[sdiy] Berfotron/Buchla now triangular, but with 4KHz offset!

[Aaron Lanterman]

Greg and I spend a lot of time today banging our head against the  
Bergfotron VCO.

http://hem.bredband.net/bersyn/VCO/VCO%20Buchla%20core.gif

The good news is that it's triangle is now triangular. It turned out  
to be a "bad" LM394. I don't know if it came mismatched, or if we  
somehow damaged it in our various explorations, but anyway when I  
grabbed a handful of LM394s and swapped them all in, the all gave nice  
even triangles, except for that bad one. (Thanks to the many people  
who wrote in with suggestions about the lack-of-triangularness issue!)  
In our explorations, we even swapped in some crappy 5% 5k1 resistors  
at the top of the OTA, and it still looks beautifully symmetric.

The other good news is it now shows an exponential change in frequency  
with CV.

The bad news is that it seems to add a 4 kHz offset on top of that!

The lowest frequency we can get out of it is around 4 kHz.

Of course, I immediately suspected the expo converter.

1) Tweaking the CVs so that we read 0 V at the base of the left  
transistor in the expo pair, we measured the voltage drop across the  
6.8 K resistor, and found that there was about 200 muA going through  
it. We measured the voltage drop across the 150K resistor, and deduced  
there was around 100 muA going through that. Then for grins we put a  
910 ohm resistor between the emitters of the NPN diff pair of the OTA  
and the collector of the right transistor of the expo pair. We  
measured the voltage across it, and deduced the current through it to  
be around 100 muA. So the balanced case seems to work right.

The freq of the oscillator should be current/(2*Vpp*C).

Plugging in 100 mA, 4.6 Vp-p (what the scope read for the triangle),  
and 4700 pF gives 2312 Hz. But instead we got something like 6 KHz out  
(didn't write down the exact number).

2) We then tweaked the CV input down to zero. (The range adjustment  
knob puts in some negative CV for us.) We then measured the voltage  
difference between the bases, and used it to try to deduce the  
expected frequency, and came up with something like 290 Hz. But  
something like 4.3 kHz ish (don't recall for sure) came out.

3) When I crank the CV volt by volt, and subtract around 4 kHz in my  
head, I'm pretty sure I see a doubling effect as I add each volt  
(again, once I subtract 4 Khz.)

4) If I turn around the freq = current/(2*Vpp*C) formula, and write it  
as

current = freq*2*vpp*c, I can plug in 4K and see what current I would  
imagine would be needed to get that 4K, and it comes out to 170  
microamps. So it's like all these magic microamps appear out of nowhere.

5) Oh, here's the wild part: we can COMPLETELY DISCONNECT THE EXPO  
CONVERTER altogether - i.e. disconnect the emitters of the diff pair  
in the OTA from the collector of the right transistor of the expo  
converter - and it still happily oscillates in the kHz range! (We  
discovered this when putting in the 910 ohm resistor mentioned above).

Any ideas? Anyone run into this mystery offset before?

- Aaron

P.S. What pains me is I have a MOTM VCO with a similar phenomenon,  
except it has an offset of 50 KHz! Paul gave me some troubleshooting  
tips, and I plan to get back to it over the summer hopefully. Anyway,  
the MOTM is a completely different design.