[sdiy] moog transistor matching circuit question

Sun Jan 10 15:58:53 CET 2016

Hi Matt,

It is basically the same method as I described, but using the internal current
source of the multimeter.

The deviation in the readings due to the "0.2%" aspect of the accuracy
specification will be practically the same for voltages that are near each
other, as a multimeter such as this uses dual-slope A-D conversion which is very
linear. So, with a 4.5 digit multimeter, if you get the same reading on two
transistors, the actual mismatch will be at most 0.1 uV and a little bit (of
course disregarding any other measurement errors, such as one transistor getting
warmer than the other due to handling them).

Best regards
Simon

> Op 10 januari 2016 om 15:18 schreef matt holland <matt at mattholland.org>:
> 
>  Thanks Simon. Consensus of everyone I asked is that it's the resistor isn't
> needed, so then the lack of one in the PNP tester isn't a "flaw". I like your
> idea of it being for current measurement, that seems very plausible. I'm gonna
> go with that explanation =)
>   
>  For the record I breadboarded up Ian's circuit last night and my whole strip
> of 3904s appeared to be perfectly matched with my Fluke 87v set to its mV
> range. 
>   
>  I then found this page
> http://www.cgs.synth.net/modules/simple_transistor_matching.html which seems
> like an even easier method for me. The 87v accuracy spec for the diode test is
> 0.2%+1 so if my math is right two with the same reading would be at most 3mV
> apart in reality.
>   
>  Matt
> 
>  On Sat, Jan 9, 2016 at 11:03 PM, Simon Brouwer <simon.o at brousant.nl
> mailto:simon.o at brousant.nl > wrote:
>    > >    Hi Matt,
> >     
> >    You're correct in your understanding of the circuit.
> >     
> >    I think the 10k resistor is just for checking the current, it should be
> > no coincidence that the "100uA" annotation is next to it. There is no harm
> > in either putting it in or leaving it out.
> >     
> >    If you are not planning to match transistors on a regular basis: a dead
> > simple matching jig consists of just a socket for the DUT, in which the DUT
> > is connected as a diode (connect base and collector together), a resistor
> > and a power supply.
> >    Use a largish voltage (say more than 10V) and choose the resistor to get
> > about 100uA. Use a DMM to measure the Vbe voltage.
> >    Since the voltage across the resistor is very large compared to the
> > variations in Vbe, the variation in current will be only small, and the
> > actual difference in Vbe voltage between transistors will be only very
> > slightly larger than what you measure using this circuit.
> >    You could make the test socket out of an IC socket (the type with
> > machined pins).
> >     
> >    Best regards
> >    Simon
> >     
> >     
> >     
> >     
> >     
> >     
> > 
> >     > > > Op 9 januari 2016 om 17:29 schreef matt holland
> >     > > > <matt at mattholland.org mailto:matt at mattholland.org >:
> > > 
> > > 
> > >     Hi All,
> > >      
> > >     I'm in the process of helping a buddy build a TTSH and as a result am
> > > being thrust into the world of transistor matching. I could just get on
> > > with it and breadboard one of the handful of circuits out there but
> > > instead I've been staring at the Moog circuit linked to here [1] and
> > > trying to understand the theory behind it. The non DUT transistor is a
> > > current soruce, with the op amp giving the ability to set the current
> > > without having to bother biasing it with a precise Vbe. The two resistors
> > > on the non-inverting input bias the inverting input to -5V, which enforce
> > > a 98uA (~100uA) emitter current through the 51k resistor through the magic
> > > of negative feedback. 
> > >      
> > >     It doesn't seem like the current value is even important at all so
> > > long as it stays the same from one DUT bjt to the next (i.e. temp is same
> > > from one measurement to the next for the tester as well as the DUT). Then
> > > we measure the Vbe across the DUT which is sort of a proxy for measuring
> > > Is (Vb = Vt*ln(Ic/Is) since it Is that models all the physical parameters
> > > that vary from one bjt to the next and is the temp dependent parameter
> > > that gets cancelled out in the expo converter to the extent that they're
> > > identical.
> > >      
> > >     The bit that I'm still having trouble with is the 10k resistor. It's
> > > there in the NPN tester but not in the PNP tester. I've seen this
> > > described as a "flaw" in the PNP tester but I'm not sure what it's doing
> > > int the NPN tester. I threw together a spice sim that steps the value of
> > > it from 0 to 10k with absolutely no meaningful change in the measured Vbe.
> > >      
> > >     Apologies if this has been covered before and appreciate any help.
> > >      
> > >     mh
> > >      
> > >     [1]
> > > http://www.musicfromouterspace.com/analogsynth_new/TRANSISTORMATCHER/images/moogmatchersmall.gif
> > > 
> > >      
> > >      
> > >     _______________________________________________
> > >     Synth-diy mailing list
> > >     Synth-diy at dropmix.xs4all.nl mailto:Synth-diy at dropmix.xs4all.nl
> > >     http://dropmix.xs4all.nl/mailman/listinfo/synth-diy
> > > 
> > >    > > 
> >     
> >  > 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://synth-diy.org/pipermail/synth-diy/attachments/20160110/4a64ab83/attachment.htm>