matched transistors

[patchell]

terry michaels wrote:

> Message text written by "Ian Fritz"
> >Hi Juergen --
>
> I agree with you about the noise issue. I was thinking mainly about
> converters. Is it possible to improve noise by paralleling transistors?
> Seems this should work, but I've never tried it.
>
>   Ian
> <
>
> Hi Ian:
>
> Paralleling transistors has some possible benefits.  I remember something
> about multiple transistors behaving as follows: the collector currents are
> additive, but the noise of each transistor is uncorrelated, so although the
> noise from two paralleled transistors is greater than one transistor, it is
> not doubled, so you gain a little in signal to noise ratio.  Also, I've
> thought about using paralleled transistors for a expo convertor, perhaps to
> increase the frequency range, but as far as I can see you don't gain
> anything.  Rbe generally defines the upper current limit, the Rbe will be
> divided by 2, so your output current can be doubled, but the collector
> leakage current will be doubled, so if you double the oscillator timing cap
> value, the overall oscillator frequency range does not increase, because
> the leakage current is doubled on the low end of the frequency range.
>
> Terry Michaels

    I have actually used parrallel transistors to lower noise.  But even in
this, it has it's drawbacks.  While you do decrease the input noise voltage of
the amplifier by the square root of the number of transistors (it takes four to
decrease by 2), you increase the noise current by the same amount.  So, if the
source impeadence of the thing you are attempting to amplify is moderate to
high, you probably won't gain any benifit.  I used parrallel 2n4403 transistors
in a preamp for a Mecury Cadmium Teleuride (sp?) IR detectors many moons ago,
but these thing had a source impeadence of something like 50 ohms, so the added
noise current did not hurt.  We were able to get the noise voltage down to
about .5 nV/root Hz doing this.  Now days, you can purchase op-amps that are
real close to that.

    -Jim



--
-----------------------------------------------------------------
Visit:http://www.silcom.com/~patchell/
-----------------------------------------------------------------