(Note: most of the text in parenthesis indicates addenda.)

Wednesday, 7 September, 1994

From: Richard L. Measures, AG6K
6455 La Cumbre Road
Somis, CA 93066

To: Paul Pagel, N1FB
Technical Correspondence Editor, QST

Re: Technical Correspondence, page 71, September 1994 QST, "Revisiting 'The Nearly Perfect Amplifier' ":
No one person knows everything about high power RF amplifiers, so there are bound to be some issues to resolve. Thus, I am grateful for Paul's column in QST. It's a good venue for separating wheat from chaff.

This discussion involves an unusual phenomenon, intermittent push-push VHF/UHF parasitic oscillation, that presently fuels passionate, sometimes odious, debate. If you are not familiar with this subject, "Parasitics Revisited" (9/90 & 10/90 QST) might be worth reading.
Readers are invited to compare statements made by the six Contributors with what was actually written in "The Nearly Perfect Amplifier."

To begin with, Fred Telewski is correct about my amplifier design goals. I am not a performance person. Long-term freedom from grief and linearity are more important to me than the last watt of performance. Murphy's Law is displayed prominently in my radio room. I try to design with Murphy in mind.

Filament Voltage
I recommended using regulated DC voltage for indirectly-heated cathode [IHC] tubes. How did Fred come up with his 5V, 30A figure? The only amplifier using IHC tubes which comes to mind that could use a 5V, 30A regulated supply is (3) 8877s in parallel.

Filament Inrush Current
On page 4 in the 3-500Z Technical Data, Eimac states that "For best tube life the inrush current to the filament should be limited to two times the normal current during turn on. This will minimize thermal stress (Tom Rauch is correct) on the thoriated-tungsten filament wire, which can cause internal geometry changes during repeated cycling." The 3-400Z/8163 carries a similar proscription. Eimac says nothing about limiting filament inrush current for IHC tubes up to and including the 290 (5kW anode dissipation), Do some people obtain their information from a higher source than Eimac?
Although it's comforting to hear that it's "extremely unlikely" that an amplifier could have excessive inrush current, neither Tom nor Fred provide specific measurements. From measurements, I know that some (models of commercially-built amplifiers) do and some do not.

Grid Protection
I have seen and heard of dozens of cases where a 0.5W carbon-film grid fuse resistor popped during a glitch. Subsequently, the tube usually proved to be undamaged. I have seen cases where a 1A 1mH grid choke on a 3-500Z failed to open during a glitch, and the tube shorted. Such chokes sometimes open during a glitch, apparently saving the tube. Since the fusing current of #27 copper wire is >15A, obviously, an unusual event took place. Such amounts of grid current can not be explained by high SWR, tube gas (the vacuum subsequently tested good), or operator error.
The transistors that are used in grid protection circuitry are typically rated at a few amperes maximum. Does it seem possible that >15A could cause a C-E short in such transistors? Would >15A be likely to open a 0.25A 3AG fuse or a 0.5W resistor? Sometimes simpler is better.
In most of the grid-shorted IHC tubes I have opened, I saw evidence of gold-sputtering on the cathode, missing gold areas on the grid, and yet no sign of heat stress on the grid structure, pretty unusual in my opinion. About half of these grid-shorted tubes came from commercial amplifiers that had electronic grid over-current protection circuitry that (reportedly) shorted during the event that shorted the grid. The remainder of IHC grid-shorted tubes I opened were damaged by over-driving the cathodes. Doing so causes the cathode coating to flake off, creating a grid to cathode short.
I have not found a bent grid in any grid to cathode shorted tube. 100% of the grid to cathode shorted directly-heated cathode [DHC] tubes I opened had bent filament helices.[1] Some of these tubes had low operating hours. I can not believe that Eimac installs bent filament helices in 3-500Zs. If they did, such tubes could not pass their rigid final inspection standards. Tungsten wire is incredibly difficult to bend. To re-straighten bent filament helices in a 3-500Z, I found that it takes a filament temperature of about 2000 degrees K and a lateral force of 10 to 11Gs applied for roughly 30 seconds. Obviously, it took an unusual event to bend the filament.
Tom, Fred, and Reid tell us that grids can not be protected by a frangible element such as a fuse or fuse-resistor. If a fuse melts before the grid it is in series with overheats, why won't the flow of grid current stop? Considering that the mass of a typical 0.25A 3AG fuse element is less than 1% of the mass of a typical grid, and the melting point of a fuse is at least 1600 degrees F less than the boiling point of gold, I don't follow their logic.

