The AM Forum

Help please on repair of a QRO Technologies 2500DX AMP

Happy New Year.


Several weeks ago via the forum we received some good advice on the repair of a QRO 2500DX
AMP. A fellow ham asked for assistance to repair the HV cards, the cap bank and the diode bank card. He was operating the amp and working properly at rated power output. Then while in standby mode, with no RF excitation, a loud pop was heard, followed by smoke and fumes coming out of the HV compartment of the AMP. Several of the caps had blown seals, ruptured, and 2 of the 16 diodes were blown and shorted. Since then, we have replaced all the caps with new 105 degree C caps, put in  proper values and replaced the 2 bad diodes. Subsequently, the AMP was turned on and HV was present, 3300V (nominal is 2600V), and at that point we assumed the meter shunt, voltage divider and meter multiplication may be in error. We proceeded to allow the amp to time out, apply filament after 2 minute time out and loaded the amp to 800 W with 15 W drive. However, the plate tune and load controls seem to have NO affect. While this was in process, he heard some popping sounds, like HV arcing and shut the amp down. Now at this point, there appears to be a short somewhere taking out the 240 V lines fuses. We have been using 15 A instead of the 20 A fuses with the notion of not applying any RF excitation and trying to track down the new problem. We know the HV supply worked, there is no sign of arcing on the diode card or cap bank card, or anywhere on the chassis. The cards look perfect. All the diodes on the diode card check out perfect. If the HV xmfr secondary, 1900V AC is removed from the HV card, the fuses do not blow. Our next effort will be to remove HV from the RF cage, or simply remove plate voltage. The assumption, there is something going on in the RF deck causing the short. We have checked the plate blocking cap cold, that is no HV, and it is fine. Any other ideas and thoughts would be appreciated. We note, the HV surge protection in the form of some R in series with the HV xmfr secondary is not used in this amp. Instead an AC zero crossing detector relay in the primary is used to control in rush current and provide for a soft start for HV. If those relays are bad, is that possible, in rush current could be an issue. At this point our next step again is to remove HV from the tube plate and see if the fuses blow again. We may add discrete R's to the primary of the xmfr under the assumption the soft start relays have gone bad. Comments? In advance, THANKS! Again, the manual and schematics are on the web at:

http://www.qrotec.com/Documents/HF2500man.PDF

http://qrotec.com/Documents/HF2500circuits.PDF

Alan

making power out but tune and load controls not effective sounds like vhf parasitic oscillation...

to fix, you need to get control of this condition and reduce its intensity so that it is non-destructive ... good discussions of this n the handbooks .... basically make high voltage supply more 'rubbery' to reduce input .... this is harder when you have common fil/hv power xfmr

Thanks Beefus, good thought, certainly possible, and your response was echoed by another fellow. So its added to the investigate list. I'll let you know.

Tnx, Alan

3.3kv is pretty hot for  standard 3cx800.  Not so bad for the cpx version,  but the standard may be breaking down being operated 50 pct over its ratings......


That is,  assuming  shunt and metering are correct.

That said,  amp supply and qro both had cpx versions of their amps that ran on 3kv plus.   Just a hunch,  check your tube types.

(added)   I had a problem in my sb220 of a crackling sound with hv on.   Ssb (higher voltage)  louder than the lower tap for CW.   KM1H,  Carl,  sent me a few bypass caps and I replaced the one on the power supply side of the plate choke.   No more odd sounds.  That wouldn't explain yours acting like an oscillator though.

Check cathode voltage for proper bias?

--Shane
KD6VXI

Thanks Shane for your reply and taking time. I agree on the HV value concerns. What I cannot understand is why the HV meter on the QRO would read 3.3 kV vs. 2.6 kV unless the meter shunt or voltage divider (multiplication) network has changed value slightly.  We re built the cards with exact replacement devices. I was going to use my Triplett meter which has a 6 kV capability until I noted its accuracy on the 6 kV scale is +/- 3.3 % of FS. I read my 1200 V supply on an amp I am building, it reads 1500 V. I need to get an accurate HV DVM. In any case, there are some good trouble shooting ideas that have surfaced and will begin the hunt this week/weekend. 

Tnx, Alan

Alan, 

The schematic you linked to is rather hard to read since the text seems to be blurred on my little laptop screen. What I do see though is they key the final tube screen voltage, and cathode return circuit. i also see that one side of the final tube filament connects to chassis ground...

