Mono gone pop (again)

[Ratcatcher]

Starting to get seriously pissed off with valve amps, especially mine :x

My amps are a sort of hybrid design with all rectification and regulation done with SS and on Friday the 900v mosfet popped amid some impressive sparks and smoke. Ive since replaced it but ive now got a periodic soft popping sound coming from the speaker. Im clutching at straws here and im quickly reaching the limit of my electronics skills but I suspect one or more of the caps on that supply have gone.

Does that sound feasible to the more knowledgeable folks here or should I look elsewhere? Which caps are more likely to go, pre or post mosfet? Please help!! :upgrade:

 

[Valvebloke]

Starting to get seriously pissed off with valve amps ... on Friday the 900v mosfet popped ...

When it works SS is brilliant. The instant anything goes wrong though it can very easily become toast, especially if high-voltages are involved. I'm afraid smoke means that something passed significant current for significant time. What it damaged will depend on the circuit topology. Do you have a link to the circuit, or at least a block-diagram description of it ? If the mosfet's failed with a short to its gate then it could have bled naughty voltages back into whatever's controlling it and that could be where your problem is. High voltage capacitors (electrolytics I assume) tend to fail catastrophically and blow fuses, particularly those more expensive three-legged silicon fuses. If this isn't happening then the HV caps wouldn't be the first place I looked.

VB

 

[Ratcatcher]

When it works SS is brilliant. VB

I suppose so but lately im thinking of dropping them from a great height! Heres a rough representation of the circuit. Im working my way through the pcb so will refine it later but this is the drawings sent from the manufacturer (out of date)

 

[i_should_coco]

Sadly, my experience is that solid state and high voltage don't mix. My friend had a PPP813 amp which used a high voltage regulator using a three-terminal reg. protected by a MOSFET. It never worked properly, and he got sick of replacing FETs and regs. every other time he switched it on.

I don't even bother trying these days. If you're going to do a reg. for a tube amps with highish voltages, best to use tubes as they are far more forgiving. Sadly, it will be hot heavy and inefficient, hence the temptation to use SS.

 

[Ratcatcher]

Sadly, my experience is that solid state and high voltage don't mix. My friend had a PPP813 amp which used a high voltage regulator using a three-terminal reg. protected by a MOSFET. It never worked properly, and he got sick of replacing FETs and regs. every other time he switched it on.I don't even bother trying these days. If you're going to do a reg. for a tube amps with highish voltages, best to use tubes as they are far more forgiving. Sadly, it will be hot heavy and inefficient, hence the temptation to use SS.

This is more or less what I was told by Mr Caswell. It seems that this is becoming a regular occurrence with my type of amps. Already made an offboard supply with the tentlabs, and looks like I may eventually have to expand it with an offboard HT supply as well....more bloody boxes. Thought id moved away from that when I ditched all my naim stuff.

 

[i_should_coco]

 

[Valvebloke]

OK, well at least the drive circuit for your IGBT is pretty simple (16 47V Zeners plus a couple of 1N4007s !). Whether the 1k resistor would have protected them against an internal short in the IGBT is still open to question though ...

If there's been an IGBT failure my first guess would be that one or more of the diodes in the voltage doubler (the FUF5408s) might have died. What test gear do you have ? Can you measure the DC voltage before the IGBT ? Can you check to see if it has any AC on it and, if so, how much ? Since you built this circuit I'm assuming you have sufficient expertise to manage the risks of a 900V power supply. These risks are pretty extreme. If you're not confident that you can work on it safely though then please, please don't risk it. A thing like that is more than capable of killing you. 40uF at 900V stores 16 joules whch puts it well into defibrillator territory. The thing about defibrillators is that they will switch your heart between the regular and the fibrillating (fatal) state. The problem is that they can and do make the switch in both directions.

