What's a good rating for a MOV suppressor?

[jmaf]

For OT flyback protection, are three series 1n4007's reverse biased frowned upon? I thought I was safe adding those, never used MOV's.

[trobbins]

As I see it, catch diodes are a good easy means to protect the OT. But they don't directly clamp the over-voltage transient that would be the cause of an insulation failure, as they clamp the voltage of the 'other half' of the primary winding in a PP stage, and so rely on the coupling performance between primary half-windings.

All methods will have their pro's and con's - I don't think its a big deal in the long run. I guess the main point is to at least include some protection if you have the chance.

[jmaf]

I always thought of the three diodes as a single use MOV. You can buy 100 1n4007's for a few bucks, thus my question if they're ok because I'm solder-happy joining them together and modding amps to include these.

The way I see it, and I may be wrong, if any of the extremities of the primary develops > 3kv, the diodes short and it's a blown fuse, and you need diode substitution. I've seen Fender Hot Rods use this, I think all Trainwreck amps do(I got this trick off Ken Fischer's tips on a book) and I've done it on some experiments of mine, I've never seen them fail so far.

The center tap on the OT comes directly from the power supply reservoir, I figured no component on the PS would stand several kv.... Having each end of the winding clamped I thought I was safe, thus the question.

[trobbins]

A PP primary CT is heavily clamped to B+ (unless an in-line fuse fails), and so won't effectively move from that voltage level.

If secondary, or primary winding currents are abruptly changed (eg. speaker plug pulled out at high listening level, or one of the PP valves fails), then the dI/dt level effectively causes a high voltage across all transformer winding inductances. One half-side of the primary would be forced to go below 0V, and the other half-side would go higher than 2x(B+), as the CT is clamped at B+. The diodes clamp the negative going end to just under 0V, and the inductive energy in the transformer is dissipated through that clamped winding as it allows current in that winding to 'continue'.

[Steve Conner]

Yes, that's true for energy stored in the transformer's magnetising inductance, or a reactive speaker load. It is safely returned to the power supply, diodes conducting in the forward direction.

But for energy stored in the leakage inductance between half-primaries, the diodes get stressed in reverse, perform like the 3kV ghetto zeners and eventually (not instantly, IME) burn out. In guitar amp OTs that leakage inductance can be substantial, and when the amp is cranked up full and abused with a Big Muff, the spikes can be pretty substantial too.

You could use diodes and MOVs. Or TVS diodes. Or a RCD snubber with the capacitor voltage monitored and hooked up to a shutdown circuit, as I once did with a heavy duty IGBT switch where mistimed switching would generate very destructive spikes, and I couldn't find a MOV big enough to eat them.

None of this will prevent you from smoking your screen grids and/or screen resistors on an open circuit load. Bean-counted commercial designs can't bear the cost of anything more than the diodes, unless you were lucky enough to get one of RG's mythical amps.

[trobbins]

Steve, with an open-circuit load condition and high signal drive level, I think you're saying that one anode saturates for 50% of time, and hence the associated screen current can be quite high and have a relatively high rms value compared to normal screen dissipation condition? The screen current would increase with drive voltage, so I guess dissipation does depend fairly strongly on drive level. Obviously not a condition to maintain for too long a time :-)

[Steve Conner]

You got it!

In my experience, every valve amp I've tried has been happy without a load, at idle and moderate signal levels. When you get past what would be the clipping point with a normal load, things start to go wrong. When I tried it with EL34s, the screen grids lit up, and glowed brighter and brighter as the drive level increased, until... :-O

OK, I stopped the experiment to avoid trashing a functional pair of tubes.

On the basis of these results I once tried to make an "open speaker detection circuit" that shut the amp down on excessive screen current. The problem was, it was vulnerable to nuisance tripping, even though it was set greatly in excess of the rated screen current. This made me suspect that EL34s are often abused in practice, it seems to be normal in the guitar amp world to have the screens merrily glowing at twice the rated dissipation.

Maybe one day I'll revisit the idea with a current clamp instead of a trip, or even hook it up to a Vactrol compressor.

[trobbins]

Sounds like an rms sensing circuit with a reasonable time delay is needed, with as you say - an attenuator for the input level of some kind that just aims to slowly pull down the rms level to a survivable level.

