[sdiy] QVCO

[JH.]

If you want to see my insane new QVCO circuit, take a look at
http://www.oldcrows.net/%7Ejhaible/matrix_fx/jh_matrix_fx.html

Insane because of the use of a bunch of expensive components.
But I found it really makes a difference here wether you use MAT
and SSM pairs, or cheaper transistors, when it comes to low offset.
And even with these, low offset discrete OTAs are not trivial.
(It's not the offset per se - it's the offset variation from altering
Iabc.)
Some details you might find interesting:
The collector of the variable gm pair that dosen't run into the
virtual GND node of the opamp integrator, is working in
a cascode configuration. In order not to draw a base current
(whith beta chanhing over Iabc), I'm not using a BJT, but a JFET
here.
The 3rd transistor in the current mirror must be a low beta (A)
type.
I'm operating the mirror from a rather high voltage (12.5V).
Contrary to my expectations, this gives lower offset variation
than running it from a low voltage.
The remaining offset (of the 20Vpp signals) is less than 30mV
from 1uA to 1mA Iabc.

A pair Vactrols switches between AC coupling (for audio rate
frequencies) and DC coupling (for sub audio frequencies).
In audio range, you want the best carrier suppression in the
Frequency Shifter application. In sub-audio, this is not so
important. 
I'm aware that a low-offset VCO core would not have been
necessary with that switched AC coupling. But I hope the
discrete OTAs will also have benefits when it comes to
V/Oct tracking. (But the tracking is still untested.)

JH.