MOV, Gas Discharge tube, or TVS Diode protection for output transformer??

[SoulFetish]

I was hoping to get some help choosing the right components to protect against transient over-voltages across my output transformer primaries. I was thinking about using MOVs, but I'm unsure as to which are the important parameters I need to consider. What the benefits are of one devise vs another etc. I know there are those of you who have a lot of experience looking at transients in OT windings when connected to a reactive load, so I will gladly listen to what ever info you can share.

[Mike Sulzer]

I was hoping to get some help choosing the right components to protect against transient over-voltages across my output transformer primaries. I was thinking about using MOVs, but I'm unsure as to which are the important parameters I need to consider. What the benefits are of one devise vs another etc. I know there are those of you who have a lot of experience looking at transients in OT windings when connected to a reactive load, so I will gladly listen to what ever info you can share.

Does your planned level of protection include the case of an open secondary (no speaker plugged in)? In this case you would have to allow for dumping energy stored in the full transformer inductance rather than just the leakage inductance. Another thing to allow for is that you have to protect against repeating outrages, at the audio rate, rather than just a single transient, such as induced by a lightning strike.

[SoulFetish]

Does your planned level of protection include the case of an open secondary (no speaker plugged in)? In this case you would have to allow for dumping energy stored in the full transformer inductance rather than just the leakage inductance. Another thing to allow for is that you have to protect against repeating outrages, at the audio rate, rather than just a single transient, such as induced by a lightning strike.

You first question is a good one! I wasnt considering no load protection but even with a switching jack there is that potential in event of a cable plugged in only at one end. So i guess it should be considered. But initially I was really just considering repeat events during signal conditions. Wasnt worried about lightning going through my output transformer, unless of course im playing Howlin Wolf's Smokestack Lightning.
So, lets talk about figuring for the full inductance. Ugh, of course, I had to go over the top and spec the primary for -1dB at 20Hz-20kHz at 10k plate to plate

[nickb]

You might find this thread of interest http://music-electronics-forum.com/t40699/

The the bottom line, as I understood it, is MOVs will only you offer so much protection for so long. It's just a matter of choosing how long is enough for you. Personally I think a better approach would be to use small MOVs together with a means of detecting the abnormal voltage condition and shutting the drive off.

[J M Fahey]

Flyback diodes are very popular because they send spikes back to the power supply itself, which by definition can handle itīs own power.

They donīt dissipate anything , just resend voltage spikes back to the supply capacitors where that energy came from in the first place..

[SoulFetish]

You might find this thread of interest http://music-electronics-forum.com/t40699/

The the bottom line, as I understood it, is MOVs will only you offer so much protection for so long. It's just a matter of choosing how long is enough for you. Personally I think a better approach would be to use small MOVs together with a means of detecting the abnormal voltage condition and shutting the drive off.

YES! I find it of great interest in fact. Some nice light reading i see (Thiele-Small, you say?). but there are some good figures to see how you calculated power equations from inductances. I'll likely have a question or 2.

As for movs, it appears that the devices which have higher power ratings and seem better suited for this application, have high capacitances. would this be problematic across the primaries of the OT? Should I use these capacitance is in design an RC conjunctive filter? I would imagine if the performance of a mob degrades with each overvoltage , then capacitance values would not be stable either but I don't know

[Wombaticus]

Here's a relevant discussion about gas discharge tubes and MOVs over at AX84: AX84.com - The Cooperative Tube Guitar Amp Project

[Mike Sulzer]

Flyback diodes are very popular because they send spikes back to the power supply itself, which by definition can handle itīs own power.

They donīt dissipate anything , just resend voltage spikes back to the supply capacitors where that energy came from in the first place..

Juan, How do you connect it to do that? This is the situation that I see:

1.You have current flowing through the transformer and through the tube to ground. It points from the power supply to the plate.

2. The tube turns off, becoming a very high impedance.

3. The effective inductance of the transformer primary wants the current to keep flowing, so the voltage at the plate connection becomes very high, higher than the power supply voltage by a lot.

4. To prevent arcing, you put a diode across the transformer, pointing from the plate to the power supply.

5. This causes curent to flow around the loop made by the transformer primary and diode.

6. Power is dissipated in both the diode and the transformer, but in this case none makes it back to the supply

[R.G.]

