Headphone Measurements Using Brüel & Kjær 5128 HATS

[JohnYang1997]

The 5128 HATS is technically more advanced than the other systems which still use iterations of the old IEC 711 ear simulator that was originally designed for hearing aids 40 years ago. Instead of just damping (masking) the unwanted high Q peaks caused by the nature of the open tube resonator they decided to redesign the simulator and bring it closer to the anatomy of the human ear. It has a smaller diaphragm being closer to the size of the human ear drum, no unnecessary long canal extension and is capable of resolving up to 20 kHz (instead of only 10 kHz) just by design. Their reference impedance for the middle ear is an average of least 20 in-ear measurements, where the old 711 is only based on two sets of data (actually those of the engineers who invented it). They also came up with their newer anthropometric ear canal, of course, averaged over 40 MRI scans.

The 45CA is more established (hence easier to compare) because it is cheaper, but certainly not state of the art.

Regards
Dreyfus

If you read through this thread. You'd see it's basically unusable in real world earphone measurement situation.

 

[Dreyfus]

What is a pretty good job?
I thought you said Harman preference curve has caused confusion. If so, and they are behind promoting it then by your definition they have not done a good job.

My point is that the Harman target is accompanied by a lot of general confusion and uncertainty because A) there is probably not enough common knowledge on the research behind it, combined with limited knowledge about the complexity of measuring headphones and B) a lot of users do not fully agree with its results in practice.

Both Oratory and Crin do not only share graphs but also explain what to expect from their measurements and how you have to treat certain phenomenons in the graphs. If I am not mistaken, they do not see it as their mission to promote Harman and explain it in full detail.

Methodically, I think it is practically more valuable to offer the possibility to compare frequency responses side by side and show the differences between headphones rather than calling for a singular (not necessarily uncompromisingly accepted) reference sound and illustrating the delta to this target (by the means of compensated graphs for instance).
But that is where everyones opinions and practical demands certainly differ. Luckily, Oratory and Crinacle do offer both approaches. So it is up to you to choose. With their comparative tools you are even free to involve a target curve, which may be Harman, their own one or even DF.

Part of the problem has been presentation of information and analysis. I am dead against throwing a bunch of measurements out there and then ending with a 10 band EQ. This is not the way to teach, nor are those 10 band EQs justified. Contrast that approach with my reviews so far where a story is told that starts with specific measurements and ends with appropriate subjective evaluation and simple equalization. It is an integrated, end to end story that holds together and builds on the foundation of Harman research but adds to it. You have to bring people along in each review.

Again, we all have differing demands, levels of knowledge and practical approaches to solve our "problems".

The way of telling a full and comprehensive story is a very valuable one, no doubt. And I do fully agree that you need the appropriate context right next to the graphs to understand what is actually going.

What I would like to see though is a broader call of attention for potential practical issues like varying pinna and canal resonances as much as a differing qualities of seal. At their extremes, such effects can be total dealbreakers for the evaluation, comparability and not the least of course for the enjoyment of sound. After my terrible K371 experience, I am very careful with intersubjective approximations of measured frequency responses. Every head is different after all.

Since such factors are hard to track and quantify with just one (not periodically modified) standard rig we should at least include something like a regular paragraph or link to an acrticle that explains the expectable relations between theory and practice. That is also where things like coupler resonances and the lack of accuracy of generalized equalization profiles should come into play.

Regards
Dreyfus

 

[DualTriode]

Hello All,

The term “circle of confusion” was co-opted, stolen by Dr. Floyd Toole.

“Circle of Confusion” belongs to optics and or photography. If CoC shows up in audio I stop reading.

Thanks DT



https://en.wikipedia.org/wiki/Circle_of_confusion

 

[amirm]

Both Oratory and Crin do not only share graphs but also explain what to expect from their measurements and how you have to treat certain phenomenons in the graphs. If I am not mistaken, they do not see it as their mission to promote Harman and explain it in full detail.

I have not made it my mission to promote Harman for headphones. I have even renamed my reference curve to something else to avoid that. Yet folks like you keep making it about Harman and want to argue that. I have a reference I use which has provided highly prescriptive to me in the way I use it. I verify everything I do with listening tests and equalization. No blind trust is put in anything much less promoting such.

If you want to say to throw out the concept of a reference and just show graphs side by side, that is not me. You are welcome to use my data to create such. You don't need me. This has been done for speakers and I am publishing the data for headphones to enable the same. In time we may develop such tools but it is not a priority right now, nor do I want to invest in the expense of a custom back-end for it.

