What techniques exist for analog modelling in the digital domain.
Like e.g. modelling a mono synth.
Is it all just some transfer function and "signals and systems" stuff?
What about its theoretical limits? Can analog modelling come close to real analog?
Analog modelling
-
- KVRian
- Topic Starter
- 1096 posts since 28 May, 2010 from Finland
Last edited by soundmodel on Tue Feb 16, 2016 12:15 am, edited 1 time in total.
-
- KVRAF
- 12555 posts since 7 Dec, 2004
Define "close".
The answer is: No.
Unless you define close as some variation of "at some considerable distance."
I'm sure you've seen recent film demonstrating modern modelling capabilities. The question is simply "is it worthwhile?"
In many cases it isn't. If you're after the particularities of a synthesizer, use that synthesizer. If you're not, use any synthesizer that satisfies you whether software, analog circuit or some combination.
Also excellent troll topic.
The answer is: No.
Unless you define close as some variation of "at some considerable distance."
I'm sure you've seen recent film demonstrating modern modelling capabilities. The question is simply "is it worthwhile?"
In many cases it isn't. If you're after the particularities of a synthesizer, use that synthesizer. If you're not, use any synthesizer that satisfies you whether software, analog circuit or some combination.
Also excellent troll topic.
Free plug-ins for Windows, MacOS and Linux. Xhip Synthesizer v8.0 and Xhip Effects Bundle v6.7.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
-
- KVRian
- Topic Starter
- 1096 posts since 28 May, 2010 from Finland
Sort of what I've been thinking.
I used to be optimistic about analog modelling. E.g. the TB-303 has been modelled fairly well by some developers (e.g. the D16 one). Then again, the TB-303 is a pretty simple synthesizer.
But then there are some things that either seem to take excessive amounts of development time, or which just don't sound like analog.
I used to be optimistic about analog modelling. E.g. the TB-303 has been modelled fairly well by some developers (e.g. the D16 one). Then again, the TB-303 is a pretty simple synthesizer.
But then there are some things that either seem to take excessive amounts of development time, or which just don't sound like analog.
-
- KVRist
- 296 posts since 1 Apr, 2009 from Hannover, Germany
Must have missed something. Which film is that?aciddose wrote: I'm sure you've seen recent film demonstrating modern modelling capabilities.
-
- KVRAF
- 12555 posts since 7 Dec, 2004
I was only generalizing regarding the processing power we have available these days. If you wanted to do a detailed circuit model of a synthesizer you could no doubt move "close" to far beyond any human ability to discern the difference.
As things are now however a single machine running code on a single core even at the high end isn't powerful enough to cover absolutely every base in a A-B comparison. It will cover most bases but under measurement by someone familiar with the peculiarities of any device nearly any emulation attempt short of a highly detailed circuit model will come up ... not surprisingly ... short.
That said, most people aren't looking for that sort of thing nor do they even know what it is. So for those people, the vast majority of us, we don't care about "how analog" something is because we know this is rooted in a misunderstanding of what the word "analog" means.
If I want "analog" it involves plugging in some cables and dealing with electronic circuits and all the associated pitfalls. Those very same issues are the things which make those circuits "analog" in the first place. Do you really want to model those?
I've got a good model: it's called physical reality. Can't be beat.
As things are now however a single machine running code on a single core even at the high end isn't powerful enough to cover absolutely every base in a A-B comparison. It will cover most bases but under measurement by someone familiar with the peculiarities of any device nearly any emulation attempt short of a highly detailed circuit model will come up ... not surprisingly ... short.
That said, most people aren't looking for that sort of thing nor do they even know what it is. So for those people, the vast majority of us, we don't care about "how analog" something is because we know this is rooted in a misunderstanding of what the word "analog" means.
If I want "analog" it involves plugging in some cables and dealing with electronic circuits and all the associated pitfalls. Those very same issues are the things which make those circuits "analog" in the first place. Do you really want to model those?
I've got a good model: it's called physical reality. Can't be beat.
Free plug-ins for Windows, MacOS and Linux. Xhip Synthesizer v8.0 and Xhip Effects Bundle v6.7.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
-
- KVRian
- Topic Starter
- 1096 posts since 28 May, 2010 from Finland
Yes it can. Digital plug-ins offer improvements over analog gear. But they don't (according to some) offer the sound.I've got a good model: it's called physical reality. Can't be beat.
I guess some sort of hybrid approach could be realizable as well. Have the sound path consist of analog circuitry, but add memory and such.