I was hoping to get some help choosing the right components to protect against transient over-voltages across my output transformer primaries. I was thinking about using MOVs, but I'm unsure as to which are the important parameters I need to consider. What the benefits are of one devise vs another etc. I know there are those of you who have a lot of experience looking at transients in OT windings when connected to a reactive load, so I will gladly listen to what ever info you can share.

As you've found by now, there's a lot of smoke around this fire. I used MOVs in a low production amplifier, and they worked and sounded fine.

But that's not a representative sample, by any means.

What you're after is suppressing voltage spikes that would puncture magnet wire insulation. How much you need depends on how well the magnet wire is insulated in the first place. In general, you don't know that, so you have nothing to base "how much voltage" on. ACK!

However, normal operation for a push-pull amp involves having B+ minus maybe 50V across the active side, and that means that transformer action puts that much above B+ on the other side. So the transformer wires had better take two times B+ even when hot and 50 years tired. You need safety margin on top of that. Just to ballpark some numbers, call it 1000V for normal(ish) operation. So your protection devices had better NOT conduct on voltages in that range, and must conduct before the wire insulation gives up. Gotta guess a number. I guessed 650 volts per side on a 400V B+. Seems to have worked.

I've had several discussions about what specs are needed, MOV capacitance, energy dissipation, short term dissipation and long term conduction. MOVs have more capacitance, but do not seem to affect the sound. I could not hear any difference nor measure any change in response when I clipped them out or soldered them into the prototype. But I've been told that TVS devices are just as capable at eating heat, but lower capacitance. Time for a listening test, I suppose.

Does your planned level of protection include the case of an open secondary (no speaker plugged in)? In this case you would have to allow for dumping energy stored in the full transformer inductance rather than just the leakage inductance. Another thing to allow for is that you have to protect against repeating outrages, at the audio rate, rather than just a single transient, such as induced by a lightning strike.

Separating the energy in the leakage and primary is good thing to worry about, but complicated. You have to take note of the difference between magnetizing current and transformed current. Magnetizing current is all that's ever in the core. In a push-pull transformer, it's quite small. Not so with SE. But in a PP, the magnetizing current is a few percent of the peak primary currents. The rest is transformed current - the stuff that whizzes through the fields into the secondary wires. Transformed current does not change the magnetic energy stored in the iron core.

But it very much does get stored in the leakage inductance. So the leakage inductance may be much smaller, but it has much larger stored energy current to work with. Which one is biggest needs some hard-core pencil and paper work.

MOVs are actually pretty good at getting rid of high amounts of heat. What matters is how long and how much heat goes into transient suppressor. I have seen MOVs melt their own solder joints and fall out, but be otherwise undamaged.

I wasnt considering no load protection but even with a switching jack there is that potential in event of a cable plugged in only at one end. So i guess it should be considered. But initially I was really just considering repeat events during signal conditions. ...
So, lets talk about figuring for the full inductance. Ugh, of course, I had to go over the top and spec the primary for -1dB at 20Hz-20kHz at 10k plate to plate

As I said, more introspection is needed. The full inductance doesn't get nearly as much E = 1/2 * L * I^2.

Tube amps are unimpressed by being shorted. The biggest nasty event is someone pulling out a speaker cable when it's going full tilt, and even then you tend to get current arcing on the plug being pulled.

A more important nasty event is when the amp is left to sit with an open secondary. I ... um, accidentally... left the secondary open for some long periods on the Workhorse protos. Never had any damage at all, and could not figure out why it wasn't a smoking ruin. Turns out the real villain here is oscillation in amps with feedback on the power amp that go into ultrasonic oscillation with no load. That will kill things, OK. The Workhorse didn't use feedback, deliberately so as to hear more of the tubes' oddities. So an open speaker cable is not sure and certain amp-death. Maybe, but not certain.

The the bottom line, as I understood it, is MOVs will only you offer so much protection for so long. It's just a matter of choosing how long is enough for you. Personally I think a better approach would be to use small MOVs together with a means of detecting the abnormal voltage condition and shutting the drive off.

