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

[Robbo99999]

Ah, you're not gonna overlay a new curve over the top like you mentioned in your review? Instead you're gonna be listening to the headphones and deciding if you want to boost 3-8kHz by a certain amount each time & you may rest at a different boost level each time perhaps, and praps even zero boost there for some headphones? I suppose that works because the level of required boost there is unknown, so it makes sense to judge that area subjectively.

I suppose it doesn't change the fact there's been a lot of love for the Harman Curve we currently use, so boosting that 3-8kHz could make some headphones overly bright I reckon. Given that danger I actually think it would make sense to also experiment with reducing 1.3-3kHz because that area is showing a consistent reduction in the study too, and that would be synergistic with increasing 3-8kHz because decreasing 1.3-3kHz will make the headphone a bit less bright & therefore make the 3-8kHz boost more palatable on overall tonality - so I reckon it's important to decrease the 1.3-3kHz area too, and maybe even bake that specific change into the curve because it seems common on all of them. I think I'd bake in the 1.3-3kHz change and keep the 3-8kHz area subjectively flexible, I think I'll try that sometime on my own headphones.

Replying to my own post here just for context because I worked out the EQ filters required to turn the Harman Curve that we use here in ASR into the "proper" Harman Curve of original Harman Research for the Hifiman Susvara headphone. I chose Susvara because didn't we say in this thread it was the most open one & therefore most consistent??

Basically it's two filters to convert your current Harman EQ into one that simulates it having been measured on the original equipment used in the Harman Research:

• Peak Filter: 2249Hz, -1.8dB, Q1.492
• Peak Filter: 6462Hz, +2.6dB, Q1.118

Here's the working in REW, and I traced the graphs Amir put up using VirtuixCAD so I could import them into REW (I also lined up the "target" on the subtle differences below 700Hz so that we could just limit it two Peak Filters for the conversion EQ):

My understanding is that this conversion will be most applicable to very open headphones like the HD800 and of course the Susvara on which this is based. @Mad_Economist (or others) please correct me if I've made a wrong assumption.

I haven't tried this conversion EQ on my own Harman EQ's I use for my headphones, but I'll try it when I'm less tired. You can basically just slap those two Peak Filters onto an Oratory EQ for instance.

EDIT 18th October: slapped those two filters on a standard Oratory EQ for the HD800 - sounds good, ended up listening at quite a lower volume than usual on my amp setting, might be related to the EQ change.

 

[amirm]

Ah, you're not gonna overlay a new curve over the top like you mentioned in your review? Instead you're gonna be listening to the headphones and deciding if you want to boost 3-8kHz by a certain amount each time & you may rest at a different boost level each time perhaps, and praps even zero boost there for some headphones?

Well, I always do that. But no, what I meant is that most of the curve has not changed so if I were to show a variation, it would only be where the deviation is noticeable which is above a few kHz.

 

[amirm]

When you say you adopted this target, which are you referring to?

The dashed blue line in my frequency response graphs.

 

[amirm]

You were the one who initially brought up the comparison with speakers highlighting solely "anechoic on-axis frequency response", so Mad's response felt more like a "you know that there is more to speakers than just that".

I brought it up in the context of having a standard for speaker measurements and presentation. Here is what I wrote:

For speakers, we have flat on-axis anechoic response. That may actually not be ideal to Nth degree either. But it has been accepted and good designers strive for it. We need the same for headphones.

I already said it is not ideal or everything one wants to know. But it is a subset of speaker measurements that is been broadly adopted by speaker designers who care. Such a standard for headphones does not exist. I suggested that it be created by AES and use GRAS-45CA with KB5000 as it is the closest, commercially available fixture.

His response was to turn this into a battle between speaker testing and headphones, throwing stones at what is done for speakers as to justify the mess we have with headphones. This is wrong. We are massively ahead of the game due to standardization in the form of ANSI/CEA/CTA-2034 than if this was research buried behind pay walls at AES.

 

[Robbo99999]

Well, I always do that. But no, what I meant is that most of the curve has not changed so if I were to show a variation, it would only be where the deviation is noticeable which is above a few kHz.

Ok, I think it might be worth taking into account the differences between 1-3kHz too, because that would be a synergistic EQ change to go with the increased treble between 3-8kHz - otherwise I think people are more likely to end up with overly bright headphones when trying it.

 

[amirm]

As a professional loudspeaker designer and full time audio engineer I can attest that Mad is not talking from a "lack high level understanding", from the numerous conversations we have had on that subject.
In fact, he's probably the first person to bury himself in literature if he feels like his understanding of a certain topic is not adequate.

I don't know what conversation you have had with him. Nor do I care in this context what papers he has read. I made a statement about where the headphone industry should be going which requires industry knowledge, wisdom and experience in setting standards. That is the conversation he jumped into for which, he lacks qualifications across multiple domains.

I consider his actions, whether on purpose or not, have helped randomize where we are today. It started with Jude at head-fi getting some deal to promote B&K 5128. And then other online reviewers thought to compete with him, they too need this fixture, damn the fact that we did not have any research to guide us on preference target for it. On the contrary, early work by Dr. Olive that straight conversion from their fixture was not possible for B&K 5128. But they ran anyway with "oh, we have a more accurate fixture." Folk's personal motivations have gotten in the way of serving consumers and industry with unified standards.

 

[amirm]

Ok, I think it might be worth taking into account the differences between 1-3kHz too, because that would be a synergistic EQ change to go with the increased treble between 3-8kHz - otherwise I think people are more likely to end up with overly bright headphones when trying it.

I don't see consistency there in the 7 measurements provided. If the difference is small, as I have post, the measurements are not dispositive anyway.

 

[amirm]

Btw, to my knowledge he is not officially affiliated nor being paid by headphones.com anymore

I know he is not anymore. The damage was done when went on their payroll.

 

[Robbo99999]

I don't see consistency there in the 7 measurements provided. If the difference is small, as I have post, the measurements are not dispositive anyway.

It's consistent in that all the headphones show less energy between 1-3kHz if EQ'd to "Harman's Original", but I agree it does vary in it's amount, but that's the same in the 3-8kHz area. If we could somehow predict which one a certain headphone would most likely follow then you could use that pattern. Or decide on a generic difference in the 1-3kHz area, and likewise in the 3-8kHz area. I just think both are synergistic EQ changes, and if a person just does the 3-8kHz EQ you're leaning more bright.

 

[amirm]

From 1 to 2 kHz, it is either the same or very small difference. So I don't think we need to mess with that.

 

[peniku8]

We are massively ahead of the game due to standardization in the form of ANSI/CEA/CTA-2034 than if this was research buried behind pay walls at AES.

I agree, CEA-2034 is a solid standard and helps a lot to demystify speakers as a whole, since the understanding and the evaluation of the DUT is on a different level compared to headphones currently, at least that's my impression.
From what I gathered, headphone acoustic impedance directly corelates to the magnitude of the tonal variance between listeners (and test fixtures) and yet, this is a metric I have never seen in any review, ever. Kinda feels like publishing speaker measurements without directivity data. You can't really know ("predict") how the device will perform in the acoustic environment you will put it in without that data.

 

[Resolve]

I consider his actions, whether on purpose or not, have helped randomize where we are today. It started with Jude at head-fi getting some deal to promote B&K 5128. And then other online reviewers thought to compete with him, they too need this fixture, damn the fact that we did not have any research to guide us on preference target for it. On the contrary, early work by Dr. Olive that straight conversion from their fixture was not possible for B&K 5128. But they ran anyway with "oh, we have a more accurate fixture." Folk's personal motivations have gotten in the way of serving consumers and industry with unified standards.

And yet... Dr. Olive himself has given presentations on the new standard's benefits to accuracy, and indeed voiced his support to such efforts in understanding it better, and in the service of moving the science forward. You should ask him what he thinks about the work that Mad Economist has done on the subject, while in our employ or otherwise. Moreover, all of this seems to corroborate what we've conjectured about acoustic impedance and the in-situ response behavior variation - at the very least, it's not surprising to us. So maybe rather than making allegations of "personal motivations" for adopting new standards, consider that some of us are genuinely interested in this topic and want to be able to show with data why certain acoustic designs may be heard differently by people. The idea that some of us want more and better information for how these devices perform should not be so shocking that the only way you can make sense of it is to attribute malicious motivations to those who want it.

 

[amirm]

And yet... Dr. Olive himself has given presentations on the new standard's benefits to accuracy, and indeed voiced his support to such efforts in understanding it better, and in the service of moving the science forward.

Sure. The question to ask him if he were a reviewer, would he jump and create a new random curve and start to publish said results.

 

[amirm]

So maybe rather than making allegations of "personal motivations" for adopting new standards, consider that some of us are genuinely interested in this topic and want to be able to show with data why certain acoustic designs may be heard differently by people.

Throwing a new measurement fixture in the mix and overlaying preference curves without a single controlled listening test as we see in the current study under discussion, shows total lack of regard for any science in this domain.

I went through the same thing with Mad_Economist with 5128 but as soon as I noticed lack of correlation with research measurements, I abandoned the fixture. You on the other hand, are running with it, science and proper protocols be damned. All you seem to need is your own opinions.

Absence of lack of logic, I have to think that you are doing this for business reasons, i.e. as I explained, because Jude has it, you must too.

 

[amirm]

Moreover, all of this seems to corroborate what we've conjectured about acoustic impedance and the in-situ response behavior variation - at the very least, it's not surprising to us.

So you knew all along that the 5128 has headphone dependency but proceeded to use the same generic target for all headphones being tested?

 

[SuperSplashBrothers]

Now as we've discussed a while back in this forum, it's actually not that hard to find a good translation between Harman's rig and the 5128 provided one uses, as you pointed out, only headphones that we know are likely (at least up to a few kHz), to translate well from one fixture to another in a desirable fashion. And if headphones that can't do so then show a "different difference" between their response and the target obtained, it should be considered a sign that the headphones' engineering leaves something to be desired, and neither be an indictment against a particular fixture nor the sign that getting a half-decent target translation, at least up to around 3-4kHz, is impossible.

Brilliant post. I'm not a massive fan of the 5128, in fact I find its treble output with over ear headphones to be almost entirely useless for me, but your post does kind of convince me that, because it's the fixture that's more sensitive to the output impedance of the headphones, 5128 should be the fixture that's prioritized when it comes to evaluating or standardizing measured performance in the region you've labeled "Definitely problematic."

Aside: It also makes me feel kind of stupid that, for a while, I thought 5128 was outright worse because it was displaying differences where the GRAS system didn't. I thought this was an issue with 5128 being inconsistent or something, but I suppose it's actually the headphones that are inconsistent.

My question here is: for defining something like a Harman curve for the 5128, do you think it would make sense for the GRAS systems to have a "high impedance" and "low impedance" version of the target? Like if we take "low variation" headphones like the HE1000 in the study and a "high variation" headphones like the DCAs or other closed back headphones, EQ them to the same target on the 5128, and then measure the headphones on the GRAS system, one would assume that the GRAS system would show similar behavior under 1 kHz for the "low variation" headphone but more energy in the "Definitely problematic" region. I think somethign like this could be one way to attack the problem of the GRAS systems' behavior in the "Definitely problematic" band, but I guess we would need to measure the output impedance of the headphone to confirm which target to use.

The whole thing kind of leaves me scratching my head about why they used the DCA headphones. Even in this study, they observed other headphones to have less variation between fixtures.

 

[SuperSplashBrothers]

Well, you need to. The measurements I have shown from the research are highly smoothed down to 1/2 octave. In contrast we use 1/20 octave for speaker measurements. The high level of smoothing is kind of needed to even out the response changes due to internal reflections. I don't do that. See my measurement again:

If I breath on the headphone, that response changes. As such, you need to learn to put less priority in that region of measurement and apply judgement. In that regard, it would not change things much if you raised the dashed target line a few dB higher. You still have to guess and then use equalization to verify.

I'm not saying that reading the tea leaves of treble is a good idea, just that the roughly ~3dB of difference shown in the paper around 5 kHz may be enough to, on average, have a meaningful effect on preference. The 95th confidence intervals of the individual listeners' preference results from the 2013 paper, if memory serves, never reached 3dB (±1.5dB).

The same is true of low frequencies where fitment on the fixture becomes harder (and much harder on B&K 5128). In the above example, that bass response can be the red, green or something else when you wear it.

The most reliable part of the measurement then is above bass to lower treble.

The data from this paper seems to suggest that the output load-specific variation can extend basically across the entirety of the spectrum (the difference in the Noire X/XO seems not to be seal/fitment related, as the bass is still fully extended), which would suggest to me that maybe a single target isn't sufficient for this fixture in the midrange either given it's not representing the variation accurately.

 

[SuperSplashBrothers]

Throwing a new measurement fixture in the mix and overlaying preference curves without a single controlled listening test as we see in the current study under discussion, shows total lack of regard for any science in this domain.

In fairness, people could also say you too are using a combination of fixture and preference curve that isn't supported by any of the existing literature, because you do not use the KB0071 used in 2013 or 2015, or the Welti ear.

I went through the same thing with Mad_Economist with 5128 but as soon as I noticed lack of correlation with research measurements, I abandoned the fixture. You on the other hand, are running with it, science and proper protocols be damned. All you seem to need is your own opinions.

I think you may have erred where I previously did in assuming the lack of correlation to be a bug instead of a feature. I was also deeply confused about 5128 measurements as recently as two months ago, but as I've learned more and spoken with people who know more than I do about headphone measurements, it's become apparent that even in the absence of a "Harman target" for the 5128, there are useful ways to evaluate the measured performance of headphones in a way that people can still understand, and the 5128 is a very powerful tool for characterizing the behavior of these devices.

 

[Resolve]

So you knew all along that the 5128 has headphone dependency but proceeded to use the same generic target for all headphones being tested?

You should read MayaTlab's earlier post because I think it gets to this point. The issue is more that for acoustic impedance sensitive devices, you're going to have more discrepencies. This is identified better by the 5128 and 4620.

I do want to also note, at no point did we ever "jump and create a new random curve". Our whole thing was always to use the DF baseline calibration with the preference tilt from Harman. Coincidentally... in a recent convention paper, even our very early efforts of DF + 10dB tilt was tested and rated similarly to Harman. Though that's not why we did it.

 

[SuperSplashBrothers]

I do want to also note, at no point did we ever "jump and create a new random curve". Our whole thing was always to use the DF baseline calibration with the preference tilt from Harman. Coincidentally... in a recent convention paper, even our very early efforts of DF + 10dB tilt was tested and rated similarly to Harman. Though that's not why we did it.

I'm not sure this is true, sir. You absolutely did introduce at least one random target at the beginning of your time with the 5128 before there was any scientific support for it (DF + 8dB tilt, if memory serves). I don't think this was unreasonable, but it did happen. Can you share the paper where the 10dB tilt was rated similarly to Harman?

Big fan of your content, by the way.