Questions - Fundamental differences in Analog and Digital synths OR Intrinsic problems in each

[hugoderwolf]

hardware digital synths are probably meant to cope with whatever they are running, so the latency is probably pre-determined, and you should never get into such a situation

Nothing stops you from implementing a sophisticated voice and load management in a plugin.

[antto]

running it in realtime with other things is another story, right?

i meant, with hardware digital synths you could have 20 and use them at the same time (if you have the money)
if you ran 20 software synths your CPU would have to deal with all of them

of course you cannot compare what's better or worse, cheaper or more expensive this way, since there lots of things to take into account
i'm just saying that hardware digital synths have a CPU each for themselves, unlike software synths where it's 1 CPU for all

[Richard_Synapse]

voices? it's obvous..

in an analog synth - each voice is basically physical copy of the same circuit, thus multiply the cost of each voice by the number of voices and you get a very expensive synth

on the digital synth it's easy, it's usually

Code: Select all

while (i < num_voices) { ... }

It's not always that simple, an important aspect is voice stealing. In analog synths a sophisticated voice management is not possible, it's just round-robin or some other simple logic. A digital implementation offers more options.

Richard

[HeavensOnEarth]

Yes, polyphony is straightforward in the digital realm but if it needs to exactly correspond to an analog synth's behavior, in some cases it's difficult; that's true in the Yamaha CS70 for example. It, like many differences, falls in the second of two categories: accurate emulation. Factors such as sawtooth shape, filter response etc. are those required for accurate emaulation of an existing analog system i the digital domain.

The first and more important category, maybe, is distortion and artifacts resulting from sound creation from discrete samples rather than a continous wave. Once also the sample resolution was limited, but that is less problematic these days.

Perhpas there are other differences that don't appear in the above two categories.

[HeavensOnEarth]

Yes, voices are easy in the digital domain. The problem with emulating some hardware implementations is what happens when there are limited voice and something like the following note sequence occurs. Say there are four voices:

C3 on, C4 on, C5 on, C6 on
c7 on <-c3 replaced round robin. Easy.
C4 off
D4 on, D5 on <- which does D5 reoplace?

Now a simple round robin no longer works. Some keyboards didn't play the full number of notes possible in all polyphonic situations, because some of the older keyboards have analog CV zones across portions of the keyboard, so the new note can only play if it is one of the separate zones. Also, in the Yamaha CS50~80, current installation have been modified since the original installations so the PCB circuit diagrams don't necessarily provide the correct information, and the only way to assure identical functionality was to perform tests on the original hardware. Sometimes the polyphonic response was changed during the production cycle also.

[AdmiralQuality]

The voice assigners in classic analog synths WERE digital. That's why we didn't have polyphonic synths until microprocessors started appearing in the mid-'70s. (What's the first one, the Oberheim 4 Voice?)

The one thing software can do that hardware couldn't do is instantiate an unlimited number of voices. But the voice limitation was often a plus in classic poly-synths, it enables certain playing styles that don't hold up without the voice stealing behavior.

And aside from round-robin, the other common polyphonic assignment mode was cycle-assignment. This would assign a new synth voice to successive note-ons of the same note. (Typical round-robin schemes would re-trigger the existing voice when the same note was repeated.)

Cycle assignment is particularly useful for percussion sounds and is even more impressive when combined with velocity sensitivity.

Any software synth can do ANYTHING, assignment-wise, that any classic "analog" could do. The voice assigners weren't analog, they were a computer that scanned every key on the keyboard, fast, over and over, to determine which are down, and then send appropriate Gate and CV signals to the internal monophonic synth modules.

[Richard_Synapse]

The voice assigners in classic analog synths WERE digital. That's why we didn't have polyphonic synths until microprocessors started appearing in the mid-'70s. (What's the first one, the Oberheim 4 Voice?)

The one thing software can do that hardware couldn't do is instantiate an unlimited number of voices. But the voice limitation was often a plus in classic poly-synths, it enables certain playing styles that don't hold up without the voice stealing behavior.

