DAC and real musical instruments

[Tony Allgood]

Here's my bit....

12 bits give you 4096 steps. We are looking for at least 3 cents
accuracy in pitch. That is 3% of a semitone step. One semitone step is
83.3mV, thus our tolerable error is 2.49mV, ie. +/-1.25mV.

Now to get 127 notes from the midi keyboard, you will need 127x83.3 =
10.58V. The DAC ref will have to be a tad higher than this to account
for the fact the lowest DAC output is zero and not 1/4096 of Vref.
However, 4096 steps will give a LSB step of 2.58mV, assuiming the DAC is
12bit accurate, but it won't be. Typical DACs will feature 1/2LSB, so we
now have a discrpency of 2.58 + (2.58/2) = 3.88mV. This was Paul's
figure he quoted. This would then rule out using 12bit DACs for midi-CV
convertors.

However, I believe that 12 bit is good enough for us, even when used in
conjunction with other digital instruments. The discrepancies at this
level, if they are audible, will ADD to the sound of your synth. We all
know that DCOs are stale sounding, and autotunes can stiffle the sound
of a good analogue (OB-8, poly600) A worst case error of 2.9mV is
acceptable to my ears and the 'randomess' of its position across the
whole keyboard will add to its anologueness. In fact, although the DAC08
is to be avoided, the DAC0800HP, a better spec'd DAC, is adequate for
most people, and that is only a 9-bit accurate DAC. Tim Orr, designer of
the Transcendent series of synths and EMS, reckoned that 10-bit accuracy
was good enough. And to my ears, at least, a musical instrument can be
built around this preposition.

And finally, I just have to wonder just how accurate our series resitor
chains are that sit under our favourite keyboards.

Regards,

Tony Allgood  Penrith, Cumbria, UK

MidiDAC Superior midi-cv, SuperLadder, TB303 clone and Filter Rack

http://www.techrepairs.freeserve.co.uk/projects.htm