assessing the linearity of an output stage

[steph_tsf]

hello,

I must be the zillionth asking if it is valid to compare apples with peers regarding distorsion figures.

Say that using LTspiceIV, you have designed an audio power amplifier equipped with an exotic output stage arrangement like an elaborate CFP. You know it is "nearly linear" when fed by a current, and "less linear" when fed by a voltage. You are too lazy to develop a dedicated text fixture under LTspiceIV, and you are intelligent enough to know that testing a component without the interactions of the device feeding it, may result in wrong conclusions.

So, you end up asking yourself what is distorsion inside an amplifier, and asking yourself how to measure distorsion inside an amplifier using LTSpiceIV.

I'm coming with an idea, and would like to know if there are people having tried it already.

Say you amp is delivering a THD of 0.000250% using a 100ms .TRAN under LTspiceIV, after 50ms for the settlement of the circuit. With all options enabled for an accurate result like double precision maths, no waveform compression, and 50ms settling time, before actually taking the 100ms .TRAN sample. The 0.000250% is is the distorsion that you read in the log after asking .FOUR (1kHz). The distorsion is measured at the output, 50Vpp, 1kHz, loaded by a 8 ohm resistor.

Put your voltage probe at the input of the output stage. Say you read a distorsion, over there, equal to 0.002500% with a voltage of 50V pp.
Put you current probe at the input of the output stage. Say you read a distorsion, over there, equal to 0.003500%, with a current equal to 1mA pp.

Now you test another output stage, with the amp delivering nearly the same distorsion at the output : 0.000195%, but that's not exacly 0.00250% from before.

Put your voltage probe at the input of the output stage. Say you read a distorsion, over there, equal to 0.002500% with a voltage of 50V pp.
Put you current probe at the input of the output stage. Say you read a distorsion, over there, equal to 3.5%, with a current equal to 10µA pp.

How to compare the first output stage with the second output stage ? Are there figures to be computed, like the square rooth of the squares, weighted by the current amplitude requirement ?

And how to correct by the effective distorsion level at the output ?

In fact, the aggregated distorsion value I'm searching for, is an objective measure of the stress the output stage puts on the rest of the amplifier, for counteracting the non-linearities, both in terms of current, and voltages, in an absolute way, and not in a relative way. If the drive current is a few microamps, and the THD of the drive current equal to 3.5%, I guess that the stress for the internals of the amp is 10x less than a few milliamps at the drive level, with a said-to-be nice THD equal to 0,035%.

You see what I mean ?
And then, combining the absolute voltage non-linearity stress with the absolute current non-linearity stress, in a proper way.

Maybe it is stupid trying to mix apples with peers.
Maybe it is mandatory to know the two parameters, individually.

Any idea welcome.

Steph