AW: VCO hacks, tweaks

[terry michaels]

Message text written by terry michaels
>Message text written by Haible Juergen
>
Ian and Terry,

what do you think about Rene's differential amp 
temperature compensation ?
It's at <http://titan.cs.uni-bonn.de/~schmitz/thcomp.gif>,
and it was discussed on the list some time ago, but
I think you joined us later. Now with all this recent talk
about tempco resistor accuracy, how does Rene's
circuit compare to the resistor method ?

Personally, I think it's very elegant, using 4 transistors 
from the same array for both, expo conversion and
temerature compensation. But I have never tried it -
it's been some time since I've designed my last VCO.

So what do you think about it ?

JH.
<

Hi Juergen:

Took a look at Rene's circuit, the concept is sound.  The T1, T2
differential amplifier along with the second op-amp has a gain that varies
directly with absolute temperature:  Vo = q/2KT times E in, times a
constant dependent on resistor values and supply voltage.  The first op-amp
forces whatever input voltage is needed to get a constant output from the
diff. pair, that same voltage is applied to the expo convertor.  Works on
paper, I'll have to try it someday.

The beauty of this circuit is it could be easily integrated onto a complete
VCO chip.

BTW, the data sheet for the Curtis CEM 3340 VCO talks about a precision
multiplier fed by a temperature dependent constant to provide temperature
compensation of the expo convertor.  Since they only show a block diagram,
there is no way to know exactly how it was done, it might have been done in
a similar way to Rene's design.

I hope I didn't duplicate whatever previous discussions there were on this
subject, if so I apologize.

Terry
<

Hi Juergen:

I've taken a closer look at Rene's circuit, and I now have some
reservations about it.  The differential pair inherently has a non linear
transfer characteristic, and has a rather narrow range of operation for
reasonable accuracy.  Based on the voltage ratio established byf R3, R4 and
R12, R13, I estimate only a few octaves of range with satisfactory
accuracy.  The voltage ratio could be changed to narrow the operating point
of the diff. pair, but as you do that, the offset voltage drift of the
diff. pair will be amplified, and may get to the point where that source of
error begins to dominate.

I think using a variable gain stage with a tempco opposite that of the expo
transistor is an excellent idea, but there may be a better way to implement
it.  Two possible alternatives come to mind:

Using a differential transistor pair with a large amount of resistive
emitter degeneration.  An example of this is a monolithic analog
multiplier, one input can be a voltage reference which varies with absolute
temperature.

Using a log/antilog conversion in tandem, to create a two quadrant
multiplier, one input of which is made to vary with absolute temperature.

I will look into both of these ideas sometime soon.  However, the
simplicity of a tempco resistor in the signal path is beginning to have
more appeal.

Terry