You stated that all hell broke loose when the amplifier was in standby. This means the screens are floating, as are the cathodes. So what would happen if one of the RF tubes developed a heater to cathode short? This would ground the cathodes when the screen grids are floating....it just might take off into a parasitic even though the amplifier is not keyed. Not being keyed, the input and output of the RF amplifier are without normal impedance loads, hence the tendency to sing somewhere at VHF.

Looking at the schematic, it might be good to provide a screen grid return path, maybe a 100K 5 W resistor to ground. This way, in the scenario described, given a heater to cathode short, the screen grids stay at ground, and no parasitic will occur.

In a similar scenario, I lost two 8877 from parasitics while in standby. This was in a 8877 converted amplifier I have. Just because Alpha on the Alpha 76 ground one side of the 8877 filament does not mean that is good engineering practice. My 3rd 8877 also had a heater cathode short, but did not oscillate. I just noticed the fans come on while in standby. I still use that tube, but now the filament is floating along with the cathode.

In the case here with the QRO amplifier, I wager the problem is an intermittent RF PA tube heater to cathode short along with being in standby when the screen grids are floating. A screen grid to ground resistor (maybe a 100K 5W) might prevent the next occurrence.

Jim
Wd5JKO

Alan,

Along with all the other suggestions, I pulled up the specification on the 4CX800's.  I believe that is the tubes this model amplifier uses.  The maximum plate voltage is 2500 and typical plate voltage is 2200.  So if your voltages are really up as high as your reading I would be concerned.  The question becomes, how did the voltage get that high...Measure it with a meter with known accuracy. 

When the amplifier just sitting there the HV will float to the highest level.  This may give you a clue on what is breaking down under unloaded conditions on the supply.   

http://qro-parts.com/images/GU-74B/4cx800a.pdf

Saying that, the amplifier obviously worked as designed so the current design should work again.  As far a inrush goes, it worked as intended and never blew the fuses so even though it may not be as good as it could be, it worked without blowing fuses.  I usually ultimately modify most of my amplifiers for better inrush control but that is just for more margin.   

In my lab I am currently working on my Ten Tec Titan and that uses the hard to find and expensive 3CX800's.  RF parts has some Taylor 3CX800's but I have no clue on how good they are.  I believe my 3CX800A7's which are Eimac's are still good...!!  Be glad your using 4CX800's!

73,
Joe-W3GMS     

Thanks Joe and Jim. More ideas and troubleshooting tips to consider. Jim, yes sir, the system schematic is a challenge to read! I have starred at it for a 1/2 day under a magnifying lens and with my other hand re drew it on a 11x17 page of paper portions that I am interested in understanding so I can read it. Fortunate, the main sub circuits and artwork of the cards are larger. Yes, I need to put to rest the real HV value, just to be sure. Going to find a HV DVM. Again, however, I can't believe an exact rebuild of the full wave rectifier card and and the cap bank is going to go above the 1.414 x Vac( 1900V) for the xmfr in this AMP. Our first point of troubleshooting upon return to this work will be to remove the tubes or at least remove the HV from the plate choke. And see if the FUSES still get taken out. If they do not and the AMP goes through its 2 minute timer, lights the filament and screen V as well HV is present, then I would be tempted to say, we an RF circuit problem, be it tube, plate blocking cap, etc...

It is interesting to note, QRO, designed the PI match to transform down to 200 ohms, not 50 and used a 4:1 xmfr to complete the match. The 4:1 double duties as the RF safety choke.

Thanks for the additional ideas again, Jim and Joe. Particularly the mods that may prevent future failures on an expensive tube!

To narrow in on your fuse blowing issue, pull both tubes.   They should really help you narrow things down. 

I would be remiss if I did not say, be very careful around the HV.  Make sure everything is off and everything is discharged prior to poking around!!   

73,
Joe-W3GMS

  Yup,

    I agree with Joe, Pull the tubes and bring it up to see
what you have.  Verify bias voltage, he should be biased off
pretty hard in stby.  I would measure the "E" across the bottom
filter cap. I bet it's >450V.    Manual says 240 VAC input.. HA Ha,
your prolly hitting it with 250 +...  Find a way to lower the screen
voltage or at least check it.. I bet its high too.