It may be useful in some situations, but ihmo a synth should by default function more like a piano, capable of sustaining many voices in parallel. Furthermore using unison in old analog hardware destroys the polyphony, so that "fat" sounds must be monophonic, limiting the sound design options.

Richard

[AdmiralQuality]

The voice assigners in classic analog synths WERE digital. That's why we didn't have polyphonic synths until microprocessors started appearing in the mid-'70s. (What's the first one, the Oberheim 4 Voice?)

The one thing software can do that hardware couldn't do is instantiate an unlimited number of voices. But the voice limitation was often a plus in classic poly-synths, it enables certain playing styles that don't hold up without the voice stealing behavior.

It may be useful in some situations, but ihmo a synth should by default function more like a piano, capable of sustaining many voices in parallel. Furthermore using unison in old analog hardware destroys the polyphony, so that "fat" sounds must be monophonic, limiting the sound design options.

Richard

Disagree, a piano should function like a piano. But pianos don't have long release times, like you can dial in on a synth. (Damper down isn't release time, it's sustain time.) One obvious example is being able to play a 6 note chord, lift your hands while it's continuing to go in a long release stage, then hit a different 6 note chord and all the old notes go away. You can even get more subtle levels of control than this where you intuitively know which voices it will steal for the new voices.

But if you didn't play poly-synths back in the day, you won't know what I'm talking about. Again, voice limitations were a FEATURE. To not model them means you can't match the behavior of classic synths.

And plenty of classic synths sounded fat without unison. Unison was for monophonic sounds exclusively on the classic 'boards. (Jupiter 8 could do duo-phonic Unison with a split/double patch.)

Sure, you can support new features that were impossible on the classics, as I already mentioned. But I just wanted to summarize how classic polyphonic voice assigners worked, and point out that they were ALL digital, even in "pure" analog synths.

[Aiynzahev]

Well, this thread has taken on a life of its own, but I really wasn't talking about any classic synth.

What I meant was what are the intrinsic draw-backs when creating synths in

1) The Analog Domain

2) The Digital Domain

For example 1) Noise, Parts 2) Aliasing

But things more in depth than that.

One particular question was about the envelopes, are there any problems getting those to be snappy in a digital synth? Sylenth1 has very snappy responsive envelopes, some digital synths (which are not emulations) are slower.

[AdmiralQuality]

Well, this thread has taken on a life of its own, but I really wasn't talking about any classic synth.

What I meant was what are the intrinsic draw-backs when creating synths in

1) The Analog Domain

You are limited by what can be done with analog electronics. No samplers, etc. FM(PM) is hard. Additive is possible but not practical.

You are limited in power, you can't simply "wish" more synths and voices into existence like you can in software.

You are limited by physics, so some noise floor will always be present. Whether this matters though, considering the listener's playback hardware ultimately must be analog too, is a matter of opinion. Some might even prefer a little bit of noise floor to make things feel more natural and organic.

2) The Digital Domain

For example 1) Noise, Parts 2) Aliasing

But things more in depth than that.

One particular question was about the envelopes, are there any problems getting those to be snappy in a digital synth? Sylenth1 has very snappy responsive envelopes, some digital synths (which are not emulations) are slower.

There is nothing you can't do in the digital domain. In theory, anyway. Aliasing, poor envelope performance, are just details of implementation that can certainly be addressed by any developer who cares to and knows how.

But I'd say the biggest plus to digital is the unlimited ability to clone your instruments to as many instances as you need at the moment, and you don't need to fit them in the trunk of your car!

That and the many other synth architectures that are available. With analog it's really just subtractive. Digital, you can do anything. You could model every molecule in a guitar and every particle in the atmosphere of the room. The only limit is what your CPU can withstand, and even then you can work around that in many ways.

In theory anyway. There's no shortage of crappy software synths out there.