Are fuses obsolete? John Fluke Co. uses a fuse to protect their DMMs. Thanks to its 2A fuse, my Fluke has so far survived every time I made a mistake, and I am well into my second box of fuses.
Re: Page 71, column 3, paragraph 1: Reid states that "We do not understand the reference to sudden bursts of VHF or UHF grid current." Then Reid issues a blanket clean bill of health to every commercial amplifier made. However, one commercial amplifier manufacturer now admits to customers that their amplifier [2] occasionally has parasitic arcing at the bandswitch, but not any more so than other manufacturers' amplifiers. This is an improvement over what they (Craig at the Kenwood service dept.) used to tell us, that burned bandswitch contacts were caused by rapidly switching the bandswitch while transmitting at full power.[3]

At the urging of Paul Pagel, Tom telephoned me after "Parasitics Revisited" was published in the Fall of 1990. Tom said that he was angry because customers who read my article were telephoning him complaining that their burned bandswitches looked like the ones in the article's photographs. Tom put the blame for the burned bandswitches on cheap coax, operator error, and bad antennas. During our conversation, Tom told me that he had repaired 400 Heath SB-220s. I asked him if he ever saw parasitic damage. He said that SB-220s have lots of parasitic problems. I asked him why the amplifiers he designed did not have parasitics. He replied: because Tom Rauch designed them. After I pointed out that his parasitic suppressor design was virtually identical to the SB-220's, it sounded like someone switched on a rather stentorian speech processor. (My wife, Sue, was standing in the doorway, about 8ft. away at the time. I was holding the phone handset well away from my ear. She could easily hear Mr. Rauch. In German, "rauch" means smoke.)

It was suggested that I discovered/invented intermittent push-push VHF/UHF parasitic oscillations in order to make money selling Low VHF-Q Parasitic Suppressor Retrofit Kits. I started selling these kits around December of 1988. It was not my idea. It was the idea of amplifier owners who had experienced the problems I discussed in the October 1988 QST article. I have not made money selling suppressor retrofit kits because that' s the way I decided to operate the business from day one.

A brief history of AG6K and parasitics: I did not discover them. They found me. Parasitic oscillations appeared in my SB-220 amplifier, quite uninvited. Intermittent arcing, an occasional big bang, tune-up vagaries and smoking suppressor resistors were the manifestations. I tried replacing the resistors, but the new ones subsequently became crispy-crittered. Initially, I had some hunches, but I was not confident what was at the root of these problems. I am a curious guy, so I began working on other amplifiers that had similar problems. The denouement began when I agreed to work on an unruly Dentron MLA-2500 in the Winter of 1985-1986. After some experimentation, it appeared that there might be something wrong with the two 8875 tubes. I telephoned Eimac. They connected me with Willis B. Foote, Eimac's Chief Specifications Engineer of the Power Grid Division. I told him what I had observed. He said that he needed to evaluate the tubes. I sent the tubes in. Subsequently, I received a phone call from Mr. Foote. He said that both tubes were found to be defective due to erratic leakage and emission problems. An Eimac tube engineer cut open the bad 8875s for inspection. He found that the gold plating on the grids had evaporated/sputtered and contaminated the insides of the tubes. Mr. Foote said that the damage was probably caused by a high frequency oscillation. I asked how high was high. He said perhaps very, perhaps ultra. He said that Eimac tube engineers encountered the gold-sputtering phenomenon during testing of the 8877. I asked him if an engineering bulletin had ever been published on the subject. He said no, adding that finding a way to improve the stability of this amplifier was my job. He said it was good we were having this conversation because he was retiring shortly, and all of the members of the 8877 team who had knowledge of this subject were gone. Later, I received a letter from Mr. Foote that outlined the oscillation/gold-sputtering problem. Mr. Foote also sent me a gratis pair of new tubes. I was surprised because the bad tubes were long out of warranty and the oscillation problem was not caused by a fault in Eimac' s design of the 8875. Power grid tubes don' t oscillate by themselves. A tuned circuit in the amplifier allows them to oscillate. A copy of the Eimac/Foote Letter is enclosed.
When I spoke with Paul Pagel on 24 August 1994, Paul said that Reid Brandon (who works for Eimac) was aware of the Foote letter. Paul told me that Reid said Mr. Foote was not authorized to release the information to me.
Did I state that parasitics bend grids, destroy tank capacitors or cause anode to cathode arcing?
Re: the list of "recognized experts in the amplifier community" A Hughes employee told me that one of their 100kW RF amplifier designs incorporates several low Q devices to thwart anode circuit parasitic oscillation. I noticed that Hughes is not on Tom Rauch's list of 'recognized experts'.

