The AM Forum

I just finished my amp and hooked it up to my single tube driver putting out around 12 watts of drive for 75 meters.  I haven't adjusted the grid tank yet for optimum and I'm testing on the air through my link coupled tuner to a dipole (tuned for resonance, and I know on the air testing is bad). Borrowing a large dummy load tomorrow.

Anyway, I started tuning the plate when the output skyrocketed to over 2kw and the SWR was pegging the meter when the plate choke literally exploded.  It sounded like a pistol went off next to my ear.

What would cause that?  Parasitics or something?  12 watts of drive shouldn't cause the amp to go over 300 watts, I'd imagine. 

More numbers:  4000 volts plate voltage (unloaded) and 4.9 volts filament. 

Jon
KA1TDQ

...another note

Plate capacitor was at near minimum capacitance at the explosion with loading fully meshed.

Did you dip the choke for resonances once mounted?  Chokes will show series resonance at one or more frequencies and those need to be outside of the ham bands or there will be fireworks like you experienced.

Exactly!  There is a "how to" somewhere in ARRL literature.  For some reason, however, I seem to recall the need to short out the choke although that didn't make sense at the time nor does it make sense.  I bet that was your problem

Wow!  I've been there, so I truly feel your pain!

So, all chokes like that will exhibit self-resonances (as has already been pointed out).  The key is to keep them out of the ham bands, and usually they are.

I'm suspecting one of the following:

1) You had/have either an instability in the amplifier itself, and it was self-oscillating at some random frequency
2) The amp has a parasitic
3) The plate circuit was tuned to a harmonic of the input frequency, and there is a self-resonance at that frequency (less likely, but possible)

Because the output suddenly jumped up, it sounds like an instability and the amp was oscillating.

I'm going to assume you had the grid at resonance, etc.

I don't remember if you have parasitic chokes on the plate and grid leads of each tube or not.  With parallel tubes, this is an important thing to have.

Perhaps this will help

http://www.w8ji.com/rf_plate_choke.htm

GL, Al

 :o  There, my ears have finally stopped ringing  :)

This happened to be the exact same plate choke that I used with my last homebrew amp, so I know it was good for 75 meters at least.  Someone also messaged me and pointed out that this turned out to be a very good oscillator.  That would explain the over 2kw output (momentarily).

For the record too, I do have another cooling fan beneath the tube sockets.  The blue RF deck fans help cool the glass surfaces and (didn't realize this until after I installed it, but I'll take credit for the engineering) help force a spray bottle effect by pulling air from below the tube sockets.  The fan below then adds to this.  You can totally feel good air flow all around the tubes on all sides, so I'm good there.

I didn't tune the input tank either to 75 meters so that's part of my problem.  I will also borrow a grid dip meter to tune the input properly before I belch out my next oscillator.  The good news is that I just have to rewind a new plate choke.  That last one had sentimental value though since it's the one my wife and I wound back in the MA trailer.

Jon
KA1TDQ

Jon,

The plate choke needs to be mounted in place before checking for resonance.  Just because it was OK in one amp doesn't mean the resonance wasn't close to being in-band and the proximity to other components in this amp could cause it to shift.  I used a commercial choke in my homebrew triple 4CX800 amp and it was marked as having a resonance at 13.1 and that is what my Measurements 59 GDO indicated in the clear and when mounted in the amp the resonant frequency was about 200 khz. lower which was still fine but certainly different.

You stated the tune capacitor was at near minimum mesh, where should it have been with the calculations you made for operation on 75?  Any chance your driver is putting out significant 2nd harmonic energy (or a spur) and you tuned up on that?  If so you might have accidentally located an out-of-band resonance in your previous choke.

I remember seeing a photo of one of the classic National chokes where the pi layers had slammed together due to choke resonance when the amp took off on a spurious frequency.

I'm with Steve, I'm thinking instability.  Out of curiosity, is there a bypass capacitor at the input side of the choke (power supply side)?  I generally use a 500 pf doorknob for this purpose and place it physically close to the input side of the choke. 

