[sdiy] 2164 Expo correction

[David G. Dixon]

> Might it be possible to feed the pin from a negative 
> impedence buffer/amplifier, tuned to cancel out the effect of 
> the 450 Ohms?

Well, the NIC is definitely very cool (and a little bit strange), but it's
not really what I need here.

I've now measured the frequency-VC response of three separate 2164s in the
same circuit, and I also re-measured the third chip in Class A mode (7.5k
from 15V to the MODE pin).  Based on these measurements, a clearer picture
is beginning to emerge.

First, there is a range of frequencies where the exponential gain is
virtually flat: between about 64 and 512 Hz.  In my circuit, this
corresponds to 1V/O CVs of about -5V to -2V.  Above this level (at high
frequencies), the required voltage begins to increase, and I think this is
mostly due to comparator delays in the VCO core.  Below this level (at low
frequencies), the required voltage drops off slightly.  This, I believe, is
the nonlinearity of the exponential response in the 2164.  Happily, all the
chips I tested behave pretty consistently.

Second, the average gain factor in the midrange (where the response was more
or less flat) is about 182 mV/octave -- decidedly less than 200mV/octave as
advertised.

Third, running the 2164 in Class A mode increases the required voltages
across the entire range by about 3 to 5 mV, but doesn't change the basic
shape of the response.

On the basis of all this, and considering what a pain in the ass calibration
would be with the additional JFET-based correction scheme, I think I'm going
to go with my design as is, tune to the frequency range around 256 Hz, use
the BJT-based correction scheme to pull up the high frequencies, and live
with the small tracking errors in the low bass, where they are less obvious.
I may drop the reference frequency from 2kHz to 1kHz (by doubling the
voltage-to-current-converting resistor in the VCO core) to increase the
range of decent tracking down another octave, depending on how "happy" I am
with the final tuning.