More on DAC stepping

[Ingo Debus]

Paul Schreiber wrote:
> Answer: the DAC isn't *accurate* enough to do that. For VCOs, we need 0.1%
> error to be "on perfect pitch".
> 0.1% of 83.3mv is 83uV. If the LSB is 83.3mv (1 bit count = 1 step) the
> linearity needs to be 0.001LSB. Nada.

IMHO that's more accuracy than needed. I think a pitch error of +/- 1
cent (= 1/100 semitone) would  still be excellent (I can't hear pitch
differences below 5 cents). For 1V/octave, 1 cent equals 1V/1200 = 833uV.


> Case #1: DAC swings all the output
> 
> If we don't range-switch (see case #2) and we want 10 octave (common) then
> we need to resolve 10V/83uV or
> 120K steps in our DAC! That's a 17-bit DAC.

If we choose the range such that one semitone equals an integer number
of LSBs (any reason why this shouldn't be done?), using a 12 bit DAC
with +/- 0.5 LSB error would result in an error of +/- 1/64 semitone =
+/- 1.56 cents for 10 octaves range (1 semitone = 32 LSBs). Still very
good IMHO.

Ingo