Tube vs solid state power amp volumes and speaker load

[J M Fahey]

J M, that wasn't a "very simplified answer."

"Flyback voltage," would have been a very simplified answer.

Oh, I started that way ... but got carried away
It's something that worries me a lot.

The "simple" solution is to use , say, 2X higher rail voltages ... which unfortunately means 4X the power (just do the Math ) , is expensive , bulky, heavy and dangerous , it's *easy* to be carried away and burn speakers.

Marshall had some tweaked Class G or H (never learnt the difference) Bass power amps, the idea was that the amp run from the main rail all the time, putting out, say, 300W RMS, but on note attack the higher voltage rail was engaged for a brief time ... claimed up to 1kW instantaneous power ... instant slapper's paradise.

The series was a commercial failure (guess Marshall customers are not exactly slappy Fusion Bass players) but the idea might have some merit for high power *guitar* amps.

[Malcolm Irving]

Oh, I started that way ... but got carried away
It's something that worries me a lot.
....

Please keep getting carried away! The longer your posts are, the more we like them!

[teemuk]

I think that such a solid state amp would have to have rather higher power supply rail voltages then would normally be required for a regular solid state amp of similar rated power output.

First of all, "flyback" is most of all a characteristic of output transformer coupled circuits, or choke-loaded circuits in general, not so much a characteristic of tubes or semiconductors as is. But admittably the high output impedance of tubes will enhance that particular phenomenon.

Anyway, I assume when you say "solid-state amp" you mean generic solid-state amps without output transformers. Yes, they will generally clip to rail voltage and any flyback transients get clipped by same process. Many amps even include diodes from output terminal to each high supply rail that clamp any "back EMF" (that would be the flyback) transient voltage to about high supply voltage level.

So you are right: In order that SS amps like that could reproduce the "flyback" transient without distortion they must never be allowed to clip to supply rail voltage limits. There are some amps that take this into concern when output power figures at x % of THD gets quoted. For example, new Quilter amps are rated for 100W but the power amp's headroom actually runs to 400W! The extra 300W is reserved for reproducing the "flyback" transients without clipping. Voltage clipping of the power amp is prevented with clever tube emulation circuit, but output impedance is artifically increased and the "flyback" transients will increase signal voltage above the former clipping limit. So extra headroom is needed. Quoted percentage of THD, however, is around the 100-watt mark. I recall there was also a Peavey 30W Vypyr amp that actually used a 60W power amp circuit (including PSU voltage) but the extra 30W was reserved for reproducing flybacks cleanly. Quoted percentage of THD again at 30W mark. Rest for headroom. Power amp clipping itself was limited by the "T-Dynamics" circuitry. I'm sure there are many more. It's not an overlooked aspect of design.

One thing that makes the job easier is that this extra headroom does not have to be "continuous power". You only need it for a brief transient, so the average effects of, say heat or supply current draw, will not pose as great design concerns as when designing for continuous power. In fact, sag characteristics of the power supply can be implemented as a protection feature.

...Then there are manufacturers who just quote a wishful estimate: It voltage clips at x volts to y-ohm load, there we have the output power. Then of course the amp will dissapoint when it tries to deliver the "flyback" transients.

But at this point it needs to be mentioned that many tube amps have diodes on output tube plates that have sole purpose of clipping excessive flyback transients. So they many not be that different from generic SS amps that there is a brickwall limit for effective output signal swing. Why? Because excessive flyback transients have a good chance of arcing through in inconvenient place in the circuit and then your precious tube amp that can reproduce all those flyback transients becomes us good as a boat anchor. So those diodes, or something similar, should already be installed as a safety feature in any reasonable design.

[Enzo]

Having said that, the flyback diodes on the tube amps generally flash over at 2kv to 3kv So the voltages at the plates can still exceed B+ by a fair amount.

[J M Fahey]

You bet.

Besides, even clamping diodes can't fully protect an OT.