Possibly the 2KW reading was erroneous cuz when parasitics are introduced to a 50 ohm wattmeter, the swr soars and can cause high readings.  But a pair a 3-500Zs can put out 2KW when pushed hard.


Over years of building, I have had my share of imploded plate chokes too. If it's a series resonance, just rewind it slightly differently and try again. The chance of it being series res on a ham band is small, but it does happen.

I can't tell clearly from the pics, but is your filament choke and filament structure "looking" at the plate tank circuitry?  The entire fil structure should be shielded from the plate. This can cause a TPTG oscillator if so.   This amplifer is good for 13 to 16 dB, right?  Hi Mu triodes need good input to output isolation, esp if they are coupling at higher freqs..

It could be many, many other things...  I'm assuming this is a GG (grounded grid ) amplifier, right?  If so:  Try loading the filament to ground with a 100 pf cap in series with a carbon 100 ohm 5 watt resistor. This will help kill filament choke resonances. Add some ferrite cores to the fil leads near the tube pins.
 
If it is grid driven, is it neutralized? It will be a bear to tame if it is grid driven, so be prepared.


Try more turns on your plate suppressors.


Try increasing the width of your plate tank connections. All leads should use 1/2" wide (or wider) copper flashing, at least.  Make sure you go directly to C1 from the tube plate caps. Then directly to L1 and then to C2.  If you run a split path, this can cause VHF parasitics to arise.

Test it at a lower voltage until you are more confident of stable operation.  1500 volts is plenty for now.

Put a 50 ohm,  25 watt resistor in series with the HV lead to limit current for initial testing. This is to get you past the blow-up stage.

Use a mirror behind the rig if there is arcing. With the lights dimmed you may be able to see arcing easier if it is occurring.

Put low drive thru the rig with the precautions listed above - and then move C1 and C2 (and input tuning) all around looking for fuzzy stuff on the scope.  You should see a stable carrier if running right.

** Get some see-thru Plexiglas sheets to construct a rig cover for safety.... if you want to continue viewing the insides.  Otherwise, scrap yard aluminium sheet is cheap and FB.

There's more, but go slowly and you will find what's troubling your patient... ;D

Post your progress.

T

It found resonance somewhere, that's for sure. I was using a tube driver (homebrew) so I might just add a low pass filter for 75 meters as an added measure.

My last transmitter project used a commercial rig for the VFO with no harmonics. My current setup ia all homebrew so I'm probably spewing harmonics all the way to 10 GHz.

Jon

I would say part of your problem is the amp is not built very well.
For interconnections, you should use strapping, copper or silver plated flat wide thin stuff.
You can flatten copper tubing, or cut sheet copper.
Thin round wire is never used for amplifiers.

Also, you want to keep the grid and plate circuits separate, grid stuff below the chassis, plate stuff above.
For grounded grid, keeping the plate and cathode separate is a good idea.

Bypassing needs to be close and direct, with the right caps.
Common point ground is a good idea.

Most amps can have issues with some combo of grid and plate tuning with really high voltage on the tubes.

Your plate tuning cap looks way too large (capacitance) and the loading cap looks too small.
The air gap on the plate tuning cap looks to be WAY too small for 4000 volts.

When I built my 2x 4x150a rf deck, I used a nice plate choke out of an FT102, and the deck worked very oddly.  I changed the choke out to a 4 pie wound old thing in the junk box and it worked fine.
Sometimes just the place bypass cap can be changed to get an improvement.

This is a GG pair of 3-500Zs, not grid driven, correct?


You said:  "I haven't adjusted the grid tank yet for optimum."  

The grids should be strapped directly to ground through copper strap.  Use a 50 watt zener of the correct voltage in the filament center tap for cathode bias. The input tuning L/C should be connected across the filaments to ground.  Maybe this is what you meant??

T
 

J,

Check this out.

http://www.tubebooks.org/Books/intro_orr_radio.pdf

Its the 'west coast handbook' ---  lots 'o gud stuff.  Better than the ArrlgL stuff. I've got the 14th ed, but this one should do.


klc

I gotta tell you.

As shown and built:

The amp is EXTREMELY UNSAFE.

