[AMPS] tube arcs - internal and external

[Rich Measures]

>Thanks for your information, Dick, on the ETO experience with tube arcs.
>
>To shed a little more experience with internal/external tube arcs, I will
>take you back to 1992. I became involved with a problem that was the blight
>of the RF plant here. The super power triodes in three or our finals were
>popping. By popping, I mean exploding open. They cost about $130,000 each,
>so we were seriously concerned. Our tax dollars at work!
>
>--------------
>
>Once I was measuring the plate dissipation with thermometers in the water
>lines, and i heard the infamous tink and saw the tell tale symptoms of
>broken ceramic. The plate tripped off, and I watched the filament voltage
>change as the tube went up to air. Nothing we could do in time to save it.
>The dissipation was 270 KW. The tube was rated for 300KW on paper. It is
>running at 200 MHz, about 50% efficiency, grounded grid, cathode driven
>cicuit. Self bias was provided with a series R in the cathode to ground.
>Cathode current was 400 Amperes, plate voltage at 21,000V DC.
>
>I remember witnessing arcs (hearing them). It was merely a tink in the tube
>envelopes. Our crowbar was functional, to limit the deposited energy to a
>pop, should you short out the B+ from the plate modulators. Using series
>tubes as a plate modulator, we have the luxary of being able to quench the
>B+ within a pulse, usually in a few microseconds. We have a pulsed RF
>system, so there is a lot of off time between HV pulses too.  After a
>lengthy investigation (months), with more broken tubes and down time, we
>believe we solved the problem -- well we did solve it, as we haven't broken
>a triode in operation since, although I knock on ceramic when I say it. By
>the way, the arcs went from plate to grid for the most part. Holes in the
>grid wires were obvious when we autopsied the tubes. I don't recall cathode
>arc marks, but the cathode was usually totally oxidized, having run
>filaments while the tube was going up to air.
>
>What was 'fixed' was:
>
>1) RF triode was produced by company run with new management -  RCA had
>sold to Burle Industries in Lancaster, PA. There were changes in the
>production, although the factory and many people were the same. Some had
>left or retired. There were changes to sources of material, QA steps,
>processes were modernized. During this time, all tubes were questionable as
>to their absolute maximum ratings.
>Similarly, we don't know how clean the ceramic was, and how good the seals
>were. Internal gas could have been a problem, as we did not have ion pumps
>connected on the tubes at that time. This was a later modification that we
>made, which is now standard in all of these tubes.
>
>2) Modulator tubes (the series switch in the B+ line) were 4CW250,000B from
>Eimac. All tubes in operation had around 20,000-30,000 hours on them. No
>buys had been made in 4 years. Someone had forgot that they will wear out,
>these tube things. As the tubes got older, they tended to have more
>internal faults, due to gas, warped grids, electron cutting on the anode.
>The cutoff bias supplies were inadequate, not insuring full cutoff between
>pulses. As the beam cut the anode, a lot of copper ions were liberated
>inside the tube. Xrays could be measured. When they faulted, usually plate
>to screen arced in the tube, and the noise we heard was the external spark
>gap (several Champion spark plugs with the electrodes modified, connected
>from screen to deck). It sounded like a shotgun,  especially if the crowbar
>was marginal. Another problem, ignitron crowbar needed maintenance. When
>the pass tubes arced, a lot of safety margin was lost, putting the B+ from
>capacitor banks directly on the final plate connection, at about 10KV
>higher than normal. All of this was basically DC breakdown problems, no
>parasitics to blame.
>
>3) We put hardware protection in place, where the dissipation is monitored
>in real time, and the HV shuts off if it exceeds 260 KW average plate
>dissipation (as seen by the cooling water).
>
>Other minor fixes were incorporated, such as new processed (debarnicled)
>4CW250,000B tetrodes,  higher cutoff bias on their control grids, and a
>lengthy test 'burn-in' for the final PA tubes, where we ran them at full
>power and duty factor into 50 Ohms, for hundreds of hours before ever
>subjecting them to VSWR of the accelerator tanks. Plus the interlocked
>Perkin Elmer ion pump appendages on the final tubes, that I mentioned last
>week here. When gas exceeded 100 uA of ion pump current, we tripped the HV
>off.
>
>---------
>
>What has this got to do with the discussion at hand, about the internal and
>external arcs of amplifier tubes, such as 8877? I reinforce the idea that
>internal tube arcs are quiet, and that circuit components go with a bang.
>And that the series R of the anode line (or a crowbar AND series pass tube
>in our example) can minimize the damage. And that these arcs and bangs are
>related to tube condition, internally: increasing gas, warped grids, broken
>filament wires, etc. Whether these experiences are applicable or related to
>the smaller tubes, I cannot be absolutely sure. But one would believe that
>the laws of scaling apply, and that it is trouble with smaller geometries
>and element spacings. Whether gas, or rocky point, or some kind of VHF
>parasitic oscillation is the cause, I cannot venture a resonable guess. But
>it sure would be advisable to put the protection R in the plate B+, and
>maybe a fuse or overcurrent relay to kill the main HV supply. Or live with
>the occasional consequences of a broken tube, with spattered gold, broken
>wires, blown grid connection to ground, destroyed meters, etc.
>
>By the way, my SB220 was up on 20 Meters Saturday cutting through the fog,
>and I decided to run the plate tune and load all over 180 degrees of the
>capacitors, while driving high speed dits into a dummy load at about 1200
>watts output. No problem, no arcing or funny stuff. 

I tried the same test on my SB-220 and on my TL-922.  I got more heat, 
less RF output, and none of the alleged 'up to triple the supply voltage' 
arcing.  

>I should go up to 10
>Meters and try it I suppose. I did put Rich's nichrome gadgets in it (the
>hairipin plate feed parasitic suppressor

Since hairpin suppressor inductors are definitely difficult to decouple 
from each other in two tube amplifiers, we currently recommend using them 
only in one tube amplifiers.  {see Figure 20 at:  www.vcnet.com/measures  
for current SB-220 low vhf Q parasitic suppressors} 

>and the grid resistors in place of
>the chokes to ground) 4 years ago, as I had to tear it open to fix the
>dreaded wafer switch. Mine had mechanically become misaligned - got the amp
>for about $150 at a hamfest cause it wouldn't play right. Those output
>wafers don't allow for  much misalignment, should the screws come loose on
>the long rods that hold the switch together. Could this be the source from
>SOME of the failed wafers in SB220s, those cheap switches Heath used?
>
-  I doubt it.  Tank v-swing is minimal when the tank is out of 
resonance.  
-  RE:  "cheap switches":  Heath used a bandswitch with a measured 
breakdown potential of over 5500v, which is a pretty fair safety factor, 
methinks. The bandswitch wafers I see with evaporated 10m contacts do not 
have an indexing problem.  Such amplifiers typically have a problem with 
the zener, and/or a damaged vhf suppressor resistor, and/or a tube with a 
bent filament helix -- problems that seem unlikely to be caused by a 
switch indexing problem.    .   
-    As I recall, during the grate suppressor debate, one chap measured 
the maximum peak tank voltage in a deliberately-mistuned SB-220 at 3600v. 
 With normal tuning, the peak tank potential is roughly 2600v on SSB.  .  

cheers
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