Analog or digital ?

[WMP]

For me it was the un natural bass on mp3 2 that gave it away first, next was the oscilator itself...far too accurate like an autiotune flute player lol.

Next it was the filter sound, which whilst being very good indeed for a vsti still gave it away.

Vsti synths all seem to have this weird bass sound to them as if it's a bass gain at 40-60hz but a very un natural one, like a cheap digital eq on a car cd player. It could all be in my head, but vsti's seem to have that sound when it comes to bass.

[WMP]

.......I'm convinced that quite a lot of that good sound vintage sound has to do with distortion in the right places, which isn't very easy to recreate digitally. A P08 doesn't sound good, but a 1980 Moog also sounds better than a 2008 Moog (at least to my ears and many others).

Yeah I'd go with that. I'm in the middle of building an Oakley Modular and it says to use 1% resistors to get the oscilator as accurate as digital oscillators The whole point of this excercise is to not sound anything like modern oscilators lol, I want the drone and drift which is why I am using 5% resistors.

The modern analogs should stop trying to be so damn clean and accurate, it must have something to do with the size of the circuit boards as well. I mean the more modern analogs weigh nothing and have tiny components... where as the circuits in my CS-30 are mentaly complex and it weighs a ton.

The Andromeda A6 has more routings etc than the CS-30 and it is polyphonic, yet the circuit boards total space is less than the CS-30's single voice! And I know which I prefer in a raw oscilator shoot out (personal preference of course).

[living sounds]

Yeah I'd go with that. I'm in the middle of building an Oakley Modular and it says to use 1% resistors to get the oscilator as accurate as digital oscillators The whole point of this excercise is to not sound anything like modern oscilators lol, I want the drone and drift which is why I am using 5% resistors.

The modern analogs should stop trying to be so damn clean and accurate, it must have something to do with the size of the circuit boards as well. I mean the more modern analogs weigh nothing and have tiny components... where as the circuits in my CS-30 are mentaly complex and it weighs a ton.

The Andromeda A6 has more routings etc than the CS-30 and it is polyphonic, yet the circuit boards total space is less than the CS-30's single voice! And I know which I prefer in a raw oscilator shoot out (personal preference of course).

Component tolerances may be responsible, but the old parts were also different. The transisors and opamps produced had more distortion and different characteristics. The crucial parts responsible for the sound of many of the Roland SH/TB/TR boxes become unavailible. It's not all vodoo. For example, the old carbon resistors from the 50s did produce more 2nd harmonic distortion in certain situations, high voltages and high voltage swings (so they only make sense in certain tube gear applications). The core material of old transformers is no longer availible, and there were heaps of other differences.
As far as analog synths are concerned I've yet to hear a synth made in the last 15-20 years that sounds as good as these old classics. Damn it.
I'm not a big fan of Andromeda either.

[Urs]

I want the drone and drift which is why I am using 5% resistors.

See, Cypher can already do that quite well, even though it may not be as audible in the mp3.

Frequency drift by temperature is about the *only* thing where analogue is less precise than digital, unless you have broken cables or large amounts of induction noise. Other than that, digital synthesis methods are all about approximation of continuity, which by principle is impossible to achieve. Analogue systems however operate in the realm of continuous signals (as opposed to discrete signals) and thus are as precise as it gets unless you have really shaky parts.

The advantage of highly stable oscillators is that you can have anything from optimal tuning to chaotic drift. A noise generator and a simple lag (both massively cheap parts in either analogue and digital world) will do the trick just fine. Any good synth (analogue or digital) should always provide for this.

[WMP]

I hear what you are saying URS and agree that it can be done with software, my only problem is that whatever you do to the sound be it tuning etc.. it always has to be resampled at the sound cards sound rate i.e 44khz.

Even the tinyest amount of aliasing has a knock on effect when it comes to the next stage of processing.

It's just me being totally anal about the sounds I like. I suppose a good anaology would be:

Vsti Synths = T Pain (autotune)

Old Analog = Tori Amos (going off on one of her wailing sessions... and I don't mean fishing)

[psychematic]

That being said, this was a great example of the differences between real analog and VA, so if you're on the fence with wanting the real deal or just a VA then listen to these two examples for some sort of guidance.

hehehe... clearly an unbiased opinion

Not really. Biased for my wants and needs, but I certainly don't think that those who buy a VA are fools. I enjoy psychedelic music and real analog's unpredictable-ness adds to the flavor. I also like the fact that my hardware doesn't sound like everyone else's (maybe) pirated VA.