You have to note the time scale. The trip voltage on a MOV decreases an infinitesimal bit each time it conducts. Do it often enough and the voltage drifts down into normal operation and the MOV then has a big heat problem. This is very, very unlikely in amp use as we're discussing, especially when we can just pick a higher voltage MOV, given the uncertainty above about what the transient voltages will be. It's easy for it to be long enough.

Flyback diodes are very popular because they send spikes back to the power supply itself, which by definition can handle itīs own power.
They donīt dissipate anything , just resend voltage spikes back to the supply capacitors where that energy came from in the first place..

That's a solid state amp consideration, JM. Tube amps have other issues, and no good way to dump spikes back into the power supply.

As for movs, it appears that the devices which have higher power ratings and seem better suited for this application, have high capacitances. would this be problematic across the primaries of the OT? Should I use these capacitance is in design an RC conjunctive filter? I would imagine if the performance of a mob degrades with each overvoltage , then capacitance values would not be stable either but I don't know

I think a quick and dirty test would tell you more than ten hours on the internet. Get yourself some 600-800V MOVs and solder them across the B+ and plate leads of an OT. Can you hear the difference? If you can, maybe you should go for TVS devices, which are reputed to be lower capacitance. If you can hear it, but like the difference, you're done.

Gas discharge tubes are a different animal, and IMHO not suitable for OT protection.

[Mike Sulzer]

Separating the energy in the leakage and primary is good thing to worry about, but complicated. You have to take note of the difference between magnetizing current and transformed current. Magnetizing current is all that's ever in the core. In a push-pull transformer, it's quite small. Not so with SE. But in a PP, the magnetizing current is a few percent of the peak primary currents. The rest is transformed current - the stuff that whizzes through the fields into the secondary wires. Transformed current does not change the magnetic energy stored in the iron core.

Well, I agree that it is complicated, but if you pull out the load, you lose that transformed current. You then have an inductor. Well, not quite. You have coupling to the other tube, the one that is turning on while the first is turning off. Since a pentode is not a perfect current source (has somewhat sloping characteristics) changing the voltage across it changes the current through it some, and maybe allows the absorption of some energy. Maybe this is why amps do not always blow up. But this looks hard to analyze!

[J M Fahey]

Sorry Mike, I was thinking about a trick I use in my own OTs where I wind an extra half-turns winding which does exactly that, clamp excess voltage (and current) not to ground but to supply, but that was not the point in discussion but instead what I wanted to emphasize is that a MOV, being resistive, *always* dissipates (most of the) power inside itself, degrading over time, while a diode is (almost) transparent, having a very low voltage drop.

But letīs start again, focusing on the main point and referring to a conventional OT and standard clamp diode wiring : reverse biased from plate to ground vs. MOVS which are in parallel with some winding.

To avoid drawing everything from zero, enter a section of Peavey Butcher:

I show both options:
1) MOVs in parallel with plate to +V winding, might also be wired from plate to plate, same basic idea.
I think of MOVs basically as plain resistors in series with back to back Zeners meaning said resistors are basically out of the circuit until some "zener" threshold is reached.
Difference is that the conduction threshold is loose, less clearly defined but the advantage being that said resistor is way more robust, can take lots more abuse than a standard semiconductor junction.
So you choose MOVs so the "zener" voltage is, say, 20% to 50% higher than V+ if plate to V+ or 2 x V+ if plate to plate or , say, 50% above expected peak voltage across load and you "should" be safe; 50% voltage overload "should" be well within transformer insulation (or it "should" be designed and built with that in mind )

Sorry for all the conditionals involved, but I mention the way foolproof transformers should be designed, not so sure about how many do that.

I bet many think: "hey, I have 450V +V , Iīll be generous and insulate for twice that, 900V or 1000V" ,while truth is when amp just reaches clipping and is loaded, one plate (almost) reaches ground and like in a see saw, the other plate (almost) reaches 2x +V (relative to ground).

*With* MOV protection itīs limited to, say, 50% more, so a hair rising 3x +V if unloaded or very inductive speaker ..... and without protection and unloaded the sky is the limite ..... or to be more prosaic, transformer insulation puncture voltage.

Problem with MOVs is that said power dissipation degrades them; I do not know what is the actual mechanism, not sure whether they short, fail open or wildly change threshold voltage, which makes them useless, guess must make some extra studying.

