I try to stay out of these discussions but this reply was spot on:
--------------
Date: Fri, 16 Jul 2010 13:20:37 +0100
From: Dave White <mausoptik@btinternet.com>
Subject: Re: [Amps] parasitic suppressor voodoo
Maybe another reason that we don't see too many worked-through rigorous
calculations is that there are so many variables in play and since
parasitics result from positive feedback the smallest variation in the
values of the large numbers of seemingly insignificant factors can cause
huge variations in the system behaviour down the line. The sheer
randomness of many factors also comes into play.
--------------
Dave may have it nailed. I add that it is difficult to predict that a
tube or transistor will act as a parasitic oscillator, before building a
circuit. Spice models of the devices are essentially worthless with
respect to stray inductance and capacitance internal to the device. Our
tubes are so large that even the electrode connector rings (for ceramic
metal) and the cones and conducting cylinders inside create lumped
circuit elements that at some frequency, might be significant
reactances. Transistor bond wires are measurable as well as the shunt C
in the device, but the varactor capacitance that happens on the drain or
collector depends on the voltage at that terminal.
If you've gone through the analysis of an oscillator in any of the
published designs (in old QST, ARRL, Orr books, new solid state books,
you can see how critical the condition of oscillation is on having the
proper phase shift and feedback voltage from output to input across the
circuit. When you build it the freq and stability is rarely exactly as
math would predict with using the known component values. So a tweaker
is installed, a variable C or L, and the thing is force to oscillate
where you want it. (except for crystals). Think about an amplifier that
turns into an oscillator the same way. In high powered industrial
heating business using RF, I used to have to add C and L to force a
triode to want to self-oscillate on a fixed frequency as a tuned-plate
tuned-grid or Harley, Colpitts or Meisner configuration. Even this
wasn't so easy, as I could get the thing to go but not where I wanted it
to (frequency). Now, think about how to force it NOT to oscillate. You
have even less knowledge of the C and L that are causing it. This is
where empirical and cut-and-try methods prevail.
There are so many ways for a parasite to occur in a high gain amplifier,
be it tube or solid state. A lot depends on the construction, the
operating frequency, the layout, the choice of power device, and all
that has been discussed here before. It is not something that simple
mathematical analysis will solve and force a solution that works.
Parasites are the bane of high power commercial amplifiers as well, they
have to be watched for, and if they are present, may cause serious
component damage. From my own experience, they are one of the last
worrisome factors that I watch for when firing up a new amplifier for
the first time. If they don't happen, and you can even exercise the
tuners and load value and still have unconditional stability, then
congratulations, don't install any extra parts. If they do happen, you
have to allot extra time, usually without knowing who long it will take,
to stabilize things. Project managers get antsy as they cannot
understand how it is not all cut and dry, straight from Ohms law and
Maxwell's equations. I have seen them take months to solve in high power
circuits to as little as a few weeks. The worst kind can be pumped by
harmonics and create appreciable voltages and currents. Some parasites
only spit during a portion of the anode voltage swing. Once one
graduates to tubes that have a circumference approaching a wavelength in
the VHF or UHF region, they gain a new respect for parasitic suppressors.
To conclude, all parasitic suppressors are there to put a finite load
somewhere in the circuit (in the leads or in a cavity) that will dampen
the high impedance at that node for the frequency of the parasitic
voltage. Some call it de-Q'ing. The suppressor may need to be broadband
enough if the freq. range of instability is wide. Others may need to
only work on specified frequencies. In all suppressors, they are made to
be frequency selective so that the fundamental desired RF power is least
affected by the suppressor, to prevent burnout and loss of efficiency as
an amplifier. This is the fundamental principle, and from here it is up
to the technician as to how to put this into practice. Passing on
tried-and-true methods like "50 ohms and 3 turns of wire for
such-and-such tube" are very useful ways to avoid spending weeks
relearning what others have. Math isn't so workable beyond knowing the
reactances of the L and R at various frequencies, from fundamental and
the parasitic range.
73
John
K5PRO
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|