The open construction looks pretty, and you said that there is no one else or pets, but that doesn't change it for a second.

All you need to do is to reach up to tune the thing and slip a few inches and you're DEAD, instantly.

Consider at least scavenging some chrome kitchen appliance grilles and re-purposing them as grounded safety shields??

Recycling yards will see you scrap metals close to scrap cost...

Never skimp on safety.

                                _-_-


Consider the plate choke as a warning...

You can buy the RF choke that's used in the Ameritron AL-82 (pair of 3-500's) for $20.

http://www.ameritron.com/Product.php?productid=10-15197

True!
All you have to do is get to close or brush up against that and get killed.

2000 volts scares the crap out of me, so I put it all in locking racks, 4000 volts is crazy!

Brett
N2DTS

The amp looks cool but agree on the safety aspect.  Put plexi shields on it so you don't lose the looks.

I'm finally at a computer and not on my phone so I can type a good reply.

Yes, safety aspects noted.  I'm building the plexiglass housing now and everything will be totally safe.  I've got a baby boy on the way and a Persian cat that I want to keep around (not to mention my wife and I). 

To eliminate  :) aspects of the whole thing, I am building a low pass filter for 75 meters to go on the input.  I will also do the 100 ohm resistor to ground with a 100 pf capacitor as mentioned above. 

I'm going to rewind the plate choke with slightly larger #20 wire.  This will still give plenty of turns and have a larger wire guage for current flow (should I need it). 

I will also drive this with just 5 or 10 watts using just my solid state driver and IRF510 amplifer and not the tube rig.  It just makes everything 50 ohms and a lot simpler.  I can very easily adjust the drive power as well.

Another thing I will do is remove the tubes during the initial test and connect a large 50 ohm resistor to the cathode.  I can then rough-in the cathode (not grid  :) tank for 50 ohms impedance at the proper frequency. 

This should all equal no blowy-uppy.

Jon
KA1TDQ

Jon,

Adding a low pass filter to the input will probably NOT have any effect on curing your parasitics.  The problem is most likely generated from within the amplifier itself, not from the exciter.

If your exciter is taking off due to poor neutralization or parasitics, then this is another story and should be addressed as a separate problem. It should be worked on and stable before driving a big linear amplifier.  

Try a commercial ricebox driver as a test once you get a handle on the big amp.

One way to find out if the 3-500Z amp has problems is to ground the 3-500Z amplifier input with a 50 ohm carbon resistor to ground. (no signal in) Then tune C1, C2 and the input tuning every which way.  If you see the plate meter bounce around, excessive, random plate color, or see some fuzziness on the scope, you will confirm it is generated from within the amplifier.

Do these tests at 1500 volts or so to minimize damage if it takes off. Slowly bring the HV up as you proof things out.

Good decision on adding the Plexiglas cabinet for safety.

T

Im at Frys now getting components so Im glad you caught me. Will do...

Jon

I wound a new plate choke using 1/4 round ceramic tile glued around a wooden dowel.  I moved the DC blocking capacitor closer to the parasitic suppresors as well. I put a couple FT37-43 cores around the filament leads going to the tubes too.  I also stuck in a 100 pf / 100 ohm resistor in parallel with the input tank to calm things down. 

Last thing I need to do is place in in-line resistor to ground with the negative plate supply lead so that I can measure plate current with a voltmeter.  I'll do testing this weekend with a dummy load and with plate voltage at around 1500 volts.  I'll start out with 5 watts drive with the solid state driver first before I add the tube transmitter in line. 

Jon
KA1TDQ

...and as a final change I am going to use copper flashing for the grid leads to ground.  Hopefully all this will lead to no plate gunfire.

Jon

Good going, Jon.

In hindsight, you will find the explosion was the best thang that could happen. You are cleaning up the rig now and will have less chance of explosions happening over and over.

Question:  Those long bare wire leads I see in the picture... are they filament leads and part of your RF input circuit? If so, they are looking at the plate structure and may cause the amp to take off again.

