🚀 go-pugleaf

On Sat, 2 Jan 1999 07:53:27 +0000, John Woodgate
<jmw@jmwa.demon.co.uk> wrote:

><76jqau$ltt$1@computer.my.domain>, Duncan Barclay
><dmlb@ragnet.demon.co.uk> wrote:
>>A BJT allows you to control a current by varying the base emitter
>>voltage.
>>
>In an extremely non-linear manner. If one adopts this concept
>exclusively, it appears that it is not possible to make a linear (non-
>grossly distorting) amplifier.
>
>>Certain circuit configurations make it appear that you can control the
>>collector current by changing the base current.
>
>It doesn't just 'appear' - you can and do.
>
>>  But as this ratio
>>varies with temperature, device, process, batch, which direction the
>>circuit board is facing, it's not very useful.
>
>So why is it quoted in (almost?) every BJT data sheet?

What an odd discussion for entirely rational folk to be engaged in!

Even odder for me to side with Gareth Alun Evans, but if you cast 
a glance back a dozen or so posts in this thread you see a lucid
and accessible post by him offering a handy-dandy model
of transistor action which can easily be exploited to explain
* why collector current is exponentially dependent on Vbe, 
  following the same law as does diode current as a function
  of Vak (because B-E junction is a diode)
* why collector current is proportional to base current, but
* why the constant of proportionality (beta/Hfe) is a very
  strong function of fine details of transistor construction
  (base region thickness and, rather less obviously, doping
  levels)
* why Ic is nearly, but not quite, independent of Vce
  if the transistor isn't saturated (Early effect)

It is truly futile to have a religious war about whether 
Ic=Hfe.Ib or Ic=Io.exp(Vbe. e/kT)  is the more appropriate
description of BJT action.  Both are correct (or, more
precisely, both are approximations that are good enough 
to be useful in a wide range of cases).  Most designers
make use of both.

However, it is VERY important to note that the first
(Hfe, current-gain) description is intimately tied
to details of the particular transistor's construction
and so if you design a circuit that depends on a particular
value of Hfe it will probably let you down when you use
transistors near the edges of the gain spread.  OTOH
it is easy enough to find interesting, useful circuits
(long-tailed pair, Gilbert cell etc) which depend on the
exponential relationship in such a way that they are
almost independent of individual transistor characteristics,
at least at low-ish frequencies.  After all, e, k and T
are likely to be rather closely similar for all transistors
in a given circuit, no matter how they are built :)
If you have access to closely matched transistors, as you 
do on any chip, then a whole slew of other circuits 
becomes accessible - current mirrors are the most obvious.

For what it's worth, when I first look at a new circuit
I generally try to decide whether it's linear or
saturating.  If linear, then I find the exponential
(VCCS, transconductance) model is often the best 
starting point.  Your mileage may vary.

None of this helps to explain HOW the damn things
work, of course.  But that's another story...

Jonathan Bromley