@amirm
RMS across any selected frequency range (like 20Hz-20kHz or whatever you want) is shown automatically in REW. Depends only on operator knowledge and skills.
Neither of which you are demonstrating seeing how you are not addressing the points I made. So let me explain it to you.
A voltage value could either be extracted using time or frequency domain. If it is the latter, which would make sense given the core functionality of REW, then it is subject to different sources of error. Key is compensation for Windowing used for FFT and spectral leakage as I mentioned. The former can be implemented as an approximation but the latter is left up to the user to avoid. Which means you have to pick frequencies that land in a single FFT bin -- something I am sure many don't know and that likely includes you. And good luck to you if you attempt to use noise as Josh mentioned. See:
https://www.audioprecision.com/news/fft-scaling-for-noise
"In Audio Precision analyzers, FFT spectra are scaled so that the amplitude axis gives the correct reading for discrete tones. This makes sense, because most often we are characterizing the performance of a DUT using sine waves. The amplitude axis, however, cannot be used to indicate the level of a spectrally non-discrete signal, such as noise, without first applying a correction factor that depends on the FFT bin width and the window used."
A single frequency tone that doesn't land in a single bin is also non-discrete.
I am actually reviewing an audio analyzer that gets the above wrong creating large amount of error.
Even outside of the above, you have to show that across the full dynamic range of the interface, accuracy is maintained.
This is all very different than how Audio Precision and multimeters work. They are operate in time domain so not subject at all to above error sources. And their accuracy is checked, specified and documented.
In a pursuit to see if voltages are the same, you need to use the same, accurate voltage measurement scheme. If you attempt to use these different methods, you can easily get wrong comparable results.
Yes, if you don't care about absolute values, you could use CHIRP/FFT based analysis but Josh's write clearly indicates a mix of these methods and hence my comment.