Marshall DSL 100 and the stupid cap that shorts

[Ronsonic]

In the Marshall DSL 100 there is a cute little cap named C46. It's a little 22pF 500V number that lives, inexplicably, between the screen and plate on one tube of the power amp. It has no reason to exist and knows it. So it dies and it dies horribly. It always dies short causing a weird glidget in the scoped waveform as one quarter of the amp behaves like a triode, or it shorts and burns and arcs and damages the board so that I have to drill a hole to decarburize it. Sometimes it waits until you have already performed necessary repairs and you are performing the final, assembled, testing before finishing the ticket. It usually, if you are lucky, takes out the HT fuse. And if you are really lucky it is obvious that it has blown up and will be seen in a visual inspection.

I hate C46.

There, I feel better now.

[Enzo]

So replace it with a 1000v disc instead of 500v.

It is there for stability, it prevents possible oscillation.

[Metro]

It's not the cause, it's a symptom

Ditto what Enzo said, (BTW, happy to hear you're on the mend! *doffs hat).
Also solder a bridge wire over the speaker out grounds. I think there is a consensus that the real cause is that the 16ohm switching socket gets corroded, goes open and causes c46 to die.

[Chuck H]

Right... Higher voltage cap.

I use a (cough) "conjuncive filter" (shunt filter) on one of my designs. The B+ is 350 on the plates so in the design phase I used a 600V cap I had on hand in series with a large resistor. All seemed fine so I thought I'd be safe with this rating, but nooooo. A digital meter can't read the spikes as fast as they happen. Not to mention that actual playing conditions are very different from a dummy load and signal generator. Three prototype amps survived initial testing and the 2009 winter NAMM show (good) only to short the cap later (bad). These amps and the half dozen I've built since now use 2kV caps for the task and no more problems.

[Ronsonic]

So replace it with a 1000v disc instead of 500v.

It is there for stability, it prevents possible oscillation.

Yeah, I know.

Just needed to vent because one just let loose on what I had thought was a completed repair.

Gotta wonder about a design that needs that one snubber on that one tube. Whassup wi' dat?

[J M Fahey]

Looks like the design wasn't still fully debugged,maybe the problem isn't even "electronic" but layout or grounding (that's why only one of 4 identical tubes needs it), they found a cheap 22pF cap "solves" it ... you know the end of the story.
Just idle speculation while I'm sipping my 3rd or 4th coffee cup of this day.

[Enzo]

It may have been a cheap shot at the problem, but it is only on one tube because that was all the thing needed. Putting them on all four tubes would have done nothing additional.

[J M Fahey]

Maybe you didn't get what I meant.
I didn't call cheap "putting it on one tube and not on the others" (which obviously is not needed) but "not spending some extra time rerouting a couple PCB tracks to make it unnecessary" (as is true on the other 3 tubes).
Why is just 1 tube unstable while the one by it, sharing the exact same driving signal and plate connection is not?
They both feed the same transformer winding.
Of course I agree with you that re-engineering the design is not the repairman's job.

[Enzo]

Nah, it isn't that one tube is unstable and the one next to it not, the SYSTEM is unstable, and sticking the cap somewhere stabilized it. It happens they put it on the end tube. COUld just as easily have stuck it on the next tube over. COuld they have designed the system better? Of course.