-
- u-he
- 28063 posts since 8 Aug, 2002 from Berlin
There's also a complementary view to this. Which is related to "close" being a numerical threshold. If you wanted "CD quality" you'd model only those effects of an analogue circuit that make a change above -84dB (14 bit). Alternatively use floating point accuracy or anywhere in-between.
Above that level you'll find there are numerous effects already that you do not wish to add to your model, such as humm, buzz, noise, and maybe some weird signal picked up from your smartphone.
In that complementary view, Analogue Modelling is defined by picking the raisins of analogue behaviour and omitting the crap. In fact, a too detailed model might be subject to lawsuits ("the software broke") while a sufficiently close approximation might just be what the customer had wished for.
That said, one of the most important ingredients of Analogue Modelling is thorough observation and the choices made based on that. A model that's solely based on theory might not convince. A model that sounds convincing needs to take effects into account that are not necessarily obvious when looking at schematics, e.g. part tolerances. Those effects express themselves as little DC offsets here and there, or a difference in level, unexpected distortion, signal bleed through whatsoever. It's up to you which of those effects you find desirable and which you don't.
Sometimes also there are properties of circuits that seem important when building a model from theory, but have no effect on the actual sound - such as the often fabled soft sync of the Prophets which in fact is plain hard sync. Or the feed forward of the dry signal in OTA cascades which is below -300dB. In such cases it's not only allowed to cut corners it's also highly advisable.
Above that level you'll find there are numerous effects already that you do not wish to add to your model, such as humm, buzz, noise, and maybe some weird signal picked up from your smartphone.
In that complementary view, Analogue Modelling is defined by picking the raisins of analogue behaviour and omitting the crap. In fact, a too detailed model might be subject to lawsuits ("the software broke") while a sufficiently close approximation might just be what the customer had wished for.
That said, one of the most important ingredients of Analogue Modelling is thorough observation and the choices made based on that. A model that's solely based on theory might not convince. A model that sounds convincing needs to take effects into account that are not necessarily obvious when looking at schematics, e.g. part tolerances. Those effects express themselves as little DC offsets here and there, or a difference in level, unexpected distortion, signal bleed through whatsoever. It's up to you which of those effects you find desirable and which you don't.
Sometimes also there are properties of circuits that seem important when building a model from theory, but have no effect on the actual sound - such as the often fabled soft sync of the Prophets which in fact is plain hard sync. Or the feed forward of the dry signal in OTA cascades which is below -300dB. In such cases it's not only allowed to cut corners it's also highly advisable.
-
- KVRian
- 1324 posts since 15 Nov, 2005 from Italy
+1 for Urs. I remember two episodes... one was a community emulation of a TB-303 made on Nord Modular G2X. It was very though to get it through all the weird behaviour of what seems a simple machine.. but with such "flaws" that makes it unique: the envelope generators, the filter which is "wrong", the oscillators and so on. That's not only a matter of following the theory and model the circuit, but you actually have to get "into" the real machine, use it and feel how does it works in order to get an approximation of the same feeling.
Another nice adventure was many years ago... I were in a need of a good analog and I were waiting for a Moog Voyager to be delivered. In the meanwhile I made a synth on Reaktor and put there all those little "tricks" I learned from studying the behaviour of analog machines... such as drifting, jitters, errors in % of values every now and then... some very subtle saturations in the signal flow.. and so on. When the Moog arrived I was so surprised to see that my little creature made on Reaktor sounded as good as the Moog. I'm not saying it was a Moog but I had the same feeling and the sound was pretty good and alive.
my 2 cents
Luca
Another nice adventure was many years ago... I were in a need of a good analog and I were waiting for a Moog Voyager to be delivered. In the meanwhile I made a synth on Reaktor and put there all those little "tricks" I learned from studying the behaviour of analog machines... such as drifting, jitters, errors in % of values every now and then... some very subtle saturations in the signal flow.. and so on. When the Moog arrived I was so surprised to see that my little creature made on Reaktor sounded as good as the Moog. I'm not saying it was a Moog but I had the same feeling and the sound was pretty good and alive.
my 2 cents
Luca
-
- KVRAF
- 12555 posts since 7 Dec, 2004
Well if "the sound" is comprised of those elements I described which many likely do not want to emulate, physical reality can't be beat as these things are impractical to model in software especially considering the fact they contribute very little to "the sound" as Urs says.Fluky wrote:Yes it can. Digital plug-ins offer improvements over analog gear. But they don't (according to some) offer the sound.I've got a good model: it's called physical reality. Can't be beat.
