[sdiy] Saw core JFET mystery

[David G. Dixon]

OK, so I went and played around some more.  First I replaced the JFET.  No
effect.  Next, I replaced the 1k reset timing resistor with 2k, and this
dropped the pitch somewhat, and gave a little bit of voltage dip at very
high frequencies.  I had to retrim the triangle corners-touching trimmer,
but the audio range looks good.

The only problem is: it doesn't track worth a damn.  With the tempco trim
set to the magic 289.45mV, I can adjust the gain trimmer (a 2k trimmer in
series with a 53.6k 1% resistor in the feedback loop) such that frequencies
below zero CV are perfectly in tune within the voltage error of my matrix
scanning keyboard.  Hence, there is absolutely nothing wrong with the 2164
Expo end of things.  That all works exactly as predicted by theory.

However, things quickly fall apart at positive CVs.  993Hz @ 1V (should be
1000), 3820Hz @ 2V (should be 4000), and 7160Hz (IIRC) @ 3V (should be
8000).  Obviously, the Franco resistor is not enough.  I guess the next step
will be to replace it with a Franco trimmer.

Honestly, I had much better luck with the 2N5485 RF JFET and custom circuit
I built around that.  I don't know why I can't get what everyone else uses
to work.  It's very exasperating.


> So I just tested the circuit a bit more, and if I put a 100pF capacitor
> across the + input and output of the comparator (in parallel with the 22pF
> one which is already there), the ramp integrator output actually dips all
> the way down to about -7V.  Note that this is not just a negative spike;
> it
> is a big J-shaped loop which adds significantly to the time of the ramp.
> The integrator - input rides ground, except for a ragged-looking spike at
> every reset which is only a fraction of a microsecond in duration.  Also,
> the comparator output never gets higher than about -4V during the reset
> even, and of course rides at -15V at all other times.
> 
> I'm really quite flummoxed by this whole thing.  I can't see how this is
> even possible.  And I don't believe it has anything to do with the Franco
> resistor.  Putting another one just like it in parallel causes an almost
> imperceptible shift in the ramp waveform, nothing more.
> 
> I guess I'll suck off the JFET and try another one next.