Kronos and OASYS (not Kross, Krome, M50, TR, M3 sans Radias, etc.) Nothing from Yamaha (they did offer it in things like the AN1X, but not in any workstations that I know of). Kurzweils have it, though.

I haven't been clear. When I say "intention of emulating" I mean both offering the equivalent functionality to an analog synth AND, most importantly, marketing or representing it as such (see page three of the D-50 manual). As you say, very few workstations claim to have any VA capabilities. So I'm basically saying that if it looks like a simulated duck and more-or-less quacks like a simulated duck, I'm comfortable calling it a simulated duck.

[Doerfler]

back in the day when Yamaha was innovative.

 

that was last century , except for digital pianos

[WillNeverPost]

What I do know is that years ago Roland claimed to be modeling drums, speaking about changing the depth of a snare, tightness of the head, etc... They lead us to believe it was all modeling technology. Then, after someone sampled the "modeled" drums Roland filed a lawsuit against the company selling samples of their "modeled" drums. The lawsuit reviled that they were not really modeled drums after all. They were modifying samples of drums. After that I'll always be skeptical of anything that Roland claims to model. Having said that, I still have a large V-Drum kit.

I feel the same. Exaggerations and even outright fabrications are not uncommon from marketing departments. Colbert called it "truthiness." But if something does what I want it to do I can get past the specious claims.

[GovernorSilver]

I think the Roland D-50 controversy came from confusion over digital oscillators vs. VA, because the D-50 used digital oscillators. Just because a synth uses digital oscillators, it does not automatically qualify as a VA. There are a number of synths out there that were made with digital oscillators that were not virtual analog - eg. PPG Wave, Yamaha DX7.

[WillNeverPost]

I think the Roland D-50 controversy came from confusion over digital oscillators vs. VA, because the D-50 used digital oscillators. Just because a synth uses digital oscillators, it does not automatically qualify as a VA. There are a number of synths out there that were made with digital oscillators that were not virtual analog - eg. PPG Wave, Yamaha DX7.

That's not why there is disagreement. Did you read the threads I linked to? It's all there.

[GovernorSilver]

I think the Roland D-50 controversy came from confusion over digital oscillators vs. VA, because the D-50 used digital oscillators. Just because a synth uses digital oscillators, it does not automatically qualify as a VA. There are a number of synths out there that were made with digital oscillators that were not virtual analog - eg. PPG Wave, Yamaha DX7.

That's not why there is disagreement. Did you read the threads I linked to? It's all there.

 

Not every single page, but I got enough of an impression that the peeps arguing that it's a VA are confusing digital oscillators with VA. One of the thread participants said that if you have an oscillator and filter, that's a VA, lol. I disagree with that.

 

I've always understood VA to be a variation of physical modeling. There's no physical modeling going on in the D-50, a great a synth as it was.

[RABid]

It is also common for VSTi's to use samples for the waveforms and digital modeling for the amps, filters and other components. Developers at KVR-Audio have stated that this can save a lot on CPU usage. Some of them are very up front about which of their VSTi's use waveforms and which are totally modeled.

[ManInTheBox]

It is also common for VSTi's to use samples for the waveforms and digital modeling for the amps, filters and other components. Developers at KVR-Audio have stated that this can save a lot on CPU usage. Some of them are very up front about which of their VSTi's use waveforms and which are totally modeled.

 

I've personally never had a problem with using samples. They're used to great effect to emulate slap bass, flamenco guitar, & hundred thousand dollar pianos. Surely they can capture an analog synth....

 

To the extent that someone who is willing to accept a flamenco guitar/saxophone/slap bass, etc played on a keyboard anyway. There are guys out there that only the real thing is good enough & a VA is close enough for some of them, where samples are not. I'm not one of them.

[mate stubb]

I've always understood VA to be a variation of physical modeling. There's no physical modeling going on in the D-50, a great a synth as it was.

 

To me they are fundamentally different. Physical modelling covers tubes (not electronic tubes but instrument body tubes) and exciters (mouthpieces, reeds, etc) and vibrating strings.

 

Virtual Analogue models electronic component behavior - oscillators, filters, amps. Control signals are not really a big part of it - LFOs and EGs have been digital even back in the analogue days.

[WillNeverPost]

I've always understood VA to be a variation of physical modeling. There's no physical modeling going on in the D-50, a great a synth as it was.

And I've always understood something else that is quite different. As I said, there can be no reconciliation between these views.