   Zener diodes don't check reliably with a DMM Diode test.
A simple curent limited supply will allow you to verify they are
doing there job.. Measure across each one for expected drop.


GL and be careful

73

/Dan



   

Does the PS have any bleeder resistors or just some balancing resistors across each 450V cap??


I just looked at the schematic,  seems there is just balancing resistors, which also work as bleeders.  I would check the resistors and make certain that there is a complete circuit to ground through the bleeders.

Seems this amp works on either 200vac or 240vac input.  Any chance the main plate xfmr is connected for 200vac input??  Maybe some connections may have been moved or reconnected incorrectly during some repair work.

Fred

Yes, there are balance-equalizer resistors across each of the HV caps in the cap bank, they return to B- and I believe, although it is not obvious, to chassis ground. The plate meter clearly shows bleed off and I have my trusty chicken stick equipped with a 20 K poker R on the end of a 4 foot pole :)

Not until the Ep goes to what appears to be ZERO, do I hit the HV cap bank with my chicken stick and lay her there for about 5-10 seconds. Then I go in. I hate getting shocked and knocked on my !@#$. The amp is set and wired via the selection block shown in the schematic for 240V. I thought the same thing at first when the plate meter displayed the 3.3kV value.

In the manual is a point of interest described in the section on PLATE VOLTAGE. There is a resistor R209, which I cannot find anywhere in the schematic or artwork, that is a protection resistor connected in series with the plate voltage. If a short develops, the resistor prevents B- from rising to B+ potential. This action protects the B+ circuit components as well the 4CX800 tetrodes. Again, I can't find this R, perhaps it is mislabeled in the manual writings. 

Build a 220vac primary "shock-absorber" using a pair of ceramic light fixtures
with two 100 watt lightbulbs ( incandescent ).  Each bulb is in series with one leg of the 220vac going to the amp.
If there is an intermittent short, the bulb(s) will brighten, but the fuses won't blow.

As always ,,, keep your hands out of the fire.


Don W4DNR

    I am still wondering about this circuit where it appears that a relay switches the amplifier screen voltage to the finals, and at the same time another contact connects the cathodes of the RF PA tubes to circuit return. The schematic is blurry, but it appears to be a dual pole SPST relay doing the switching.

    So with one side of the filament grounded, what happens if one or both RF tubes get a heater to cathode short/leakage when the screen grids (G2) are floating? Remember that in Standby the filament voltage is often high..at least this is the case in many big amplifiers.

Jim
Wd5JKO

Thanks for the additional tips and questions Jim and Dan. At this point we clearly need to separate HV and perhaps screen V, cathode etc...from the supply side of the unit. I am pulling together a HV DVM with 2% or better accuracy and proceed with measurements on the unloaded supply side of the unit. Need to know exactly what are the current supply readings now that the HV card (diode stack and CAP bank) has been "repaired". 

     This approach works so long as the light bulb wattage is sized to the load. You want the light(s) to be dim when everything is normal, and bright during a fault. With a QRO (high power) amplifier, that might mean the bulb needs to be high wattage. Perhaps a 600 watt movie projector bulb shielded so it won't blind you should it come on full brightness. Even here, the approach might only be viable to the amplifier in standby, or when keyed with no drive. You surely don't want the 4CX800 filaments running too low.

Jim
Wd5JKO

Yes, you do have to size the incandescent bulbs to fit the rig.

I have a plywood board with nine 750 watt bulbs to use on three phase broadcast transmitters. 

At these power levels and tripping 100 amp breakers, I'm looking for a shorted plate transformer or a faulty cable in the conduit.

On my tube ham rigs, a single 100 watt bulb seems to allow a slow ramp up of the voltages... at least slowly enough to see or hear an arc or to measure different voltages without taking out a breaker or fuse. 

You need to understand what the bulb is suggesting.... A full bright may mean a shorted transformer or shorted diode stack ... a repetitive bright bulb flash could mean a HV arc ... discharging a capacitor and then cycling again.   Those are usually easy to spot or hear.    I've used a small wireless lapel microphone taped to components to isolate the sounds.     Again, just be safe with the HV interlocks defeated.



Don W4DNR

Its pretty easy to build an auxiliary DC voltage divider for measuring the HV:

Place five (5) 1 megohm, 1%, 1W* resistors in series... and then add an additional 5kohm 1/2W resistor in series with the five 1 megohm resistors.

