[sdiy] CV from soundcard

[Richard Wentk]

On 8 May 2009, at 00:22, Paul Schreiber wrote:

>>
>> Most soundcards have *much* better converters than a typical MIDI/ 
>> CV  box, so aside from that minor scaling tweak you should get  
>> better CV performance overall.
>
>
> Ummmmmm.....not hardly. #bits does NOT equal *accuracy*. Most audio  
> DACs are quite crappy for DC accuracy (that's why they are designed  
> to be *audio* DACs). The ear is quite forgiving having adjacent  
> samples "off" DC-wise at a high (say 48Khz) sample rate. You are  
> more likely to hear transient *artifacts* from audio DACs (such as  
> clock jitter) than absolute sample-to-sample, DC accuracy. Besides,  
> if sample N is "off" but is N+1 is "correct"  21us later, you are  
> not going to leap up out of your chair because it "didn't sound  
> right".

Depends now much N is off by, and what happens next, surely.

> It's a different story altogether when forcing a 'static' value  
> (albeit one updated at the sample rate), in fact I will bet that  
> voltage will "wander" sample-to-sample but probably below making a  
> VCO shift enough to hear (it's integrated anyway).

What does the hypothetical static performance of systems which are  
clocked, and always have been clocked, have to do with real world  
performance for this particular application?

Has anyone ever, in the entire history of the universe, tried to sell  
a non-clocked mainstream commercial soundcard or audio interface?

> Many sigma-delta DACs are sort of like DRAMS: they are designed to  
> be a *clocked* system: they themselves are not 'static' in the  
> sense a traditional DAC is. Also, what about the DC accuracy/drift  
> of the components "downstream" of the DAC itself? Are you going to  
> convince me a JRC4556 or 4580 headphone op amp that cost 11 cents  
> is "good enough" as a CV source?

Considering that many commercial MIDI-CV boxes think 12 bits is  
adequate and 16 bits is extra super special, and not a few of them  
use budget op-amps to follow, I don't think there's an issue there.

Richard