Acoustic Loudness - Is Scaling Required?

Hello all - hope you are all well?

I'm looking for some help using the built-in Matlab command for calculating acoustic loudness (AcousticLoudness) and other subsequent psychoacoustic metrics.

We capture acoustic data using Dewesoft measurement hardware and software - this raw data is then exported in .mat format. The raw data is unweighted Pascal values at 50kHz sample rate.

We want to complete the analysis using Matlab - however when using the built-in AcousticLoudness function, the results we get are ~2x higher than the results we get using Dewesoft and other analysis software (including SQAT which runs in Matlab).

I'm wondering if there is some scaling required - however this seems odd as the input data we are providing is physically measured Pa values (rather then us trying to infer the correct values).

I've included an example below to illustrate - same input data is used in all cases, and the same calculation parameters ("Method","ISO 532-1","SoundField","diffuse","TimeVarying",false).

Dewesoft result: 210 Sone

SQAT result: 210 Sone

Matlab result: 452 Sone

I hope the above all makes sense, but please let me know if not. I'm hoping someone can help us solve this issue :)

Andy

[andytodd.msp]

Just to add some further information on this - we have input a 1kHz tone at a level of 40dB (which should result in a loudness result of 1 Sone).

The matlab AcousticLoudness function returns a result of 1.9 Sone.

Dewesoft and SQAT both return the correct value of 1 Sone.

I have attached the .mat file in case it is of help!

Andy

[Mario Malic]

Apparently, Dewesoft uses A-weighting curve to calculate loudness. You might have to weight your sound first and then calculate loudness with mentioned MATLAB function.

https://forum.dewesoft.com/software-discussions/dewesoft-software/dewesoft-x3-loudness-in-sones-vs-decibel-levels

[andytodd.msp]

Hi Mario - thanks for the reply.

If you look at the example I posted in my second post, I dont think the A-weighting curve would make any difference (as the A-weighting at 1kHz in 0dB.

Hope that makes sense.

Andy

[Mario Malic]

Yeah, I missed out on that it's a single tone on 1kHz. Unfortunately, I can't think of anything else.

Answer 1

[Jimmy Lapierre]

Are you using audioIn and FS (the audio signal itself) or SPLin (28 one-third-octave bands between 25 Hz and 12.5 kHz) ?

If you have the recordings (not the 1/3 SPL levels), you can use a 1kHz tone and the calibrateMicrophone function to get the right calibration level (the 3rd input parameter of acousticLoudness).

Answer 2

[andytodd.msp]

Hi all, thanks for the input.

Yes the issue was related to the calibration factor. As we are inputting Pa values, using a cal factor value of 1 results in the correct result.

Thanks all!