Princeton build with mod

[Helmholtz]

Measuring power output with clipping doesn't make much sense.
Results are neither reliable nor comparable.

Once the output voltage hits the clipping threshold, its peak value is limited, so can't increase any further.
With even stronger drive signal the flattened portion of the output increases and so does the area below the curve.
This in turn makes the output RMS voltage and power increase until the signal finally looks like a squarewave.

The power produced by a perfect squarewave is exactly twice the power of a sinewave of the same peak value.
So the theoretical output limit with clipping is twice the power measured at the onset of clipping.
In reality it will be somewhat less, mainly due to power supply sag.

In summary, clipped power depends on drive level and can vary between max. clean power and almost twice that number.

[Tom Phillips]

Measuring power output with clipping doesn't make much sense.
Results are neither reliable nor comparable.

Once the output voltage hits the clipping threshold, its peak value is limited, so can't increase any further.
With even stronger drive signal the flattened portion of the output increases and so does the area below the curve.
This in turn makes the output RMS voltage and power increase until the signal finally looks like a squarewave.

The power produced by a perfect squarewave is exactly twice the power of a sinewave of the same peak value.
So the theoretical output limit with clipping is twice the power measured at the onset of clipping.
In reality it will be somewhat less, mainly due to power supply sag.

In summary, clipped power depends on drive level and can vary between max. clean power and almost twice that number.

All true.
However, measuring the the clipped response of a guitar amp is useful to me because that is one of the modes that actually happens when many players use the amp and sometimes, strange behavior shows up. The clean power is always measured first and I don't "cheat" by specifying the power at 5 to 10% THD like Fender does. Recording both numbers for the customer also solved the problem of them feeling that their amp was under performing when they compared the number to the advertised manufacturer's power rating or the internet lore number. One example of internet lore is that every amp the used two 6L6 power tubes is a 50W amp.

[Helmholtz]

However, measuring the the clipped response of a guitar amp is useful to me because that is one of the modes that actually happens when many players use the amp and sometimes, strange behavior shows up.

I did not say or mean that scoping the clipped output is useless to the expert.
It's just that clipped power values depend on input signal and controls' settings and are not comparable between different amps.

I don't cheat" by specifying the power at 5 to 10% THD like Fender does.

Quite the contrary, I think that specifying the THD is a serious attempt to better define the measuring conditions as something like "before clipping" or "at the onset of clipping" is all quite vague.
BTW, something like 5%THD doesn't require clipping and might actually be almost invisible. Just takes a little compression or expansion.
Tubes are inherently non-linear and each preamp stage may add something like 1%THD even when set for "clean".

To avoid preamp effects I inject my signal to the PI input.
This also allows for testing with square and triangular signals as the frequency response of the power amp is essentially flat.

[The Dude]

BTW, the difference between 20W and 16W is less than 1dB.

^^^^^^ That! Spending money and time only to gain a few watts is futile, IMO. The difference will be hardly noticeable. Also, one of the easiest and most noticeable changes would be swapping to a more efficient speaker (of course, this assumes the existing speaker efficiency is not that great).

[Chuck H]

The way I've always figured it for "clean tones" from a guitar amp, they're seldom really that CLEAN. So I usually figure amp output by using the scope to see where the waveform starts to flatten/clip and then back off to where it's rounded again. I determine RMS based on that. Which seems entirely appropriate for a guitar amp. The actual distortion content would be way beyond any considerations for something like a PA or reference amp. But that's not the job at hand with a guitar amp and "I" think a much higher distortion content is not only acceptable, but often preferred. No players I know like a flat, bland clean tone.