My mission is simple: I want to be in a position to answer at all time what some headphone is about quickly, reliably and to the point. I am doing it for myself as I do the testing and am simply sharing that with the rest of you. I couldn't do that with B&K 5128 and waited a year to do the whole thing for headphones because of the challenges that existed in the past. But now I feel comfortable that I can tease out what makes a headphone tick. This is an approach that has worked for all other categories of products I have tested. And this is what the membership asked me to do: replicate what I have done elsewhere for headphones.

For those of you who live and breath other people's measurements, know every nuance of every version of each headphone, etc., my reviews may not be for you. Continue using the other sources. But please don't tell me to follow them. I know about them. Despite all of that existing, I put huge investment behind doing things differently than them. I also know the research inside and out, including the researcher himself. So please don't say I should know this and that. This is what I did for over a year before taking this plunge.

 

[Dreyfus]

@amirm
I am sorry. But neither do I have accused anyone to be a promoter of Harman nor do I have stressed the point that the comparative method is the only valid method to approach things. What I did is answering to your statement that there is no Harman curve for the 5128 and to the question why it might cause confusion and uncertainty in practice plus what makes a pretty good job for headphone measurements and reviews (in relation to the mentioned works of Oratory and Crinacle).

The way of presenting the data to the public is your choice. I respect that.

The only thing I generally regret in the headphone measurement community is that there is rather little consideration of things like leakage effects or individual PRTF which - together with seating and unit-to-unit variation - can practically corrupt the integrity of the described sound, its relation to a called (p)reference curve and the cross-comparability to different heads and ears.
From a user standpoint, headphone FR graphs are often (mis)taken as absolute measures of sound and irrevocable sources for buying decisions and discussions about taste. I have seen a lot of threads on the internet which follow such beliefs. And I am pretty sure that sooner or later they will also refer to ASR, link to your graphs and claim the same. That is why I constantly underline my critique here on ASR and call attention to the factors of uncertainty. And that is not because I doubt the competence of engineers like you. It is because I respect the captiousness of the technology and fear that we have to deliver much more clear and direct clarification on how headphone measurement works, what it realistically can achieve and what it cannot achieve.

If I consider your statements about headphone equalization and what divides sheer data from reputable science, we are probably on the same fundamental wavelength.

Regards
Dreyfus

 

[HRTF_Enthusiast]

@GaryH regarding the flat bass discussion we were having, my preference is flat bass on 5128, not 45CA.

 

[GaryH]

@GaryH regarding the flat bass discussion we were having, my preference is flat bass on 5128, not 45CA.

That would still be bass-deficient in comparison to the response of a neutral speaker (anechoically flat on-axis) measured by the HATS in a studio-like room. The 5128 underepresents bass in comparison to the 45CA, but not enough for flat to be anywhere near neutral.

 

[HRTF_Enthusiast]

That would still be bass-deficient in comparison to the response of a neutral speaker (anechoically flat on-axis) measured by the HATS in a studio-like room. The 5128 underepresents bass in comparison to the 45CA, but not enough for flat to be anywhere near neutral.

Harman target isn't simply the response of an anechoically flat on-axis loudspeaker measured by HATS. Even Amir boosted bass on 8361A.

The difference between Stealth on 45CA and 5128 is about 4.5 dB in bass. If we subtract 4.5 dB at 20 Hz for the "all listeners" preferred response and an additional 2.5 dB for the trained listeners response, the bass response on 5128 would be very close to flat.

"

 

[GaryH]

Harman target isn't simply the response of an anechoically flat on-axis loudspeaker measured by HATS.

I didn't say it was.

Look carefully at where the two responses are aligned with the Harman target - the top graph at 425 Hz, the bottom at 1 kHz. If they were both aligned at 500 Hz as they should be (the standard in the scientific literature), the difference in bass would be less. (Note the red 'fixture difference' exaggerates the bass difference even further by aligning at 3 kHz.)

and an additional 2.5 dB for the trained listeners response

You can't do that. The differences between trained and untrained listener preferences are not the same for headphones as for speakers, as I showed in this post. For headphones, the large majority (70%) of trained listeners fell into the 'Harman target lover' class, and 30% preferred more bass than Harman. The class that preferred less bass was comprised entirely of untrained listeners, and predominantly over 50 (so likely presbycusis caused them to prefer less bass to balance out high-frequency hearing loss).

 

[Robbo99999]

Harman target isn't simply the response of an anechoically flat on-axis loudspeaker measured by HATS. Even Amir boosted bass on 8361A.

The difference between Stealth on 45CA and 5128 is about 4.5 dB in bass. If we subtract 4.5 dB at 20 Hz for the "all listeners" preferred response and an additional 2.5 dB for the trained listeners response, the bass response on 5128 would be very close to flat.

