VCO temp compensation ideas

[Kimmo Koli]

You are quit right !

I've made a tempererature compensated VCA (not for a synth, though) with
the same temperature compensation principle. My divider was a log/antilog
multiplier/divider but otherwise the principle is the same. Because my
circuit was a full-custom analog integrated circuit, I had no chance to
use any tempco resistors. Fortunately, I had as many matched transistors I
needed and I could choose the level of matching by using more or less unit
transistor in a mathced array.

So the principle is this:
Basically there is two temperature dependent term in the Vbe, namely 
Vt =k/q T and Is, so that 
	Vbe = Vt ln(Ic/Is). 
A difference of two Vbe:s gets rid of the Is, so that
	Vbe1-Vbe2 = Vt ln(Ic1/Ic2).
And two of those two-Vbe differences divided don't have the Vt-term
anymore: 
	(Vbe1-Vbe2)/(Vbe3-Vbe4) = ln(Ic1/Ic2)/ln(Ic3/Ic4).
If the ratio of the currents Ic3 and Ic4 is contant we have temperature
independent operation.

In my quite traditional VCA the attennuation depends on a Vbe-difference
and the division was done by scaling the control voltage by the
fixed 1/dVbe. So the VCO exp-converter is quite the same, I think.   

Best regards, Kimmo Koli

On Thu, 23 Oct 1997, Rene Schmitz wrote:

> Hi all!
> 
> During the last days I thought about temperature compensating VCOs.
> There are several methods which all have drawbacks, such as the 
> heating of the diff pair, which uses a "high" current, or the
> tempco resistor which only reduces the temperature coefficient.
> (3300ppm/K vs. 3500ppm/K, 200ppm remaining) 
> 
> I started with the Idea of changing the reference current into the 
> diff pair with temperature, which soon turned out not to work. :-(
> 
> Then I had the idea to replace the resistor /tempco resistor voltage
> divider by a VCA, that is controlled by a thermo sensing transistor.
> Again this doesn't work, since I'm multipling where I should be 
> dividing. :-{   (old trap!!)
> 
> Then I had the Idea of using a divider, consisting of a multiplier
> in the feedback-loop of an opamp. It is obvious that if you use a 
> two quadrant multiplier consisting of a differential transistor pair,
> and an differential opamp amplifier, you get the same annoying 3300ppm
> term, right? 
> Now in the feedback-loop of the opamp this turns into 1/3300ppm, hmmm!?
> Using two of the three "spare" transistors in a CA8046/86 makes a fine 
> differential amplifier on the same temperature like the expo. converter 
> transistors. This all combined looks like temperature compensation could 
> be archeived by introducing a divider (preset to divide by one) into the 
> CV path between the summing amplifier and the V/oct trimpot.
> The parts count is pretty much the same as with the heater approach,
> it uses two opamps and maybe ten resistors, but doesn't need as much current.
> 
> And it should work at the whole temperature range. The heater just can't 
> make the chip cooler than it is, I mean when you preset the chip temperature
> to 40°C what do you do when the ambient temperature is higher??
> (Anyone that takes a synth to the beach?! :-)
> 
> A further advantage is that one could add a temperature dependant voltage
> (remember there is still one trannie left in the CA3046/86)
> at the multiplier pair to over-compensate, maybe this can compensate the
> temperature drift in other components as caps, discharge transistor...
> One would have a trimpot to set compensation.
> 
> I have not tried it, because I first wanted to know what you experts think
> of it, so...
> What do you think?
> 
> Bye 
>  Mr. Am-I-right-or-am-I-wrong Rene
> 
> 
> 

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
   Kimmo Koli                         Helsinki University of Technology 
   kimmo at ecdl.hut.fi                  Electronic Circuit Design Laboratory
   http://www.ecdl.hut.fi/~kimmo      P.O.Box 3000
   Tel:  +358 9 451 2273              FIN-02015 HUT
   Fax:  +358 9 451 2269              Finland 
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=