expo converters

[Martin Czech]

> I have just been experimenting with a LM13600 based VCO (sawtooth/pulse
> circuit in figure 16 of the National datasheet).  I used an expo
> converter similar to the one in Tom G's VCO5a, (two 3906's with the
> emitters tied to V+).  I can get decent V/oct performance at low
> frequencies, but the oscillator goes flat as the frequency climbs.  I
> understand this is common in VCOs and can be compensated for ("high
> trim").  I guess I have a few questions:
> 
> Does anyone know of an LM13600 VCO circuit that has high trim circuitry
> that I can look at?  I haven't been able to find one.  Or maybe it can
> be done in the expo converter circuitry.
> 
> I substituted a 500pf cap (actually 2 1000pf in series) for the
> specified 50pf item (smallest ones I have - time to order some more
> parts).  Do larger cap values increase the high frequency "droop"?
> 
> Is there a general or common approach to implementing "high trim" or is
> it a circuit-specific problem?  
> 
> Why do some VCOs not have high trim (for instance the ASM-1)?  Is it
> because these VCOs have a very fast discharge time and thus there is
> less error at high frequencies?
> 

Many questions:

High frequency droop has two sources of error:

1: comparator feedback loop delay gets relatively stronger to period.
2: high currents make you feel the BJT parasitics as well as 
   beta saturation.

For 2 the integrationg cap should be a good compromise between leakage
(too small cap) and too high currents (too big cap).

There are two possibilities for compensation:

1.: reduce amplitude with higher frequ, so charging needs less time.
    most saw osc. use this method, the famous trim at the cap.  I think
    it is no good. amplitude loss makes waveshaping (tri) difficult.

2.: Measure current and feedback to expo converter. CEM3340 does this
with a current mirror trany. In discrete applications you don't have
the mirror.  (With the exeption of LM13700 otas, steer both with the
same expo current, use one for integration, the other as current miror,
voila!)  3080 design would measure the voltage acrosss the Iabc junction
(electrometer op amp). Anyway, this way one will be able to get a voltage
that is a measure for the expo current. Now feedback this voltage in
the expo input summer.

I guess, this will only fix the last octave. If you have problems
in more octaves, better get a better BJT pair and/or
faster comparator.

m.c.