Masking, Sones and BSR: Is this a new methodology? I have questions!

[Fredygump]

Context: I have recently tried to subjectively evaluate my speaker build, comparing it to previous stages/ iterations. I have measurements saying the new build is better, but I'm trying to understand how the improved measurements translate to what I hear. (Multiple subwoofer integration with a twist.) It looks good, but is it good? Is it better, or just different?

Then this morning this video showed up on my youtube recommendations.

Is this measurement of "SONES" established in the industry? If I'm understanding correctly, they're processing THD measurements and filtering out any distortion that is masked by the human ear. So they're trying to quantify perceptible noise and distortion.

Before this I knew that frequency masking is a thing, but I didn't know anything about the specific relationship between the "strength" of the masking effect vs the relative frequency of the sounds.

I'm interested in this, because it seems to explain my subjective analysis of my speaker project. My perception was that my "optimized" version sounded more detailed. Without optimization, the subs had more impact, but some details were obscured.

Applying this expanded idea of masking, it seems clear that a peak in the frequency response widens the range of frequencies that are being masked...pushing the masking curve up in dB level. My measurements show low frequency peak when I remove my subwoofer optimization, and these peaks, along with this idea of masking, would explain why I noticed some detail was missing when I disabled the optimization.

Am I getting this right? Is there a way we can access this type of tool? I feel like this could be like the holy grail of audio measurements? At the end of the video there were comments about getting rid of the all knowing "golden ear"...

 

[mhardy6647]

BSR?

source: https://www.worldradiohistory.com/Archive-Catalogs/Lafayette-Catalogs/Lafayette-1968-680.pdf

source: https://www.worldradiohistory.com/h.../HiFi-Stereo-Review-1987-12-OCR-Page-0129.pdf

 

[Fredygump]

Interesting, but not exactly what I was referring to.

The "BSR" stands for Buzz, Squeak, and Rattle. This comes from the automotive industry, and goes hand in hand with "NVH" (noise, vibration, and harshness).

The linked presentation is describing a methodology to identify what noise and distortion is audible. The presenter seems to be working in the automotive industry, but the methodology being described works for any audio system.

In the presentation he gave an example of a "good" speaker that had a lot of 2nd order distortion, but very little high order distortion...which gave it a high THD even though that distortion was not audible. To contrast this, he provides an example of a "bad" speaker that has much lower THD, but it had a lot of high level distortion that was audible. And he demonstrated how masking prevents us from hearing 2nd order distortion, but higher level distortion can be very audible.

Did I post in the wrong section?

 

[alex-z]

Not a new concept at all, just unpopular. Earl Geddes and Lidia Lee did a paper on perceptual weighting of distortion in 2003.

http://www.gedlee.com/Papers/Distortion_AES_I.pdf

http://www.gedlee.com/Papers/Distortion_AES_II.pdf

There is a corresponding GedLee metric you can use, unfortunately the only software I know that integrates it directly is Virtins Multi Instrument Pro.

The Lab Bench of 2021: Virtins Multi Instrument Measurement Software

Stuart Yaniger shares his more recent experiences with Virtins Multi Instrument Measurement Software, which he combined with his RME ADI-2 Pro FS R audio interface and the Linear Audio Autoranger II to create what he defines as the Lab Bench of 2021. This is a very complete solution for anyone...

audioxpress.com

 

[DonH56]

Sones are a standard unit for subjective loudness, akin to phons (also a measure of loudness). Commonly used for SPL measurements and comparisons, though most charts use phons converted to dB. See e.g. https://en.wikipedia.org/wiki/Sone

 

[Fredygump]

Not a new concept at all, just unpopular. Earl Geddes and Lidia Lee did a paper on perceptual weighting of distortion in 2003.

http://www.gedlee.com/Papers/Distortion_AES_I.pdf

http://www.gedlee.com/Papers/Distortion_AES_II.pdf

There is a corresponding GedLee metric you can use, unfortunately the only software I know that integrates it directly is Virtins Multi Instrument Pro.

The Lab Bench of 2021: Virtins Multi Instrument Measurement Software

Stuart Yaniger shares his more recent experiences with Virtins Multi Instrument Measurement Software, which he combined with his RME ADI-2 Pro FS R audio interface and the Linear Audio Autoranger II to create what he defines as the Lab Bench of 2021. This is a very complete solution for anyone...

audioxpress.com

Thanks.

Before I posted I searched for key words, and "sones" and "BSR" returned zero results. So it's almost like it was never mentioned before on this forum? Seems unlikely, but I guess my result correlate with "unpopular".

You seem to be speaking for the audio community as a whole, but can you say why it is unpopular? Is it that people don't like to think that there is a limit to their hearing and perception? Or are they afraid that engineers and designers will get lazy, and instead of perfecting gear, they just shoot for "inaudible"?

The concept seems to have a lot going for it, specifically for manufacturing and production. The presenter in the video is applying the concept to an automated measuring device used in automotive assembly lines that can pinpoint the location of audible noise and correlate the audio's location to a position on an image, so the issue can be quickly addressed. I'm not sure if it is just for car speaker systems, or if it is also for the regular BSR test they would normally conduct.

And he also mentioned that the company's founder did original research on the subject in 2017, I believe, and their process is based on that. I didn't try to find if that research is published anywhere...

 

[NTK]

As I understood from the video, the metric is general and should be applicable to home audio speakers. However, as far as I can remember, the HBK presenter didn't show anything about the relative sensitivity of the HBK BSR system. What I mean is that he didn't show side-by-side comparisons between different speakers' harmonic distortion FFT spectra and their BSR scores to let us know how sensitive the BSR score is and what type of resolution it can give.

What is very interesting to me is that the HBK scientist Dr Woo-Keung Song seems to have devised a method to extract the loudspeaker output but without nonlinear distortions with an arbitrary input signal (music in this case, see below screen grab, seen as the time-varying loudness value Nx). This enable the comparison between the speaker output with (Ny) and without (Nx) the nonlinear distortions, similar to how we extract HD information using the logarithmic sine chirps. The HBK presenter did mentioned that the method is currently kept as a HBK company secret.