My original request didn't mean opamps, ICs or even DSPs are not
available, as I said, it is more a challenge between technicians at
my work. Maybe a foolish one
Indeed opamps are a much better way
to go but the line has already been drawn.
You have pointed out an important difference I was not really aware
of, I mean the difference between AGC and compression. So far the
circuits I have seen use a feedback from the output and I wonder if
compression occurs with closed loop circuit and AGC with open loop
circuit.
A guy came up with a circuit (open loop) that changes the current flow
through a 1N4148 diode as a control element.. The input signal is
first divided by about 50 with a resistor to that diode, such small
signal (few millivolts) at the diode seems to be necessary to
maintain low distorsion because of the diode non-linearity at higher
signal level.
The current through the control diode is directly controlled by the
input itself (1Vrms max) through an 1N34A diode, an RC time constant
and a transistor. As the voltage rectified from the input increased,
it drives the transistor in a constant current configuration and its
collector drives the control diode. When some threshold has been
reached, if the input signal doubles, the current through the control
diode also doubles, reducing by half its "apparent" resistance and
reducing by half the output signal thus maintaining it to about the
same level. The signal at the control diode is then reamplifed by
about 50 to be usable.
The guy says the transistor has to be biased near conduction point at
no signal to be sensitive enough and it makes the circuit somewhat
instable. Here opamps would cure the bias instablility problem and
this is exactly what the guy is about to try (cheating..), just to
see how good this design could be.