Tom states that it is impossible to perform failure analysis of a tube and determine if the failure resulted from excessive VHF or UHF grid current. Eimac was doing it before 1986. You should try looking at the insides of a bad tube yourself, Tom. It' s fun and educational. A low power microscope is helpful. Details and photographs are provided in "Parasitics Revisited."

Tom tells us it's possible to damage an 8877 in "only a moment" by applying 100W drive. Normal drive is around 80W. In a vacuum, gold boils at roughly 2000 degrees F. The 8877's grid is made from bar-stock, not wire. Thus, its mass is substantial, probably on the order of 40g. How can an extra 20W of drive BOIL GOLD off of the grid? (This is simply not possible with a mere 20W of surplus drive--also, see below: "VHF Stability".)

Glitch Protection
I agree with Fred that it's a matter of designer choice. For long-term freedom from grief, planning for unusual events hardly seems like a matter of choice. Using a glitch protection resistor to limit fault current is not Rich's idea. Beginning sometime around 1978, Eimac began stressing the importance of limiting fault current. In the Eimac 8989 Technical Data (1985), page 3: "A protective resistance should always be connected in series with each tube anode to help absorb power supply energy if an internal arc should occur." More information is provided in Eimac Application Bulletin #17, "Fault Protection". For their 5CX1500B, Svetlana Electron Devices in St. Petersburg, Russia says "The tube and associated circuitry should be protected against surge current in the event of an arc with a current limiting resistance of 10 to 25 Ohms..."
While it is true that glass-coated wirewound resistors are not as high-tech as pulse-energy absorbing resistors, in practice 10W 10 Ohm glass-coated resistors seem to hold up pretty well in legal-limit amplifiers, and they are much less costly. For above-average (over 3000V) anode voltages, using two (or more) such resistors in series is indicated.
Tom is correct about the ESR of electrolytic filters. For a typical HV supply and HV-RFC, a figure of roughly 9 Ohms total is representative. However, 3000V divided by 9 Ohms is still a substantial amount of peak current (over 300A). From my experiences, as well as Svetlana' s and Eimac' s, a series resistance is needed to reduce the peak discharge current.

Power Supplies
Transformers, Re: Potting: Polyester laminating resin is designed to allow air bubbles to escape from fiberglass laminates. I use the stuff, without a vacuum chamber, for potting transformers. I see air bubbles cease rising to the surface before the resin gels. For specifics, send me a SASE.
Rectifiers: I agree that rectifiers should be tested. I covered this subject in a QEX article I wrote on building and using a HV breakdown tester. Manufacturers of HV rectifiers stopped using so-called "equalizers" at least a decade ago. Old habits die a slow death. (The 1995 ARRL Handbook advises not to use "equalizers" because they can cause rectifier failure.)
Fred said he did not understand my comments about choke filters. A question: Why don' t Collins Radio, Gates, Henry Radio, Continental, Harris, Hughes, and other manufacturers use swinging choke filters or non-resonant fixed choke filters in the HV power supplies for their linear amplifiers? Such filters try to maintain a constant current into the load, not a constant voltage across it. To amplify SSB linearly, an amplifier needs a fairly constant supply voltage. The only type of choke filter that tries to maintain a constant voltage across the load is the resonant-choke.
Once I asked an old timer, Mac, W6SDM, what he thought about using choke filters. He told me that choke filters have poor load-transient V regulation. He recommended a test. Mac said to connect a DC oscilloscope, not a meter, across the output of a choke filter power supply, change the load resistance, and watch the scope. I tried it. When the load R decreased, there was a temporary dip in output voltage. When R increased there was a temporary surge in output voltage. But isn' t that exactly what an inductance is supposed to do, i.e., maintain constant current? (That's just basic AC Theory, Fred.) The larger the change in load current, the larger the voltage transient. When the load went from zero to 100%, the output voltage temporarily dipped more than 50%. The 'old goat' was right, as usual. Although transient regulation can be somewhat improved by increasing the C of the filter capacitor, the best solution is to resonate the choke, like the big boys.

Electrolytic Capacitor Equalizing Resistors: Tom says that the resistance of such resistors is very important due to leakage current considerations. I can' t find anything like that in Sprague' s Engineering Bulletin 3431D for electrolytic capacitors. It seems to me that it would be nice if the equalizer resistors did not bake the life out of the capacitors. Currently manufactured 450V 300 microfarad electrolytic capacitors typically exhibit a leakage of well under 300 microamps @ 25 degrees C and 450V. With 100k Ohm equalizers, the ratio of the equalizer and leakage currents is more than 15 to 1. In wide practice (more than 6000 units), Matsushita 100k Ohm, 3W, MOF resistors have, to the best of my knowledge, worked flawlessly in this application.

