MDAQS For Beginners With Jacob Soendergaard From HEAD acoustics
Feb 29, 2024 at 12:08 PM Thread Starter Post #1 of 10
Ask two Head-Fi'ers to name their favorite headphones, and you may get two very different-sounding headphone models. This is normal within our community. As we know, it's also common for a single Head-Fi'er to equally love multiple headphones with vastly different sound profiles. These examples highlight why evaluating headphones solely based on their frequency responses can be too reductive, and sometimes even misleading. Human hearing and perception are incredibly complex, and that's where HEAD acoustics excels.

As HEAD acoustics notes, audio quality extends beyond linear frequency response. Our perception of sound subconsciously considers many other dimensions. MDAQS (Multi-Dimensional Audio Quality Score) addresses these complexities, evaluating the key attributes that shape overall quality. This creates the world's first binaural perception-based measurement tool, offering a scientific and quantifiable way to assess device audio quality.

We previously interviewed Dr. Hans W. Gierlich from HEAD acoustics about MDAQS, and that discussion was at times technical and complex for a first look. So we invited Jacob Soendergaard from HEAD acoustics to visit Head-Fi HQ so that we could have a discussion more at the introductory level. Here is that discussion:

 
Feb 29, 2024 at 7:25 PM Post #3 of 10
Jude,

Please do a deeper dive on timbres, a bit on how it's measured, and MDAQS scores that critical data point.

The lack of “subjective” measurements of how a headphone performs in reproducing timbres is one of the reasons measurements have low value in my personal evaluation of headphones. If it doesn't sound like music, it's game over, but then we get into the variances of experience listening to live, mostly amplified music.
 
Mar 11, 2024 at 8:55 AM Post #4 of 10
Sorry this MDAQs thing is useless. The HD 800S scores 3 in immersiveness, that is supposed to indicate soundstage and imaging, the NDH 30 scores 2.8 , while IEMS score way above these two, which is because they picked random 100 people to produce the data, on average a person is more likely to use iems and they probably set their preferences based on the iems which got fed into the model.

Sorry MDAQs does not bring back the human factor in measurements, but quite the opposite, a human (with a minimal interest in sound quality) would term the above headphones to be some of the most immersive ones while traditional measurements would point out their flaws in frequency response. Sorry the RTINGS rating gives much more data.

For 6 years of research data 100 random people is just too small. EDIT (My mistake it is 250 , i probably read the number 100 in a forum)

https://www.soundguys.com/sennheiser-hd-800-s-review-88114/
https://www.soundguys.com/neumann-ndh-30-review-111124/

Compare the scores regarding immersiveness with those of IEMS (I did not mark the red circle, it was from a forum)

1710162085340.png
 
Last edited:
Mar 12, 2024 at 11:49 AM Post #5 of 10
Sorry this MDAQs thing is useless. The HD 800S scores 3 in immersiveness, that is supposed to indicate soundstage and imaging, the NDH 30 scores 2.8 , while IEMS score way above these two, which is because they picked random 100 people to produce the data, on average a person is more likely to use iems and they probably set their preferences based on the iems which got fed into the model.
Hey @Trinkle, I feel like you're making some assumptions here, that I can maybe help clarify.

First of all, I don't know how you ended up with the a sample size of 100.
- I'm not sure if we have actually posted the total sample size in writing, but throughout the various videos HEAD acoustics have released or participated in, I believe we reference that n = ~250 people.

Secondly, I'm not sure how you can make the claim that the "average person is more likely to use iems".
- If you look at numbers like sales volume, TWS products are probably the most popular and wide-spread. But to me, iems do not mean TWS. And their general tuning profiles also tend to differ drastically. But this might be a matter of semantics.? Also, a lot of people listen to music in their cars or on home stereos or small BT speakers. Making the claim that "they probably set their preference based on iems" is a bit of a leap.
The people that participated in the listening tests, were all "average consumers" with normal hearing and no special training in audio. Doesn't mean they aren't audiophiles, just means they didn't have official training in audio. They were presented with 90 different comparisons where each comparison was to rank if they preferred sample A or sample B on 4 different parameters (Timbre, Distortion, Immersiveness, Overall Quality). That's 90*4*250 = 90,000 data points for the algorithm to work with.

Also, the definition of Immersiveness is roughly an indication of soundstage and imaging.
- For your reference the people in the jury study were briefed to pay attention to the feeling of being enveloped by the sound, being able to localize instruments, hearing sound coming from different directions, etc. They were not told, for instance, that a wider sound stage is always better.

