Marshall DSL question - design flaw?

[AdmiralB]

A guy brought me a 2000 model DSL 100 last week - he'd been playing it and smelled smoke, so he shut it down and brought it over.

I opened it up and one of the screen resistors was burnt, cracked apart. After pulling the main PCB I noticed a burnt ceramic cap next to the screen resistor.

l looked up the schematic, and the cap is a 22pF/500V that bridges pins 3 and 4 on the last EL34 socket. It apparently burnt and shorted, and it took out the screen resistor too. Since the current is passing from screen to plate (and vice versa) the 1-ohm cathode resistors didn't blow. Nor did the HT fuse, although it was the correct value and type.

I own a DSL 50, and a few other DSLs have passed through my hands, but this 2000 model was the most recent (my 50 is a '97). This cap doesn't exist on my '97, so it must have been added on a later rev (they're up to rev 6 on the main PCB).

I just left the cap out, replaced the screen resistor, and it was good to go.

Fast forward to this morning - a friend brings over his new DSL 100 - bought at the GC Green Tag sale this weekend - to have me check the bias. I do, and fire it up. After a few minutes...I smell smoke.

I pull the chassis, and that cap is looking a little toasty. I fired the chassis up and let it sit, energized, but with no input. Everything's fine.

I plug in a guitar and hit a chord, and the lead on the plate side of the cap lights up. Mute the strings, it cools off. Lather, rinse, repeat.

I pulled the cap (what was left of it) and it's good.

What the hell is going on? Why would you shunt ultrasonics to the screen supply? And why only on one side of the OT?

And more importantly, why does it fail with apparent regularity under normal operating conditions?

[Enzo]

The 22pf cap is a common enough failure, but how did you determine it does it with regularity? In other words, how many of these amps did NOT burn up the cap? That would be most of them. Granted WHEN one of this series of amp fails, I do check that cap first off, but many amps come in for service and the cap is fine.

MAybe one side is all that is needed for stability. Power supplies have low impedances, and having the cap netween screen and plate means there is a lot less voltage across the cap compared to grounding the non-plate end. Plate already has close to 500 volts on it, add to that several hundred volts of signal , and possibly more for a glitch, and you'd need 2000 volt caps.

One question comes to mind, where the tubes OK after the repair, or did you replace the tubes? A screen to cathode or heater short would leave the little cap in the high voltage distress I just described. In other words, a screen the ground - or close to it - short would blow up the cap and take out the screen resistor. If the cap alone shorted, the screen resistor would be under not all that much stress.

If you get no parasitics without the cap, fine, leave it out. But I have certainly fixed my share of MArshalls needing a new screen resistor, both in the cap models and the non-cap models.

[AdmiralB]

It just seems rather coincidental that I fix one, and then a couple days later a brand new, out of the box amp does the same thing.

In both cases the tubes were fine - well, I guess I can't speak with 100% certainty in the first case as the tubes in the amp were brand new, but I don't know how long ago the damage might have started. For sure the second amp had only been out of the box for about six hours.

[Enzo]

Stuff happens. I had not seen nor heard about an Echoplex for several years a little while back, then in one week I had three of them sitting on the counter. And a Watkins Echo to boot. Coincidences can be spooky.

I think it important to keep in mind the difference between likely to fail, and likely to be what's wrong WHEN it fails.

[AdmiralB]

I think it important to keep in mind the difference between likely to fail, and likely to be what's wrong WHEN it fails.

I think that's the crux of my question - WHY does it fail? DC between plate and screens should be minimal. The only thing I can think of that would stress the cap would by flyback...but if that's the cause, then wouldn't these fail pretty reliably in any amp that gets cranked?

If my thoughts above are incorrect, then what would be wrong with these two amps that toast the cap but work fine without it?

[fixitrightthefirsttime]

Coincidence

Not to be philosophical, but coincidences are intriguing at least.

Recently I saw several new (warranty) Kustom Coupe 36's and 72's with serial numbers within a week or so of each other, and each had a somewhat different problem.

[Enzo]

Well I delivered my thoery - the screen shorting to ground or close to it. It is also possible we had an arc, but the socket never seems harmed, and after repair, I never see them again, so that seems to rule out arcs. The max DC across the cap should normally be roughly the amplitude of the plate signal, assuming they are pretty close at rest.

