[sdiy] Temperature Compensated Exponential Converter Using SSM2164

[HOSHUYAMA]

Hello list,

I'm so sorry that I was off in the weekend.  
I've been so surprised by the long thread of discussions.

Thanks a lot to those who gave the comments on my idea,
especially, to David G. Dixon for the experiments.


I know Jim Patchell's approach.  As was pointed out in the discussions, 
the possible advantages, which motivated me, were:

  - Thermal coupling is almost ideal. 
  - Temperature coefficients can be trimmed.
  - DC performance of SSM2164 seems to be better than LM13700.

The circuit is simple, but its simulation was not easy, 
because there is no model for SSM2164.  First, I simulated 
Blackmer variable gain cells, which is supposed to have similar 
characteristics, and got good results, then moved to simulation 
with SSM2164-like VCAs.


I'm very happy to see that my idea is confirmed by experiments.
(I don't have sufficient facilities in my condo now.)
I guess the error that David pointed out when the CV is positive
is caused by the saturation of the output transistors in the 
exponential converter.  We should shift the input CV range and 
increase the reference input current of the VCA for exponential 
conversion to solve the problem.  (The purpose of the 22k 
resistor at the summing opamp's output was to avoid the problem, 
however, it's not a good idea.  I was wrong.)

I hope I answered all the questions.

Anyway, thanks a lot again.

Sam HOSHUYAMA
Saitama, JAPAN


"David G. Dixon" wrote:
> Excellent point, Ian.  I was just trying to demonstrate the difference
> between relative and absolute error, but perhaps that is not terribly
> appropriate for this particular problem.  I'm going to have to do more
> analysis to wrap my head around all of this, and preferably not at 1:00 in
> the morning, after nine hours on the road.
> 
> In any case, the maximum relative errors I'm getting with the 2164 analysis
> are in the neighborhood of 0.2%, and significantly less if one uses the
> "proper" values of tempco trim and feedback resistance, and doesn't stray
> too far from the reference temperature, particularly if one errs on the
> warmer side (cooler temps give more error).  Hence, I think, with a little
> care, this expo can be adjusted to give very slow beat frequencies
> everywhere in the audio spectrum.
> 
> > >Missing the next octave by 1 Hz is no big deal
> > >if your target is 1760 Hz, but it is a very big deal if your next octave
> > is
> > >55 Hz.  Capiche?
> > 
> > Ummm ... no.   :-)
> > 
> > What is most objectionable about mistuning is the beating it creates.  In
> > your example there would be a 1 Hz beat rate (against a perfect
> > reference)  for either case.  Not bad at all.
> > 
> > This why I set my VCOs up so that deltaVbe at the converter pair
> > corresponds to something around 2 kHz.  (deltaVbe = 0
> > --->  exp(a*deltaVbe/kT) = 1, so there is no T dependence at this
> > point)  Drift above this point isn't too important, because you rarely use
> > fundamental frequencies much above 3 kHz.  Drift below this point becomes
> > less and less critical as you go down, because a 1 Hz beat represents
> > smaller and smaller fractional error.
> > 
> > Just my personal philosophy, I guess.  :-)
> > 
> > Ian
> 
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at dropmix.xs4all.nl
> 
> http://dropmix.xs4all.nl/mailman/listinfo/synth-diy
> 
>