[sdiy] VCO Tuning goals

[David G. Dixon]

> David, I suggest a footage select for the 0V frequency so that it can
> be set where the oscillator should be performing the best. Maybe if
> you want to only use very high pitches, set the footage very short,
> and you're OK. Or if you want it to perform best at the bass range,
> set it there. Generally, a lot of sounds are not going to happen
> across the whole range of the keyboard, and will only be played in one
> or two octaves.

This is easily done by changing the reference expo current, the integrator
cap value, or both.  For example, if instead of a fixed 1M resistor on the
-15V rail supplying reference current, one used a 1M 18-turn trimmer, one
could dial in any reference current, and hence, any zero-CV frequency. 

> With regards to the HF trim: isn't the diode a critical component?
> What can happen when the diode is less than perfect?

Diodes are less than perfect by definition.  However, to me, the key
question is, how does a given diode's I-V response correlate to the needs of
the circuit for a certain amount of correction current vs frequency (CV);
i.e., the I-V requirement for HF correction?  Is the exponential response of
a forward-biased diode the right curve?  What about the square-law
transconductance of a JFET?  And of course, the curves need to coincide at
the correct voltage (frequency), which means that the key adjustable
parameter here is the zero-CV frequency.  I don't know the answers to these
questions because I have not investigated this yet beyond the "daydreaming"
stage.  However, it is on my list for the weekend, because I want the fog to
lift in my own mind before I commit my "Sowa-Hoshuyama" VCO to PCB.

> Wouldn't it be possible to have multiple HF trims, say - two or three
> - by using precision voltage comparators? Maybe one trimmer per
> octave?

This would seem to be overkill, but I guess it would be feasible.

> David, with regards to the integrator reset 'problem': are you
> referring to the error introduced by the comparator being off by some
> percentage, say A? In that case the output frequency will be 1/A f'
> where f is the frequency we're trying to get. Which in turn means that
> a simple offset voltage fed into the linear voltage input in the EC
> would be good enough I guess?
> And if so, then it's a problem only in the (quite probable) situation
> when you want to run multiple oscillators from the same EC.
> 
> Then there could be the consideration of the switching lag in the
> comparator, which I think might introduce a linear frequency shift and
> then a small exponential freq shift if the switch speed depends on the
> voltage across the comparator.. but that can't be bothersome until
> we're in HF range, right? Definitely not in the audible range... I
> think.. but that's just guessing, not backed up with any real
> knowledge or experience =)

I think (again, don't quote me here) that I'm referring to the finite
on-resistance of the cap-shorting JFET, which imposes a finite time
requirement to shunt the charge on the cap to ground.  This time is
inconsequential at low frequencies, but substantial at high frequencies.  I
don't believe that the comparator per se has anything to do with it,
presuming that an actual comparator (ultra-fast) or a reasonably fast opamp
(TL07x, with a slew rate of 13V/us) is used.  If I'm wrong about this, I'm
sure I'll be corrected...