Transformers show 2 inductance values, under 2 different conditions.
Most care about the first and ignore the second, but both are important.

a) you connect the transformer primary, all other leads floating, to your bridge/L meter.
That's the main primary inductance, is in parallel with the load, so it determines the low frequency cutoff.

Say nominal load is 4k ohms? ... when that inductance also shows 4 k ohms impedance, half the current fed into that transformer will go to the speaker, half will be lost in that parallel inductor so you'll have a -3dB point at that frequency.

That's why Bass amp transformers are *huge* , lots of iron, lots of copper, to be able to get high inductance and lose little bass.

But there is also another inductance value:
b) you short transformer secondary and measure again.

You *should* measure zero inductance ... but you will not, you'll still measure some.
That residual inductance is called parasitic or shorted inductance (freely translating from Spanish) and by definition is NOT shorted by clamped diodes, so you can still see some hairy spikes on plates (diodes clamp "the other end" of primary, but between "there" and "here" we still have the parasite inductance).
Admittedly, situation is far better than no diodes at all anyway.

By the way, that parasitic inductance is in series with the load and kills highs.
No big deal with guitar amps, but a problem with Hi Fi ones, that's why they interleave windings (that lowers parasitic inductance).

A worse problem is that parasitic inductance and parasitic capacitance resonate at some frequency and create a nasty peak there.

Parasitic capacitance also lowers with interleaving which sends the peak well away from audible frequencies.

[Malcolm Irving]

But there is also another inductance value:
b) you short transformer secondary and measure again.

You *should* measure zero inductance ... but you will not, you'll still measure some.
That residual inductance is called parasitic or shorted inductance (freely translating from Spanish) ... .

Yes. That parasitic inductance is also called 'leakage inductance'. It is the result of magnetic flux from a winding which is not linking with the other winding(s).

[Mike Sulzer]

Yes. That parasitic inductance is also called 'leakage inductance'. It is the result of magnetic flux from a winding which is not linking with the other winding(s).

And the main effect of that is to limit high frequency response. It is the enemy of the broadband transformer. Make the inductance higher to extend the low frequency response, but then yo run into problem with too much leakage flux, meaning that the leakage inductance rises..

[Mike Sulzer]

Having said that, the flyback diodes on the tube amps generally flash over at 2kv to 3kv So the voltages at the plates can still exceed B+ by a fair amount.

Enzo, I hope you do not mind if I state the obvious. After reading through this discussion, I am not sure everyone is on the same page, and I am responding to your post because it is the most recent.

First, in a stage with choke or transformer load, the plate voltage must go both above and below the B+ by amounts approaching the level of the B+ in normal linear operation. That is simply how they work. Excessive voltage from tubes shutting off is another matter.

Also, the diodes used from plate to ground are intended to stop the plate from going negative (perhaps caused by energy from the transformer if the current through the tubes is abruptly shut off). The 1N4007 or R3000 are high voltage diodes. They are connected to conduct when the voltage goes negative, and a high enough voltage rating is used so that they do not break down (with possible destruction) when the voltage goes positive. This might appear to be not good enough to dump energy, but I think when current from one half of the primary winding is abruptly interrupted, energy couples to the other winding attempting to move the other plate negative, but the diode conducts and the energy gets safely dumped.

Or at least this is how I think it works, but I will certainly listen if anyone else knows better.

[Enzo]

No, of course everyone is not on the same page, in fact this particular discussion is impossible to have. As soon as the topic is breached the regular factions all chime in with tangential views. Not saying anyone is incorrect, just not all talking to the same purpose. I try to corral the "a watt is a watt" people because that was never the point. Then there are the tube and solid state are the same people, or at least, nothing one can do can't be done by the pother sort of position. But creating a tube circuit out of a pile of op amps and such may be quite possible, but no manufacturer does it so it is moot. And so on. So we just bat it around a page or two or three, until it sort of peters out, and then it comes up a gain later, and we do the same dance all over.

