[sdiy] Temperature stable lin-exp converter with a CA3086 or CA3046

[Scott Bernardi]

One reason might the value of the tempco available.  To divide 1v down to the
approximately 18mV needed is a ~55:1 divider. With 100K inputs, that means about
1.8K.  If you have 2K tempcos, that works great because you just use a multiturn
trimpot after the opamp to adjust the V/octave.  I got a bunch of 1K trimpots from
EFM; I wanted to keep the 100K inputs (I guess I could have gone to 49.9K), so I
used the tempco in the voltage divider.

Ian Fritz wrote:

> Hi Rene --
>
> A good reason for putting the tempco in the feedback path.  I guess the
> voltage divider scheme is more popular, but I have never quite understood why.
>
>    Ian
>
> At 01:37 PM 5/27/2002, René Schmitz wrote:
> >Hi all!
> >
> >The maths for this can be found in the expo tutorial at my website.
> >Infact there is an approximation involved for a divider which has the
> >tempco in its lower leg. So in some sense that not truely a perfect
> >compensation, however one can argue that the error is vanishingly small.
> >(I.e. smaller than the other imperfections.) When you put the tempco in
> >the feedback loop of an opamp the compensation would be exact.
> >
> >Cheers,
> >  René
> >
> >At 12:31 27.05.02 -0600, Ian Fritz wrote:
> > >Sorry, but Hal is wrong on this (As is John Simonton, who makes the same
> > >incorrect statement).
> > >
> > >The temperature dependence of the converter goes as exp(qV/kT).
> > >
> > >V is proportional to the tempco resistance. If the tempco resistance is
> > >proportional to absolute T, then the converter is properly compensated.
> > >
> > >This is very simple physics.
> > >
> > >What seems to confuse people is that this tempco resistance R = AT gives a
> > >tempco of (1/R)(dR/dT) = 1/T. This indeed changes with temperature, but
> > >nevertheless is what is needed to provide correct compensation.
> > >
> > >Hal Chamberlin is not a physicist.
> > >
> > >   Ian
> > >
> > >
> > >At 12:02 PM 5/27/2002, Tim Ressel wrote:
> > >>I must disagree:
> > >>
> > >>--- Ian Fritz <ijfritz at earthlink.net> wrote:
> > >> > A standard tempco resistor (with resistance
> > >> > proportional to absolute
> > >> > temerature over the operating range) will in fact
> > >> > give *exact* compensation
> > >> > (cancels the 1/kT factor in the exponent). The idea
> > >> > that compensation is
> > >> > correct at one temperature only is a common
> > >> > misconception, unfortunately
> > >> > promulgated by some people who should know better.
> > >>
> > >>I quote from Chamberlin:
> > >>
> > >>"Note that the compensation is exact only at 27C
> > >>because the exponential converter temperature goes as
> > >>1/T rather than as KT, which the resistor provides."
> > >>
> > >>This was the line that convinced me to go with AN299.
> > >>I don't go against Hal Chamberlin, no matter how much
> > >>its been discussed here.
> > >>
> > >>--TR
> > >>
> > >>__________________________________________________
> > >>Do You Yahoo!?
> > >>Yahoo! - Official partner of 2002 FIFA World Cup
> > >>http://fifaworldcup.yahoo.com
> > >
> > >
> >--
> >uzs159 at uni-bonn.de
> >http://www.uni-bonn.de/~uzs159
> >
> >

--
Scott Bernardi
sbernardi at attbi.com