AW: VCO hacks, tweaks

[Ian Fritz]

Haible Juergen wrote:
> 
> Ian and Terry,
> 
> what do you think about Rene's differential amp
> temperature compensation ?


I've actually experimented with it a bit -- in fact I have it on the
same breadboard as the VCO I just described and I plan to keep working
on it. So this is a *preliminary* report.

I studied the design quite carefully to be sure I understood it
completely. I agree with what others have said about it, it is certainly
an elegant design. It is also clearly a very carefully thought out
design. But the most impressive thing about it is that it seems to be
*new*, at least as a VCO feature. 

My first shot at implementing the circuit was just a bit different from
Rene's. For the diff amp I used a 3080 OTA followed by an opamp I-V
converter. At first I had a lot of trouble with spurious oscillation in
the servo loop, but this was easily tamed by replacing the 3140 opamps I
was using with the BB OPA132's. Feedback caps of 30pF around each opamp
give rock-solid stability at all signal levels.

As far as the circuit performance, so far I'm not quite satisfied with
it for my needs, but I haven't given up on it. To me its main drawback
(and of course this was acknowledged by Rene) is that the nonlinearity
of the diff amp is transfered to the VCO tuning curve. Since the diff
amp transfer function is divided into the expo converter, the
sublinearity of the diff amp shows up as a superlinearity in the VCO
frequency. In other words, as you go up to high frequencies, the octaves
become progressively sharper, and as you go down to low frequencies they
become flatter. 

The high-end errors can be partly compensated by increasing the
oscillator reset time -- in other words by deliberately making the
oscillator itself droop at high frequency. This can only be taken so
far, though, as the frequency dependences of the two errors are
different. And since I worked so hard at making the oscillator as fast
as possible, I couldn't see making it deliberately worse!

As a second shot, I tried reducing the input to the diff amp by a factor
of two. With the consequent improvement in linearity, I got a tuning
curve that was not too bad over about 100Hz to 4kHz. I'm sure that by
playing with the dead time setting and with different HF tracking
schemes the tracking over this range could be improved even more.

I didn't look at the temperature compensation carefully, but it seemed
to be working OK. The concept is sound, so I don't worry about that
part. As far as noise and jitter from the servo, I didn't hear anything
that bothered me. Since I mostly play from my wind controller, which
always has some breath noise, this isn't a big issue for me. 

My next mod will probably be to try to make a very low noise diff amp
and to attenuate its input even more.

Rene -- do you have any numerical data on your tracking (V/oct) behavior
that you would be willing to share with us? And congratulations on a
great new idea!

  Ian