AFAIK, the HD800S are generally considered a great headphone amongst this audience, but I think it's fair to discuss how the average consumers would appreciate its sound. They're not really Harman-ish and their lack of bass is very likely to result in overall lower scores across the board. The SoundGuys review you reference sums it up similarly.

To flat our dismiss MDAQS as useless because you selected two example headphones that are rather unique seems a little premature. However, I would love to have you (and this community) participate in the jury study, so we can capture your preferences and also see if you can beat the average amount of circular triads (A>B>C>A). No pressure 😉

Either way, the goal is to provide a slightly more holistic perspective on Audio Quality. I'm a big fan of the Sound Wheel by FORCE Technology (download for free here: https://forcetechnology.com/en/articles/gated-content-senselab-sound-wheel) when talking about sound, and MDAQS tries to capture more than just frequency response and timbre.

Hope that helps address some of your concerns and questions. And maybe gets you interested in how MDAQS works and can be interpreted - rather than dismissed.
 
Mar 12, 2024 at 2:05 PM Post #6 of 10
Hey @Trinkle, I feel like you're making some assumptions here, that I can maybe help clarify.

First of all, I don't know how you ended up with the a sample size of 100.
- I'm not sure if we have actually posted the total sample size in writing, but throughout the various videos HEAD acoustics have released or participated in, I believe we reference that n = ~250 people.
Thanks a lot for your detailed and patient answer.

I am sorry, I have written my comment in a dismissive way (Its easy to let loose on a forum and I am guilty of that). I probably picked up the sample size of 100 from some forum which was not too responsible of me and I fixed it in my comment. But the thing is the few headphones Soundguys measured using MDAQs all gave numbers which I am sure will be highly disputed by majority of headphone enthusiasts, you can look at the numbers the HD 490 pro scored too.

https://www.soundguys.com/sennheiser-hd-490-pro-review-111184/

The thing is if HD 800, NDH 30, and even HD 490 score low on timbre it is acceptable, but if they all score much lower than multiple IEMs in terms of immersiveness then it should raise some eyebrows. It is only a sample size of 3, but note Soundguys only measured a handful of headphones with MDAQs, so the percentage is too high. If you look at RTINGs, they do have some form of measurement for soundstage and imaging, they may not be super accurate but more or less the numbers they provide align with how a human would feel. It would not give open backs like the mentioned ones such low scores in soundstaging/imaging. But then again to be fair you mentioned immersiveness only roughly relates to soundstaging and imaging. But my point is then what indicates what we care about .... soundstaging and imaging when we speak of immersiveness ?

I think it has to do mostly with the input data with which the model was built and am pretty sure if it was built with people from HeadFi it will be very very different. At least the HD 800 and NDH wont tank in immersiveness. I have a guess (may not be correct), that the bulk of work is in the algorithm and so it would be easy to integrate new data to your model and hope you could involve communities like HeadFi to provide the input. I hope your model keeps on evolving with new input data.

And again I apologize for the dismissive comment and thanks for your answer.
 
Last edited:
Mar 12, 2024 at 4:55 PM Post #7 of 10
When I saw immersiveness I imagined that isolation would rank high. Along with just having more bass as that's a fair way to feel some sound being sort of everywhere around. I also considered that I might want to rate higher a small wireless earbud that's easier to forget.

You on the other hand read immersiveness and thought HD800. I sort of understand where you're coming from. But it clearly didn't match my idea of it, and even now, I still wouldn't consider the HD800. If we were to talk about 3D simulation, as it's a pretty clean headphone, then I would put it high in the immersiveness rankings.

How many people also had a different interpretation, or just different priorities or impressions? IDK. Perhaps a model made from asking many people could tell us? :stuck_out_tongue_winking_eye:
I have a hunch, that when thinking statistical average Joe's point of view, perhaps the accepted norm for elite audiophiles isn't it? I would be interested to find out how deviants we are(if we are). I suspect we are for several reasons, starting with a culture of refusing the common man's audio gear as being enough.
 
May 26, 2024 at 6:31 PM Post #8 of 10
@jude

Thank you for this video. I want thanking you specially for the types of questions you asking AND for really trying, as much as is possible, for keeping this complex topic more understandable. You conducting a very, very good interview, asking very relevant questions.
 
Last edited:
Sep 16, 2024 at 5:52 PM Post #10 of 10
Watched all of the videos linked about MDAQS. It seem like an interesting way to calculate a preference score without humans having to test and compare each headphones. Is there a list of headphones rated by MDAQS anywhere? Seem like this product has been released for a couple of years now?
 

Users who are viewing this thread

Back
Top