"

As an interesting aside, the 2013 Harman Curve is in fact an exact representation of the solid "All Listeners" line when you apply that slope to the HATS measurement that Harman did - the HATS measurement that Harman did was with speakers that had been EQ'd to a flat in-room response (not anechoic) at the listening position. I know this because I used REW to apply that "All Listeners" slope to the that flat in-room measurement - I used VirtuixCAD to trace the lines so I could import them into REW, and it accurately followed the 2013 Harman Headphone Curve.

 

[HRTF_Enthusiast]

I didn't say it was.

Look carefully at where the two responses are aligned with the Harman target - the top graph at 425 Hz, the bottom at 1 kHz. If they were both aligned at 500 Hz as they should be (the standard in the scientific literature), the difference in bass would be less. (Note the red 'fixture difference' exaggerates the bass difference even further by aligning at 3 kHz.)

You can't do that. The differences between trained and untrained listener preferences are not the same for headphones as for speakers, as I showed in this post. For headphones, the large majority (70%) of trained listeners fell into the 'Harman target lover' class, and 30% preferred more bass than Harman. The class that preferred less bass was comprised entirely of untrained listeners, and predominantly over 50 (so likely presbycusis caused them to prefer less bass to balance out high-frequency hearing loss).

That paper is quite misleading.

AES E-Library » A Study of Listener Bass and Loudness Preferences over Loudspeakers and Headphones

In order to study listener bass and loudness preferences over loudspeakers and headphones a series experiments using a method of adjustment were run. Listeners adjusted the bass and loudness levels of multiple genres of music to their personal preference in separate listening sessions over...

www.aes.org

 

[HRTF_Enthusiast]

 

[solderdude]

Were the DT990 equalized or used as is ? When the latter there would have been a considerable lows boost to start with.
Were peak levels also measured (70dBA average says little about peaks, they may not have exceeded 95dB) ?

 

[HRTF_Enthusiast]

Were the DT990 equalized or used as is ? When the latter there would have been a considerable lows boost to start with.
Were peak levels also measured (70dBA average says little about peaks, they may not have exceeded 95dB) ?

No specifics are given on peak SPL

 

[solderdude]

So EQ'ed...
That makes a difference as mid bass would have been lowered, the treble peak lowered and subbass might have had a slight boost.
DT990 is not very sensitive to seal issues and fits most heads so it isn't a poor headphone to do some preliminary testing and preference research with as long as it isn't done at high SPL (which doesn't have seemed to be the case)

 

[HRTF_Enthusiast]

So EQ'ed...
That makes a difference as mid bass would have been lowered, the treble peak lowered and subbass might have had a slight boost.
DT990 is not very sensitive to seal issues and fits most heads so it isn't a poor headphone to do some preliminary testing and preference research with as long as it isn't done at high SPL (which doesn't have seemed to be the case)

The frequency response of the DT990 wasn't what I was pointing out

 

[solderdude]

Clearly the distortion plot from Amir was and is what I replied to.
EQ'ed DT990 at 70dBA average would be fine to use during such research.
It will not have been used at 114dB SPL which is it's maximum power rating (0.1W) as it start to softclip there.

 

[HRTF_Enthusiast]

Clearly the distortion plot from Amir was and is what I replied to.
EQ'ed DT990 at 70dBA average would be fine to use during such research.
It will not have been used at 114dB SPL which is it's maximum power rating (0.1W) as it start to softclip there.

If the test was designed to determine bass preferences with volume adjustment, why was such a limited headphone used?
OG HD800 would have been a far better choice.

 

[solderdude]

If the test was designed to determine bass preferences with volume adjustment, why was such a limited headphone used?
OG HD800 would have been a far better choice.

When the testing was done at 100 to 110dB SPL then yes.. HD800 would have been a better choice.

However, I don't think the DT990 at sensible listening levels is such a limited headphone.

Below the DT990 and with a very simple EQ (1 band and 2 shelves). Horizontal is audible flat with Harman-ish bass already applied.
I don't see what is so limited about this headphone for some preference testing. The fact that Amir tests headphones near it's extremes is another matter.

Below distortion at 90dB SPL (not EQ'ed)

H

 

[HRTF_Enthusiast]

When the testing was done at 100 to 110dB SPL then yes.. HD800 would have been a better choice.

However, I don't think the DT990 at sensible listening levels is such a limited headphone.

Below the DT990 and with a very simple EQ (1 band and 2 shelves). Horizontal is audible flat with Harman-ish bass already applied.
I don't see what is so limited about this headphone for some preference testing. The fact that Amir tests headphones near it's extremes is another matter.

Below distortion at 90dB SPL (not EQ'ed)

H

imo the fidelity is audibly compromised at elevated levels even if not audibly clipping. Who knows how this could have affected the test?