VB

 

[Ratcatcher]

OK, well at least the drive circuit for your IGBT is pretty simple (16 47V Zeners plus a couple of 1N4007s !). Whether the 1k resistor would have protected them against an internal short in the IGBT is still open to question though ...If there's been an IGBT failure my first guess would be that one or more of the diodes in the voltage doubler (the FUF5408s) might have died. What test gear do you have ? Can you measure the DC voltage before the IGBT ? Can you check to see if it has any AC on it and, if so, how much ? Since you built this circuit I'm assuming you have sufficient expertise to manage the risks of a 900V power supply. These risks are pretty extreme. If you're not confident that you can work on it safely though then please, please don't risk it. A thing like that is more than capable of killing you. 40uF at 900V stores 16 joules whch puts it well into defibrillator territory. The thing about defibrillators is that they will switch your heart between the regular and the fibrillating (fatal) state. The problem is that they can and do make the switch in both directions.

VB

Cheers VB,

Although fairly close, the drawing is only a rough representation so ill try and post a better version up later. The 1k resistor on the gate is/was actually a 220ohm which ended up charred and there's also a 1ohm on the emitter leg which was none to healthy either! The fet in question is a IRFPG50 and not a IGBT. Ive been fighting for the correct drawings for a while but the dealer/manufacturer aren't playing ball.

I cant lay claim to building it either....way beyond my level! Test equipment is only a dvm, megger and a big hammer. (Ive just ordered a capacitance tester off eBay though) On the actual board, the FUF5408 you mention are actually two diodes in series on each leg. If your certain that it could be these, i'll just nip down to RS tomorrow and replace them. What would be your next best guess? May as well buy the components if im going down to Gatwick.

(No problem with high voltage....mess with the big stuff on my ship!)

PS trying to work out what actually caused this and was having a bit of a blond moment: what drives an amp more?

amp on 8ohm tapping driving a 4ohm speaker, or amp on 4ohm tapping driving same speaker?

 

[Valvebloke]

The 1k resistor on the gate is/was actually a 220ohm which ended up charred

Sounds like it did fail with a short-to-gate then.

I cant lay claim to building it either....way beyond my level! Test equipment is only a dvm, megger and a big hammer. (Ive just ordered a capacitance tester off eBay though).

OK, that does worry me more. I note what you say about working with high voltage on the ship. But making and interpreting live measurements on electronics requires a different skill/experience set. If you don't have experience with it then honestly I'd suggest that you find someone who does. The best scenario is if you can get a competent friendly person to stand alongside you while you and he sort the problem out. That way, assuming you want to, you can learn the approach needed to keep yourself alive in case you ever want to try this again. It's how most of us have picked it up. You can do quite a lot with a DVM up to several hudred volts. But 900V is right on the limit and there are certain fault modes that could even put a few more peak volts onto the existing components. The test leads will only be rated to 1000V and frankly if they've come from somewhere a bit cheap-and-cheerful I wouldn't trust them up there. I'd start by building myself a 10:1 divider that I had complete faith in and then I'd be doing all the measurements at 100V rather than 1000V.

On the actual board, the FUF5408 you mention are actually two diodes in series on each leg. If your certain that it could be these, I'll just nip down to RS tomorrow and replace them. What would be your next best guess? May as well buy the components if im going down to Gatwick.

I can't be certain at all. But if the FET has gone short then the diodes might have been stressed more than normal. They could still be fine though. A few hundred milliamps could have been enough to fry your 220R but could also have been well within the capacity of the diodes. That's why we need a few measurements. But now I'm back to saying that we really can't diagnose this fault over the internet. I'm afraid it does need a high voltage electronics engineer to get his hands on it.

trying to work out what actually caused this and was having a bit of a blond moment: what drives an amp more?amp on 8ohm tapping driving a 4ohm speaker, or amp on 4ohm tapping driving same speaker?

If you attach a lower-impedance speaker than the amp is nominally matched for (say you put a 4 ohm speaker onto an 8 ohm output transformer tapping) then for the same output power you'll have to draw more peak current from the HT rail (compensated by a smaller voltage swing across the transformer primary). I suppose it's possible that the increased peak current could have been enough to see off the MOSFET. It would depend on how far the amp was driven into Class AB though, which would depend on how loud you were playing and what the quiescent current through the valves was, and it would also depend on the smoothing cap's ability to handle the peak current demands. Frankly the design would have to be a bit marginal for a mis-matched speaker to tip it over the edge. I think I'd be more concerned about the switch-on surge that Coco-san mentioned. At switch-on the second capacitor looks more-or-less like a dead short to the MOSFET. Even if it's only that way for a few tens/hundreds of milliseconds it's still a hard load to cope with. Did the device fail at switch-on or when the amp was running normally ?

