I'm getting sucked in (sorry!). This is a long one, so bear with me, or hit
DELETE now:
----------------
PARASITICS
>>I am
>>not sure of how a transient current condition in the tube can produce a
>>VHF ringing.
>The same way that striking a bell with a hammer produces a damped wave
>oscillation. This is the principal behind a spark transmitter.
>Rich...
I mentioned this before, (back in Dec. 1997) but the 1935 article in
Proceedings of the IRE, by G. W. Fyler of GE Company titled "PARASITES AND
INSTABILITY IN RADIO TRANSMITTERS" is a comprehensive report on many of the
types of parasitics that can plague amplifiers. Fyler must have worked on
the WLW 500 KW medium wave TX as he mentioned about a DOZEN different
parasitic oscillations that occured and had to be exorcized before it
worked well. Some even in the class B audio stages. I need to scan this
thing and OCR it, but haven't the time right now. If someone wants to do
it, I'll snail mail it to them. It has lots of diagrams making OCR a bit
more difficult.
He said "Parasites or spurious oscillations cannot be fully anticipated in
building radio transmitters of a new type. [AMEN!] If the best results are
desired it is usually necessary to eliminate any existing parasites by
special tests after a transmitter has been constructed. In many cases the
determination of the parasitic circuit may require considerable study and
use of cut-and-try methods....."
They didn't have email and this mail group then, but the similarities
between what Fyler described (later Terman referenced him) and what you
folks are talking about is.... [twilight zone theme]... astounding.
More from Fyler:"The tubes and associated circuits in a transmitter may
have damped and undamped parasites depending upon the circuit losses, the
feed-back coupling, the grid and plate potentials, and the reactance or
tuning of the parasitic circuits. The damped oscillations or "trigger"
parasites as they are sometimes called occur as the result of a shock or
transient with peak modulation, keying transients, or flash arcs in vacuum
tubes [ROCKY POINT EFFECT]. "
And this:"A capacity must have connections made to it and at a very high
frequency the capacity circuit may actually be inductive with negligible
series resistance."
As G3RZP said:
>Maybe I'm just getting old, but it sure seems that we can spend a lot of
>time re-inventing the wheel, and you can still learn a lot by going back to
>Terman ( and 'Millman and Taub') and maybe a few others.
Yep. see above and below examples....
--------------------------------
MODELLING AND WHAT IS HAS TO DO WITH MODERN AMPLIFIER DESIGN:
>Of course, those guys did a few sums and a few experiments: these days, you
>have to simulate evrything instead.
>Peter G3RZP
We hear a lot about modeling, and component models for all those teensey
chips and circuit boards these days. So we decided to try it with the big
tubes. It worked. We created a complete tube model for several jugs, the
Eimac 4CW150,000E, the Thomson TH537, and the TH555A. In our file we had
not only the capacitances, but the dynamic operation of each element, based
on fitting curves to the constant current plots. The model is made of
equations with diodes to kick on different curve generators. This gave us
real operating points, where we could exercise the amplifier with a sine
wave (or any wave for that matter) using SPICE. The RF part was based on as
many measured components that we could make, like bypass capacitors (mica,
ceramic), using impedance meter to 100 Mhz. The tube inductances had to be
a guess, because no one has those tabulated. The socket stray inductance
also. The stray capacitances came from real measurments in amplifiers.
When we ran the model, the CPU would crash many times. It took overnight to
run a SPICE simulation of only the first 10 microseconds of operation. We
bought a more powerful PC, with 250 MHz CPU. It ran in minutes. The
software caused lots of grief: SPICE 2G engine running in the commercial
program Microcap IV. This is extremely touchy about things like zero
resistance components, zener parameters, etc.
It ran and predicted the amplifier tuning right on, which I designed Pi
networks already for. The transient analysis proved that it would start up
OK, and also showed some effects like oscillation if we put a lot of
inductance in the screen grid bypassing. Also, showed that the output
impedance was lowered with class A over class B and other interesting
tricks. All in all, it was an impressive model, and I even plugged in
neutralization capacitors and saw the effect on the feedthrough power.
So I cannot knock modelling now, as it even will work for high power tubes,
if you spend the trouble making the device models. Still, the old linear
network analysis does fine for most RF amplifier circuitry, along with a
little Fourier analysis of the tube current to get the DC values.
-----------------------------
ANOTHER NAME FOR VARIABLE CAPACITORS
>is true for perfect components on a schematic. It needn't be for real
>components in a real box - the inductance of a 'bread slicer' cannot
>necessarily be disregarded at a couple of hundred megs.
>Incidentally, I'd call it a 'toast rack', but a Californian friend states
>that only Limeys are crazy enough to put toast in a rack to get cold!
>Peter G3RZP
I still call it a meat slicer, but bread slicer may be more appropiate.
--------------------------------------
WARPED
>In a grounded-grid amplifier, how could an anode arc to the grounded-grid
>damage the filament?. I have never autopsied a tube that had a bent
>grid. Bent filament, yes. Gold-sputtered grid, yes.
>Rich...
Want to see a warped grid? Here's an interesting failure. Take a
4CW250,000B tetrode, and run the cutoff voltage on the grid at about -350.
Now, take 30 KV DC on the plate, and leave it there. Set the screen at
about 1800 VDC, pulsed (bring the grid to ground during the pulse). You
have a big switch that will conduct about 100 Amperes with about 8 KV drop
across the tube. After some time, the E-beam in the tube cut grooves in the
plate. The beam was focused. It created lots of loose ions (copper?), which
of course slammed into the filament. Boom, pop. After some time, we
autopsied such a tube, and found the grid basket was warped. Once it
starts, it's a runaway, as the beam radially around the tube is nonuniform,
and the grid/plate spacing changes and causes more beam in that locale,
which then gets hotter, etc.
Concentricity of the grids to the filament and plate are important to keep
hot areas from forming.
Solution was to raise the cutoff bias to -600; something the original
designers neglected to remember, take those tube curves from Eimac as only
representative for 1 tube. Always take the cutoff for 0.1 Amp or less plate
current, and then double it, in my experience. No more warpem gridem.
---------------------------------------
ARCS AND WARPED FILAMENTS?
>IMO the only thing that could bend a filament is the almost instantaneous
>discharge of the unprotected PS. There is absolutely no way ( at least
>for today) that I will believe that a parasitic can generate that much
>energy.
>Carl KM1H
I am not sure about all the causes, but I would say that a high current
discharge, for whatever the reason, will bend a filament, for sure. The
paper that I scanned and offered here last week from Philips "A DISCHARGE
PHENOMINA IN LARGE TRANSMITTER VALVES" proposed the type of forces from a
power supply on a particular filament bar construction, and the deflection
it caused. The importance of using a series R in the anode line is
discussed.
As for parasitic developed mechanical stresses, I cannot comment as have
not studied all the writing and theories of this manifestation.
By the way, another oldie, that Philips paper. Yes, they did a lot in the
1930-40s that we are just relearning. I need to spend more time in the
technical library at work, cause there is a lot to learn from the old
timers who wrote about their work.
john
K5PRO
--
FAQ on WWW: http://www.contesting.com/ampfaq.html
Submissions: amps@contesting.com
Administrative requests: amps-REQUEST@contesting.com
Problems: owner-amps@contesting.com
Search: http://www.contesting.com/km9p/search.htm
|