I think I need to go back...

[eschertron]

... and measure the output power of my 2xEL84 amp again. Measuring across a 4R load gave me (P=E^2/R) about 8-ish Watts, roughly half of what I wanted to see with a sine wave just at/below the point of clipping. I may have forgotten to move the impedance selector switch from the 16R tap to the 4R. This brings up a host of questions:

Impedance bridging allows 1/2 of a supply power to be transmitted to a load. So if I have a PA delivering, say, 16W into a OT, is half of that 16W being consumed by the OT? Or being a reactive source, the 'rule of 1/2' doesn't apply?

I've learned that using shorting jacks on the output protects against high voltage spikes from damaging the OT winding insulation. Also learned that OTs function on an impedance RATIO, so I induce that if the secondary is shorted and the OT gets a reflected load of ~0R, the primary should also be a very low impedance. Right? Low impedance, output voltage also falls toward zero, output power goes toward zero and the OT is safe. OK. But looking at the simulator on Nick B's website, I see that the formulas used here show the output TUBE power to skyrocket as OT impedance goes to zero. Is this an actual effect, or a limitation of the model?

Regards.

[Dave H]

Impedance bridging allows 1/2 of a supply power to be transmitted to a load. So if I have a PA delivering, say, 16W into a OT, is half of that 16W being consumed by the OT? Or being a reactive source, the 'rule of 1/2' doesn't apply?

I don't think the OT consumes half the power. It's just that the power isn't generated because the tubes can't provide the current required for the lower impedance load.

When the load impedance is zero there's no voltage dropped across the load. The tubes limit the current but now all the voltage is across the tubes so their power skyrockets.

[J M Fahey]

Agree.
1) transformers are usually around 90/95% efficient.
Tube OTs are somewhat lossier, generally because they are wound with thinner wire than needed, so higher resistive losses, they are also used WAY above 50/60 Hz mains frequencies which increases magnetic losses, but even so they should be 75/80% efficient ... or better; no way they lose 50% of power going through them.

Besides, tubes are self "current limited" , as simple as that, so going down in impedance is a certain way to lose power.

[eschertron]

So transformers are fairly efficient. I would hope so, but I'm still confused about rules for impedance matching and power transmission. Not something to worry about today. Off the table.

I went back and ran my test again, still mystified. Here's what I know:
OT impedance selector definitely on the 4R tap, with a 4R 25W load resistor. OT secondary signal starts to flatten out at ~18..19vpp on scope, reads 6.85vrms on DVM. (6.85^2)/4 = 11.7 watts into the resistor. So I started probing the circuit path. At the grids of the EL84s there's no sign of flattening. The PA is not grid limiting yet. Grid limiting comes after about another 5vpp, where I'd expect to see output tube clipping (and another 5 or 6 Watts!). Other parts of the circuit will clip, but later than the output tubes.

I took my trusty fluke and measured the plates directly. 340vdc at idle, 164vac signal just at clipping, or about +/- 232v from B+. The model suggests I should get another +/-50 or 60v before running into grid limiting, which is what I think I'd get if the waveform didn't flatten out before then.

Looking at the model I suspected that if the screens were dropping a lot - down into the low 200s - then I might lose power that way. So I measured all the PSU nodes at Idle, at the start of clipping, and at nearly full bore. The amp is cathode biased at ~14v (Chuck H zener mod - license pending).

_____________Idle Clip Full
B+___________348 325 310
Screen node___344 319 300
Screens_______341 314 285

With individual 1k screen resistors, I see the screen current go from 3ma at idle to 15ma at full. Should I allow the screens to deliver more current by reducing the size of the screen stoppers? Looking for but not finding what to plug into a screen dissipation calculation. 15ma x what? Screen-Cathode voltage is 270vdc. That would make 4W, twice the rated limit? That calculation can't be right?

The power supply is definitely sagging some. If the plates drop to 300vdc, and there's a 230v swing, the model shows the tube gets real close to the grid limit. But the scope shows I can increase the signal at the grids by several more volts before current starts to flow! Frustrating! What am I not seeing here?

Nick B's tube model page
set up for idle conditions (B+ 330, OT 8kaa, screens 325, bias 13..14 ish) it looks really pretty.

side note: after all the above, I temporarily clipped OUT the 14v zener on the cathode of the output tubes, just to see what effect I'd get. The clipping still happened at 18..19vpp on the 4R load, but a very obvious crossover developed early as the MV was turned up. I put the zener back in

[nickb]

Taking the information given and plugging it in I see about 15W out with Vgk just touching zero. You are seeing 11.7W. Allowing for model approximations, tube tolerances, measurement errors and losses that seems reasonable. If the load impedance were slightly higher that you expect then that will produce the effect you describe. I used 8.7K which I calculated from the figures above. 11.7W vs 15W is about 1dB i.e only just perceptible.

