ASR Headphone Testing and BK 5128 Hats Measurement System

[Mad_Economist]

Aren't all pinnas flexible? I thought the difference is the canal extension part?

All pinnae are flexible, but the flexibility and deformation characteristics of the pinnae vary. In the old days, the main difference was between the Shore OO-55 pinnae that were suitable only for free field or hearing aid stuff, and the soft OO-35, but Todd Welti published a paper about low-frequency headphone response roundabouts of 2013 or 2014, and while I have some gripes about the takes therein, it prompted GRAS and B&K to make new pinnae, so there must be some merit.

An example of the difference between the anthropomorphic pinnae and an older, stiff model can be found in Brent Butterworth's Susvara measurements - the free field HRTF of the mannequin is essentially unchanged by the pinna difference, per GRAS, but you can see that the headphone response is meaningfully impacted in cases where the pinnae are meant to bend and collapse. This effect would be less significant if Butterworth's old ear was an OO-35 model (which is what really should be the minimum for any headphone measurements), but anecdotally I've measured differences between my OO-35 stock B&K pinnae and custom made more flexible/collapsible ones in a headphone specific way.

The canal extension also differs with the anthropomorphic pinnae - I have very little commentary on this subject, as IEMs are outside of my purview, but GRAS' rationale for it seems sound to me in theory.

 

[bobbooo]

Aren't all pinnas flexible? I thought the difference is the canal extension part?

Some are more flexible than others. The KB5000 has a hardness of 35 on the Shore OO durometer scale, closer to a real human ear than the KB0065, which has a higher hardness of 55, and so doesn't deform correctly compared to a human ear by the headphone pads.

Edit: @Mad_Economist beat me to it

 

[JohnYang1997]

Some are more flexible than others. The KB5000 has a hardness of 35 on the Shore OO durometer scale, closer to a real human ear than the KB0065, which has a higher hardness of 55, and so doesn't deform as much as a human ear by the headphone pads.

Edit: @Mad_Economist beat me to it

Now I have no experience with these pinnas. But i don't know if that's the cause of the difference especially when susvara shouldn't touch the ears as it's very large and deep. They do have different shapes don't they?

 

[Mad_Economist]

Some are more flexible than others. The KB5000 has a hardness of 35 on the Shore OO durometer scale, closer to a real human ear than the KB0065, which has a higher hardness of 55, and so doesn't deform as much as a human ear by the headphone pads.

Edit: @Mad_Economist beat me to it

Fastest nerd gun in the west

Notably, Shore OO-35 pinnae aren't new - my 4128C's pair predates the modern trend of anthropomorphism - and GRAS has the KB1060/61 and KB1065/6, but the GRAS fans I know have lead me to understand that there's something to how the pinna has been constructed that's more than just the durometer to make it more ear-like in its compression.

 

[Mad_Economist]

Now I have no experience with these pinnas. But i don't know if that's the cause of the difference especially when susvara shouldn't touch the ears as it's very large and deep. They do have different shapes don't they?

You might be surprised by how much pinnae protrude - and how small of a deviation can produce a meaningful response change at the ear drum/sim.

I'm unsure if this will be allowed by the site's security, but attached is a zipped album of all Butterworth's KB0065/KB5000 overlays made by a peer of mine; I can turn it into an imgur album if zipped files are a no-no here. Edit: Imgur album here

The HRTF of the mannequin in free field conditions is not meaningfully changed by the two pinnae, from what I've been told. Certainly, externally, they seem extremely similar.

 

[JohnYang1997]

You might be surprised by how much pinnae protrude - and how small of a deviation can produce a meaningful response change at the ear drum/sim.

I'm unsure if this will be allowed by the site's security, but attached is a zipped album of all Butterworth's KB0065/KB5000 overlays made by a peer of mine; I can turn it into an imgur album if zipped files are a no-no here.

