Tube vs solid state power amp volumes and speaker load

[loudthud]

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.

Yes, the time/date stamps should have survived the zip process. Those are just the file names the camera assigns sequentially.

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?

I've never been able to detect a difference in tone with or without the diodes. Try it yourself.

When an inductor discharges into a diode, it should take longer than if the inductor is allowed to produce a large Voltage and discharge into a resistor or other circuit element, even a zener or MOV. You can see this in simple switching power supplies where the inductor doesn't fully discharge on the first couple of pulses and current spikes up until the output cap can charge up. I expect it to sound different, but have never heard it.

Could it be that even with interleaving, either side of the primary is well coupled to the secondary, but each side of the primary is not so well coupled to the other side? If you measure leakage inductance on one side of the primary, wouldn't you get a different number if you shorted the other side of the primary as opposed to shorting the secondary?

[J M Fahey]

I've never been able to detect a difference in tone with or without the diodes. Try it yourself.

Weird; I hear huge difference between 640 and 641.
The first one is middier, flatter and not *that* distorted, the second is way more aggressive, balls to the wall, louder, definitely better.
Won't comment on the third (642) because it shows way less gain and distortion than others; apples and oranges comparison.

[uneumann]

Weird; I hear huge difference between 640 and 641.
The first one is middier, flatter and not *that* distorted, the second is way more aggressive, balls to the wall, louder, definitely better.
Won't comment on the third (642) because it shows way less gain and distortion than others; apples and oranges comparison.

The clip to clip audio is hard to gauge without knowing more how it was created. I've tried the diodes on one amp, and I too could not hear the difference. BUT - looking at the plate signals makes me wonder why I didn't hear it. Is there something I'm missing as to why a signal so different (w/wo diodes) doesn't sound different? Next time I try it I will scope the spkr signal too. Something about all of this doesn't quite make sense to me. How can the primary look so different without the spkr signal being affected? And if it's affected, why don't I notice?

[loudthud]

The audio on the clips is just the microphone on the camera, it has no external audio input. The amp is pointed under a table and then through some shelves that hold some test equipment accessible to the table top, then into a large area, perhaps 15X20 feet with a 15 ft ceiling. I was sitting behind the amp. The scope is on a cart and the camera is on a tripod about a foot from the scope. I have to press the button on the cam, then start strumming.

The camera compresses quite a bit, on the third clip you just hear me strumming the SG into a dummy load. Amp settings were not intentionally changed between clips but can't rule out them being bumped. Volume control is only about half way up.

It's hard for me to A-B changes unless I have a footswitch setup and can switch back and forth. That would take a high Voltage relay in this case. By the time I put down the guitar, pickup a soldering iron, make changes and pickup the guitar again, my frame of reference is gone. Note that the sweep speed is 200uS/div. I ran it that fast so you could see the inductive bumps. One division is the period of 5KHz signal, so some of what you see is beyond the ability of the speakers to reproduce.

[R.G.]

Y
When an inductor discharges into a diode, it should take longer than if the inductor is allowed to produce a large Voltage and discharge into a resistor or other circuit element, even a zener or MOV. You can see this in simple switching power supplies where the inductor doesn't fully discharge on the first couple of pulses and current spikes up until the output cap can charge up.

The discharge time of an inductor is pretty much completely determined by the voltage across the inductor as it discharges: di/dt =V/L . If it's being clamped by a diode, then the voltage is the diode's forward voltage of 0.5 to a few volts (for large currents), so di/dt is small, and the time to discharge is long. Discharge into a zener or MOV is the same, di/dt = V/L, but the zener or MOV voltage is bigger than a diode forward drop. Discharge into a resistor yields a time constant, of R/L.

An oddity is when you make the resistor equal to zero. When you short an inductor externally, the current theoretically lasts forever; actually it is discharging into its own internal resistance. A shorted inductor made with a superconductor wire would continue to conduct current forever.

I expect it to sound different, but have never heard it.

Could it be that even with interleaving, either side of the primary is well coupled to the secondary, but each side of the primary is not so well coupled to the other side? If you measure leakage inductance on one side of the primary, wouldn't you get a different number if you shorted the other side of the primary as opposed to shorting the secondary?

Things get funny when you're dealing with leakage. Yes, it can be, and is, that both sides of the primary can be coupled to the secondary, but not so much to each other. This is in fact common: wind on a half-primary, then the secondary, then the other half-primary. Much better coupling from secondary to each half-primary than from H-P to H-P. That's a cheap way to wind a transformer, and a bad way to wind it for good audio, as it also includes badnesses for imbalanced wire resistance and parasitic capacitances. In general, you DO get a different number for leakage from any winding to any other shorted winding in a multi-winding transformer.

The precautions in winding a high quality, flat response output transformer can be severe. Sometime read the magazine article on winding the OT for the "Williamson" amplifier. for an intro to this. The way to minimize leakage is to wind the wires for the various windings literally side by side: multifilar winding. When you wind separate wires literally side by side, you maximize the inductive coupling between them. This is why tight coupling of half-primaries requires bifilar winding.

