Low output on JTM45 build

[mozz]

This is what I did. I expected most power at half the load impedance of the selected output impedance. However, this was not the case, measured output was roughly the same (~ 1 VRMS variation).

Did you calculate wattage or just measure and compare voltages? 10vrms across 8 ohms is 12.5 watts, if it's only 1 volt less across 4 ohms it's over 20 watts.

[tubby]

Did you calculate wattage or just measure and compare voltages? 10vrms across 8 ohms is 12.5 watts, if it's only 1 volt less across 4 ohms it's over 20 watts.

I had set my load always at 8R and would change (reflected) output impedance with just flicking the amp's impedance selector between 8R and 4R.
That would make masured Vrms voltages and thus output power directly comparable if I'm not mistaken?

[Helmholtz]

I had set my load always at 8R and would change (reflected) output impedance with just flicking the amp's impedance selector between 8R and 4R.
That would make masured Vrms voltages and thus output power directly comparable if I'm not mistaken?

With 8R load and the impedance selector at 4 Ohm, the Raa is doubled (13.2k) - not a desirable operation.
(Never flick the impedance switch at high output.).

You can compare voltages at the same load resistor. 10% more voltage means 20% more power.

[tubby]

With 8R load and the impedance selector at 4 Ohm, the Raa is doubled (13.2k) - not a desirable operation.
(Never flick the impedance switch at high output.).

You can compare voltages at the same load resistor. 10% more voltage means 20% more power.

Sorry, that was a typo. As you pointed out that mode of operation makes no sense. I meant to write flicking the amp's impedance selector between 8R and 16R!

I would power the amp down in between switching, so I wasn't hot-switching.

[pdf64]

...And from what I understand power output rises with increasing bias current. At least up to a certain point and provided the power supply can keep up.

Power output is determined by peak anode current rather than its idle current.

[g1]

I'm quite surprised changing the impedance tap only affected full power out by 1V. Did you readjust the signal level so you are just at clipping in both cases?

Idle current (bias) will increase or decrease output for a given signal, but max. power should remain the same as it is limited by the supply.

[Helmholtz]

At least in theory hotter bias may somewhat increase clean output, i.e. as long as B+ doesn't drop significantly.

I considered that effect when I restricted my formula above to class B.

Comparing the loadlines for class B and class AB with the same B+ shows that the usable length of the loadline increases with hotter class AB.
That means larger available voltage swing, thus more power.
This effect increases with larger Raa.
Another way to look at it is that higher idle current shifts the loadline upward. This increases the area of the triangle below the accessible part of the loadline.
And output power is proportional to that area.

Also hotter bias increases power tube gain and requires less grid drive for full output.

In reality the operating point tends to shift towards class B at large grid drive, so the difference between class AB and class B should be small.


Edit: My explanation above is a bit tentative. Need to think about it some more.

[pdf64]

I take your point, but practically, my perception is that with valve guitar amps, if we're playing loud, there's likely to be at least some degree of power amp clipping. And unlike hifi or perhaps PA, that's fine, the speakers we use add harmonics, so a bit more from the amp isn't a problem per se. Hence I don't think that the linearity of high power sine waves is really significant, so the x watts at y % THD way of assessing power output isn't very helpful, eg amps with an NFB loop should stay cleanish right up to clipping, whereas those with little or no NFB may have far more distortion by the point of clipping, and would have passed a nominal 5% THD perhaps some way before they reached clipping.
I think it's better to note the Vp at clipping, and if necessary for comparison purposes, derive a equivilant RMS from that (ie multiply it by 0.71), rather than use actual VRMS at 5% THD, or even VRMS at clipping. Because distortion can reduce or increase the VRMS with respect to the Vp.
Vp seems the only reliable metric by which we can compare say an AC30, JTM45 and a 5150.

[Helmholtz]

.....so the x watts at y % THD way of assessing power output isn't very helpful, eg amps with an NFB loop should stay cleanish right up to clipping, whereas those with little or no NFB may have far more distortion by the point of clipping, and would have passed a nominal 5% THD perhaps some way before they reached clipping.
I think it's better to note the Vp at clipping, and if necessary for comparison purposes, derive a equivilant RMS from that (ie multiply it by 0.71), rather than use actual VRMS at 5% THD, or even VRMS at clipping. Because distortion can reduce or increase the VRMS with respect to the Vp.
Vp seems the only reliable metric by which we can compare say an AC30, JTM45 and a 5150.

Not sure what you're referring to, as no one mentioned THD.

Vp can be used for power calculation, but only if the signal is a sine. If not, power results will be wrong.
E.g. with increasing clipping Vp stays constant while real Vrms and power increase.