The HRTF of the mannequin in free field conditions is not meaningfully changed by the two pinnae, from what I've been told. Certainly, externally, they seem extremely similar.

The new pinna measurements definately look a lot like headacoustics' measurements. Ananda's measurements look very similar to rting's.
And I remember headacoustics was talking something similar ages ago.

 

[Robbo99999]

The site says order now to receive by next Wednesday. Just saying.

Also, do these measurement rigs have the Harman target curve or others included? If not, that means one has to create their own just to know what true flat measurements look like for that specific rig.

That is something we need to create. They do come with some calibration but I suspect they are free-field or diffused.

Surely the calibration file is just something that offers a bit of a fudge factor to massage some of the measured anomolies to what they "think" the frequency response actually is at that point on the curve......then surely what we do with the frequency response is our business, in terms of fitting it to a Harman Curve or Diffuse Field or other? I mean that's the way I imagined it fitting together.

EDIT: although thinking about it the calibration would also include the HRTF of an average head discluding (maybe) the ears, because they're already physically simulated by the "rubber ears". So in conclusion I'm thinking the calibration file would include HRTF of an average head (discluding ears, maybe) and then on top of that would have the "fudge factor" I mentioned in the previous paragraph to account for measurement anomalies which may try to account for specific limitations of the kit including perhaps calibration for the actual microphone itself. Is that right, have I mapped that out right? If that's the case then the calibration files are gonna contain some influence as to whether it was decided to go after Headphone Harman Curve or Diffuse Field, etc, because they all have different measurement/simulation protocols in terms of speaker layout in the room which influences the HRTF that is measured. Yes, so it would be critical to know which one they were targetting, and I imagine they could provide different calibration files to target different 'simulations' for speaker layout in the room during HRTF calculation (eg Headphone Harman Curve or Diffuse Field, etc)?

EDIT#2: If my above paragraph is correct, then it's absolutely necessary for us to know which simulation/target they are using! Because that will affect how we use/interpret the results.

 

[solderdude]

The real issue with making any correction curve is that when you are seeking correlation between how the headphone sounds and measures this won't be perfect with all headphones simply because you would have to end up with an average one. Average headphones don't exist so you will always be 'off' when basing exact EQ on it. Besides there is a substantial delta between real ears and an average pinna.
Having a good 'average' is best of course. I think all manufacturers strive for this using years of experience and research. Yet, none of these rigs will measure all headphones in the same way (at least above 1kHz).
Each rig is a compromise. One is better than the other for sure, mostly depending on what is being measured.

 

[Robbo99999]

The real issue with making any correction curve is that when you are seeking correlation between how the headphone sounds and measures this won't be perfect with all headphones simply because you would have to end up with an average one. Average headphones don't exist so you will always be 'off' when basing exact EQ on it. Besides there is a substantial delta between real ears and an average pinna.
Having a good 'average' is best of course. I think all manufacturers strive for this using years of experience and research. Yet, none of these rigs will measure all headphones in the same way (at least above 1kHz).
Each rig is a compromise. One is better than the other for sure, mostly depending on what is being measured.

We all know that by now, and that's been repeated loads of times, but that's not what I'm referring to (if your response was to me?).

 

[solderdude]

It is when one doesn't want to use just one of the few supplied correction files but make your own.
That is if Amir seeks correlation between his perception and measurements using just a few headphones.
When the goal is to test the rig and see how it correlates to other rigs all that is needed are the same headphones.

On top of those correction files indeed there are various 'room corrections' (I consider Harman one of those) which would have to be incorporated.
One can simply apply a few of them and see (hear) which seems to be the most correct one.
This implies that Amir's room should be a treated or 'proper' one (which will probably be the case, picky as he seems to be).
An experienced listener should be able to set personal preferences aside for the most part.

In the short period Amir has the test rig I would say measure as much as possible and play with all kinds of corrections later. Maybe even by using a lot of smoothing as well. This way one can come to a workable correlation, providing a certain (standard) average listening SPL is used.