[Malcolm Irving]

... might appear to be not good enough to dump energy, but I think when current from one half of the primary winding is abruptly interrupted, energy couples to the other winding attempting to move the other plate negative, but the diode conducts and the energy gets safely dumped.

Or at least this is how I think it works, but I will certainly listen if anyone else knows better.

Yes, that sounds reasonable, but I don’t think it’s easy to confirm.

A push-pull output transformer is really a three-winding transformer (with two primary windings connected together at the B+ terminal).
This means that there are 7 flux paths to consider, when working out various inductances. Namely: flux that links all three windings, fluxes that link only one winding (3 of those) and fluxes that link pairs of windings (3 of those). Complicated!

To be sure of what is going on, either we need accurate 3-winding transformer models in Spice, or some comprehensive bench tests. (Might have a go at that sometime!) Once you throw in the nonlinearity of the magnetization (including possible saturation) it gets very complicated!

Although we do tend to ‘go around repeating circles’ in these kinds of threads, I often find that I understand something a bit better each time!

[J M Fahey]

Also, the diodes used from plate to ground are intended to stop the plate from going negative (perhaps caused by energy from the transformer if the current through the tubes is abruptly shut off). The 1N4007 or R3000 are high voltage diodes. They are connected to conduct when the voltage goes negative, and a high enough voltage rating is used so that they do not break down (with possible destruction) when the voltage goes positive. This might appear to be not good enough to dump energy, but I think when current from one half of the primary winding is abruptly interrupted, energy couples to the other winding attempting to move the other plate negative, but the diode conducts and the energy gets safely dumped.

Yes, that's how it works ... and also shows the problem with leakage in ductance:
One plate saturates, reaches ground (in fact will stop some 50V above it) but transformer inductance forces it to go on, reaching negative voltages.
"The other end" , meaning the opposite plate, will rise by the same amount above 2X +B voltage.

Now, if you clamp the saturating plate to a couple diode drops below 0V (what's needed to forward bias protection diodes), the opposite plate "should" not risem ore than negligible couple diode drops above 2 X +B .

Problem is that between clamping diodes and opposite plate(s) you have the largest "leakagiest" (sorry) winding possible in a transformer: both plate windings, to boot in series, 2X the turns and 4X the leakage inductance.

IF there is an unexpected voltage peak (and there is), this is where it comes from.

For a numarical value we can measure all inductances and combine them in a transformer model as suggested above, no doubt about that, , but without reaching there, I think the concept is clear: we still do have an inductor, and it can and will produce flyback pulses, even if "clamped".

Again: not as destructive as an unclamped one, transformers last way longer when diode protected, but there must be some mechanism which destroys such diodes now and then (we have all replaced them one time or another), even if they seem to work in a fail safe mode , as in: peaks "should" never reach above 2X +B and current "should" never be above what tubes are passing when saturated ... yet "something " kills them sometimes.

Some (properly attenuated) scope captures should clear things beyond what we are discussing here.

Loudthud anybody?

[R.G.]

Problem is that between clamping diodes and opposite plate(s) you have the largest "leakagiest" (sorry) winding possible in a transformer: both plate windings, to boot in series, 2X the turns and 4X the leakage inductance.

IF there is an unexpected voltage peak (and there is), this is where it comes from.

Yes. You can think of any transformer wire as having an invisible inductor in series with it. This is the leakage inductance of that wire's winding to every other winding. I amounts to magnetic field that is not coupled to the other wires, and cannot be, by definition.

This is the origin of my preference for plate-to-plate clamps like MOVs (and Steve would correct me here by saying that TVS are better; that's probably true ) and why I do not like the silly stacks of clamp diodes from ground to the plates.

The diodes, as noted, do not clamp the flyback pulse of the "off" winding. They can't. They can only clamp the two halves of the primary that are tightly coupled inside the transformer, but they rely on the internal transformer action to try to clamp the leakage that is NOT coupled to the two primary windings. So they fail. Actually, it's not as bad as all that, because the leakage inductance is a few orders of magnitude lower than the primary inductance where the really big energy is stored, but the leakages can still carry a lot of voltage-whallop.

