Resolve's B&K 5128 Headphone Target - you can try the EQ's.....

[solderdude]

A small recap on how I see this:

A thread got started about Resolve's attempts to involve the headphone crowd in some things he would like to know by getting some responses.
Resolve did not ask participants here... it was on his own website.
Kuddos for Robbo for making ASR members aware.

Resolve has a 5128 that due to the shape of the pinna and earcanal has some potential. Having made such an investment and no headphone target available it would have to be created. No such a thing from Harman nor B&K, just the well known FF and DF which of course are there as that seems to be the primary purpose and is easy to calibrate.

For headphones a different target is needed. We all agree.

Amir played with his loaner and found some issues which are known to exist. Also Dr. Olive recognizes those. Resolve and Jude know as well.
For Amir these 'limitations' of the fixture were no option to base measurements on. The GRAS makes much more sense for this AND a target is already made for it (well close enough).
Amir seems to have no interest in spending time on efforts he sees as futile.

Resolve and co still like to try and come up with something and seem to suggest that with careful placement and perhaps some 'aids' something potentially good may still come out of this.

In the end they will end up with A target and 'compensated' plots that they hope have, over the whole, have equally good, marginally less or even better 'correlation' between measurements and perception.

Amir's objections of involving the public while NOT being present to administer the tests is quite valid from a 'creating a standard' point of view.

AFAIK the involvement Resolve asked is more about finding out preferences in sound which was a 'question mark' for him. It was not intended to invalidate Harman or an amateur attempt to 'redo' Harman research.

Then things get side stepped and the word 'science' is brought in and hell breaks loose.

Me... I don't really care about GRAS vs 5128 or any other fixture. They are all indicative and that's it. 5128 and 45CA will just indicate something different. The final targets that will be used will just mean the computer generated EQ will differ too. So fixture + target differences. Then we have production spread and differences between actual heads (and the way people wear/use headphones) and reality. The 4th factor is recording tonal balance. The fifth factor is preference of tonal balance.

So 5 factors that will create differences between 'a single squiggly' and personal perception.

I say... let the guys try. And no... it won't be a new 'standard'. It will be something Resolve will be using on his website. Some will like/prefer it, others won't.
People can choose where to look and look at different measurements. Even at Headphones.com as CA43 + 5128 seem to be used so one can compare themselves.

It's fun for some that they are trying. Thumbs up from me. I get Amirs standpoint and see why he stands by his points and why he would not see any benefits in participating, not even in discussions. That's it.

carry on..

 

[IAtaman]

What I understand as the primary objection from Amir is that 5128 is not good enough progress to warrant the development of a new reference target, which the industry desperately needs to standardize audio recording and reproduction to achieve some level of tonal consistency. Seems like the discussion is revolving around whether 5128 good enough or not. I am in no way qualified to comment on that, so I will not.

What I can comment on however is the fact that progress is desirable and also unavoidable, and regardless 5128 represents that progress or not, at some point we will have to move on from GRAS measurements and targets based on those. I am not sure what would be to best way to transition, but I am sure people who know much more can comment on that if they like as well.

Which brings me to my question.

If I understood correctly, the problem with headphone measurements seems to be that the FR measured is heavily dependent on the measurement rig and therefore any reference target needs to take the effects of the rig on FR into account; and anyone who intends to compare their results to the reference target, need to use a similar rig. Therefore, every time we think we have a better measurement device, we have to re-establish the target. It seems a bit like using a different rooms to measure the speakers, and B&K seems to claim that they have a better room - a room that is a better representation of the average listening room which is fine I suppose. But would it not make more sense in the long run to develop a way measure the headphone FR in a way that is head and ear canal independent, a Klippel NFS for headphones if you like, instead of trying to approximate a theoretical average head and ear canal?

 

[solderdude]

Therefore, every time we think we have a better measurement device, we have to re-establih the target.

The 'issue' with the 5128 is not the pinna nor the coupler.
It is the variability of seal which is headphone dependent.
Because of this variable (seal due to the shape of the fixture) the debate comes down to:
What is more representative... the fixture with a seal that may be somewhat similar to actual heads (not counting hairs, glasses, pad size, ear shape/size) or a FP (which also can have seal issues) but is more consistent and thus more repeatable.

I would propose to measure on a 45CA (established performance), add (documented) seal testing, also measure on 5128, or what is similar or still to come, and maybe measure actual on-head responses to get an idea of variability on the head.
Draw conclusions based on that and stop the silly promotion of 'exact multiband EQ' against some overly smoothed 'target' based on 1 plot/fixture but instead base it on several measurements and look for common 'deviations' seen in various measurements.
As Mandalorians would put it... this is the way.

