[sdiy] Thought experiment...

[James Patchell]

The Mike Sims VCA has the main advantage of having a very high 
performance/price ratio.  It may be posible that using a CA3280 in the same 
circuit you might be able to squeeze out some more linearity, but the 
LM13700 has proved its worth in that I can't seem to see any problems that 
it causes...plus, the LM13700 is only what, 50 cents, and the CA3280 is $5+ 
(unless you buy them from me :-)...I originally used an AD633 for the 
multiplier, but it did not perform as good as the LM13700, in this application.

At 10:06 PM 3/22/2005 -0600, Ryan Williams wrote:
>hi Dave,
>
>I also cannot see how you mean for this to cancel the Vt variable. If you 
>are calculating otherwise, can I see the math?  This type of circuit, with 
>error compensation or pre-distortion fascinates me, my favorite type.  I 
>haven't spent a great deal of time on this yet, so excuse me if I have any 
>mistakes.
>
>About a month ago, I decided to solve this problem. The only solution I 
>could think of involved multiplying the input CV by the Vt temperature 
>variable.  This should be the same as in the circuit Jim Patchell has 
>posted although I have only just glanced at it. The way I've done it, 
>(again which is probably similar to Jim's) is to use the second pair to 
>generate the Vt. I have an opamp holding the first transistor at a 
>constant current, and another opamp holding the other transistor's current 
>constant. the second opamp's output after a voltage divider connects to 
>the base of transistor 2. the base of the transistor 2 sits at 
>ln(Ic2/Ic1)*Const*Vt. I had it setup as ln(50)*10*Vt. the 10 came from the 
>voltage divider. I just made up these numbers. Haven't put any thought 
>into what would be best. I really need to spend some time learning how 
>Jim's circuit is better (just asuming it is). I planned on trying the 
>THAT300 quad npn, don't have them yet though.
>
>The difficult part I think is the multiplier. I expect when I actually 
>build this I'll use the two OTA circuit (also used in Jim's circuit, Mike 
>Sims VCA in EDN magazine, and other places) although I've not yet measured 
>it's performance. Incase Jim or anyone else who is reading is familiar 
>with this circuit; is there any advantage to using the CA3280 in place of 
>the LM13700. I was under the impression that it's transistors were matched 
>well, better perhaps? What about a multiplier IC? If i remember correctly, 
>the two LM13700 VCA doesn't completely remove it's temperature error, but 
>only makes it much smaller. I don't really care about a couple of dollars 
>more if performance is better, atleast not for my purposes.
>
>-Ryan Williams
>http://www.sdiy.org/destrukto
>
>
>D A F wrote:
>>Perhaps I did a crappy job explaining clearly what I was trying to
>>accomplish.  When I mentioned a 'positive-going' and 'negative going'
>>version of the CV input, what I meant that the CV coming in would be first
>>put through a "phase inverter" (to borrow an audio amplifier term) such that
>>when you feed a CV into the input, you make two CVs - one "in phase with"
>>and one "180 degrees out of phase with" the input.  This could for example
>>be accomplished by feeding the CV into two op-amps, one set up as an
>>inverting amp, the other non-inverting, but both stages having the same
>>(abosolute value) gain.  The outputs would then be a -pair- of CVs that
>>track one another, one going up as the other went down, etc..  This pair of
>>'inverted phase' signals drive the two pairs of transistors, so that when
>>you feed the outputs from the two log amp stages into the dif amp at the
>>end, you do _not_ cancel the CV (since itself has been made differential
>>prior to the transistor stages), but _do_ cancel the temperature effects
>>from the transistors themselves (which, given matched devices and good
>>thermal coupling, would track one another).
>>Then again, perhaps I should just build one and see if it works..  :-)
>>
>>Dave
>
>         -Jim
>***************************************************************
>http://www.oldcrows.net/~patchell
>
>***************************************************************