[sdiy] Expo converter ??

[Ian Fritz]

At 05:41 AM 1/18/2003, jhaible wrote:

>Big advantage: no servo loop required.
>
>Ian, I think it was _you_ who discovered that the "standard" expo converter
>with opamp and npn pair (or pnp pair) is prone to having a tiny bit
>of HF oscillation that is hard to detect on a scope, but which might
>cause a little bit of HF mistracking. I think you suggested to add some
>extra compensation to get rid of this problem, but of course this will also
>make the expo converter slower for audio rate modulation and such.
>(from some old EN article, as I remember it.)

Right, and I'll never for get that experience!  I was trying to improve the 
tracking of my ultrasonic VCO and for some unknown reason stuck a scope 
probe onto the converter output.  Suppressing the oscillation that I found 
helped with the tracking.  Then I looked at all my other VCOs and VCFs and 
found the same thing!

Terry Michaels also found an oscillation when he reevaluated his original 
VCO a couple of years ago.  I believe he posted his fix here, as well as in 
EN.  (Are you AMS-bangers listening?)

I was concerned about the extra capacitance I put in as a brute force 
solution hurting the modulation response, but I couldn't see any effect, at 
least up to a few kHz.

These days you can just use a modern opamp that is optimized for servo 
applications and avoid the problem.

>Now with a npn + pnp configuration, you get a fairly constant current
>in your "reference" transistor just with a resistor, without an opamp
>and servo loop. (As Don pointed out, it's not ideal. But it's close,
>because the expo transistor just eats current from the reference transistor
>with its base, not with its emitter. And the voltage change across the
>resistor is small compared to the DC bias, so it's almost constant current.)

Good points.

The question about how to go about matching different transistor types 
still remains, though.  I say there's more to matching than just checking 
Vbe at some current level.  Since it is the cancellation of the temperature 
dependence of I_es that is important, I think that matching should be done 
by actually measuring how the Ic ratio changes with temperature at zero 
differential bias. I think this can be done with simple test circuits, both 
for devices of the same type and for PNP/NPN pairs.

   Ian