[sdiy] 2164 Expo correction

[David G. Dixon]

> << Anyway, based on the numbers I measured, I found that I 
> could model the exponential response (which should be linear 
> on a log-linear plot) more or less perfectly with a quadratic 
> function.>>
> 
> This may be correct but I would suggest checking the 
> measurements with other 2164s to make sure this is typical of 
> the parts and review your current Expo section to be sure it 
> is not introducing math functions that are clouding the issue.

Your definitely right concerning the specific 2164, as I believe they're all
slightly different.  I will definitely try another one tonight.  The rest of
the expo circuit, however, is simplicity itself (basically just an opamp
inverter with CVs fed through another 2164 connected to a constant voltage
to cancel the temperature coefficient).  And, in any case, I'm only using
the CV circuit to set the desired frequency -- I'm actually measuring the
voltage right at the VC pin of the 2164, so I can't see how that could be
affected by anything upstream in such a way that it could change the
internal gain factor of the 2164 itself.

If this quadratic error is common to all 2164s (and the one in there now
appears to be one of the crappier ones I've seen), then my calculations
suggest that the quality of the correction will not depend on the specific
JFET used, and (with proper biasing) should reduce the tracking error to <
0.03% across the musical range.