But that's probably just me.
The 'wish' for 'better correlating' measurements is understandable as people believe this will lead to a more 'exact' computer generated EQ (an utopia)
That may be the goal Resolve and friends are after, at least that would have been my goal if someone bought me several high $ fixtures...
I sure as hell won't be doing such investments just to please others with a slightly differing EQ.

 

[amirm]

The 'wish' for 'better correlating' measurements is understandable as people believe this will lead to a more 'exact' computer generated EQ (an utopia)

It is just such silliness to expect exactness or anything close to it. I have a saying that headphones measurements are like a compass. If you are lost, it is a heck of of a lot better than nothing. But you best not expect it to be GPS. It isn't and it will never be. We need to learn to live with fuzziness.

BTW, often I search for measurements when I review a headphone and find yours. Almost always I see the correlation between what you have and what I measure! To see this just requires stepping back and looking at the big picture and differences in fixtures.

 

[IAtaman]

What is more representative... the fixture with a seal that may be somewhat similar to actual heads (not counting hairs, glasses, pad size, ear shape/size) or a FP (which also can have seal issues) but is more consistent and thus more repeatable.

That is the part I don't understand (and I do realize it is probably just me) why the representation of an actual head and ear canal needs to be built into the measurement device? Isn't it preferable to avoid all the complication at the measurement device and build that into the reference target?

A silly analogy maybe, but if I wanted to measure the circumference of a circle, I do not look for a circular meter, I use my "linear" meter to measure the diameter and calculate the circumference. Instead of trying to come up with a meter that represents circles the best, is it not better to come up with the most precise and reliable length measurement method?

 

[MayaTlab]

If I understood correctly, the problem with headphone measurements seems to be that the FR measured is heavily dependent on the measurement rig and therefore any reference target needs to take into account the effects of the rig on FR into account; and anyone who intends to compare their results to the reference target, need to use a similar rig. Therefore, every time we think we have a better measurement device, we have to re-establish the target.

Yes and no. For headphones insensitive to coupling issues, up to around 4-5kHz if you allow for some wiggle room (which will happen anyway between a 45CA and real human heads), it's not that difficult to transpose the Harman target to the 5128 (and not at all below 1kHz). Most of these headphones will more or less successfully follow the average transfer function between the test rigs up to a few kHz. That's of course if you want to make as straight a transposition as possible, and are ready to accept that it will limited to only a part of the spectrum. Personally I'm quite open to LMG / HP.com's attempt at going for a more theory-driven approach to design a 5128 target than a simple transposition, and curious to see what will come out of it, but you then have to make some sacrifices in terms of being able to claim that it's supported by listening tests, particularly in the case of HP.com's current preliminary target in my opinion.

Headphones sensitive to coupling issues will differ, sometimes markedly so, but a newer 5128 target should not be accommodating for these headphones as it's a problem on the headphones' behalf, not the 5128. If the resulting measurements don't align with what you'd get on a 45CA, or, given the 5128's propensity to leak more easily than flat-plate style fixtures, are more inconsistent, it's just that the headphones have coupling issues, something a good review should try to detect anyway, even better characterise their extent, and in my view quite severely punish. Whether the 45CA or 5128 is more representative in these cases... is a bit immaterial. The main problem becomes the headphones, not the fixture, and whether you use one or the other, you'll need more than a simple FR curve measurement on any such fixture to assess the extent of the issue (which isn't really a binary "sensitive / insensitive to coupling" question but rather better expressed in degrees and more or less prevalent on different parts of the spectrum, depending on the headphone model).

It seems a bit like using a different rooms to measure the speakers, and B&K seems to claim that they have a better room - a room that is a better representation of the average listening room which is fine I suppose. But would it not make more sense in the long run to develop a way measure the headphone FR in a way that is head and ear canal independent, a Klippel NFS for headphones if you like, instead of trying to approximate a theoretical average head and ear canal?

It was found with speakers that "independent of room" measurements can still provide insights as to which speaker would be preferred (hence the anechoic / Klippel measurements), but right now I don't think that we have found any evidence yet that an "independent of headphone to head coupling" methodology can be effective. That's above my station but I'm wondering what some members here would have to say on the question of whether or not it's even feasible, which I sense is the meat of your question .