2) clamping diodes (wired as shown here) , being reverse biased, are out of the circuit most of the time, even if amp clips but is loaded .
Now if load is very inductive or plain absent , so transformr inductance takes over, seesaw action means that when one plate tries to go beyond 2x +V, the other one tries to go below ground, what forward biases the diode and safely discharges excess voltage and remaining current to ground.
The diode itself dissipates little, thatīs what I wanted to emphasize compared to MOVs, and in heory lasts forever.

They do fail now and then anyway, but many suspect itīs through a different mechanism: very high peak plate voltages appear anyway (although much narrower) because of parasitic inductance, which prevents the seesaw mechanism to be perfect, that extra voltage "zeners" protection diodes by force, like it or not.
There must be something to this, because in theory 1200 or 1500 PIV diodes should be more than enough for most standard amplifiers, yet Manufacturers often use 3x 1N4007 in series or 3kV diodes .

[Mike Sulzer]

Sorry Mike, I was thinking about a trick I use in my own OTs where I wind an extra half-turns winding which does exactly that, clamp excess voltage (and current) not to ground but to supply, ...

I like that! But I think that for protecting an existing amp MOV devices (or something that behave in a very similar way) are the right way to go. I just do not feel competent to decide which of the available specific devices listed in a catalog is the right one.

[R.G.]

Well, I agree that it is complicated, but if you pull out the load, you lose that transformed current. You then have an inductor. Well, not quite. You have coupling to the other tube, the one that is turning on while the first is turning off. Since a pentode is not a perfect current source (has somewhat sloping characteristics) changing the voltage across it changes the current through it some, and maybe allows the absorption of some energy. Maybe this is why amps do not always blow up. But this looks hard to analyze!

You're right. It gets complicated fast.

My response to "complicated" was drummed into me by my early mentors: if you can't predict it, find a way to make the wobbles insignificant in the greater scheme of things. I'll usually fall back on estimates of boundary conditions, like "well, how big can the [energy/voltage/current] really be, anyway?" I figure that if I can handle that, I have the subtleties covered.

In this case, the energy in the transformer core is the unbalanced current in the primary. A push-pull core with no signal has a residual magnetic field of zero. The bias currents cancel out, leaving only the degree to which the tube currents don't match. That changes with signal.

For a voltage signal, the current in the core is the integral of Vdt/L from zero up to the voltage. Translated into human words, the current ramps up at a rate of V/L until stopped by something. The something that stops it is the output tubes. One significant case is with one tube dead. The core is then conducting the full bias current through the primary inductance. A worse condition is if the amp is motorboating. If this is happening, then it's swinging subsonically and there may be large currents in the primary. But that doesn't happen much in real amps.

I suspect that if you could catch it on an instrumented amplifier, transformer death would tend to be around an event like the amplifier oscillating at high frequency and overheating an output tube till it opens. (a shorted output tube would clamp its half-winding and stop any transients) When this happens, the current in the leakage inductances is unclamped and makes a voltage that punctures the insulation on the magnet wire in one place. Later events work on that same place until an arc forms that welds the wires there. Death ensues.


JM says:

2) clamping diodes (wired as shown here) , being reverse biased, are out of the circuit most of the time, even if amp clips but is loaded .
Now if load is very inductive or plain absent , so transformr inductance takes over, seesaw action means that when one plate tries to go beyond 2x +V, the other one tries to go below ground, what forward biases the diode and safely discharges excess voltage and remaining current to ground.
The diode itself dissipates little, thatīs what I wanted to emphasize compared to MOVs, and in heory lasts forever.

They do fail now and then anyway, but many suspect itīs through a different mechanism: very high peak plate voltages appear anyway (although much narrower) because of parasitic inductance, which prevents the seesaw mechanism to be perfect, that extra voltage "zeners" protection diodes by force, like it or not.
There must be something to this, because in theory 1200 or 1500 PIV diodes should be more than enough for most standard amplifiers, yet Manufacturers often use 3x 1N4007 in series or 3kV diodes .