VB

 

[Ratcatcher]

Sounds like it did fail with a short-to-gate then.OK, that does worry me more. I note what you say about working with high voltage on the ship. But making and interpreting live measurements on electronics requires a different skill/experience set. If you don't have experience with it then honestly I'd suggest that you find someone who does.

Well, plenty of experience with big voltages but as you say, electronics is another ball game. I have the rudimentary skills and was hoping to be able to sort the amp out quickly. Especially as I have a couple of pairs of speakers on dem here. I agree, live diagnosis would probably reveal the problem quicker but I really don't have the tools here to do that and when I phoned around the repair guys, I was quoted 2 weeks minimum!

If you attach a lower-impedance speaker than the amp is nominally matched for (say you put a 4 ohm speaker onto an 8 ohm output transformer tapping) then for the same output power you'll have to draw more peak current from the HT rail (compensated by a smaller voltage swing across the transformer primary). I suppose it's possible that the increased peak current could have been enough to see off the MOSFET. It would depend on how far the amp was driven into Class AB though, which would depend on how loud you were playing and what the quiescent current through the valves was, and it would also depend on the smoothing cap's ability to handle the peak current demands. Frankly the design would have to be a bit marginal for a mis-matched speaker to tip it over the edge. I think I'd be more concerned about the switch-on surge that Coco-san mentioned. At switch-on the second capacitor looks more-or-less like a dead short to the MOSFET. Even if it's only that way for a few tens/hundreds of milliseconds it's still a hard load to cope with. Did the device fail at switch-on or when the amp was running normally ?

VB

As suggested by the dealer, id been running the speakers out of the 4 ohm tap, but really didn't like what I was hearing. I switched the amps off, changed the speaker outputs, turned the amps back on and......fizz, bang, pop.

I'd been running the speakers the previous day and evening no problem out of the 8ohm tap.

At switch-on the second capacitor looks more-or-less like a dead short to the MOSFET. Even if it's only that way for a few tens/hundreds of milliseconds it's still a hard load to cope with.

VB

Do you mean C125 on the drawing?

 

[Valvebloke]

As suggested by the dealer, id been running the speakers out of the 4 ohm tap, but really didn't like what I was hearing. I switched the amps off, changed the speaker outputs, turned the amps back on and......fizz, bang, pop.I'd been running the speakers the previous day and evening no problem out of the 8ohm tap.

If the bang happened before you turned the music on then I think the 8ohm/4ohm choice is unlikely to be the cause. The amp's output stage doesn't 'know' what load impedance it's trying to drive until the music starts. The fact that you ran it the previous day with the 4ohm sepaker on the 8ohm tap also suggests that the amp doesn't care much (which, frankly, is what I'd have hoped).

Do you mean C125 on the drawing?

No, I mean the 40uF after (i.e. to the right of, on the circuit diagram) the MOSFET. But looking in a bit more detail at the diagram I see that there is a slow-charge circuit fitted to the MOSFET's gate. C125 (10uF) charges through a 150k resistor so the time-constant for that is 1.5 seconds. That means that the FET's output voltage will only rise slowly. So it won't be trying to dump a huge current into the 40uF after all. The downside is that the voltage-doubler capacitor stack (the two 60uFs in series) will charge quickly. And with relatively little current being drawn, because the MOSFET output stays quite low for several seconds, the total off-load voltage across those two 60uFs may be pretty high. The MOSFET has to withstand that large voltage difference between its drain and its source and according to its datasheet it's only rated for 1000V there. That may not leave much headroom ... I think if it were mine I might consider replacing it with a more modern, higher-voltage device - something like the STW9N150.

But actually we've got away from your intial query, which was about the popping noise from the amp. We've been concentrating on the power supply but of course it may not be coming from there at all. It could be that the short through the MOSFET has over-volted some component in the main body of the amp. Have you tried taking the valves out of the 'bad' amp and trying them in the 'good' one just to eliminate them from the scenario ?