[AdmiralB]

Well I delivered my thoery - the screen shorting to ground or close to it. It is also possible we had an arc

If it were the screen, wouldn't it happen regardless of whether there was input signal?

The new amp worked fine until there was input AC present. Then the plate-side lead of the cap went orange (the screen resistors didn't even get warm).

And if the screen shorted, that would pretty much mean the end of the tube, would it not?

[Peter Clothier]

Admiral B.
If you had'nt posted this thread, then sooner or later, I or some one else would have.
A few years ago my TSL decide to give that little pop and a wisp of smoke. I thought to check the bias first and saw +25volts on the molex pins on one side???????????Argh.

Sure enough C46 22pF had obliterated itself. But, R76 1K screen res had risen to a perfect 100K with no physical sign of degradation. In fact I thought that maybe the resistor had been a duffer from new. No defects measured on the EL34 though?

I installed a new resistor and upped the cap to 1000wvdc. End of problem.

I have tried to work out the real purpose of this 22pF and have no definitive conclusion. Why is it not required on the other half of the PA? It is on 1 or 2 other Marshalls.
Has anyone had this trouble with a 50W jcm2000? Hmmme.
Is it just the usual Marshall component quality thing do you think?

When I get the chance I'll scope pin 4 and see if I can record any unusual transients.

BTW. Next time that you have your amp in bits checkout R48 (100K) on V3. If it's a pre 2005 model then you may notice that it is looking slightly discoloured. The current in this cathode line needs a 1/2W rated resistor!!!!!!
Also.... Junk the BR62 rectifier. These are ok in tranny amps but they dont like the huge amount of smoothing capacitance and the cold load of the 2 preamp valves (especially if you use Mullard ECC83's rather than the stock russian garbage Marshall favour)

Keep on rockin.

Peter.

[AdmiralB]

Actually, if I had been smart enough to do a search I would have noticed two older threads on this same issue. But they didn't arrive at any conclusions beyond "clip it out and it's OK" either.

[AdmiralB]

The more I think about it the more convinced I am the reason it's there is specific to that socket. Since the cap is tied directly to the pins, it's isolated from the screen supply by the screen resistor. If it was to globally (or at least hemispherically, since it is only on one side of the OT primary) swamp oscillations to the screen node, it should be outside of the screen resistor.

Likewise, it doesn't seem to be present on any of the 50W models - which only populate the center two sockets.

So I gotta conclude there's something iffy about the PCB layout around that socket.

[Enzo]

Possibly. My thought is that if something is shaking, you don't necessarily have to hold both sides of it to make it stop. In other words, damping one side of the circuit is enough to stanch any oscillations.

[AdmiralB]

But why grab it inside of the screen resistor? That way you're only grabbing one socket. If they had connected it from the screen supply to that side of the OT primary, it would better illustrate your point.

Of course, it could be that they DID intend to do that and just screwed up. It wouldn't be the first (or second, or third) time.

[fixitrightthefirsttime]

Schematic

OK, I gotta see the schematic.

Enzo, send me one?

This mystery now keeps me up at nights, and I gotta see it on paper.

[Enzo]

Thanks for considering the theory at least. It may be sufficient to damp off just one tube. As to inside versus outside, if you connect it outside the resistor, then your 22pf cap has a 1k resistor in series with it, making it less of a rolloff.

I can believe poor engineering choices, I can believe changes made with good intention that sap tone (Fender back a few years), I can believe wrong parts installed, I can believe I have to resilder heater rectifiers in so many Marshalls. I just replaced a 100k resistor that should have been 100 ohms in some store stock new item last week. (Fender Blues Junior) But I have a hard time believing the schematic capture software misinterpreted the circuit that bad that the layout was so wrong. And that NO ONE in the entire engineering and production team spotted it. The amp agrees with the drawings, after all.

The disagreement here seems to be whether the little cap would be more effective across the tube or from plate to ground. (through the powr supply)

SInce we are not sure just what problem the cap is there to solve in the first place and under what conditions, we are both guessing.