 

[solderdude]

That is the part I don't understand (and I do realize it is probably just me) why the representation of an actual head and ear canal needs to be built into the measurement device? Isn't it preferable to avoid all the complication at the measurement device and build that into the reference target?

A silly analogy maybe, but if I wanted to measure the circumference of a circle, I do not look for a circular meter, I use my "linear" meter to measure the diameter and calculate the circumference. Instead of trying to come up with a meter that represents circles the best, is it not better to come up with the most precise and reliable length measurement method?

It is needed because pinna interaction is different with different headphones and not really easily predicatable.
One can do without when the goal is to asses driver performance (quality control) or is willing to allow inaccuracies in tonal balance in the frequency range the pinnae affects.

I personally prefer to separate the 'correction' part and the 'target' part but in the end a raw measurement needs to be corrected (measurement values minus changes the pinna makes which is fixture dependent) + a 'target'.
A target is comparable to a 'room curve' of sorts as most people want to hear the tonal balance of speakers in a room. The headphone sound should have a similar tonal balance as a speaker in their living room (for most).
I prefer the headphone sound to be similar to a (calibrated) near-field and don't really 'need' the Harman bass boost. I agree that that kind of boost is 'impressive' and more 'immersive' but not on all recordings to me.

That 'speaker in room' target would mean more bass and perhaps a little less upper mids and overall treble level.
Once you have those you can convert the weird looking RAW plot into a more common (amplifier/DAC alike) line which represents the 'tonality' as it is heard meaning a flat line is 'neutral' in sound.

But as this is a combination (correction for 45CA + perception research by Harman) it is always shown as a 'highly averaged' target line in raw plots.

 

[IAtaman]

It is needed because pinna interaction is different with different headphones and not really easily predicatable.

Thank you for entertaining my what I suspect to be naive questions, appreciate that.

How is it different than a speaker in a room though? Speakers and rooms do also have compex and hard to predict interactions, yet we can come up with "room-agnostic" measurements and predict in-room response. Why is that not possible for headphones and ears?

 

[solderdude]

Speakers in a room measured with a HATS is very different from a Klippel or measurements in an anechoic chamber.
From the latter you can calculate an expected resoponse in a room (directivity and room reflectivity) and the results can be somewhat close to 'an average' living room (whatever that may look like) but there is some relevance for sure.

One can use a HATS with gated measurements though but then positioning of speakers and the room is a factor.
That would be great IF the room acoustics in situ need to be included for (perception) research and stuff.

With headphones the circumstances are very, very different. Sound not coming from the front (pinna effects are thus very different) but the sides, different driver-ear difference and 'coupling' of the driver to the ear. A few meters and mixing L and R is very different from a small closed front volume.
Then the seal becomes important. There is no seal with speakers but there is a room instead.

For this reason speaker measurements are a bit more predicatable and can be measured with a bit more (and different) accuracy compared to headphones.

Both speaker and headphone measurements aren't as exact as most think they are. That said they are just results measured according to a standard and could be valid for that standard (depends on external factors)
Headphone measurements are simply even less 'exact' than speaker measurements.
5128 is not going to help here either I am afraid. Maybe in some cases 5128 is 'better' than GRAS and in other cases the GRAS will be 'better' with respect to 'sguiggly/pereceived sound' relation but both are not as accurate as most 'squiggly afficionados' think/hope/believe they are.
In other words.. a standard and 'ones actual ears' are not the same thing.

 

[Scgorg]

Thank you for entertaining my what I suspect to be naive questions, appreciate that.

How is it different than a speaker in a room though? Speakers and rooms do also have compex and hard to predict interactions, yet we can come up with "room-agnostic" measurements and predict in-room response. Why is that not possible for headphones and ears?

Headphones and IEMs are more load-dependent than a loudspeaker is. That is, changes in their load impedance is more meaningful than what it is for a loudspeaker. The difference in load impedance for a loudspeaker measured anechoically and one measured in a room is quite small, and so anechoic measurements have a good predictive ability for loudspeakers in rooms too (barring room modes, which are highly individual).

For headphones and IEMs this is less true. The acoustic output impedance of two headphones or IEMs can be enormously different. While it is typically quite low for vented open-back electrodynamic headphones, the opposite would generally be true for a closed-back, closed-front electrostatic headphone, and this is where the challenge lies. Because the output measured from the headphone depends on the relationship between its output impedance and the load input impedance - which is frequency-dependent - we want our load to accurately mimic the load impedance the headphone will be presented to. There's no single compensation that can be applied to a measurement system to compensate for wrong acoustic impedance, because the difference of the difference will vary between rigs.

