All I was saying is that if the SPL levels are the same, then the wattage as seen on the carver meter will be more or less accurate as representing the watts the Grundig is producing.
To do an actual measurement of watts output at the speaker terminals, you need to set a few more conditions, and this is typically done on a test bench.
1. Fit in a dummy load on the speaker terminals instead of an actual speaker. i.e., an 8 ohm resistor on the 8 ohm tap. The resistor needs to be beefy (usually a wirewound) so that it can handle the heat produced for prolonged tests.
2. Run a pure sine wave at say 1KHz through the amp.
3. Use a scope or a good true RMS meter to measure the voltage output across the load. If using a scope, convert the peak reading to RMS. (divide by square root of 2).
4. Calculate the power in watts delivered across the load by taking the RMS voltage value, squaring it, and dividing by the resistance value.
Note this is exactly the same procedure one would use with complex program material on an actual speaker load IFF you could measure the actual RMS voltage delivered across the actual speaker load. Real program material is a composition of many superimposed sine waves at numerous different frequencies, and a real speaker load is a combination of a pure resistance and complex reactance. It is represented as a vector that contains real (resistive) and complex (reactive) vector components. When plotting on a 2-D plane, these are represented as the X and Y components on a cartesian plane. Equivalently, you can represent complex impedances as amplitude and phase. Both representations are identical. The reactance portion of the impedance of a speaker is created from the capacitive and inductive components in speaker coil and crossover. You can get an idea of this complex load by looking at a specifications that hifi magazines sometimes publish for impedance of speakers. They are often cited as impedance-amplitude and impedance-phase plots from 20 Hertz to 100 KHz or so. In the amplitude/phase method of describing impedances, you will note these plots are rarely straight lines. They fluctuate, sometimes greatly, with frequencies across the load.
So the catch in doing this approach with real program material on actual speaker loads would be to use a true RMS meter to capture the actual RMS value across the complex load. But the process would be identical as I described above. To make this more reasonably doable on a layman's test bench, we replace the complex speaker load with an 8 ohm resistive load, and we replace the actual program material with a single sine wave at a given frequency. Then you can at least vary the frequency while keeping the load purely resistive, to give a glimpse into how the amp responds to changes in power at different frequencies across the load.
