[sdiy] MIDM-CV DACs, Part #3 (last)

[Seb Francis]

Theo wrote:

> > e.g. DAC714P.  The datasheet also gives another parameter "monotonicity"
> as 14bits.
>
> This just means you don't want to listen to this DAC ;)
>
> Don't know if I put it "technically" right, but you could think of this as:
> The 15th bit step is smaller than the error.
> As a result the output voltage is no longer guaranteed to increase
> "monotonically".
> This is math geek speak for: the output voltage might even get lower with
> the next step up!

Right, so the max error is <= the 14th bit (which is consistent with the quoted differential linearity error of +-4LSB)

But Paul's stated the rule "for every 1/2LSB of INL error, you effectively give up 1 'perfect' bit".  This would suggest that this DAC is effectively a perfect 8 bit DAC.  I'm not sure how a DAC with an error > 0LSB could ever be 'perfect'.  Surely a 'perfect' DAC has exactly equal voltage steps.

The (not differential) linearity error is also quoted as +-4LSB.  If my understanding of the previous emails is correct the error in this DAC when used over a full MIDI scale is:
10.58V / 65536 * +-4 = +-646uV
or
12700 cents / 65536 * +-4 = 0.78 cents
which puts this DAC well within limits for an accurate full-scale VCO controller.  Or does it?  I've probably overlooked something again

And it seems like there's absolutely no point in even setting the last 2 bits since it's not guaranteed whether the voltage will go up or down .. seems a bit pointless to me.  I'm guessing this is because of the way they are graded after manufacture - all the really bad cases get labelled "DAC714P".  This kind of suggests that they are almost guaranteed to be worse than the next grade up (DAC714HB) which is +-2LSB.  Not sure if this thinking is correct.

Seb [on a quest to build an accurate MIDI2CV]

P.S. Thanks to everyone who keeps posting all this useful knowledge.  I would be lost without you guys (I'd have to go to university or something!)