Were you using an optocoupler across each screen resistor? I think the linear analog versions would be a good start - an interesting design exercise.

Ciao, Tim

[jmaf]

Sounds like an rms sensing circuit with a reasonable time delay is needed

Or a shorting output jack

[jmaf]

In my experience, every valve amp I've tried has been happy without a load, at idle and moderate signal levels. When you get past what would be the clipping point with a normal load, things start to go wrong. When I tried it with EL34s, the screen grids lit up, and glowed brighter and brighter as the drive level increased, until... :-O

OK, I stopped the experiment to avoid trashing a functional pair of tubes.

On the basis of these results I once tried to make an "open speaker detection circuit" that shut the amp down on excessive screen current. The problem was, it was vulnerable to nuisance tripping, even though it was set greatly in excess of the rated screen current. This made me suspect that EL34s are often abused in practice, it seems to be normal in the guitar amp world to have the screens merrily glowing at twice the rated dissipation.

Maybe one day I'll revisit the idea with a current clamp instead of a trip, or even hook it up to a Vactrol compressor.

The impedance on the OT primary is higher without a load connected, right? So what you observed, in your conclusion, was a consequence of what? The plate was swinging lower than the screen, so the screen attracted more electrons? How would it swing lower than in loaded functioning, if when unloaded the OT has higher reactance? Or was the tube switching action causing a flyback voltage to return through the screen due to the secondary being open? Maybe obvious questions, I'd just like to try and understand better.

Directed beam tubes don't seem to abuse their screens as much, maybe this is obvious, but the EL34 and EL84 seem to require more attention to their screens in most amps I've seen, the 1k resistor cookbook recipe doesn't always cut it.

[Steve Conner]

Yes, it's exactly the too-high load impedance that causes all the current that ought to go to the plate to hit the screen instead.

Think of some time you went to pick some object up and thought it was going to be a lot heavier than it was. It goes shooting up in the air and if it was something really big, you might even fall on your ass. Or you're pedalling a bike, the chain snaps or comes loose, and you prang your nads on the handlebars.

The same thing happens, the lack of a load on the OT secondary makes the primary taps very "easy to move" with respect to the B+ centre tap. As soon as a tube pulls on one end (down, not up, if you subscribe to the positive-is-up school of schematics) then it collapses to ground and the tube does the electronic equivalent of pranging its nads on the handlebars.

Flyback pulses from stored energy are a separate issue to this.

It's not an issue below clipping because most tube amps have negative feedback. This responds to the abnormally easy load by reducing drive to the grids. But once you're into clipping the NFB loop is broken.

[jmaf]

Couldn't get any clearer after your explanation. We've sorta accidentally hijacked the thread so I'll leave some other lingering questions about this for a future chat.

[NateS]

Just a point of clarification - wouldn't the power switch eat into your MOV each and every time the amp is switched off? Wouldn't 150V be cutting it kinda close for a 120V primary? I was contemplating using a gas discharge tube @ 150V paralleled with a 200V MOV. One of the fun bits about building stuff - you can do stuff for the price of a hamburger the bean counters would never allow in a million years.

[R.G.]

Not every time, but yes, sometimes. However, the number of switchoffs that punch a lot of power into the MOV isn't really all that large compared to what the MOV can take.

The power switch cycle isn't a big number compared to transients on the power line.

MOVs are rated for the nominal RMS power line they go on. They used to use 130Vac MOVs for 120Vac lines. That was a bit close. That's why the recommendation became to use 150Vac MOVs on 120Vac power lines.

There is a problem with not cutting it close enough. If the breakover point is too high, it saves the MOV, but exposes the protected equipment to what the MOV is there to protect against. They're in direct conflict, as so many design issues are.

[NateS]

Not every time, but yes, sometimes. However, the number of switchoffs that punch a lot of power into the MOV isn't really all that large compared to what the MOV can take.

The power switch cycle isn't a big number compared to transients on the power line.

Wow - I really had no idea the line was THAT noisy. But that's a big honking inductor, and 1/3-1/2A primary current at idle for a 50W tube amp by my figuring (using 4.5A@6.3V + 20w idle bias), and a really really short dt. Divide by almost zero gives almost infinity. I would expect that each and every time you cut it off, you'd get a big spike. I believe you, maybe L or dt arent as extreme as I thought.