Separate/different simultaneous output signals for L and R.

[trobbins]

Jan, perhaps not important for your testing scenario, but be aware that the single tone generator in REW can have its harmonic levels manually adjusted (using the Adjust harmonic levels tickbox). In a typical loopback scenario around a soundcard, adjusting the harmonics can realise an RTA response with harmonics suppressed into the noise floor. Although that pedantically doesn't mean the output of the soundcard has similarly suppressed harmonics, it does a pretty good job of making a distortionless signal source if the soundcards ADC and front end is blameless. I have tried that scheme to include an external power amplifier, but not to include a speaker and mic, although I'd anticipate if the speaker harmonics were dominant in that loop then they may be able to be similarly suppressed so as to derive an audible speaker signal with lower harmonics.

Perhaps another possible concern is that REW generates a tone frequency that may not be 'exactly' at the chosen Hz, especially if the 'Lock frequency to RTA FFT' box is ticked.

 

[trobbins]

One possible alternative path to derive separate signals is to use an existing dual tone signal from REW, and then split the signal into two paths, each path with a notch at one/other of the IMD tones. Recombining the two paths and making a loopback RTA may indicate how well that baseline approaches the original dual tone, or perhaps be useable as a calibration response to then use when the two paths go through separate speakers and one MIC. Suitable notching may be the devil in disguise I suspect, but using the raw power of REW and RTA IMD measurement, and using the RTA spectrum to separate the wood from the trees, may have its benefits too.