Those people who say software can't reproduce "that sound" should go ahead and use "that equipment" if so desired.
I absolutely agree with them in fact. I'm neither for nor against. When I want "that sound" which includes all those potentially undesirable things which are a part of "analog", I use the circuits themselves. I consider it a waste of time to do anything else, of course!
When I want to fool around with software, I fool around with software without any significant care for modelling or emulating anything at all.
There is a point at which it doesn't make sense to attempt to reproduce one thing with the other... it is easier just to use the original if you want to go that far.
Otherwise as Urs said, when you don't want what "analog" actually means (a lot of negative stuff!) and you want to focus on producing a track, you use software when the advantages are useful to you.
Mixing and matching is also definitely a good idea.
One example of a synthesizer which is both digital and analog is the alpha juno. The oscillators are digital with analog waveshapers / filters. The envelopes and LFOs, glide and everything else are digital. The amplifiers / mixer are analog, the chorus is analog.
Such stuff is old-hat to the extreme.
Free plug-ins for Windows, MacOS and Linux. Xhip Synthesizer v8.0 and Xhip Effects Bundle v6.7.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
-
- KVRist
- 37 posts since 13 Jan, 2016
Fluky wrote:What techniques exist for analog modelling in the digital domain.
Like e.g. modelling a mono synth.
Is it all just some transfer function and "signals and systems" stuff?
What about its theoretical limits? Can analog modelling come close to real analog?
Look up for "system identification" . its just that. Is just signals and system.
Limits? You can get as close as you want and still.wont be close enouhg
-
- u-he
- 28063 posts since 8 Aug, 2002 from Berlin
So, now that "close" is out of the way, are you interested in the mere number of methods (I know just one, with variations in workflow) or are you merely interested in "how does it work" in general?Fluky wrote:What techniques exist for analog modelling in the digital domain.
Like e.g. modelling a mono synth.
Is it all just some transfer function and "signals and systems" stuff?
What about its theoretical limits? Can analog modelling come close to real analog?
-
- KVRist
- 60 posts since 7 Aug, 2013
broken record here, but what exactly does "real analog" entail? there are great digital emulations of specific analogue instruments that are still distinguishable when both are played live. when recorded onto a consumer level DAW, this distinction can prove unimportant. then again, some DAWs are also developed to sound "more analog", while there are excellent plugs that have been developed to create the sensation that digital recording contain aspects of traditional tape recording. and yet, in those still bricks and mortar music shops that still manage to stay afloat, CDs are losing ground to a sudden interest in old tech records. on a very limited local level there are musicians who are recording onto refurbished 4 track machines to produce tapes for immediate record pressing. there's a audio equipment show near DC coming up where very expensive components will be showcased playing, mostly, records and reel to reel. crazy times, really. the 16 bit, 44.1 standard will not disappear soon, but even at 24/96.4 are we really zeroing in on analogue? at some point technology will make a jump to a more sophisticated regimen of Fourier transform.
-
- KVRAF
- 12555 posts since 7 Dec, 2004
Broken record indeed. Vinyl fetishism is the ideal example to use.
Full stop.
Full stop.
Free plug-ins for Windows, MacOS and Linux. Xhip Synthesizer v8.0 and Xhip Effects Bundle v6.7.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
The coder's credo: We believe our work is neither clever nor difficult; it is done because we thought it would be easy.
Work less; get more done.
-
- KVRer
- 3 posts since 2 Nov, 2010
Of course, none of it even matters because you could spend $1bn building a measurable, exact model in the digital domain, warts and all, and some nerdlinger would STILL claim that 'it just doesn't sound quite right to my ears, I can still hear the 0s and 1s' 
-
- u-he
- 28063 posts since 8 Aug, 2002 from Berlin
Thing is, as tiring as the debate gets, and understandably one can't hear analogue-here-analogue-there anymore, it simply doesn't make non-analogue modelled filters sound any better.
One can't discuss away that certain analogue filters are simply the bar of reference. Neither can one discuss away that abstract or dsp-by-the-book models have a lesser chance to reach the ideal set by such reference than filter models based on an analogue model of that filter.
One can't discuss away that certain analogue filters are simply the bar of reference. Neither can one discuss away that abstract or dsp-by-the-book models have a lesser chance to reach the ideal set by such reference than filter models based on an analogue model of that filter.
