[sdiy] 2164 gain response variability

[David G. Dixon]

OK, so I rebuilt one of my friend's 2164 Expo VCOs according to my new
design, but I changed the current limiting resistor such that the frequency
at CV = 0V is roughly 600Hz (should be 606Hz).  Here's the tracking I got at
octaves and fifths, using (more or less) Yves Usson's tuning method from
37.5Hz to 600Hz, and HF trimming to 4800Hz (absolute and relative pitch
errors in parentheses):

-5.417V    14.00Hz (  -0.06Hz, -0.44%)
-5.000     18.68   (  -0.07  , -0.37%)
-4.417     28.10   (  -0.02  , -0.09%)
-4.000     37.50   (   0.00  ,  0.00%)
-3.417     56.35   (   0.10  ,  0.18%)
-3.000     75.15   (   0.15  ,  0.20%)
-2.417     112.7   (   0.2   ,  0.18%)
-2.000     150.3   (   0.3   ,  0.20%)
-1.417     225.2   (   0.2   ,  0.09%)
-1.000     300.0   (   0.0   ,  0.00%)
-0.417     450.0   (   0.0   ,  0.00%)
 0.000     600.0   (   0.0   ,  0.00%)
 0.583     900.0   (   0.0   ,  0.00%)
 1.000    1800     (   0     ,  0.00%)
 1.583    2400     (   0     ,  0.00%)
 2.000    3600     (   0     ,  0.00%)
 2.583    4800     (   0     ,  0.00%)
 3.000    7180     ( -20     , -0.28%)
 3.583    9513     ( -87     , -0.91%)
 4.000   1403X     (-41X     , -2.6 %)
 4.583   1830X     (-90X     , -4.7 %)

Tracking is between 0.1 and 0.2% from 28 to 300Hz, perfect between 300 and
4800Hz, and begins to fall apart beyond the +20dB point of the 2164 VCA, as
expected.  That's seven good octaves, four of them perfect.

I'd still like to know how this performance stacks up against conventional
VCO designs with unmatched discrete transistors, matched discrete
transistors, and matched transistor pairs.

Remembering that the whole point of this design is to eliminate the need for
a tempco resistor and thermal coupling of parts, is it worth it?