If you knew the acoustic impedance of the DUT, as well as for the test rig, you could compensate for errors in a rig's acoustic impedance, but we generally do not know this. It is also vastly more work than just using a rig with a proper impedance and measuring the frequency response in the first place.

To sum it up cleanly - putting a loudspeaker in a room means the room creates a series of constructive and destructive interference that mucks up the response, but the output from the loudspeaker itself has not changed much. Changing the load of the headphone, on the other hand, changes how the headphone itself behaves.

 

[IAtaman]

To sum it up cleanly - putting a loudspeaker in a room means the room creates a series of constructive and destructive interference that mucks up the response, but the output from the loudspeaker itself has not changed much. Changing the load of the headphone, on the other hand, changes how the headphone itself behaves.

Right. Thanks a lot for this, makes it a lot more clear for me.

So a headphone is like an amp with an unknown and hard to measure output impedance. In order to understand what will its FR be on a person's head, we would like to use as close to the actual load it will face out in the wild as possible. Speakers on the other hand are like low impedance amps, barring room modes, their FR is lot less load dependent. Is this fairly accurate?

Somehow everything becomes clearer for me when its modeled as an electrical circuit

 

[Scgorg]

Right. Thanks a lot for this, makes it a lot more clear for me.

So a headphone is like an amp with an unknown and hard to measure output impedance. In order to understand what will its FR be on a person's head, we would like to use as close to the actual load it will face out in the wild as possible. Speakers on the other hand are like low impedance amps, barring room modes, their FR is lot less load dependent. Is this fairly accurate?

Somehow everything becomes clearer for me when its modeled as an electrical circuit

That, and the fact that for loudspeakers the measurement condition and usage condition has relatively similar input impedance.
The impedance of an anechoic chamber and that of a room of reasonable size is not too different (both are very low, while the loudspeaker's impedance is very high). So it's not that loudspeakers are immune to impedance changes, it's rather that we generally don't encounter such an impedance change in practical use.

 

[Resolve]

The 'issue' with the 5128 is not the pinna nor the coupler.
It is the variability of seal which is headphone dependent.
Because of this variable (seal due to the shape of the fixture) the debate comes down to:
What is more representative... the fixture with a seal that may be somewhat similar to actual heads (not counting hairs, glasses, pad size, ear shape/size) or a FP (which also can have seal issues) but is more consistent and thus more repeatable.

Actually, so far I haven't found this to be quite the issue that some have made it out to be (including Sean's slide on this... I have issues with that). If you're diligent with the positioning/seating, you can get a seal, and I was able to do so on just about every headphone I've tested with it so far. Others who've spent a lot of time with a 5128 will likely confirm this as well. It's just that getting there is a more finnicky process.

I think there will be exceptions, but there will also be exceptions on real humans. So in my view it's less of an issue in practice.

 

[Robbo99999]

Holy Hell, this thread had become a squabbling mess, thanks to those who walked away from the squabble! (better over the last page)

 

[Resolve]

Holy Hell, this thread had become a squabbling mess, thanks to those who walked away from the squabble! (better over the last page)

And I think a lot of it is down to misconceptions and misunderstandings. Like the more I read the more obvious that becomes. And it's not like we're not considering Amir's objections to all of this either. These are valid, understandable concerns. But once again I think many of them go away once everyone understands the various positions/intentions.


From our perspective, the goal here is primarily to show additional data to the information we already provide from our GRAS system. And, as Amir has pointed out, there needs to be a compass for that information, and a suitably tested one doesn't currently exist. We'd of course love to conduct or contribute to that research if possible, but that takes time - and I believe Dr. Olive has said that they're also trying to better understand the 5128 since that's the direction things are going, and it would be great if they were to do that research.

In the meantime, the proposed alternative is, in my view, on solid theoretical ground, with a key benefit being broad compatibility applications for past and future measurement equipment. Additionally, this method will allow us to show how headphones behave across different 'heads'. I feel this is highly relevant information and including it provides a more complete picture of headphone performance. So far the counter-argument to this seems to be rooted in the idea that additional data points like this obfuscate 'the truth' of the matter, and if this is indeed the locus of the debate (maybe there's something else I haven't seen), then yeah I'm going to strongly disagree with that. If anything, in my view, adding this information gets us far closer to the truth about various products because it lets us understand how they perform in different conditions - which are realistic situations for varying head/ear shapes.

