Circuit Board Update 4/23

[gstopp at fibermux.com]

     Buried somewhere in Electronotes I think there is an explanation of 
     these things, but I can't remember where. One statement I did find is 
     that the matched-pair packages like the MAT-02 are "optimized by the 
     manufacturer for optimum log conformance". This would be in addition 
     to the fact that all parameters are deliberately matched and thermal 
     equivalence is guaranteed.
     
     On my circuit board I placed the two transistors next to each other, 
     and the tempco resistor location is next to the transistor pair. This 
     way if the builder does use two NPN's and a tempco, he can touch them 
     all together and hold them in place with a tie-wrap and some heatsink 
     goop or whatever method is easiest.
     
     I've found that the +3300ppm/C' term that is left over after all other 
     effects are compensated for does not really cause a lot of problems in 
     a studio environment. I suppose it would be a good thing to cancel out 
     for live use, if you were for example on stage at an outdoor festival 
     and needed to reach up and play a beat-free lead line on your 
     synthesizer without having to stop and tune it....
     
     Joachim, perhaps you know of a good procedure for matching transistors 
     for a VCO application (beyond selecting from the same manufacturing 
     lot). That would be a good reference for the DIY list.
     
     - Gene
     gstopp at fibermux.com


______________________________ Reply Separator _________________________________
Subject: Re: Circuit Board Update 4/23
Author:  Joachim Verghese <jocke at netcontrol.fi> at ccrelayout
Date:    4/24/96 4:06 AM


On Tue, 23 Apr 1996 gstopp at fibermux.com wrote:
     
>      The exponential 
>      converter uses a pair of NPN's, so that means that a couple of 
>      hand-matched 2N3904's should work fine. That's what I'm using right 
>      now, and they're not even matched.
     
What does Electronotes say about this matter? According to my experience, 
the two transistors of an exponential voltage-to-current converter need 
not be matched in the normal sense, i.e. for equal Vbe and/or Beta.
     
The first transistor is used to provide an offset voltage for the 
exponentiating (second) transistor. This offset voltage contains
the temperature dependent factor of the transistor's reverse collector 
cut-off current. If this factor is equal for the two transistors,
and they're always at equal temperature, the effect will be completely 
cancelled. (Note that this does not eliminate the temperature dependence 
of the exponential scale factor (0.33%/K) which is why you still need
a tempco resistor or similar).
     
In other words, the two transistors should ideally be matched for similar 
temperature charateristics, and have good thermal coupling.
This is perhaps easiest to achieve by using integrated transistor pairs
or arrays. If you use discrete devices, I'd imagine you'd get good results 
as long as you provide adequate thermal coupling. (ARP used to put brass 
clips around their TO-92 pairs). It might also help if the two transistors 
are off the same manufacturing lot.
     
Any differences in Vbe or Beta will be corrected for by the frequency 
offset trimmer of the VCO.
     
-joachim