Ideally you'd like your source signal to be as free of harmonics & noise as possible. That's what the HP8903B does, i.e., it's output signal has very low noise and THD so that when measuring a DUT the readings are attributed to the DUT. That just makes everything easier. But you could measure your source and determine it's harmonics. When you apply that signal source to the DUT, you could calculate (knowing its gain) how much of the harmonics at the output was due to the harmonics in the input signal. This is complicated to do by hand, but a computer could do it easily.
But have a pure signal source is the best of all possibilities.
But have a pure signal source is the best of all possibilities.
Hopefully I'll find time this coming Monday to look at this.
Like you say, the sine wave should be spectrally pure.
The DUT produces the harmonics, and THD is the ratio of the sum of the powers of all harmonic components to the power of the fundamental frequency.
Wish I had more free time to pursue this, I find it really interesting. Thanks!