Normally, all this fil/input stuff is BELOW the chassis and shielded away from the plate tank.  You should probably put an alum shield around the bifilar choke and put some coax braid around the filament leads, using heavy insulated wire for the fil leads.  Ground the braid and make sure all fil and RF in leads are shielded from the tube plates and plate tank circuitry. 

Even your antenna relay that switches input to output needs to be carefully looked at to be sure these coaxial connections are not coupling to each other.  I usually use two separate relays and think out the potential situation for coupling input to output.

Once you make the mods, take some more detailed shots of your amp so we can see other problem areas. (especially under the tube sockets)  If you do things right this time, there is a chance that amp will be very stable on the second try - just like a pair of GG 3-500Zs should be.

T

Actually, that's a duplicate ground for the grid. Im still going to put copper flashing on though.

The filament leads are under the chassis and steel ground. Should be fine.

No antenna relays. Beverage antenna for rx and dipole for tx.

Jon - on cell with small screen. Im getting a huge Samsung Mega tonight! The wife decided to treat me.

Good on the fil leads being already shielded.

Yes, that LONG, thin grid lead to ground could be a reason for the TPTG oscillation.  When you later ground it with wide copper strap to ground, make the connection as short as possible. (1" long) Do it separately for each tube. You can run a grid strap between the tubes too, if you wish.   Under the right conditions, even a few inches is enough to cause a VHF parasitic to come to life.

I don't remember, but if there is more than one common grid pin internal tube connection in a 3-500Z, strap them all to ground.  Since there is no screen to bother with, the grid is the critical element to worry about.

I see you are using an aluminium plate for the chassis. Look at all RF paths and make sure they ALL get grounded directly to that plate with short copper flashing connections.  All fil caps, grid leads, RF plate bypass cap, C1, C2, input tuning caps, safety choke, swamping resistor/cap, *everything* related to RF needs to go to that common plate groundplane.  Connect your exciter to it thru a thick braid too, as well as the shack common ground for audio, HV power supply, etc..



You should have much better luck next time... ;)

T

The sooner the better!  In an instant, your wife can be a widow and your child without a father, my friend. Do you have that guard around the PA yet?

When I was working near the pi network on my homebrew 833C amplifier earlier this year, I found myself waaaay too close to live 3500 VDC. I decided it was time to stop right then and install some hardware cloth screen around it.

And you need a chicken stick, if you don't already have one. Voltmeters and their dropping resistors can fail without warning.

Don't let us see your name in the Silent Keys section of QST!  :)

I'm selling an old Admiral 399-6M (working!) tabletop radio on eBay right now, and when it sells I'm and going straight to Home Depot to get my HV section plexiglass.  When that plate choke exploded, it reminded me instantly about how much power is sitting there.  That should be an exercise for newbies... blow up a plate choke.

Jon

Sounds like a plan, Jon.

The following experience should be required reading for every tube amp builder. :-)
https://www.eham.net/ehamforum/smf/index.php?topic=72496.msg490120#msg490120

That's a great looking amp. I like how everything is accessible and you can work it hot if you have to. Mine, like an iron maiden I have to clamp an interlock to work it live and you can't see anything.. but once the bugs were out that has not been necessary to have it open.

It would be nice with Lexan covers so its beauty could still be seen and the dust would stay out better. Another member here built an amp with clear panels, although it was not so interestingly illuminated. You could also engrave the covers/panels with the text or ornate patterns of your choosing and light those, as were the old ruled oscilloscope graticules lit.

A suggestion if you do want to cover it for example with Lexan or Plexiglass to show it off while preventing dust getting on it, might be to put magnets on the frame and epoxy little steel discs to the lexan and 'stick' the covers on to enclose each side as you like.Easy on/off.. not pesky.

Regarding electrical safety, everyone working around high voltage should view this video...

http://www.youtube.com/watch?v=hfnEuRA7-vo

Most radio people working around high voltage should not be. 99% of the work can be done with the power off.

Thanks for the safety reminder.  I just pulled a power supply from the shelf last night to begin restoring it.  A piddly 800v compared to  2300v, but I promise not to take short cuts.  b.

Jon,

From where I sit you are far too cavalier about the HV and the danger.