[Dave H]

My 2 x EL84 amp measures -

Idle
B+ 336V
Plate 333V
Screen 318V
Cathode 11V
Bias 35mA

15W
B+ 324V
Cathode 13V

The waveform @ 15W, 10R load (on the 8R tap,10k p-p), 5% THD is as below -

[eschertron]

I scoped the voltage directly on the screen pins and I see that the screen voltage drops significantly (screen conduction increases) at that point where the output voltage flattens. Surprisingly, when I jumpered out the 1k screen stoppers, the output waveform looked no different even though the screen voltage didn't sag the way it did with the 1k stoppers.

RMS measurements on the load resistor indicate 11W just before clipping, and about 16.5W full-on square wave (all my testing was done at 400Hz). The PT and OT do not get hot - just barely above room temp - so I'm not sure where the (current/power) limiting factor is.

Earlier design had the bias hotter, right at 100%. I've swapped cathode components from (shared) 120R resistor and 12V zener (and bypass cap) to 300R and 14V, because the model looked OK and the amp sounded just fine running cooler.

If there's no design flaw, and if this discussion doesn't indicate evidence of a failure somewhere else in the amp, I'll accept it and start looking at adding to my PA knowledge.

Thanks for the replies, I value all of your contributions here, to MEF, and the community in general!

[eschertron]

So I can see clearly from Nick's model that the EL84 starts to compress as it nears peak voltage. I guess my conceptual error lay in that I thought I would still see the clipping as a result of input headroom being reached. The light is starting to dawn on me now.

I'm sure this 'early' compression is part of the EL84 (and EL34) tube's charm. I'll have to work on a design that takes advantage of it.

I'm still fumed that I only get 11 clean Watts out of an 18W 'design'. Is it worthy of note that the OT is a Hammond 1760E rated at 15W? I figured I could exceed that number by a little, with impunity.

[Dave H]

I'm still fumed that I only get 11 clean Watts out of an 18W 'design'. Is it worthy of note that the OT is a Hammond 1760E rated at 15W? I figured I could exceed that number by a little, with impunity.

How are you deciding when the 'clean' output limit has been reached? The waveform I posted is well rounded on the peaks of the sinewave but it's still only 5% THD. You then need to measure it with a true RMS meter. Using a scope peak to peak reading will result in a low power calculation because of the compression.

[J M Fahey]

And who says itīs an "18W" design?
Marshall and other guitar amp makers?
Meaning their Marketing depts?

Only "printed" rating I believe is that on the tube manufacturers datasheets, who are *very* boring Engineers who have nothing to win and a ton to lose if they are not accurate; all others stand and be measured.

[Tom Phillips]

It's also a good idea to independently measure the actual resistance of your 4 Ohm load. Everything in your test setup must be within calibration limits for your ultimate measurement to be correct.
And ++ to what Juan said Re: "...who says itīs an "18W" design?"

Edit: Individual parts do not always live up to their advertised specifications. It's also a good idea to measure the actual turns ratio of the output transformer. It's another thing on the long list of variables.

[g1]

It's also a good idea to independently measure the actual resistance of your 4 Ohm load. Everything in your test setup must be within calibration limits for your ultimate measurement to be correct.

I'd also be curious what a bit of mismatch would do for the power levels.

[Chuck H]

That IS a pretty small OT (under two inches high and about two x three inches wide). It's also probably straight wound. Small core and the least efficient winding is going to mean poor LF. If you consider that it takes A LOT of power to make LF, well, that lack of efficiency could be where your lower measure is. But I wouldn't worry about it. My 2Xel84 with a Vp of 355 makes about 14W (Dave seems to be the winner for watts s far ) but there's no guarantee that all of the LF watts being measured by Dave and I ever get heard!!! Guitar speakers are only good down to about 70Hz. In other words, it's entirely possible your amp is just as loud with just as good balance of LF (for guitar) as mine or Dave's when your amp is plugged into an actual speaker cabinet.

EDIT: Perhaps try taking measurements with an input of 800Hz or 1000Hz. It's entirely possible that OT is rolled off a half a dB or more at 400Hz.

[J M Fahey]

Notice that nobody (so far) has posted an actual 18W amp ... at most almost 15W ... what the old Philips/Siemens/Telefunken datasheets promise

Now going to buy some coffee to stay the night, when I come back Iīll check the datasheets and post 2 maximum possible power designs, one with conventional grid voltage drive (and fixed bias low idle current of course) and the other a Music Man cathode drive type.

Not predicting anything, will post just what I find.

[Justin Thomas]

But, but, but, MY PLATES ARE DISSIPATING 12W <EACH>! IT HAS TO BE PUTTING OUT 24W OF POWER, right???

Eschertron, that's not a dig at you -- just a joke at the expense of the "plate dissipation = power out" crowd. I know you know better.

Justin