[Richard_Synapse]

Disagree, a piano should function like a piano. But pianos don't have long release times, like you can dial in on a synth. (Damper down isn't release time, it's sustain time.) One obvious example is being able to play a 6 note chord, lift your hands while it's continuing to go in a long release stage, then hit a different 6 note chord and all the old notes go away. You can even get more subtle levels of control than this where you intuitively know which voices it will steal for the new voices.

But if you didn't play poly-synths back in the day, you won't know what I'm talking about. Again, voice limitations were a FEATURE. To not model them means you can't match the behavior of classic synths.

Voice limitations were not intended as a feature- they existed because making polyphonic analog synths was expensive as hell, as each voice needed its own voice board. The polyphonic stuff still sells for silly prices on ebay.

About playing styles, it's surely clever to exploit limited polyphony in a creative way, but that doesn't mean it should be like this by default. A synth is typcially played via a keyboard, and that is a polyphonic input device like the piano. So it shouldn't be crippled to monophonic, duophonic or six-voice playback unless the user explicitly wants that. Hence I'm not disagreeing with you, just pointing out its the options that make the difference between analog and digital. In analog you have few of them, in digital anything is and should be possible (ideally).

Sure, you can support new features that were impossible on the classics, as I already mentioned. But I just wanted to summarize how classic polyphonic voice assigners worked, and point out that they were ALL digital, even in "pure" analog synths.

Yes, you're right of course.

Richard

[Ichad.c]

From another perspective - analog (especially discrete designs) can be hopefully fixed - when they break 10-20years down the line, and they will - most probably retain some of their re-sale value. Would you buy an old license from somebody of some of those first 16bit plug-ins, would it even work on a modern PC? Probably not! IMHO, Digital has a poorer reputation - not because it's inferior in any way - it's just that analog sticks around for longer - giving users more time to explore it. Digital is quite 'fast-paced' - todays rocket-science will be tomorrow's toaster.

P.S. No disrespect to anything digital - I'm a coder myself, have old digital and analog gear - as well as software that I love. My remarks are only an observation.

Regards
Andrew

[AdmiralQuality]

Actually I think, at least on PC, software has a longer lifetime.

And there were 16 bit plug-ins? I don't recall computers being capable of real time audio processing back in the Windows 3.x days. They could barely feed a 16 bit SoundBlaster!

And MAYBE analog can be fixed. But what happens when there's no more voice chips for your Juno or no more Curtis chips for your Prophet? There are a lot of non-discreet components in most classic synths, particularly that digital voice assigner I was talking about.

Most of the time old gear is broken just because a circuit trace is literally broken. But when it's a dead component, you better hope you can find one in stock somewhere.

I doubt there'll ever be a time you can't run Win32 software. Even when the platform is long gone there will be emulators forever. And my old 16 bit code from before NT/95 still ran fine until they dropped it in Windows Vista! 32 bit code from Win95 still runs fine on the latest versions of Windows. That's approaching a 20 year lifespan!

And then there's the Mac.

[Tzarls]

And my old 16 bit code from before NT/95 still ran fine until they dropped it in Windows Vista!

Funny... I've run Cakewalk 3 (1994) under Windows 7... I think it was made for Windows 3.1?

[AdmiralQuality]

And my old 16 bit code from before NT/95 still ran fine until they dropped it in Windows Vista!

Funny... I've run Cakewalk 3 (1994) under Windows 7... I think it was made for Windows 3.1?

Maybe. I could have sworn they said they no longer supported 16 bit apps after Vista, but I could be wrong.

I was using Cakewalk back in 1989, on an IBM XT with a Roland MPU-401 in it! Of course, it was MIDI only back then, no hard disk recording (for that matter, I don't even think that machine HAD a hard disk! It was all on floppy!) I remember syncing it to a 4-track reel to reel for at least one of these numbers I used it on...

http://soundcloud.com/admiral-quality/camouflage

http://soundcloud.com/admiral-quality/words-in-the-sky

http://soundcloud.com/admiral-quality/aperfectmachine