Did my article say that there was anything wrong with electronic bias switches? I said that there were problems with RF-actuated electronic bias switches when using SSB. I use an electronic bias switch, but it is not controlled by the amplitude of the RF driving the amplifier. Thus, it is not possible for the amplifier to switch into non-linear bias during transmit. (If you guessed that one of the critics uses RF-actuated bias switching in the amplifiers he designs, congratulations!)

High Speed Relays
Fred says he isn' t sure the relays are sequenced properly. I don' t see how anyone could be sure unless he measures (no pun intended) the make/break timing with a dual-trace oscilloscope. I did.
Steven Katz wrote to me extolling the virtues of PIN diodes. It sounded pretty good. I suggested that he write an article for QST. How about it Steven? (QST rejected Steven's article) Incidentally, Steven lives in the same area of So. CA I live in. We don' t have much lightning. A number of people who live in areas that have severe lightning told me of problems with PIN diodes. A legal-limit amplifier develops 800Vp-p into a 50±j0 Ohm load. With a greater than 1 to 1 SWR, the p-p voltage can go down or it can go up depending on the Z of the load presented to the output terminals of the amplifier. If the load was 10±j0 Ohm,[4] the SWR would be 5 to 1 and the potential would be 350Vp-p. If the load was 250±j0 Ohm, the SWR would be 5 to 1 and the potential would be 1600Vp-p. If the inverse rating of the PIN diodes is 1000V, there might be a problem.

VHF Stability
Did I say anything about striking gongs? Spark gap transmitters operate on a principle that can be compared to ringing a bell. Both generate damped wave oscillations. If there is a better analogy, I would like to hear it.
If you have doubts that the anode resonant circuit in HF amplifiers generates energy at its VHF resonance, loosely couple a spectrum analyzer to one, key the amplifier on and off, and watch the analyzer.[5] If there is no damped wave VHF signal present during anode current transients, then AG6K is a contemporary version of the infamous Larson E. Rapp!
The main obstacles to a better understanding of push-push VHF/UHF parasitic oscillations are their propensity for transience and lack of cooperation. When a parasitic decides to appear, the observer doesn' t have a chance to observe and make scientific measurements. He hears a (typically loud) noise, perhaps sees a flash in the corner of his eye, and it's all over. What follows is like investigating the scene of a fire.
Tom assures us that there is little difference between silver and nickel-chromium alloys at VHF, but that nickel-chromium alloys produce dire consequences at 28MHz [sic]. In an HF amplifier, my dipmeter says otherwise.
In my opinion, parasitic oscillations in HF amplifiers can not be completely eliminated by lowering the Q of the VHF resonant anode circuit, but they can be made less likely to take place. And if one does occur, glitch protection can prevent damage.

Is More Gain Always Better?
On the air, in SSB mode, ALC does not perform as well as it does on paper. The intrinsic problem is finite attack time and subsequent overshoot. One look at the typical ALC path in a transceiver shows why, a series of decoupling resistors and bypass capacitors. Does a "systems view" (Telewski) obviate T=RC ?

Adjustable Tuned Inputs
If what Tom and Fred are saying is correct, i.e., that the radio' s output filter reactance and the amplifier' s tuned input reactance do not interact, then it should make no difference what length of coax is used between the radio and the amplifier. (But it definitely does.)

I believe that technical disputes are best resolved by hands-on measurements and the laws of physics, not philosophy or opinions, even if they be "expert."
At the risk of plagiarizing McGeorge Bundy, I would like to close by saying that there is no safety in technological hubris.[6]

R. L. Measures, AG6K

Copies to:
Al Brogdon, QST
David Newkirk, QST
Paul Pagel, QST
Mark Wilson, QST
Reid Brandon
Bill Clemow
John Fakan
Steven Katz
Tom Rauch
Fred Telewski

Notes [...]
[1] Photographs of grid-shorted IHC and DHC tubes appeared in "Parasitics Revisited", September-October 1990 QST.
[2] Uses silver-plated parasitic suppressors.
[3] If the RF relays in an amplifier switch slower than the transceiver, non-parasitic related bandswitch arcing can occur during the R/T transition.
[4] To be able to match such an impedance, an amplifier's tank would need a higher than average Q. A pi-L tank would be helpful in this regard.
[5] Use a short probe wire and start with high input attenuation. The probe should not be within 5cm of the anode resonant circuit.
[6] "There is no safety in unlimited technological hubris" (McGeorge Bundy).