Imo, if you are not scared out of your wits each and every time you are near that stuff, then you're
a prime candidate for a pine box. You should be scared silly, especially at 4kV.

That's serious HV and hyper dangerous.
It can and will reach out and grab you.

Plexi/lexan is ok up to a point.
In my view at minimum the tubes + local parts should be in a faraday cage.
It could be as simple as wire cloth/window screen (it comes in different weaves), held up
by a frame.

There is a REASON that amps and other HV gear uses interlocks and double shielding.

The POWER SUPPLY MUST BE GUARDED and be unavailable to be touched by anything or anyone at any time, regardless of what unexpected events are taking place that hurtles them into the amp!!

I'd buy the magnet hold down ONLY if there were also pins that hold the plexi in place, so they can not be removed by simply sliding them off.

Ok... think about this more carefully, please.

Nah........... When I first built my 4X1 rig, I was testing it into a dummy load. I was slowly bringing up the power and the audio. At about 500w of carrier, I socked an atomic yaaaaeeellooo into the mic and blew the plate blocking caps to bits! It went off like a blockbuster!

After I cleaned my drawers out, I looked to see what blew. When I saw the caps blown to bits, I sat back with a big shit-eatin grin and laughed. That crapout was a good thing! ! ! !    ;D  ;D

Frank

heheheh....

Speaking of blowing up caps, I think we all make this mistake in the beginning: We try to use a string of fragile 500 pF TV doorknob caps as a padder for C1 in the plate tank. They cannot handle large RF circulating currents as found in a KW resonant tank circuit.

A friend of mine built a 4-1000A linear and didn't have a plate tuning cap. He made a series stack of five 500 pF door knobs that were selected by a bandswitch - using a roller inductor as the tuning.  He fired it up and BAM!  I was gagging from laughter. There was tan doorknob material all over the inside of the amp. What a mess. He's lucky he didn't blow a hole thru his chimney and 4X1 tube glasswork.  The amp worked FB once he replaced the doorknob garbage with a conventional air breadslicer cap.

T

Funny you should mention this. About the time you posted this, a 90 pF 5 kV disc ceramic padder across C1 in my homebrew amp failed. It just shorted and smoked, nothing spectacular. I was loading the amp into a high SWR with ~3500v on the plates when it happened. I wonder what the best type of cap for that purpose is, other than one with a higher voltage rating?

from the pics it looks like the plate blocking cap in this amp is one with a 5kV rating. I would wear safety glasses around that amp.

One has to consider both the DC blocking capability of the capacitor, and (separately) the ability of the capacitor to handle AC (i.e. RF) current.

For example:

An RF bypass capacitor from the modulated B+ line to ground on the power supply side of the plate choke has to handle the peak value of the modulated B+ DC across it, There is essentially no DC current flowing through the capacitor, unless it is defective... and therefore no associated heating due to DC current passing through the series resistance of the capacitor. In this application, the RF voltage across the capacitor, and the associated RF current passing through it is very small. Therefore, there will be very little heating caused by the capacitor's RF dielectric losses or RF current passing through the capacitor's series resistance.

On the other hand, as Tom pointed out, a padding capacitor in parallel with the plate tuning capacitor will have a large amplitude (roughly equal to the value of the B+) RF voltage across it, and a correspondingly high RF current flowing through it. Since the padding capacitor is not ideal, the high RF voltage across it with cause high dielectric losses (large amounts of heating), and the high RF current flowing through the capacitor's series resistance will also cause large amounts of heating.

The situation regarding a plate blocking capacitor is in between the extremes of: a modulated B+ line bypass capacitor (full modulated B+ across it, very little RF voltage across it, and very little RF current flowing through it); and a padding capacitor placed in parallel with the output tuning capacitor (no modulated B+ across it, but very high RF voltage across it, equal in peak amplitude to the peak of the modulated B+; and an RF current passing through it that is some significant fraction of the total current circulating around the output tank circuit).

The plate blocking capacitor will have the modulated B+ (audio frequency modulated DC) across it; but it will have an RF voltage across it whose peak amplitude is much less than the peak value of the modulated B+; and the RF current passing through it will be approximately (1/Q) x the RF current circulating around the output tank circuit.