Yes, and I think that's the idea behind the clamp diodes. The diode that helps you is the one that conducts from ground in the forward direction. No doubt some situations are prevented by just this. The problem that lurks here is the leakage inductance. The leakage cannot, by definition, be clamped by transformer action, so it acts like an inductor in series with each transformer lead. This much smaller inductance is charged up to the same current as the lead and plays V=L * di/dt when the current changes.

I think that the clever guys who figured out protection by transformer action with diodes was dismayed to find out that some spikes leaked through, and used the easiest to find cheap diodes to catch the remaining spikes by avalanche/zener action. Then they added enough of them in series to not break over in normal operation.

You're right that a diode going into conduction is not terribly dissipative. But the zenering action is. So in zenering mode, the clamp diodes do heat. Whether they die or not eventually is problematic, but they will last many more cycles than a MOV. On the other hand, MOVs degrade quite slowly. In AC power circuits, they only die when they drift down into the normal AC line voltages. This is years of constant use for 130V MOVs on a 120V line, and much longer than the equipment life for a 150V MOV on the 120V line. So a MOV with an additional 100% of B+ as a breakover in a tube amp is going to be essentially immortal; well, no more mortal than the resistors and caps.

In an amp with a 500V B+, one might put 1000V breakover MOVs direcly across the transformer leads from CT to ends. That means that the MOVs would clamp the entire primary at about 2kV, still less than the 3KV and more that three 1N4007s try to catch, but that they would take a long, long, .... long ... time to degrade down to being only 1*B+, which is where they'd start burning out. So MOV drift can be made a non-issue.

TVS devices apparently don't have the drift-down effect, and also have low capacitance. These have been developed after MOVs, and recently (...um, last decade? I'm getting old) have been scaled up to higher energies for clamping. So a TVS may be like a MOV, but better. I haven't tried them.

[SoulFetish]

However, normal operation for a push-pull amp involves having B+ minus maybe 50V across the active side, and that means that transformer action puts that much above B+ on the other side.

For whatever reason I was confused about this statement earlier, but I just realized you are talking about the anode voltage swing during full drive and that potential difference across that side of the winding.

Separating the energy in the leakage and primary is good thing to worry about, but complicated. You have to take note of the difference between magnetizing current and transformed current. Magnetizing current is all that's ever in the core. In a push-pull transformer, it's quite small. Not so with SE. But in a PP, the magnetizing current is a few percent of the peak primary currents. The rest is transformed current - the stuff that whizzes through the fields into the secondary wires. Transformed current does not change the magnetic energy stored in the iron core.

But it very much does get stored in the leakage inductance. So the leakage inductance may be much smaller, but it has much larger stored energy current to work with. Which one is biggest needs some hard-core pencil and paper work.

I would really like to get a deeper understanding of the magnetizing current, leakage inductance, and the relationship of energy to the electromagnetic field in a transformer. I apologize for sounding so incredibly generic, but I don't know how else to say "There is more for me to know with these things".

As I said, more introspection is needed.

Well I just went through all the threads and links to other threads and at this point, it's probably a full blown existential crisis.

The biggest nasty event is someone pulling out a speaker cable when it's going full tilt, and even then you tend to get current arcing on the plug being pulled.

I've never seen the effect of current arcing in this way. What is the effect and why would it happen in this case?

I think a quick and dirty test would tell you more than ten hours on the internet. Get yourself some 600-800V MOVs and solder them across the B+ and plate leads of an OT. Can you hear the difference? If you can, maybe you should go for TVS devices, which are reputed to be lower capacitance. If you can hear it, but like the difference, you're done.

Ah! I can't wait to get to this part! I'm not there yet. My chassis is drilled and ready; I'm drilling my turret board tonight and am placing what is hopefully my last order of bits and pieces I need for parts.
On that note, I'm struggling to find 2AG sized cartridge fuses rated for 500V-600V with current ratings suitable for my secondary windings. Anyone have a good recommendation?

[SoulFetish]

Sorry Mike, I was thinking about a trick I use in my own OTs where I wind an extra half-turns winding which does exactly that, clamp excess voltage (and current) not to ground but to supply, but that was not the point in discussion......

Whoa, whoa, whoa... Not so fast, Fahey. You just got mike all excited about this (probably, because he already knows what this would do), but I kinda' want to know what all the fuss is about this. I like to learn to tricks. How does this one work?