Am I using REW RTA to measure SNR correctly?

[microbailey]

Hi. Electronics experience but relative REW newbie here so go easy

I'm using REW with a Native Instruments soundcard to make some connected measurements of audio equipment I am designing and refining.
My Soundcard OUT plugs into DUT IN (device under test), Soundcard IN from DUT OUT. REW on Windows is being used to make the measurements.
I've calibrated the REW Signal Generator for RMS voltage using a bench AC voltmeter on True RMS
REW SPL meter has been calibrated for 0dB using looped-back 1kHz sinewave generated by REW at 0dBu (the level I've chosen to use for my measurements).

Frequency sweeps, gains etc are all fine. I'm now attempting to measure A-Weighted SNR.

After calibration my process is:

• open the REW Generator popup and play a 1kHz sine at 0dBu
• set gain on the DUT to unity using the previously calibrated REW SPL meter
• open the REW RTA window, press REC button and observe the 1kHz tone at 0dBu + noise floor (including harmonics etc)
• enable REW RTA "Show distortion" popup (top left corner of RTA window)
• clear down the RTA plot by toggling REC off and then back on
• after a few secs disable the REC to leave a 'watermark' of noise (this time window seems finger-in-the-air)
• read A-weighted “SNR” value directly from the value in the REW "Distortion" popup

I have assumed in the above that REW has calculated the RMS value of the noise floor, subtracting the 1kHz tone, and A-weighted it.

• I don't know if using a time window with RTA is the right thing to do.
• Did I understand the REW report in "Distortion" popup correctly?

 

[John Mulcahy]

REW's SNR figure is unweighted, and the N component excludes harmonic distortion. There is an A-weighted N+D figure that is the level of noise + all distortion but that is an absolute figure in the current graph Y axis units, not relative to the signal.

 

[microbailey]

Ah thanks for clarifying that John - i didn't realize that.
So it sounds like my basic use of the tool is correct.

It's just I can't directly read off an A-weighted SNR figure.
If I understood you correctly I could, however, read off the absolute A-wieghted N+D figure if I set the RTA Y-axis to dBu (to align with my reference) and then calculate the ratio against 0dBu? In that case I guess I'd quote my SNR as X dB(A) @ 1kHz 0dBu.

 

[John Mulcahy]

With the axis on dBu the figure could be used directly since your test signal is 0 dBu, but note that the SNR definition excludes harmonic distortion so the result would be worse than an SNR figure unless all the harmonics were below the noise floor. You can apply A weighting via the mic cal file by loading an inverse A weighted response, there are example files in other posts on weighting.

 

[microbailey]

Thanks for the suggestion re mic cal file.
Yes of course you're right about the SNR figure. In my case the 2nd and 3rd harmonics sit well above the noise floor.
Really appreciate your support.