An idea for replacing tempco resistors ??

[Haible Juergen]

Hi !

While I was thinking about temperature compensation in OTAs for my
frequency shifter project, I suddenly found something else, which I have
been looking for since a long time.

I was wondering how the temperature compensation in the CEM3340
could work. The Data sheet only speaks of a multiplier and a temperature
correction, but no details.
Now, I still have no idea what's inside the 3340, but maybe I found a
circuit
that does a similar thing.

Now, if you want to get rid of a tempco resistor, you need an amplifier
that
linearly increases its gain with temperature. The amplifier also has to
be
very linear in order not to degrade the V/8ve courve of the VCO.

Now, Doug Curtis was famous for his highly linear gilbert cells, as in
his VCFs and VCAs, so chances are that he used one in the VCO as well.

A gilbert cell is temperature compensated (as opposed to a normal ota
which looses gain with increasing temperature). So the only thing we
have to
do is increasing the bias current of the differential pair with
temperature.
This could be done by not using a constant current source, but a
constant
voltage source which controls a current by driving a resistor and a pn
junction
in series. The pn junction could be the input of the Iabc current mirror
in
a 3080-type ota, or maybe even the emitter of the differential pair
directly
(not sure about the last one). The voltage drop over the pn junction
decreases
with increasing temperature, so the voltage across the resistor
increases,
and so we have a positive temperature coefficient for Iabc and for the
amplifier gain.

So far, these are just thoughts - I haven't done any experiments yet.
If someone wants to try it, I'd start with a gilbert-type ota like the
LM13700 or CA3080, use constant current sources for the diode current,
and a fixed voltage / resistor control at the Iabc pin.
The fixed voltage must be larger than the voltage drop of the Iabc pin,
but
not too much larger. Somwhere there might be a voltage / resistor
combination that gives 3300ppm.

What do you think?

JH.