Unexpected offset with arithmetic trace when calculating Signal-to-Noise Ratio

[Tatane_Wilson]

Hello all,

First of all thanks to John for a tremendous software (beats a lot of very pricy industry package I've used). Probably gonna get myself a pro license when I get around to it. Now to the matter at hand.

I am using trace arithmetic (A/B) to compute the SNR of two measurements I made (speech and noise) but when I do so, the response seems offset down by 80-90dB. See a screenshot below (I haven't applied any smoothing to remove the impact of that).

I am getting the same issue when I try to divide a spectrum by itself :

I get a flat line as expected but it should be on 0 dB, not -108dB(?).
I am using version 5.20.5 on windows 11. These measurements are imported wav files recorded with reaper and a steinberg ufc44 interface.

Am I doing something wrong?

Any help would be greatly appreciated,
Antoine.

ps : a few people here seem to have had difficulty interpreting this operation, particularly when the phase influences it. Maybe an option to do arithmetic operations on the magnitude of the spectra would be useful when one just want to "substract the dB" to get an attenuation or something of the sort.

 

[John Mulcahy]

There is an issue with the result of division of imported audio data which I'll look at.

You can make a magnitude-only response by selecting a single measurement on the All SPL graph and clicking the Average the responses button or using the RMS Average button.

As a workaround you could try importing the captures as impulse responses, but you will have to manually set the IR windows to cover the portion you want to analyse.

Edit : division of magnitude only responses made from imported audio data is OK.

 

[Tatane_Wilson]

Thanks so much for the prompt reply.
I tried using the RMS average and it worked like a charm.

Cheers!
Antoine.