well, first of all you can never neglect the effect of the physical parts in the audio chain.
So if you listen to an mp3 audio example from a (say) Nord Modular, the 'sound' has passed a specific converter at a specific samplerate through specific op-amps, capacitors, resistors (some parts also have inductive side effects).
Regular tolerances of parts (add the respective cabling...) make it almost unpredictable, as you have to consider the same chain of tone mangling (in case the analog output is sampled) before an mp3 is generated.
Of course the afforementioned parts (mostly) have a good (and constant) quality today, so it will not change the sound completely, but on a detail level there may be noticable differences.
(sidenote) Some of the most crucial parts are capacitors, for one because their 'audio capabilities' have a wide range and second because circuit board layout for mass production is usually driven by pure numeric figures, part size and costs, ignoring audio quality aspects.
Look at the link in the 'Gristlelizer' thread - on the circuit board you see 3 Styroflex capacitors, living fossils, so to say...
as they are out of production and there's only new old stock - in an electronic shop you'll pay at least 3 Euro for a single item, while a 'regular' capacitor is 10 cent or so...
Needless to mention that Styroflex caps have legendary audio properties, but those 'rolls' also have an enormous size and cannot be handled in machines...
...In order for the sharc technology to truly mimic another synth like the ones mentioned, it would have to go deeper than just creating the plug in exactly the same...
regarding '100% analog synths' and some hybrid ones like the 'Oscar' it's already difficult to build a 'common' model, as no 2 specimen will be exactly the same, aging of parts has an additional influence to what was already mentioned.
for example the discrete Moog ladder filter requires that each transistor is measured individually and a 'matching' one for each filter stage is found.
The thing is simple and dead cheap to build, if you just solder the parts together - but it will not work (or not as you expected)...
you need a 'stack' of at least a few hundred transistors to find the perfectly matching ones - and of course time (and patience)
the approach Creamware took for the Moog and Sequential models was to add the influence of the physical circuit to the 'idealized' math models of the respective filter and amp stages.
Obviously they succeeded, according to reviews, but it's not just a time consuming process - it also requires a lot of measurement gear and knowledge (!) - and of course vintage gear in mint shape...
A Virus, Nord or Novation are entirely digital synths - as mentioned, there's an analog conversion stage that influences the sound, but afaik all those have digital outputs, too.
As such the 'true modelling' would be much easier, as each sound is 100% reproducable.
It would somehow break down to model-the-math-model, imho a completely ridiculous approach, but a great way to waste time and energy
(and I don't think the 3 devices discussed here feature that approach)
It's perfectly okay to rebuild the structure in Scope and (possibly) add some 'soundshaping' to match the originals character.
One should not neglect the influence of parameter 'access' - that's what makes almost every Scope synth that's focussed on the (same!) Waldorf Wavetables still individual.
Regarding the Sharc math model - well, afaik this DSP was designed from the ground up with high end audio in mind, and it got the appropriate library right from the beginning. That's what makes it so good in it's domain.
Consider that the folks responsible for that part of the code are from a 'one out of tenthousand' elite - you just don't write that sh*t at home... and if you could, you probably wouldn't waste your time with stuff for some audio geeks, but would be a highly sought expert in industrial, medical or military (oops) developement...
Actually I think most Scope developers aren't even 'real' programmers at all...
DP or SDK are developement systems that generate the output that will finally execute on the chips by a graphical tool - you cannot even code in DSP assembly within that environment.
That makes it easy to focus on the result instead of dealing with code blocks.
You can easily tell by the output of MacCyrano, Shroomz, Sharc and others how effectively this works - once you get beyond the basic aquaintance with the system.
cheers, Tom
The Knowledge!!!