[GovernorSilver]

I've always understood VA to be a variation of physical modeling. There's no physical modeling going on in the D-50, a great a synth as it was.

And I've always understood something else that is quite different. As I said, there can be no reconciliation between these views.

 

I presume your view is, at the very least, a bit more rigorous than "if it's got an oscillator and a filter, it's virtual analog!".

[WillNeverPost]

I've always understood VA to be a variation of physical modeling. There's no physical modeling going on in the D-50, a great a synth as it was.

And I've always understood something else that is quite different. As I said, there can be no reconciliation between these views.

 

I presume your view is, at the very least, a bit more rigorous than "if it's got an oscillator and a filter, it's virtual analog!".

I didn't say exactly that. But I do not believe that VA is a form of physical modeling, mainly because I have studied the math behind both. Have you?

[GovernorSilver]

I've always understood VA to be a variation of physical modeling. There's no physical modeling going on in the D-50, a great a synth as it was.

And I've always understood something else that is quite different. As I said, there can be no reconciliation between these views.

 

I presume your view is, at the very least, a bit more rigorous than "if it's got an oscillator and a filter, it's virtual analog!".

I didn't say exactly that. But I do not believe that VA is a form of physical modeling, mainly because I have studied the math behind both. Have you?

 

And I didn't say that yousaid that. Didn't mean to put you on the defensive.

 

It's just the sort of thing that makes me stop reading GearSlutz threads.

[WillNeverPost]

I presume your view is, at the very least, a bit more rigorous than "if it's got an oscillator and a filter, it's virtual analog!".

I didn't say exactly that. But I do not believe that VA is a form of physical modeling, mainly because I have studied the math behind both. Have you?

And I didn't say that yousaid that. Didn't mean to put you on the defensive.

Really? Rolling smiley and all? I'm not sure there is any other way to take that. And regarding studying the math, that was to inform you that I do understand these techniques, not as well as someone actually programming them (there's a real art to that) but enough to understand the differences. My question was not rhetorical. If you have studied them too then we can segue into a discussion about why you think VA is a variation of PM.

 

[GovernorSilver]

I presume your view is, at the very least, a bit more rigorous than "if it's got an oscillator and a filter, it's virtual analog!".

I didn't say exactly that. But I do not believe that VA is a form of physical modeling, mainly because I have studied the math behind both. Have you?

And I didn't say that yousaid that. Didn't mean to put you on the defensive.

Really? Rolling smiley and all? I'm not sure there is any other way to take that. And regarding studying the math, that was to inform you that I do understand these techniques, not as well as someone actually programming them (there's a real art to that) but enough to understand the differences. My question was not rhetorical. If you have studied them too then we can segue into a discussion about why you think VA is a variation of PM.

 

I meant no disrespect towards you personally, nor your opinion. I don't know how I can state this more clearly.

[WillNeverPost]

I meant no disrespect towards you personally, nor your opinion. I don't know how I can state this more clearly.

And I really didn't assume that what's you did. I just assumed that it meant you were disagreeing with me. Which is fine too. It's all good! Friends?

[GovernorSilver]

SilverI meant no disrespect towards you personally, nor your opinion. I don't know how I can state this more clearly. /quote] And I really didn't assume that what's you did. I just assumed that it meant you were disagreeing with me. Which is fine too. It's all good! Friends?

 

I'll buy you a beer (or other drink) if we ever meet.

[JAD]

Beer seems to solve everything.

[CEB]

my Roland FA-06 has virtual analog

 

I don't see any mention of the word "analog" on their product page - on either the Features or the Specs tabs. They probably would have mentioned VA if they wanted you to buy it for its VA features.

 

http://www.rolandus.com/products/details/1298/features/

 

It does have VA. SN synth tones have 3 oscillators, 7 waveforms etc.... Fully editable .... But I use it like a ROMpler. It comes loaded with 2000+ sounds plus all the available Axial downloads including Integra 7 libraries.

[J. Dan]

I think the most basic distinction that needs to be made for something to be VA is to be able to manipulate the wave form. A sample playback may have a 25% and 50% pulse! but VA can modulate the pulse width. Of course more features like hard sync and modeled filters can sweeten the pot, but that's the basic distinction for me.

 

[Kenny Ingram]

I always considered my SH-201 to be a perfect example of Virtual Analog. The sounds are FULLY editable and shapable. Wonderful little keyboard.