The free end of the 5k ohm resistor must be connected to ground.

Using an ohm meter, verify that the resistance between the non-grounded end of the 5k ohm resistor, and chassis ground, is 5000 ohms. Verify that the resistance from the free end of the 1st 1 megohm resistor and ground is 5 megohms

The free end of the 1st 1 megohm resistor, should then be  connected to the HV point.

Verify, once again,  that the resistance between the non-grounded end of the 5kohm resistor and chassis ground is 5kohms.

Use this voltage divider, in conjunction with a plain vanilla multimeter to measure the voltage between chassis ground and the non-grounded end of the 5kohm resistor. This voltage will be 5000ohms/5,005,000 ohms x the high voltage being measured. I.e. 1/1000 x the high voltage

You can calibrate this voltage divider by using it to measure a DC voltage that you believe you accurately know the value of. I.e. if you use this voltage divider to measure the voltage of a voltage source whose value is known to be 153VDC, and the voltage you measure across the 5kohm resistor is 160mV, then you know that this voltage divider produces an output that is 0.160V/153V x the input voltage.

*Since the total resistance of the five 1 megohm resistors, in series, is 5 megohms, the total power dissipated by these five resistors, plus the additional 5kohm series resistor... when measuring a DC supply whose value is 3500 volts will be 3500 x 3500 /5,005,000 ohms = (approximately) 2.5 watts. Almost all of this power dissipation will be distributed, equally, among the five 1 megohm resistors. I.e. 0.5W dissipated in each of the five 1 megohm resistors.

Therefore, each of the five 1 megohm resistors should be rated (for safety) at 1 watt (or more).

Also, each of the five 1 megohm resistors will have 3500V/5 = 700V across it. For safety, do not substitute a single 5 megohm resistor for the five 1 megohm resistors in series.

Stu

Hi Stu. Yes, agree and I was going down that path. :) Then a colleague caught me by the collar and said here, use this, bring it back when your done.  ;D

that fluke hv probe is calibrated for 11meg input Z dmms .... a vom is usually 20k unless otherwise noted

Actually 10 Meg, but you can use other values of Zin, either add a shunt if possible or calculate the required correction for Zin values not equal to 10 Meg. It is basically a 1000:1 voltage divider, with it's Zin I believe at 1000 Meg.

Update...Repair of a QRO Technologies 2500DX AMP

Again, TNX for advice on the troubleshooting of this amplifier. :)

At this point we know what the failure is NOT, but not what IT IS!

Fuses 250V, 20A still blow, either with a healthy SNAP or a sometimes a repetitive SNAP-SNAP. WE NEVER SEE ANY ARC OR FLASH OVER!!
 
Tubes are pulled, HV removed from the plate choke so the RF side is totally out of the picture.
Measurement of the now unloaded HV showed about 1 kV, before the 20 A fuse goes. This is an
estimate, cause the event which takes the fuse out is pretty fast and the meter response time lags. Not an Oscope!

Ok, that said as a review, we had good HV (well it was HIGH at 3.3 kV?) from the get go after repair of the CAP bank card and the diode bank card. All the CAPS were replaced, 2 diodes were bad and replaced and that is it. We did not change out all the other diodes, however they measured fine out of circuit. We did not check out or measure all the Resistors on the cap bank card or the diode bank card. Is it possible that a HV (450V) cap is breaking down cause of un equal voltage division... hard to imagine, were at ~1 kV and we have 8 series caps.

My sense is we either have an in rush current problem (bad in rush control relay) or a dead short that is not simply on the HV card as the resistance to gnd on the HV card is not a short. That is there maybe a HIPOT short which develops and the LIGHT BULB idea in the primary may be the next best trick to find this short. Any other ideas? Again, the power xmfr is fine. Also note, WE SEE NO ARCING WHATSOEVER when the bang occurs and the fuse blows.

HV Capacitor break down? Diode(s) going short circuit, INRUSH current event??

Tnx, Alan

You are still getting "snaps" after disconnecting the RF section ?

I once found a 30S-1 doing something similar.... snap sounds but no flashes.

It turned out to be an arc INSIDE a ceramic HV stand-off insulator.