VB

 

[Ratcatcher]

If the bang happened before you turned the music on then I think the 8ohm/4ohm choice is unlikely to be the cause. The amp's output stage doesn't 'know' what load impedance it's trying to drive until the music starts. The fact that you ran it the previous day with the 4ohm sepaker on the 8ohm tap also suggests that the amp doesn't care much (which, frankly, is what I'd have hoped).

That's good to know, this is the first time this failure has happened. There have been others, but this fet hasn't blown before.

No, I mean the 40uF after (i.e. to the right of, on the circuit diagram) the MOSFET. But looking in a bit more detail at the diagram I see that there is a slow-charge circuit fitted to the MOSFET's gate. C125 (10uF) charges through a 150k resistor so the time-constant for that is 1.5 seconds. That means that the FET's output voltage will only rise slowly. So it won't be trying to dump a huge current into the 40uF after all. The downside is that the voltage-doubler capacitor stack (the two 60uFs in series) will charge quickly. And with relatively little current being drawn, because the MOSFET output stays quite low for several seconds, the total off-load voltage across those two 60uFs may be pretty high. The MOSFET has to withstand that large voltage difference between its drain and its source and according to its datasheet it's only rated for 1000V there. That may not leave much headroom ... I think if it were mine I might consider replacing it with a more modern, higher-voltage device - something like the STW9N150.

Before you spend too much time analysing that circuit, ive quickly scribbled down what I think is the actual circuit. Have to entertain the damn family today and would rather be messing with the amp. Let me know if theres any glaring errors and ill go back and check. As for suggesting improvements, then please do, i'm all for upgrades at this point in time. As ive moved the heater supplies offboard its left me with a spare 10v tapping of the mains tx, so ive started to consider putting a fan running off this supply but I don't know how successful (or quiet) that may be.

But actually we've got away from your intial query, which was about the popping noise from the amp. We've been concentrating on the power supply but of course it may not be coming from there at all. It could be that the short through the MOSFET has over-volted some component in the main body of the amp. Have you tried taking the valves out of the 'bad' amp and trying them in the 'good' one just to eliminate them from the scenario ?

VB

Yes, tried both the driver and output valves. No difference unfortunately.

 

[Valvebloke]

... ive quickly scribbled down what I think is the actual circuit ... Let me know if theres any glaring errors and ill go back and check.

I don't think the diodes and capacitors on the left hand side of the circuit are quite right. But the important thing is that it's a voltage doubler, and you do have all the parts you need to make one there.

Yes, tried both the driver and output valves. No difference unfortunately.

That's good news in a way. They would have been among the more expensive components so you don't really want them to have failed.

VB

 

[Ratcatcher]

. But the important thing is that it's a voltage doubler, and you do have all the parts you need to make one there.

Sorry VB, not sure I understand you here.

 

[Valvebloke]

A voltage doubler circuit uses two diodes and two capacitors to 'double' the voltage out of a transformer winding. There are two ways of achieving this:

The first circuit is the one the manufacturer used in post #3 above. The one you've drawn doesn't correspond exactly to either of them and it won't act as a voltage doubler I'm afraid. But it has got (essentially) two diodes and two capacitors (plus a couple of surge limiting resistors) so I'm sure that in reality the amp contains a voltage doubler. Your drawing's just gone a bit awry, that's all.

VB

 

[i_should_coco]

I assume the high voltage FET is simply a soft-start circuit?

 

[Valvebloke]

Is it also a voltage regulator (zener stack in the gate circuit) ?

VB

 

[i_should_coco]

Not sure. TBH, normally, you would sample the output and compare it to a reference, but I don't see that being done here unless I'm mistaken.

Edit: maybe some kind of crude stabilisation...

 

[Ratcatcher]

Ok, just spent another interesting (?) hour tracing the circuit and I think I have it nailed. Does this look right? Even if we, the wam collective that is, cant solve this problem, having a relatively correct circuit diagram can only help the repairer if I end up taking the amp to him. I did a bit more digging around the zenor stack and it looks like the feedback section gets involved as well as another feed from the mains tranny: