AES 2025 Paper: New targets for the B&K 5128 GRAS 45CA-10

[MayaTlab]

n.b. can anyone explain what the "_7" and "_20" means in the graphs and whether the GRAS system Sai uses at unhearldab matches any tested?

"_7" means the average transfer function found after measuring the seven headphones mentioned in this study between the different fixtures.
"_20" means the average transfer function found after measuring 20 headphones between these fixtures, done in a previous study presented on B&K's Youtube channel in 2021 :

When neither "_7" nor "_20" are mentioned, only the headphones under test were used to derive the transfer function, hence the very similar scores.

 

[MayaTlab]

There is nothing remotely racist to state the fact that Danish population was used to develop 5128. They should have spot checked rest if the world as a minimum to quantify differences.

We have discussed this ad nauseam already, but let's repeat it : this argument is moot since the original data that lead to the development of the 711 coupler came from four ears from two Danish males only. Besides we simply don't know the ethnicity of the test subjects in B&K's study and it isn't like the Danish population is exclusively composed of caucasians. If the results were biased, and they could be, it's at least miles better than the data from which the 711 couplers were made from. And we've had since then a few studies using various means to estimate the response at the eardrum showing similar trends anyway in other regions. If there is a bias, it is, in all likelihood, not as important as how unrepresentative of the average ear canal 711 couplers are.

 

[Robbo99999]

Yes the GRAS Welti and KB5000. I think amir earlier did say he was considering a boost to 3-8kHz. For the other region I interpreted his description of "either the same or very small difference" to mean within the limits of measurement error. I personally agree and would not worry.

n.b. can anyone explain what the "_7" and "_20" means in the graphs and whether the GRAS system Sai uses at unhearldab matches any tested?

It's certainly not within measurement error because the study showed a difference in the 1-3kHz area on all the headphones (apart from Sennheiser 518 but even that one if you didn't align at 1kHz would be below in that area slightly because you may well choose to line that headphone up in the bass because showing some difference in bass, but yeah even then not much difference between 1-3kHz but that's just one of the seven, the other 6 headphones show less energy between 1-3kHz). Of course it's up to Amir how he incorporates the results of this study but I think both areas should be included, and like I've said before the depression between 1-3kHz is synergistic with an increase in 3-8kHz in terms of balancing tonality - if you just increase the 3-8kHz area then increased chance of headphone being too bright.

 

[amirm]

We have discussed this ad nauseam already, but let's repeat it : this argument is moot since the original data that lead to the development of the 711 coupler came from four ears from two Danish males only.

Proving that the path to better predictivity of measurements is NOT to chase more participant scans in the design of a new fixture. Excellent results were achieved with the modified 711 coupler by Harman over a decade of research. Yet no one has shown any benefit to using 5128 in the same regard. At best, it is kind of close to 711. At worse, it is highly variable resulting in less accurate predictions.

If the goal was to improve predictive power of the measurements using the new fixture, B&K should have conducted similar studies to Harman to demonstrate that. Without it, it is a hypothesis which has been demonstrate to fair extent to have been wrong.

 

[amirm]

Besides we simply don't know the ethnicity of the test subjects in B&K's study and it isn't like the Danish population is exclusively composed of caucasians.

This should have been the first thing documented in the white paper for 5128! That its most fervent promotors of the fixture don't even know what the make up was, is quite damning. How could you run with this being the major benefit of 5128 and key to its "accuracy" when you don't even know how the population sample correlates with the rest of the world?

And this is not the half of it. From what I recall, they discarded a few of the participant scans because they didn't fit the average model they wanted to build!

This notion of using more of the population to get better accuracy reminds me of the first time we bought a house with a lawn. I bought a new Honda lawnmower and got a bag of fertilizer. Put the bag in a spreader and it poured out like nobody's business. I stopped and went online and asked if that was OK. Folks said to run and pour water on it before it burns the grass. I did but by then it was too late and I had a nice wide row of dead grass right in front of our driveway! This is what we have with 5128. A fixture that costs three times as much as GRAS 45CA yet has not shown any benefit whatsoever for consumers in any formal study.

 

[listener650]

There is nothing remotely racist to state the fact that Danish population was used to develop 5128. They should have spot checked rest if the world as a minimum to quantify differences.

Not at all, it was more that he was (seemingly) intentionally ascribing an argument to me that I didn't actually make—I did not claim that he was hearing things differently than the graphs because of "mongrel genes."

Regardless, I've worked to put it all behind us and I hope Mark can do the same, because I do genuinely value his input in the discourse.

 

[markanini]

We have discussed this ad nauseam already, but let's repeat it : this argument is moot since the original data that lead to the development of the 711 coupler came from four ears from two Danish males only. Besides we simply don't know the ethnicity of the test subjects in B&K's study and it isn't like the Danish population is exclusively composed of caucasians. If the results were biased, and they could be, it's at least miles better than the data from which the 711 couplers were made from. And we've had since then a few studies using various means to estimate the response at the eardrum showing similar trends anyway in other regions. If there is a bias, it is, in all likelihood, not as important as how unrepresentative of the average ear canal 711 couplers are.

If the issue matters to 711 is doesn't automagically stop mattering to 5128.

 

[amirm]

I disagree because it's from 1-3kHz and it's a fairly significant change if we look at the Susvara for instance, you can see the EQ filter used to correct that area is pretty significant being Peak Filter 2249Hz, -1.8dB, Q1.492.

I have said these differences are audible. Question is whether they get us closer to the truth. Take a topic that you are familiar with: channel consistency. Such variations easily occur and then some in the treble region. The research targets are highly smoothed and averaged. We can't take them and use them as gospel to this level of accuracy.

 

[listener650]

We have discussed this ad nauseam already, but let's repeat it : this argument is moot since the original data that lead to the development of the 711 coupler came from four ears from two Danish males only. Besides we simply don't know the ethnicity of the test subjects in B&K's study and it isn't like the Danish population is exclusively composed of caucasians. If the results were biased, and they could be, it's at least miles better than the data from which the 711 couplers were made from.

I actually think we do know the ethnicity of the subjects in B&K's study; if memory serves there were 2 subjects of Sri Lankan descent and one of South Korean descent. Absolutely not characteristic of the vast tapestry of human ethnic diversity, but certainly not nothing.

 

[MayaTlab]

Proving that the path to better predictivity of measurements is NOT to chase more participant scans in the design of a new fixture. Excellent results were achieved with the modified 711 coupler by Harman over a decade of research. Yet no one has shown any benefit to using 5128 in the same regard. At best, it is kind of close to 711. At worse, it is highly variable resulting in less accurate predictions.

If the goal was to improve predictive power of the measurements using the new fixture, B&K should have conducted similar studies to Harman to demonstrate that. Without it, it is a hypothesis which has been demonstrate to fair extent to have been wrong.

Well the great news is that B&K's goal was simply to measure insert devices more correctly, which is sort of exactly what you'd expect from a metrology company.
Instead of writing a meaningless empty word salad like this, why don't you actually try to answer my earlier question ?
This would make it instantly apparent to you why you won't get anything predictive by assessing the Anker A40 with ANC turned on mentioned a few posts earlier against the Harman IE 711 target.

This should have been the first thing documented in the white paper for 5128! That its most fervent promotors of the fixture don't even know what the make up was, is quite damning. How could you run with this being the major benefit of 5128 and key to its "accuracy" when you don't even know how the population sample correlates with the rest of the world?

And this is not the half of it. From what I recall, they discarded a few of the participant scans because they didn't fit the average model they wanted to build!

*sigh* No that isn't the reason why. Read their articles again.

 

[MayaTlab]

I actually think we do know the ethnicity of the subjects in B&K's study; if memory serves there were 2 subjects of Sri Lankan descent and one of South Korean descent. Absolutely not characteristic of the vast tapestry of human ethnic diversity, but certainly not nothing.

To my knowledge we know their country of origin, which is not the same as their ethnicity / descent.

 

[listener650]

This should have been the first thing documented in the white paper for 5128! That its most fervent promotors of the fixture don't even know what the make up was, is quite damning. How could you run with this being the major benefit of 5128 and key to its "accuracy" when you don't even know how the population sample correlates with the rest of the world?

This is simply a good point. That the 5128 didn't come with more comprehensive data to justify it's existence, claims, and price at the outset is something we shouldn't excuse. That people like us took to it very quickly without this data is also not excused by the fact that later efforts to characterize the 5128's behavior have seemingly confirmed what the original assumptions were.

From what I recall, they discarded a few of the participant scans because they didn't fit the average model they wanted to build!

I believe they discarded the impedance measurements because they couldn't get the probes to fit reliably, but their anthropometry was still included in the shape modelling of the ear canal, I could be misremembering though.

 

[amirm]

Well the great news is that B&K's goal was simply to measure insert devices more correctly, which is sort of exactly what you'd expect from a metrology company.

And there is where the plot is lost. No consumer cares about that unless it predicts their preference. No study has shown B&K 5128 to be able to do that as good as the GRAS fixture.

And that is giving you benefit of the doubt that B&K has achieved that correctness. On what basis do you say that? What is the reference for correctness that it is compared to when it comes to in-situ measurements of headphones?

 

[listener650]

To my knowledge we know their country of origin, which is not the same as their ethnicity / descent.

Ahhhh yep I just checked, you are correct. Thank you for the correction.

For those curious (from Wideband impedance measurement in the human earcanal; In vivo study on 32 subjects by Søren Jønsson, Andreas Schuhmacher, Henrik Ingerslev):

 

[amirm]

Here is a bullet from B&K marketing material on 5128:

"Measurement of the audio response of headphones in exactly the way a human would perceive the quality."

Who here can prove what is bolded? What research has B&K performed to correlate exactly the measurements on 5128 vs human perception of quality?

 

[markanini]

To my knowledge we know their country of origin, which is not the same as their ethnicity / descent.

This is only one detail in a bigger question. Do you know enough to strongly advise BK5128 as upgrade to GRAS 45CA? If a leap of faith is needed due to lack of independent validation, or knowledge gaps, maybe it's worth holding off. Other players might join the P.58 space and satisfy valid concerns in the interim. This is applying long-game strategy to objective audio metrics.

 

[MayaTlab]

And that is giving you benefit of the doubt that B&K has achieved that correctness. On what basis do you say that? What is the reference for correctness that it is compared to when it comes to in-situ measurements of headphones?

LOL I've already linked in this thread another study to that effect showing corroborating evidence. It's not the only one.

And there is where the plot is lost. No consumer cares about that unless it predicts their preference. No study has shown B&K 5128 to be able to do that as good as the GRAS fixture.

Please read again ten times if needs be Harman's validation article for the methodology used for IEMs.

Now I would agree that regardless of whether you're using 711 couplers or 5128 ones, you'll get a lot of noise anyway when trying to derive predictive models given the in situ, inter-individual variation, and that the only viable solution to reduce that noise for IEMs is to use active systems that can compensate for the undesirable HPTF variation.

 

[Somafunk]

This argument/discussion/disagreement reminds me of a mate a number of years ago who began to use a new style/design of vehicle dyno cell in his race prep workshop, he was utterly derided by the greats and knowledgeable in the scene and told he was pissing up against the wall but he persevered as he believed it to be more accurate, quicker in use with more reliable recorded outputs as his real world testing (heavy right foot) aligned with his measurements.

Long story shortened - his use of the dyno cell/information/atmospheric controls gleaned along with improved analysis proved him absolutely correct, his dyno cell and data is now one of the few in the UK that is accredited to test engine bhp/torque conformity for race series.

There's always room for improvement, the effort and time expended to get this accepted by others is long and tiresome but ultimately worth the struggle.

 

[MayaTlab]

This is only one detail in a bigger question. Do you know enough to strongly advise BK5128 as upgrade to GRAS 45CA? If a leap of faith is needed due to lack of independent validation, or knowledge gaps, maybe it's worth holding off. Other players might join the P.58 space and satisfy valid concerns in the interim. This is applying long-game strategy to objective audio metrics.

For IEMs, hell yes. Paradoxically, even more so for active ones, since you can't even use the Harman IE target for them, even when the load impedance should be of a lesser importance for them in the range where their active systems operate.
For over-ears, not a necessity, but not a bad thing either. Having both is even better .

 

[amirm]

There's always room for improvement, the effort and time expended to get this accepted by others is long and tiresome but ultimately worth the struggle.

There is only room for improvement if you target the most significant variables. Here, the B&K 5128 has been out some 5 years if not more. No benefit has come about despite folks trying (poorly I might add). The marketing material talks as if the benefits are right in front of you as soon as you write the $35K check. I chose to evaluate the thing before writing said check and found the results more variable and less able to predict listener preference.

Given this, why would you keep hope? I mean just look at the variability between channels in a headphone. Or seating positions. Or variability in music production. What makes folks think that the high order bit was the ear simulator?