The insulator had a threaded hole at either end and the arc was from the ground screw and the screw that was holding the B+ to the insulator.

Don W4DNR

Hi Don and thanks for the reply. I understand your idea. However, the tubes are pulled, the HV from the HV B+ of the card (this is the diode bank card) connection to the RF plate choke is also removed. So this takes tube failure out of the picture, takes plate blocking cap and all of the PINET and safety choke etc... out of the picture. The problem focuses us to the three cards on the right side of the chassis, namely the HV CAP BANK, the DIODE BANK and the screen supply PCB. There is also the meter switch card as well maybe the RF I/O switch. Finally, The in rush relay (AC zero crossing detector) which feeds the primary of the HV XMFR. The XMFR is off the hook, since if the VAC (1900V) secondary is pulled from the HV diode PCB, no fuses blow. AGAIN, WE SEE NO ARCS, WE SEE NO FLASH OVERS. Something is forcing a 20 A fuse to blow with no apparent visible sign. Again, HV CAPS and HV DIODEs look pristine. I might mention that the screen supply CAPS have not been replaced, however, they look like new.

I guess what I am wrestling with is what on the diode bank card or the cap bank card could force a 20 A fuse to blow without letting the smoke out? I like the in rush current theory and maybe I need to come up with a simple experiment to check it out, like stick a nice size set of series R's in series with the HV diode bank card fed by the secondary of the HV xmfr. Easier said then done, just some effort involved but possible if you look at the schematics.

P.S. Although the schematic of the diode bank card shows transient caps (noise suppression) in // with all the diodes, .01uf, there are NONE on this card? Looks like they were never used or placed.

Tnx, for the BANDWIDTH and rambling.

Does the HV go through any shielded cable ?
IF not, can you put a piece of plastic between the hv cable and the chassis where it touches the chassis ?

Cut up pieces of milk jug or coke bottles for insulation.

Don W4DNR

An inrush problem would blow the fuse when you first turn on the HV supply.  Is that what is happening or do you get the bang after the supply is on a few or more seconds??

Fred

Hello Fred. I would say instantly. And when first turned on one bang and the fuse blows. The first time today while the HV is being measured and the HV removed from feeding to the plate choke, the FUSE blew after say 500 ms. This was long enough for both of us to watch the front panel HV meter as well the Fluke DVM with a 1000:1 divider and see the HV hit about 1 kV. After replacing the fuse, pulling both tubes, a bang-bang occurs and for that case, hard to get any meter reading. But... direct to your question, for the most part instantly. Recall, when first repaired the HV it sat idle for minutes at 3.3 kV and was tuned up to the tune of 800 W. Then, the big bang and/or bang-bang stuff started occurring and now continues.

Sorry, I missed Don's comment on the HV via a shielded cable. Interesting, NO.

The HV wire(s) in the QRO, might be HV wire rated, not sure, but slipped over ANY HV wire, QRO slips a clear plastic tube, does not feel like teflon, and does not feel like heat shrink. However, more like fat spaghetti tubing. We were careful in their dressing, must confess to get these boards out and to do rework we had to cut very carefully the tie wraps for the multiple wire harness and nylon clamps removed then replaced but not all bolted back down to the common wall between RF section and HV supply section. I can't imagine aggravating any HV cross talk among those wires in the harness, as they were all initially in very close proximity of one another.

A new input from an outside source says... CHANGE ALL THE DIODES!!

Hard to believe but certainly possible after a decade or so, multiple diodes could be breaking down, despite having 10 diodes in each of 2 strings in a full wave bridge. Gee... is that possible? 1N5408's are cheap.

And perhaps not all 1N5408's are created equally. The ones I used in replacement, not sure what their country of origin or manufacture was?? I did notice upon inserting them, that their lead wire diameter was less than what was present for the others. Do 1N5408's vary that much in quality and can they degrade w/ time? 

You are almost there ....

Elminated the RF deck

Elimintated the Transformer

I'd eliminate the capacitor bank next.  One could be internally breaking down.


Don W4DNR

I wouldn't worry too much about meter readings.  There is no way to make any accurate or meaningful readings in that short a time.  No meter reacts that fast to give an accurate reading.  I would take a close look at the HV wiring and all connection points along the HV circuit.  Could be a HV wire is too close to something and it may be arcing, maybe in a harness of wires.  Also, ceramic insulators passing through the metal chassis are another common problem (someone already mentioned this).  They arc inside and you can't see that happen.