 

[Robbo99999]

And I think a lot of it is down to misconceptions and misunderstandings. Like the more I read the more obvious that becomes. And it's not like we're not considering Amir's objections to all of this either. These are valid, understandable concerns. But once again I think many of them go away once everyone understands the various positions/intentions.


From our perspective, the goal here is primarily to show additional data to the information we already provide from our GRAS system. And, as Amir has pointed out, there needs to be a compass for that information, and a suitably tested one doesn't currently exist. We'd of course love to conduct or contribute to that research if possible, but that takes time - and I believe Dr. Olive has said that they're also trying to better understand the 5128 since that's the direction things are going, and it would be great if they were to do that research.

In the meantime, the proposed alternative is, in my view, on solid theoretical ground, with a key benefit being broad compatibility applications for past and future measurement equipment. Additionally, this method will allow us to show how headphones behave across different 'heads'. I feel this is highly relevant information and including it provides a more complete picture of headphone performance. So far the counter-argument to this seems to be rooted in the idea that additional data points like this obfuscate 'the truth' of the matter, and if this is indeed the locus of the debate (maybe there's something else I haven't seen), then yeah I'm going to strongly disagree with that. If anything, in my view, adding this information gets us far closer to the truth about various products because it lets us understand how they perform in different conditions - which are realistic situations for varying head/ear shapes.

I agree with your sentiment in your first two paragraphs, but just on that last paragraph re your methodology & approach - as I understand it you've not fully decided on how you're gonna formulate & structure your entire project, so I'm not gonna attempt to add any opinion of mine to your methodology, because as I understand it you're still in the process of creating your project.....and I'm also not asking you to outline your entire project right now.....my understanding is that you're in the process of determining what your approach is gonna be.

 

[Resolve]

I agree with your sentiment in your first two paragraphs, but just on that last paragraph re your methodology & approach - as I understand it you've not fully decided on how you're gonna formulate & structure your entire project, so I'm not gonna attempt to add any opinion of mine to your methodology, because as I understand it you're still in the process of creating your project.....and I'm also not asking you to outline your entire project right now.....my understanding is that you're in the process of determining what your approach is gonna be.

Yup, but at the very least it'll include data from GRAS and B&K. I'm hopeful we can have more to show folks in the near future.

 

[ObjectiveSubjectivist]

So in my view it's less of an issue in practice.

It is an issue.
You see, as tool we need consistency, even if seal 'problems' are more in line with humans it's nothing good when we talk about measurements and gathering data.

If you don't see a problem... Oh well another youtuber.

 

[amirm]

If someone thinks we don't place high value on measurements in our reviews they clearly haven't been paying attention hahah. I get criticized on a regular basis for putting too much emphasis on measurements.

I have read and watched many of your headphone reviews. I also read the comments in your youtube videos. From what I can clearly tell, showing measurements to bulk of your audience is to sing the praises of Tofu at a Steak Eating competition! They are as purely subjectivist as one can get, certainly relative to what we do here.

This brings us to a key question that I still don't have the answer to: why? Why did headphones.com decide to go and get this expensive fixture if what you say about your audience is true? You already had capable measurements courtesy of Gras fixture. What possible need do you have to migrate to a new fixture, or add it for that matter? Clearly it is not to serve your audience. So who is it for?

Can you start at the beginning and answer this simple question? Without it, my mind goes to getting a sweet deal on the fixture and now trying to do something with it. Or trying to keep up with LTT as far as appearances.

Bulk of my concerns have been around this. I have explained the issues I have at very high level and still see no answers.

 

[amirm]

That is the part I don't understand (and I do realize it is probably just me) why the representation of an actual head and ear canal needs to be built into the measurement device? Isn't it preferable to avoid all the complication at the measurement device and build that into the reference target?

A silly analogy maybe, but if I wanted to measure the circumference of a circle, I do not look for a circular meter, I use my "linear" meter to measure the diameter and calculate the circumference. Instead of trying to come up with a meter that represents circles the best, is it not better to come up with the most precise and reliable length measurement method?

Some answers have been given on why not to do this. But let me say that I personally like to do it. It would be so revealing to know what the transducer is doing in free field. Just like the way I perform near-field measurements of the drivers in a speaker as diagnostic tool, I think we should look at what the transducer is doing.

I have often thought about using the Klippel NFS to measure one channel but can't yet figure out how to build a fixture to do this without damaging the headphone itself. Maybe I do it on a throw-away headphone just to see what we can discover.