[sdiy] interesting expo design

[Scott Bernardi]

I've been looking at your circuit, and one thing puzzles me about using a derived temperature
dependent current to multiply by the control voltages - doesn't the OTA itself have a temperature
dependence in the gm term?  Iout = Vin * 2q/kT * Iabc (the 2q/kT term turns into the familiar 19.2
value at 25deg C). Q16C's collector is running at 100 times Q16A's, so you get a delta Vbe of
kT/q*ln(100) = 4.6*kT/q. This voltage is amplified by U43A by 5K/620, so you have 37.1*kT/q at the
output of U43A (about .965v). This is converted into a current by U45A to drive the input OTA.
Ic17 = 37.1/10K*kT/q or 3.71e-3*kT/q.

Iout = Vin * 2q/kT * (3.71e-3 * kT/q) ---> the kT/q terms cancel, so instead of a temperature
dependent term to multiply the input control voltage by, you have canceled it!  Then you run it
through a second OTA, U44B running at a constant current (adjusted by the Scale control) - and it
introduces a kT/q term in its gm, but its a different chip at a different temperature.
Am I missing something here?

patchell wrote:

>     Been there...Done that...sorta...
>
>     I saw that article some months back, and the solution I came up with is here:
>
> http://www.silcom.com/~patchell/seqproj/1001013.pdf
>
>     On page 10.  I used a single LM13700 (although both OTAs are used) as the VCA.  The boards
> that this is on will be in my greasy little hands in a couple of days.  Earlier experments say
> this should work really good.  I will keep you posted as to the results.
>
> Michael Ruberto wrote:
>
> > this is interesting, check it out:
> >
> > http://www.e-insite.net/ednmag/index.asp?layout=article&articleid=CA178103&industryid=2281
> >
> > the problem I see with this approach is adding in external CVs. I suppose it
> > could be done using a very precise VCA. the external CV's can be summed into
> > the VCA's CV input and the reference from IC3 connected to IC1's + input
> > through the VCA? hmmm I imagine there would have to be some offset
> > correction too.
> >
> > may have to experiment with this...
> >
> > ~M
> >
> > _________________________________________________________________
> > Chat with friends online, try MSN Messenger: http://messenger.msn.com
>
> --
>  -Jim
> ------------------------------------------------
> * Visit:http://www.silcom.com/~patchell/
> *-----------------------------------------------
> *I'm sure glad Merry Christmas comes just once a year
> * -Yogi Yorgensen
> ------------------------------------------------

--
Scott Bernardi
sbernardi at attbi.com