>In a grid-driven amplifier, using tetrodes, one bypasses the filament leads
>for RF, to ground. It is printed in every book I've seen. Some even go into
>more detail, like the Philips RF Heating book by Dittrich says to be sure
>and decouple the cathode/filament connections in an RF heater/oscillator,
>to prevent damage to the filament by RF.
If the cathode is not at RF-gnd potential, the resultant feedback
decreases amplification.
>I spoke with my coworker who is a
>reputable high power RF engineer going back many moons in the big tube
>business. What I heard surprised me. He suggested that one only should put
>bypasses there to prevent RF from getting to the filament transformer.
I agree, and the bypasses need to be low Z. multiple groups of several
different values in parallel seem to work ok for one-guy tetrodes..
> Yet,
>(and I showed him the quote) Seimens and Philips both say that a good RF
>capacitor should bridge the filament connections below the socket, and
>Eimac, ARRL, Terman, you name it, all show the typical 0.01 uF across the
>filaments and from one leg to ground. Of course that one is mandatory if
>you have a center tapped filament XFMR as the RF should get a path to
>ground besides through the winding. What my friend said is that the
>filament has "RF all over it inside the tube" and that you don't need to
>worry about it on the leads (differentially), but only the common mode
>stuff needs to be bypassed. Now, the European books say to be sure and kill
>all differential RF on the two connections, with good low L capacitor.
>Where does this lead me?
If both sides of the filament are well bypassed to chassis-ground, then
there is minimal differential potential across the filament And minimal
cathode (k) RF potential to chassis ground.
. .
>I know from experience with industrial and home oven magnetrons that
>capacitance is required to prevent RF backheating in these tubes from
>shortening the filament life.
>
>I spent the past two days measuring the filament impedance differentially
>on a tetrode, a 250/300 KW device. Using the trusty HP 4193A impedance
>meter, it was swept over and over (Crimson and Clover) from 1-100 Mhz or
>so. By installing the right amount of shunt C across the socket, sure, I
>can parallel resonate the filament. Where to park this resonance? I settled
>on between the fundamental and the second harmonic. On frequencies below
>the resonance, one sees the inductance of the filament structure, and above
>it, one sees capacitive. About 10-15 MHz or so, there is series resonance,
>at which time the phase goes back inductive, (the caps themselves resonant)
>and this continues all the way to 100 MHz (inductive).
>
Capacitor self-resonance is why one needs different values of C in
parallel. For an 8171, I use (4) each 0.01uF, 0.005uF, and 0.002uF @
3000VDC ceramic disk caps. This looks like about 1.3 ohms at 1.8MHz.
The peak cathode current is typically 12a, so this means that the peak RF
potential developed across the cathode would be c. 15v, which is about 5%
of the g-k bias potential.
>I tried the following mounted radially around the filament connections in
>the socket:
>
>combinations of silver mica leaded caps, 3900 pf
>combinations of ATC 100E high power ceramic chip with ribbon leads, 2200 pF
>each
>combinations of the above, 4 ATC, 4 silver micas, soldered in parallel
>combinations of 0.01 ceramic disk, 3 KV
>combinations of Centralab 585, or High Energy HT58 1000 pf 5 KV ceramic
>knobs (too big to mount there)
>Cornell Dublier CDM 272 mica cubes, 0.02 uF, rated for 1 MHz (NASTY
>RESONANCE APPEARS AT 40 MHZ)
>
>My tests ran from A-I last night and A-D tonight, so 13 combinations and
>values were swept, with the tube in the socket, upside down, ass end
>sticking up, the socket sitting on top. It takes three of us to move the
>tube without straining.
>
>All caps had their charateristics, and the best (cleanest, low inductance)
>was the ATC 100E series. Similar to the Murata and Unilator/Matroc ribbon
>leaded high power ceramic caps. They cost about $8 each, and I got 50 in my
>order. So I was happily soldering and punching the ribbon leads, etc. What
>I did was to parallel two in a sandwich, put it under the screws between
>the F and K connections on the socket. There are 6 metric screws in this
>tube socket (itself costing $10,000), and I used one pair of caps between
>each set of hot and ground screws. After it was built a few hours ago, I
>swept it, with the filament leads connected, and now it doesn't change
>whether i connect or disconnect the 300 amp transformer. Aha! I consider it
>decoupled. The impedance at my operating frequency and the harmonics is
>around 3 - 5 Ohms. F is 2800 KHz for this amplifier. I have 0.0022 x 12, or
>0.0264 uF. So, I am happy that I did bypass the leads for the filament, and
>am working to make sure K to ground is also low enough impedance for the
>RF.
>
Sounds like u are headed in the right direction, John, however, 3-5 ohms
sounds like a bit much Xc for a three-guys tetrode. For a 4cx250,000A,
with 2kV on the screen, the peak cathode current is c. 100a. Therefore,
the peak RF potential drop across a 3-ohm filament bypass would be
several hundred V. .It is my opinion that this would substantially
affect tube performance. I would try to limit the peak cathode potential
to no more than 5% of the g-k bias -- i.e., the need for more k-bypass C
is indicated.
>I was wondering if some of your parasitic-obsessed contributors out there
>care to comment on your experiences bypassing filaments. While the ARRL,
>and others have treated it mundane and ho-hum, for many years, as has my
>friend, who has many successful cavity amps under his belt (not literally),
>I could imaging really tearing up a filament in a big tube if one casually
>placed the standard 0.01's across the leads, and parked a big ol parallel
>resonance on the operating freq or a low harmonic. The inductance and the
>lossy nature of the filament do a heck of a job de-Qing the resonance, so
>I read about 30 Ohms at the peak. But it's still there. Grid dipper was
>used to verify, yessir-ree. I could imaging developing some RF potential
>across 30 Ohms. Would this also be on the filament structure, which is up
>to 6 inches away inside the tube, at HF?
>
6-inches at 2800kHz ain't much.
>John
>K5PRO
Rich...
R. L. Measures, 805-386-3734, AG6K
--
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
|