There's no general relation between Vp and power.

BTW, what power output did you measure with your JTM45 build?
I plan to take some measurements on my JTM50 in a week or so.

[tubby]

At least in theory hotter bias may somewhat increase clean output, i.e. as long as B+ doesn't drop significantly.

On that note...Here's a nice little video which shows exactly that in a practical application (Fender DR).
Includes actual measurements!

You may want to skip to 41:35

[tubby]

I'm quite surprised changing the impedance tap only affected full power out by 1V. Did you readjust the signal level so you are just at clipping in both cases?

That's exactly what's bugging me the most at the moment and what I can't get my head around!
I can live with a clean power output of "only" 25 Watts or whatever for a JTM45 build with a 6,6k primary and the voltages measured above. This might all be attributed to measurement accuracy, parts quality, transformer loss, etc.

But halving the reflected impedance to 3,3k, those voltages and just +1Vrms more at the onset of clipping contradicts theory and the datasheets a little bit too blatantly.

[Helmholtz]

But halving the reflected impedance to 3,3k, those voltages and just +1Vrms more at the onset of clipping contradicts theory and the datasheets a little bit too blatantly.

Do you know this great calculator: http://bmamps.com/ivds.html ?
It lets you see the effects of different Raa and supply voltages.

It should be noted that a dedicated 3.3k OT can be expected to put out more power than a 6.6k OT with halved secondary load.
Reason is different OT efficiencies (losses) caused by different DCRs.
A 6.6k OT requires more turns of thinner wire than a dedicated 3.3k OT.

If a 6.6k OT has an effective loss resistance of 10% of the Raa, the percentage will double with halved Raa.
This means that OT efficiency drops from 90% to 80%.

[tubby]

Do you know this great calculator: http://bmamps.com/ivds.html ?
It lets you see the effects of different Raa and supply voltages.

It should be noted that a dedicated 3.3k OT can be expected to put out more power than a 6.6k OT with halved secondary load.
Reason is different OT efficiencies (losses) caused by different DCRs.
A 6.6k OT requires more turns of thinner wire than a dedicated 3.3k OT.

If a 6.6k OT has an effective loss resistance of 10% of the Raa, the percentage will double with halved Raa.
This means that OT efficiency drops from 90% to 80%.

Yes! I've been using this calculator from time to time, in addition to this one:

https://www.vtadiy.com/loadline-calculators/loadline-calculator/

I like the graphs on the latter a little bit better since they're from the actual data sheets.
Between those two calculators, a lot of tubes are covered.
They both have their inaccuracies but definitley can help in designing a tube amp.

Thanks for hinting at increased transformer loss with mismatched reflected impedances!
As mentioned in post #30 I've already tried another OT with a dedicated lower primary that didn't yield more
rms voltage output at onset of distortion which makes me wonder if I'm still missing something here.

[Helmholtz]

I like the graphs on the latter a little bit better since they're from the actual data sheets.

What do you think the other calculator uses for graphs?

Just noticed that the vtadiy calculaor shows a wrong power limiting parabola with the KT66 curves (30W instead of 25W).

Still having a hard time believing your results.
Your full power plate and screen voltages look rather high for the measured output.

Is your DMM accurate and the battery fresh?
Did yu measure screen voltages directly at the tube pins?

[tubby]

What do you think the other calculator uses for graphs?

Just noticed that the vtadiy calculaor shows a wrong power limiting parabola with the KT66 curves (30W instead of 25W).

Still having a hard time believing your results.
Your full power plate and screen voltages look rather high for the measured output.

Is your DMM accurate and the battery fresh?
Did yu measure screen voltages directly at the tube pins?

I'm sure the bmamps calculator also uses data from the tube datasheets, what I meant was that the
actual graphic representation of the curves in the vtadiy calculator look closer
to what I'd expect from a data sheet, eg the diode line is clearly visible.
Well, it's probably just a matter of taste and won't matter for calculating purposes.

I do believe my DMM is accurate and the battery is fresh. Just to be sure I've borrowed another DMM
from work and the readings were within +/- 1V.
Plate and screen voltages were measured directly at the corresponding pin of the tube sockets.

Thanks for bearing with me so far y'all!
At this point, I don't know what else to do to get to the bottom of this issue.
However, I'd REALLY like to understand what is going here since the results I'm getting
seem to contradict theory and some experiences from other builders.

So, if there's anything else I can do from my side (e.g. measurements, oscilloscope screen grabs, pics) to maybe sort this out, pls let me know what you need.