even 'worse' ... any part of the system can apply any kind of math, int, float, variable data length...garyb wrote:yes, and no.
...
it's not either/or, any hybrid form works. the math will be done in the system given the task. Scope is really external hardware to the computer, in that it's a seperate processing unit. there's really no way for one to be on the other. they can share info, but not tasks(for the most part).
it's up to the respective processing block, wether in the DAW, a VST plugin, or Scope.
The programmer should choose the math format according to what's being coded - and there IS plenty of options...
Someone posted (in the Modular section here) a link to a VST clone (a very simple one) of the Modular.
It's certainly well programmed, but the author doesn't get along with audio dynamics inside his system at all
(which is a bit irritating as he mentioned 'intelligent' scaling of audio level within the patches as a particular strength of the device)
Anyway... this is a good example for (failed) range management/control, which is one (important) aspect of scaling numerics and choosing an appropriate format.
If you tweak similiar modular patches in Scope, you will immediately notice a more balanced handling, a smoother way of dialing, as there's obviously more headroom in the whole setup. The numerics fit the functions, so to say...
The afforementioned VSTi will drift into distortion very quickly and you often have to manually 're-set' a dial to avoid the red part of the meter, but then find yourself in a situation that the 'initial' part becomes too thin.
But... (except for distortion) this doesn't influence fidelity of the sound too much - the filter output of high resonant parts was nice indeed, but you couldn't do specific (continuous) tweaks, as you had to interrupt and set back levels on the way.
More math savy folks will be able to properly explain how much the frequency content (which we perceive as 'sound character') of a signal depends on bit deepth, but imho its way more about dynamic range.
My usual 'hands-on' hint is to render a file into an 18bit and a 24bit version and try to tell the difference in a random sequence...
I fully get your point that sometimes you want to keep what you recorded in the most pristine form.
Me too - but I've amended this attitude somewhat more into the 'practical' domain.
To be honest the biggest driving force behind this was sound, which simply was 'too good to be true'.
Kensuguro recently gave an example with Garritans Brass library (in the Music section).
These horns were recorded with world class gear (probably), but they just don't sound like what 'brass section' triggers in our heads.
My suggestion was to 'spoil' the individual tracks to give it back some 'blood' or funk or whatever you call it...
Now how great is that ? I bet those horns would have been way more cutting it in 12 bits
If I needed a Piano on a track, one of my hottest candidats would be from the Korg DSS-1 sample library
Hear what you can do with a 12 bit Piano...
My point is quite simple: 'sound character' is defined mostly in the 300-900 Hz range, 'precision' is faked frequently by extended highs beyond 8khz, or even that 16khz 'airy' bs...
The focus is way too much on technical figures than on the sound itself as it reaches our ears.
That's why I stopped worrying and started to love the bomb
The little synth mentioned above gets raving reviews by some of it's fans at KVR - its fine for an occasional (!) line of notes or an fx, but the tone is so boringly balanced and precise.
ok, lot of bla beneath the topic (question), so at least a rule of thumb (...or what my self resorted to...) :
make a kind of prototype setup that covers Scope and VST sources.
If you monitor via Scope, compare the final DAW rendition to the original sources.
In case you notice any significant differences, try to find out how it's been altered and adjust processing accordingly.
That way you learn to anticipate what your specific system will do and you compensate in between (if required).
You won't find any solution in contemplating about numeric formats - there are too much and you cannot(!) know how they were applied internally.
keep it simple, Tom