And they don't need much. They can punch through wire insulation or carbonize dust trails on the tube sockets, or flash over the tubes inside without a huge energy punch.

Stacked diodes DO prevent things from getting too out of hand, because they clamp the leakage inductance flyback pulse to no more than the difference between B+ and their avalanche voltage. That's good as far as it goes, because the plates are going to see twice B+ minus the 50V "saturation" voltage of the power tube every other half cycle. A diode with only a 1kV rating will die pretty quickly there, as it will be broken over a lot by the 2xB+ minus 50V and the additional unclamped leakage inductance and will degrade fast.

That's why there are stacks. A "1kV" diode does not avalanche at 1000V; it's guaranteed NOT to break at 1000. How much over that it goes is one of those little unspecified things. A 1N4007 may break over at 1100, 1200, 1500?? it's unknown until you test it. So an amp maker that loses a lot of output trannies and 1kV diodes puts in another 1kV diode. It gets better. But you still lose some diodes, as the 2000-3000V breakover of the combined stack can still be punched by that unlimited flyback voltage of the leakage diode.

So you can get THREE "1kv" diode in a stack. Or more. At some point, the avalanche rating of the stack stops the diodes from avalanching entirely - and they may as well not be there as far as leakage inductance is concerned, because they can't clamp the leakage pulse, and the pulse breaks something else.

So yes, there are unclamped leakages even with diode stacks. They can kill the diodes until they start killing something else.

I know this sad story from my history designing high frequency switching flyback power transformers. Nothing will clamp a leakage except a snubber or clamp right on that wire.

And Enzo is right - this has nothing to do with tube watts and solid state watts.

[loudthud]

Finally got a chance to do a little work on the questions raised in this thread. Here's what I did. On my 5F6A re-issue a soldered a stack of three 1N4007s on each plate. They each measured about 1600V breakdown. A 100X Tek probe was attached to each plate. The scope was setup as follows. The ground reference for each trace is set to 2 divisions above the bottom of the screen. The deflection factor is 200V per division. The Vertical mode is chop so both traces are at the same time, but they can get a little fuzzy at 200uS/div sweep speed.

The attached file is in Zip format because the forum software doesn't recognize the video format. I know it will play on Windows Media Player, Mac users may be SOL.

The first file is with the diodes in place. You will need to pause the video to study what is going on. You can see the when on trace hits the diodes, the other trace can exceed 2X the B+ Voltage. The second file is the exact same setup but the diodes were disconnected. The third file is the same as the second except a 2 Ohm dummy load is connected instead of the speakers.

[Steve A.]

loudthud: Just confirming that the first file as mentioned in your post is MVC-640W.MPG, with the 2nd and 3rd increased to 641 and 642, respectively.

Steve A.

[uneumann]

Finally got a chance to do a little work on the questions raised in this thread. Here's what I did. On my 5F6A re-issue a soldered a stack of three 1N4007s on each plate. They each measured about 1600V breakdown. A 100X Tek probe was attached to each plate. The scope was setup as follows. The ground reference for each trace is set to 2 divisions above the bottom of the screen. The deflection factor is 200V per division. The Vertical mode is chop so both traces are at the same time, but they can get a little fuzzy at 200uS/div sweep speed.

The attached file is in Zip format because the forum software doesn't recognize the video format. I know it will play on Windows Media Player, Mac users may be SOL.

The first file is with the diodes in place. You will need to pause the video to study what is going on. You can see the when on trace hits the diodes, the other trace can exceed 2X the B+ Voltage. The second file is the exact same setup but the diodes were disconnected. The third file is the same as the second except a 2 Ohm dummy load is connected instead of the speakers.

Thanks for that... Do you also have a trace of the speaker signal? It would be great to see if there's any difference between the diode/no diode cases. Do you hear any difference?