For the tuning capacitor padding application you need a padding capacitor specified to handle the peak RF voltage that will be across it (and the associated peak RF current that will be passing through it) at the highest frequency that it will be used. The RF peak voltage across the padding capacitor will be equal in amplitude to the peak value of the modulated B+; which is equal to: the B+ at carrier x (1 + the plate modulation index). The peak RF current passing through the padding capacitor will be a fraction of the total peak circulating RF tank current. The total peak circulating RF tank current will have an amplitude of (roughly) the average DC plate current (at carrier) x the Q of the output tank circuit x (1 + the plate modulation index). The fraction of this total current flowing through the padding capacitor will be: [(the padding capacitor's capacitance) / the total tuning capacitance, including the padding capacitor's capacitance, the "bread slicer" capacitor's capacitance, and any other padding capacitors' capacitances)]

Stu

Well, for C1, (plate tuning cap)  the very best method is to use a 500 pF @ 10 to 15 Kv  vacuum variable cap with counter, of course... ;)   In the end, this is sometimes the cheapest and best quality way to go. Worth the extra $.

As for fixed padding caps, the Russians sell a few lines of very good quality fixed mica caps for reasonable prices that are physically large and more apt to handle high currents. I've seen some 15 KV units the size of lemons or even oranges which should handle big RF currents.  They can be found on eBay as well as a few importing American dealers with websites.

The common 500 pF TV door knobs are rated at 30KV, but at currents I would guess to be less than an amp. I have used them for padding in loading circuits for antenna tuners, but even that is pushing it.  

For amplifiers, the Q * I  = total circulating current in a plate tank circuit is the problem. At a Q of 12, 1 amp of plate current becomes 12A of RF that is exchanged between C1, C2 and L1... yikes.

Once in a while big fixed caps show up at flea markets for cheap.  The common 1000 pf @ 5KV fixed tan mica caps should not be used in the plate tank or for plate coupling caps in rigs bigger than say 1000w pep.  I know it is commonly done, even in commercial 1500w pep linears, but it pays to stay conservative. For 1KW AM rigs, we need to go bigger in both voltage and current ratings. (size matters)

I'll bet the guys here can suggest some more high quality fixed RF caps that are available on the web.

T

At freqs. Up to 50mhz or so, ceramic capacitors with leads are perfectly acceptable provided sufficient number are put in parallel and leads are short.   Alpha for example uses a gang assembly of five  75pf/18kv ceramic caps for the 160 meter pad in their PI-L output .

The blocking cap in my dual 833C amp has two .001 uF/10 kV axial-lead disc ceramics in parallel. It was the 5 kV padder next to them that failed.

Thanks for the other helpful posts.  :)

Yep, the wisdom in this approach is that they are using only 75pF caps at 160M, which splits a relatively small RF current thru each individual padder cap. Very high voltage rating too.

T  

In this case, the impedance of each of the parallel 75pF capacitors, at the highest frequency in the 160m band, is:

-j / [ 2,000,000 Hz x 2pi x .000000000075F] = -j1061 ohms

If (for example) the B+ is 3000V, and if we assume that the amplitude of the RF voltage across each of the parallel tuning capacitors is equal to the value of the B+...

Then the amplitude of the RF current flowing through each of the 75pF capacitors is: 3000V/1061 ohms = 2.8 amperes. The rms value of the RF current flowing through each 75pF capacitor is 2 amperes.

Spreading the RF current among 5 padding capacitors in parallel (as well as the bread slicer tuning capacitor) is a very good thing to do... but keep in mind that each of the 5 parallel padding capacitors still has to handle 3000V peak RF across it (in this example), and 2.8A peak flowing through it (in this example).

Stu

Wandering off topic a bit......

Any quick way to test padding caps in ckt? My VK2 has a bad cap somewhere, and I don't want to shotgun the whole thing by replacing all the caps. Nor do I really want to unsolder everything, and add caps to find the bad one. If I can help it. And all I have are some chinese scope probes.....


klc