What you'd LIKE to do is to tri- (or quad- or penta-) filar wind the secondaries with the primaries. That would get the lowest possible leakage. But that's hard to do when the turns ratio between primary and secondary is in the 20's.

The next best is interleaving: wind a layer of primary, then a layer of secondary, then another layer of primary, then another secondary... It's not as good as multifilar, but better than non-interleaved.

BUT - looking at the plate signals makes me wonder why I didn't hear it. Is there something I'm missing as to why a signal so different (w/wo diodes) doesn't sound different? Next time I try it I will scope the spkr signal too. Something about all of this doesn't quite make sense to me. How can the primary look so different without the spkr signal being affected? And if it's affected, why don't I notice?

I haven't seen the difference in traces you're talking about yet, but remember that when leakage is discharging on one half-primary, it's the OFF half-primary. The ON half-primary is the one that's conducting, and so is the one that's driving the speaker load. The discharging leakage is by definition not coupled to the other windings, so it's going to be very hard to hear.

[Malcolm Irving]

.... Discharge into a resistor yields a time constant, of R/L. ...

Minor correction: the time constant is L/R

[uneumann]

I haven't seen the difference in traces you're talking about yet, but remember that when leakage is discharging on one half-primary, it's the OFF half-primary. The ON half-primary is the one that's conducting, and so is the one that's driving the speaker load. The discharging leakage is by definition not coupled to the other windings, so it's going to be very hard to hear.

Thanks for that... it helps. The trace differences are those with/w.o. the diodes - they are clear in LT's clips. What seems to be happening (and unintuitive for me) is that some portion of the primary inductance is uncoupled and creating signals on the plates that are visible, but those signals don't couple to the output... That's just weird.

[g1]

Just to clarify, the amp does have NFB loop ?

[nickb]

If leakage inductance alone is the source of the overshoots then why are none present in the resistive load trace?

It seems that the total inductance reflected to the primary is what matters i.e. Lmag//(turns^2 *Lspkr) + Lleak

[loudthud]

Just to clarify, the amp does have NFB loop ?

Yes. 27K from a 2 Ohm secondary to the presence control.

[Mike Sulzer]

The discharging leakage is by definition not coupled to the other windings, so it's going to be very hard to hear.

In modeling the effect of leakage flux, the leakage of flux is replaced with an inductor in series with a perfectly coupled winding. The energy in the leakage inductor is a function of the total current through the winding. When the winding is switched off abruptly, energy in this inductor is coupled to the other windings, but the rate at which this can happen is limited to that determined by the value of the leakage inductance. Thus a spike develops across the winding, and the spike is not seen on the other windings precisely because the high frequencies are not transferred.

[J M Fahey]

Yes, that's the very definition of leakage inductance, and precisely implies non transfer.
Otherwise the load would .... well .... load it and damp the peak.

When this does not happen, we have to retort to snubbers.

[Mike Sulzer]

Yes, that's the very definition of leakage inductance, and precisely implies non transfer.
Otherwise the load would .... well .... load it and damp the peak.

When this does not happen, we have to retort to snubbers.

In theory, you could use a transformer with very good high frequency response. The transient would be shorter, less total energy (but still high in instantaneous voltage). Or you could slew rate limit the drive to the output stage, slowing down the tube's transition from on to off. I think you could do this without affecting the output at and below 5 KHz very much while reducing the energy in the transient quite a bit. The key here is how much is "very much": you would alter the response in your bandpass some, and you might not be able to keep the sound you want.

[Malcolm Irving]

Well yes, the current through, and the voltage across, the leakage inductance (by definition) does not couple, by transformer action, to any other winding. But, as Mike implies in post #71, the leakage inductance is in series with the ‘ideal transformer’ part of the transformer’s equivalent circuit. It will therefore have an influence over the voltage and current into that ideal transformer.

Energy stored in an inductance (including a leakage inductance) is potential energy in the magnetic field of that inductance. As long as current through the inductance remains constant, the energy stays put in the magnetic field. If we try to reduce the current, then we get a voltage across the inductor. The product of the current and voltage is of course power (i.e. rate of energy transfer) and this will either be dissipated in any resistive part of the circuit, or transferred to a new energy store (e.g. into a capacitance).

Perhaps the winding stray capacitance is in play here, and might account for the spike not getting through to the secondary. (That’s perhaps another way of saying that the bandwidth of the transformer might not be wide enough to pass the spike through, as already mentioned by Mike.)

[Mike Sulzer]

...

Perhaps the winding stray capacitance is in play here, and might account for the spike not getting through to the secondary. (That’s just another way of saying that the bandwidth of the transformer might not be wide enough to pass the spike through, as already mentioned by Mike.)

Stray capacitance is not essential in preventing the spike from getting to the secondary. At low frequencies the load on the primary is the load on the secondary altered by the turns ratio squared. For correct power transfer, you can consider the leakage inductance in series with either the primary or the secondary, adjusting the value as required. But to look at the spike it is most convenient to think of it as in series with the secondary. Then you can see that the load on the primary rises with frequency and so a voltage spike appears across the primary, but the voltage reaching the load on the secondary is reduced by the action of the series inductor.