DAC "error"

[Paul Schreiber]

> I believe 12 bits would suffice for 10 octaves of semitone-step control
> if the DAC had 1/4 LSB relative accuracy.  Using the seven most
> significant bits for semitone selection and holding the 5 least
> signifacant at zero:
>
> 127 * 1/12V = 10.583 full scale DAC voltage
>
> 10.583 / 4096 = .002584V LSB step size
>
> .00258 / 4 = .000645V 1/4 LSB error
>
> So the error would be +/- 645uV.
>

No, the problem is that the "error" is even if the DAC is "perfect", you can
hit in steps of 2.584mv. The error is 2.584/83.3 or 3,1%! Barf!

People that don't 'see' this error usually:

a) control 1 VCO
b) play fast passages

If you have 3 VCOs all fed by this arrangement, all tuned in unison at 50Hz
at 0.0V control,
then feed then say 7.000V to go up 7 octaves, you can hear the effect!

For most VCOs out there, the *total error contribution*, which is the DAC
RESOLUTION and the DAC
ACCURACY together must be less than 600uV. For "perfect pitch" the number is
more like 100uV!!

Paul Schreiber