> I just finished 2 MOTM300 VCOs last night and spent some time playing
> around with the calibration. I started at A440 using a very accurate
> frequency counter. If I get the trimmer set exactly right for the
> octave interval between 440 and 880, the next octave which should be
> at 1760 is invariably a little flat at about 1758 with another octave
> up flatter still at around 3495 instead of 3520. Going down from 440
> the octaves are very close. If I adjust over several octaves I can
> get the 1760 and 3520 closer, but the 880 octave ends up a little
> sharp at about 882 and the 220 octave a little flat at 219 with 440
> right on.
Ah, the age-old problem :)
The way I do it at the factory is set the VCOs for the best tracking where the ∗ear∗ most
sensitive. That is the range from 100Hz to 800Hz.
I use a 5 octave keyboard (starts at 'C') and set the 'C's for:
50Hz
100Hz
200Hz
400Hz
800Hz
1600Hz
I go back-and-forth between 100 and 800Hz, setting them to within 0.1Hz. At 800Hz, that is an
error of 0.01%. That's 10 TIMES better than any other VCO :) My freq counter is a Philips PM6666
with accuracy to 0.00001Hz, has oven-stabilized time-base.
The other intermediate frequencies (200, 400) are then well within 0.1% as well. Yes, you will
see errors "creep up" beyond 800Hz. But the ∗ear∗ (as opposed to the freq counter) is not nearly
as sensitive.
You can "warp" the response by using smaller values, like 2M instead of 2M2. If you are off say
more than 15Hz at 1600Hz, you may find lowering R50 will get you closer.
The reason HFT is not adjustable is that is a source of drift.
Also, what CV source is ∗driving∗ the VCO? It had ∗better∗ be within 50uV (yes, 50uV) if you want
errors less than 0.1% across the range. That requires a 6 1/2 digit DVM to calibrate :)
Heck, 5Hz off at 3520Hz is still <0.2% error. My Roland D-50 is about that, and it's DIGITAL.
Paul S.