Traynor YGM, sparks flying!

[hylaphone]

Hello folks,
I've been servicing some iteration of the YGM-3 with the three inputs. I believe the attached schem is right.
It had some noise issues and needed filter caps, after addressing that the amp behaved/sounded fine.
However, I was seeing 430V on the EL84 plates and thought I'd try the amplified zener trick on the CT as shown here:
http://www.geofex.com/Article_Folder...osfetfolly.htm

After passing a bulb limiter test, I went to check voltages and a big spark jumped from the filter node to my meter probe as it approached, and ruined the meter.
Totally freaked out by this I removed the whole affair on the CT and reverted back to stock.
Tried with another meter and different probes, same thing happened.

What on earth could be going on?
Thank you --

[g1]

I was going to ask about the meter being set for current readings, but you tried a different meter, so don't think you would make that mistake twice.

[hylaphone]

I was going to ask about the meter being set for current readings, but you tried a different meter, so don't think you would make that mistake twice.

No, I'm very cautious about that.. just baffled

[g1]

My next guess would be that there is oscillation and the voltage there is much higher than what is showing as just DC.
So the caution about not probing plates when signal present (due to excess HV) also applies to oscillation.

[Chuck H]

My next guess would be that there is oscillation and the voltage there is much higher than what is showing as just DC.
So the caution about not probing plates when signal present (due to excess HV) also applies to oscillation.

This is a super intuitive assesment. Not entirely speculative, intuitive rather.

And hylaphone,.. I would have been freaked out and posting here on the matter as well. I hope they weren't expensive meters. From now on take HV readings from the OT CT node rather than the actual plates. The difference would be negligible and the OT impedance protects that node from spikes that may be caused by oscillation. And as noted B+ reading at the actual plate can cause oscillation so such readings can be less reliable than reading at the CT. Huge spark gaps not withstanding

In all my testing I've never seen an arc greater than maybe half a millimeter. There must be some really high voltage at some peak frequency there to make the arc you describe. Probably not enough current to electrocute you, but no promises and getting a bad deep burn is an eminent possibility. So still spooky at the moment.

[Helmholtz]

As has already been said, B+ is best measured at the OT center tap or the filter cap it connects to.
The filter cap shorts any AC or oscillation to ground.
Plate voltages should be only a few volts lower.
Also, oscillation can be avoided by pulling the PI tube.

When probing a power tube plate with the PI tube in, the resulting oscillation may cause peak voltages of 2kV and higher.
Such high voltage can kill the meter but also the OT.

[Chuck H]

Such high voltage can kill the meter but also the OT.

Right! Less a matter of current than voltage these spikes can punch through the enameled wire insulation.

But to be fair this potential oscillation circumstance is only recognized by the occurances here. At least where things stand. That isn't to say there could be safer/better means of disovery.

[hylaphone]

Thank you for the thoughts! Before I roll up my sleeves for another round of testing, here are a few more anecdotes:
The spark also occurs probing the OT CT/Reservoir node.
I haven't scoped the signal path for oscillation, in fear of damaging my scope..
There are 1ohm bias check resistors on the EL84s. I did not see any rise of current with the PI installed (which I assume would occur if it were oscillating this wildly)
Each side of the OT measures around 200 ohms cold, but of course that doesn't rule out windings shorting with voltage applied... I suppose I could tack in another OT as a test...

Still unsure what a safe and practical next step should be. Any ideas?

[Helmholtz]

Is the reservoir cap good?

What is the voltage limit of your meter?

Do you have a dummy load at the speaker output?

[hylaphone]

Is the reservoir cap good?

What is the voltage limit of your meter?

Do you have a dummy load at the speaker output?

Reservoir cap was new and tested fine, but I just replaced it again to be sure.
I switched over to a handheld meter, 1000VDC 750VAC
And yes, got a load on there.

[Chuck H]

I switched over to a handheld meter,..

Oh crap! Did you blow up two benchtop meters? That would be a true bummer.

[hylaphone]

Oh crap! Did you blow up two benchtop meters? That would be a true bummer.

No just one, and got that sorted - it cooked one of the divider resistors in the input

[hylaphone]

Wow, spoke too soon. Handheld meter just died during this procedure:
1. Hooked up scope across dummy load
2. Powered on with variac and power tubes only
3. Adjusted variac for 400VDC at reservoir node, with bias set to 60%. No sparks while checking.
4. Installed PI tube. No sign of oscillations across the dummy load.
5. Went to check voltage at reservoir node again, BOOM!

I drained the caps and went back in with my good HP meter to check DCR at all the ground points and did notice one thing, not sure if relevant. This standby switch lifts the PT CT, and reads about 1ohm to ground in the off position. All other ground points are around .1ohm, more or less the probe resistance...

I'm committed to solving this but really can't afford any more gear losses...
Any other wild ideas?

[Helmholtz]

The kind of oscillation I meant is induced by a meter probe touching a power tube plate.
Otherwise you won't see signs of oscillation at the speaker output.

The high voltage at one of the plates might have damaged your meter's overvoltage protection.
Consequently the meter might draw a significant current above some voltage threshiold.
You should test your meter by measuring a comparable B+ of some other amp.

[mozz]

The switch reads 1 ohm to ground or the CT reads 1 ohm to ground ? Easy enough to unsolder the switch and see if it's shorted to case or CT shorted to case?