IIRC mo purely
linear system can produce stable oscillations (any
small change will either cause said oscilaltiosn to colalpse to zero
or to grow ithout limit (or more practically untyil the signal hits
the supply rails).
On the one hand, that feels right.
On the other, I'm considering an AGC circuit that doesn't use a bulb,
but instead uses a FET as a variable resistor, and trying to figure out
where it's got anything nonlinear in it (assuming the amplifier is
running class A).
I was using 'linear' in its mathemtical sense. A linear sysnmte is one
such that if input A gives oputput B, then input kA (k a constant
multiplier) gives output kB (same k). And if input C gives outptu D, then
input A+C must give output B+D.
It is clear form this that any aplifier iwth any form of AGC is not
linear. If we increase the input signal (multiply it by a suffiicenlty
large 'k'), the gain drops, so the output is less that you might expect.
I am sure we
all know the stroy of Fred Terman and a certain research
student.
Doesn't sound familiar to me, for what that may be worth.
Oh dear, oh dear...
We need to think back to the early 1930s. The problem was to produce an
adjustable audio oscilaltor. LC oscillators were impractical because
the L and C needed were large, and it's difficult to make a 1uF variable
capacitor (say). And the indcutoir is likle to be iron-cored, which is
likely to be non-linear.
The normal way it was done at the time was to have 2 RF oscillators
(such thigns couple be made adjustable fairly easily) -- one fied, the
other variable. The outputs were mixed and the differnvee frequency
extrcted. That was the audio output. One big problem was that if the 2
oscilators were at similar frequencies (to get a low freqeuncy otuput),
they had a habit ot interfering with each other. THis either led to them
jumping into lcok (so the output frequency became 0), or at least
producing a very distorted differnce signal.
The solution was to make an RC oscilaltor. Large adjustable resisotrs are
perfectly possible, and can be used in such a cirucit with fixed capacitors.
Fred Terman was a professor of electrical engineerign at Stanford (IIRC).
He had a research student working on this, and said student made an RC
oscilaltor using a Wein bridge network as the frequency-dependant part,
the gai nwas stailbied by having a small filamanet bulb as one of the
gian-control resistors in the amplifier. It worked well, in fact it
backame a commercial product.
The name of this student was William Hewlett. The product was the model
200A audio oscillator.
-tony
