[sdiy] Transistor help

[Roy J. Tellason]

On Wednesday 27 April 2005 12:22 am, James Patchell wrote:
> When I was in college...we used to use a curve tracer to match
> transistors...it had two transistor sockets on it, and there was a switch
> to flip between the two parts...basically, you were matching HFE doing it
> this way...seemed to work...
>
> One thing I have noted about modern batches of transistors is they sure
> seem to be all very similar...I generally match about 10 pairs at a
> time...out of 20 transistors (generally 2n3906) I always get 10 pairs...and
> all of the transistors are within about 2mV of each other....30 years
> ago...it wasn't like that...as I recall.

I have a curve tracer, a B&K unit that somebody gave me because they didn't 
have any use for it (!)...

It has a couple of sockets,  and also test leads for both sides,  and a switch 
to select one side or the other as you describe.  Some time back,  when I was 
using "ECG" replacement parts,  I took what they called a "matched pair" of 
transistors and hooked them up.  The resulting curves were *very* different 
on the two parts,  so I'm wondering what exactly they were matching in that 
case.

I'm not sure which part number it was,  perhaps ECG130MP,  their 2N3055 
replacement or similar.

> At 10:01 PM 4/26/2005 -0600, Ian Fritz wrote:
> >At 09:05 PM 4/26/05, Bob Weigel wrote:
> >>But the way suggested is more proper since it takes into account the
> >>actual things which are going to affect the performance of the circuits
> >>in question usually. -Bob
> >
> >Well, we've been around on that one many times here.  If the mismatch
> >between two units is due to different mesa areas (say) then the
> >temperature dependences will cancel properly and the mismatch will not
> >matter (just an unimportant current ratioing).  So the standard method
> >could give false negatives.
> >
> >OTOH, if the differences are due to, say, defects, then these would also
> >affect hFE.  I think that is why I have observed very good correlation
> >between Vbe and hFE.  And why I think simply matching hFE makes sense.
> >
> >   Ian
>
>          -Jim
> ***************************************************************
> http://www.oldcrows.net/~patchell
>
> ***************************************************************