But that's just my preferences and I certainly don't expect others to follow suit.

[james0tucson]

There are certain qualities and characteristics in real analog that digital cannot match.

On the other hand, those qualities and characteristics were preserved in an MP3, which is completely digital.

Different thing altogether

If you say so. I say the lines are not as absolute as they are made out to be.

oh, I agree with that. I am only talking about current practice, not ultimate reality!

Recording the analog synth is not nearly as complex as digitally generating the sound of an analog synth. It is just a matter of realtime cpu power. If cpu's were 100 times faster developers would close the gap fast.

I don't know about that. What we're talking about is modeling hysteresis -- a specific type of chaos. It may be a hard problem (probably an NP-complete one), and with that in mind, 100x or 1000x faster computing won't get you significantly closer to solution.

These tend to be the class of problems where "closing the gap fast" would require exponentially increasing computational capacity, which of course isn't happening.

[hakey]

These tend to be the class of problems where "closing the gap fast" would require exponentially increasing computational capacity, which of course isn't happening.

But Moore's Law, which is closely related to (bang for the buck) computational performance, is an exponential rule.

[uvacom]

I don't know about that. What we're talking about is modeling hysteresis -- a specific type of chaos. It may be a hard problem (probably an NP-complete one), and with that in mind, 100x or 1000x faster computing won't get you significantly closer to solution.

These tend to be the class of problems where "closing the gap fast" would require exponentially increasing computational capacity, which of course isn't happening.

Absolutely. But even if these inherent deficiencies in analog modeling (feedback modeling, for example) is an NP-complete problem, there is still a wealth of territory to be explored in approximations which only try to model or approximate the resultant output and not the underlying behavior, and there doesn't seem to be any shortage of creative software synthesizer programmers.

[Mans]

I hear what you are saying URS and agree that it can be done with software, my only problem is that whatever you do to the sound be it tuning etc.. it always has to be resampled at the sound cards sound rate i.e 44khz.

Right... And @ 44khz an analog signal can never be faithfully represented. You'll just have to invite Tori over each time you'll want to hear her sing. Not that that's a bad idea of course.

[WMP]

You misunderstand me, resampling to be done well rquires a program like Voxengo's rate brain pro which uses a lot of cpu and is slower than realtime depending on your cpu.

Vsti's have to do it in realtime and use as small amount of cpu as they can. This is why oversampling can kill your cpu.

Sure 44khz is fine for recording, try your eq's and vsti's at a higher sample rate,... you'll be suprised. But that's another argument altogether.

[Urs]

Vsti's have to do it in realtime and use as small amount of cpu as they can. This is why oversampling can kill your cpu.

The process of oversampling in a VSTi is is really cheap actually, as it usually is an integer factor (2x, 4x), has to be latency free and thus not phase linear. It doesn't take up much cpu and fits the bill perfectly well.

The actual algorithm that is oversampled is a different story though. It has to be processed twice or four times as often, or even worse. That's what takes up cpu, but it isn't really the upsampling/downsampling algorithm itself.

When doing sample rate conversion though one has to deal with odd factors, not just 1:2 or anything, and one usually wants to yield linear phase response. In that case one commonly goes with convolution, which is much more cpu intensive than simple upsampling/downsampling.


Another story then is samplerate conversion while preserving pitch. That's basically time stretching and it is a really cpu intensive task - depending on the quality.

[Mans]

You misunderstand me, resampling to be done well rquires a program like Voxengo's rate brain pro which uses a lot of cpu and is slower than realtime depending on your cpu.

Vsti's have to do it in realtime and use as small amount of cpu as they can. This is why oversampling can kill your cpu.

Sure 44khz is fine for recording, try your eq's and vsti's at a higher sample rate,... you'll be suprised. But that's another argument altogether.

Isn't that the reason that many plug-ins allow you to set a different internal sample rate for real time and offline?

[zerocrossing]

...so if you're on the fence with wanting the real deal or just a VA then listen to these two examples for some sort of guidance.

I disagree. I think that Synth Squad, and probably other decent VAs could actually do a much better job at nailing the first sound than the patch we heard suggests. Maybe not perfect, but better.

[pdxindy]

...so if you're on the fence with wanting the real deal or just a VA then listen to these two examples for some sort of guidance.

I disagree. I think that Synth Squad, and probably other decent VAs could actually do a much better job at nailing the first sound than the patch we heard suggests. Maybe not perfect, but better.

and more importantly, sound lively and good. That Cypher audio was sterile for Cypher. I guess Andy just copied the technical elements to match the general sound.