You should try disconnecting all load wires right at the HV capacitor output.  If it still blows the fuse, then there is something breaking down is the power supply.  If the 20 amp fuse also fuses other power supplies or filament xfmrs, you may have problem in one of those circuits.

Fred

1N5408s are pretty good diodes,  I use them, never had one fail and I've made a lot of diode strings with them.  I always use a 470K resistor and a .01ufd 1KV ceramic cap across each diode.  Many will say you don't need these added parts but I use them anyway.

Replacing all the diodes might be a last resort if you can't find anything else.  I can tell you that once one shorts or breaks down it stays shorted, they don't recover.  A good diode will read open in one direction and read around 900 ohms in the other direction.  You must use a digital meter on the 2k range to make these resistance checks.

Fred

I agree on the diodes and I meant to say 5 not 10 in each leg. It is a full wave bridge, so their are 5 series diodes in each of 4 total legs. Anyway PIV is then 5 kV, which at 2.6 kV forward, seems a little light. I would want a bit more headroom. Also, after talking to the fellow ham I am helping on this repair, he indicated we have been using 20 A fast blow, while the manual states, 20 A slow blow. Again, I would like to believe that in the ideal sense, should not matter. However.... just maybe!

One other clue, when first turned on the unit gave out a single snap or bang. The unit had been idle for week, cap bank fully discharged. Then, the second time turned on, we do NOT get a single bang, but a set of multiple bangs. The difference, the cap bank was partially charged. This would have had an effect in the magnitude of the inrush. Just thinking out loud...

If you have 5 diodes in each of the four legs of the FWB circuit, that is 10 diodes in series.  With that high a voltage (2-3KV) supply, five diodes in series would short.

Fred

If the caps are partially charge, that would lessen the inrush current.  The fuse should be a slow blow, quick acting fuses could blow just from the sudden inrush current.  But, this doesn't explain the banging noise you're hearing.

One thing to consider are the filter caps,  never liked strings of electrolytics, I only use HV oil caps.  Electrolytics, even though exactly the same, don't always have the same internal resistance or charging rate.  The equalizing resistors, if are too high of a resistance, will not divide the voltage equally across each cap.  Each resistor is in parallel with differing resistance of the caps, so the voltage does not divide equally.  The lower the resistance of the equalizing resistors. the better the voltage division occurs at any instance in time.

Maybe the caps are breaking down,  They probably do recover from an arc over internally.

Fred

Yes Fred, that is possible. On the PIV, the requirement for the bridge is just the magnitude of the secondary V, ~ 2 kV. The 5408 is rated at 1kV, so 5 puts us at 5 kV, about 2.5 x... I would personally like more and usually go 4 x. In any case, not sure I understand the comment on the diode short. Yes, the bang is really quite loud...i.e. when the fuse blows, why do I hear as well as FEEL vibration although slight from the chassis. I failed to mention that earlier, but my colleague ham who threw the switch, said he actually felt a slight vibration delivered to the ON/OFF rocker switch front face.

Quick comment on the diodes, with 5 diodes in each of the four legs of the FWB, you do end up with a string of 10 diodes. You have to think about it and look at the FWB.  On each half of a cycle there are 10 diodes conducting, they are just not next to each other in the series circuit of the xfmr winding, the load, and the diodes.

The vibration is not unusual, most likely the plate xfmr under strain of a heavy overload.

On the PIV,  remember the 2KV xfmr voltage is the RMS voltage.  The peak voltage is 1.414 times that number.  The diodes are 1KV PIV (peak inverse voltage).  You always design for a big margin of safety with the PS diodes, more is better.

Fred

Does this amp have a 2.5mH Choke to ground at the output of the PI net?

It could be the output coupling capacitor(s) are leaky or shorting.


Phil - AC0OB

Hi Fred, I understand your comments on the PIV, and I believe you are saying it is unlikely that the PIV is the problem, as 5 diodes would HAVE to short before an inverse breakdown would possibly occur.
So the bad diode(s) idea is unlikely.

As you and your friend have observed, a High Voltage arc is really a High Voltage SHORT.

3000 volts    Lets say the arc was a 10 ohm short.   I=E/R          3000v / 10r =  300 amps

Power =  Volts x Amps    3000volts x 300amps = 900,000 watts  ( however briefly )

Enough to blow a 30amp 220vac primary fuse   ( 6,600 watts )  and make the plate transformer emit a huge
pulse of energy into the chassis  ( observed as a slight vibration ).

At this point, I'd use the lightbulbs in the primary  or change the primary 220vac  breaker
to something like a 5 or 10 amps.

Reconnect the diodes ( if you cant read a dead short on each diode with your meter, they are good )

Disconnect the capacitors from the diode bridge but leave everything else connected.

Make sure the transformer is connected to the diodes correctly  ( just saying )

and turn it on again.     if it still arcs, you have ( at least ) eliminated the capacitors ...
and one source of the energy in the arc.


Don  W4DNR 

Understood Don, thanks. Will try this approach in addition go over to slow blow fuses. And finally, need to look into the value of the resistance the AC side of the xmfr is presented.  Had not checked this and would like to think this is a reasonable size resistance. This would involve both cards, CAP and DIODE card present.

Most likely the DC resistance of the plate xfmr primary is only a few ohms.  You're almost right about the diodes.  You have a total of 10 PIV for each half cycle and you don't want any of them to be shorted.  If it's not too difficult you can recheck them to be sure none are shorted.

Don made a good point, try disconnecting the caps from the diode bridge.  This leaves just the xfmr and the diode bridge, no caps or any other load.  If you don't blow the fuse the problem is the caps or a short in the HV circuit somewhere.

Fred

A cheap Hi-potter.


At this point, to save the  $$$ transformer, it might be worth a trip to the local neon sign shop and
beg for an older ( non solid-state )  neon transformer that they have taken out of service.   

A small beer commercial window logo transformer will do,.      small transformer .... not the beer.

I'd still run it through a variac and add a couple of microwave oven diodes in series with the output.

Hook the output of the diodes to the output of the amp's diode quad.

Another way would be to lift the transformer's leads and let the diode quad rectify the neon transformer's HV AC.

Connect the neon transformer case or ground wire to your amp's negative returm or the capacitor negative wire.



The neon transformer is self-limiting  ( low amperage ) , but it should start arcing at your problem area fairly steadily.    Makes a pretty mean "Jacob's Ladder"  to impress the grandkids too.

This may give you enough time to spot or hear your failure point,
 

Tnx, Fred, Don and others for additional thoughts!

We measured the HV xmfr secondary winding R, it's about 150 ohms. The HV cards total Rin (cold...no Vin) appears to be a high Z, no real shorts here. There is a second winding on the same HV xmfr and so we will measure the screen supply bias card just to be sure its Rin is also high, no shorts. Next step, it is easy to clip out the CAP bank from the DIODE bank. The tubes are pulled, the HV is off the plate choke, will fire this up and see if fuses go. And with that will report back. Hopefully will find its either the diode card OR the cap bank card breaking down.

Problem found.

First, thanks for all the advice from the forum.

Lesson learned, when replacing the CAPS in the cap bank for HV supplies, don't stop with just the caps. Check and replace if needed the equalization R's. Obvious after the fact. HALF OF THE 5W 51k flame proof metal film ? R's were open or in the several meg ohm range. The result of this event was that two of eight replaced caps went to dead shorts. The remaining 6 of 8 have high leakage! Either a result of the open R event, or all poor caps from the start.

I suspect the HV read while the unit was in operation is possibly traceable to the meter shunt and its value change. Of course double check of the R's in the voltage divider network of the supply will be done to be sure that is not the culprit for the slightly higher HV reading.

Alan

That was unusual...I guess resistors wear out too. What is the wattage rating of the failed resistors? Too much stress?

Fred

Your failure was similar to some of the Ameritron amplifiers of years back.  The equalizing/ bleeder resistors were seven watt WW @ 50 K.  I have an AL 1200 that I use regularly but not contesting.  It began to snap which indicated a HV problem so I removed the capacitor deck and checked the resistors after checking the diodes. 

One resistor was open and after removing the capacitors from the board, the one that was under the open resistor had spewed some of the internal paste. Ameritron changed the resistor supplier and, to my knowledge, have not had the problem since.

Had I continued to use the amplifier, I would have had several capacitors do the same, you did good in finding your problem and probably saved the transformer.