 

[Jimbob54]

Unless there's a new ground breaking standard that solves the issues. The only way to get satisfying result is through human ears. 3-5 very good ears is better than hundreds. Use human ears as equipment is the way out. Or developing a new standard. The whole hi-res thing is just bs to begin with. Wasted too much energy of the two most well known companies in the field.

In other news, mad trouserless man arrested on the streets of Seattle after trying to slice off human ears. "I only wanted good ears though , it's for SCIENCE!" he cackled at police.

 

[Mad_Economist]

Surely the calibration file is just something that offers a bit of a fudge factor to massage some of the measured anomolies to what they "think" the frequency response actually is at that point on the curve......then surely what we do with the frequency response is our business, in terms of fitting it to a Harman Curve or Diffuse Field or other? I mean that's the way I imagined it fitting together.

EDIT: although thinking about it the calibration would also include the HRTF of an average head discluding (maybe) the ears, because they're already physically simulated by the "rubber ears". So in conclusion I'm thinking the calibration file would include HRTF of an average head (discluding ears, maybe) and then on top of that would have the "fudge factor" I mentioned in the previous paragraph to account for measurement anomalies which may try to account for specific limitations of the kit including perhaps calibration for the actual microphone itself. Is that right, have I mapped that out right? If that's the case then the calibration files are gonna contain some influence as to whether it was decided to go after Headphone Harman Curve or Diffuse Field, etc, because they all have different measurement/simulation protocols in terms of speaker layout in the room which influences the HRTF that is measured. Yes, so it would be critical to know which one they were targetting, and I imagine they could provide different calibration files to target different 'simulations' for speaker layout in the room during HRTF calculation (eg Headphone Harman Curve or Diffuse Field, etc)?

EDIT#2: If my above paragraph is correct, then it's absolutely necessary for us to know which simulation/target they are using! Because that will affect how we use/interpret the results.

I must admit, I can't entirely follow what you're thinking the components of the calibration file will be here, but it seems quite complicated relative to what it actually is: HRTFs!

As an example, my 4128C came with a floppy disk (that's how you know this stuff is high tech ) containing the unit's free field HRTFs from many angles, as well as a diffuse field HRTF - both were specific to the unit in question, having being measured in some anechoic chamber presumably in Denmark (and possibly a reverb chamber, although sometimes it's easier to just infer DFHRTF from anechoic data...). These are simply saying "with a flat equalized speaker in an anechoic chamber, at this position relative to the HATS, the frequency response was this". A 0/0 (that is, frontally located and neither above or below the canal entrance) free field frequency response and a diffuse field frequency response are on page 2, figure 2.

In principle, our HATS and ear sims and suchlike are equivalent to population averages, and their HRTFs ostensibly match population averaged HRTFs, so we could also use averaged HRTF data from populations, and this is what some people do with 43AGs and other ear sims where placing the test device into a representative sound field won't actually show us a proper HRTF. At the end of the day, however, the compensation data is a frequency response reflecting the ear drum/ear simulator frequency response of whatever your "reference point" acoustic situation is - historically either a free or diffuse field, nowadays people like the Harman target, which is a bit of a mix of diffuse and free field (direct and reflected speaker sound), plus some preference filters.

Edit: In hopes of making what I'm saying clearer, have this visual aid taken from Audio-Precision's somewhat cursory but nonetheless quite good application note on headphone measurements:

 

[Robbo99999]

I must admit, I can't entirely follow what you're thinking the components of the calibration file will be here, but it seems quite complicated relative to what it actually is: HRTFs!

As an example, my 4128C came with a floppy disk (that's how you know this stuff is high tech ) containing the unit's free field HRTFs from many angles, as well as a diffuse field HRTF - both were specific to the unit in question, having being measured in some anechoic chamber presumably in Denmark (and possibly a reverb chamber, although sometimes it's easier to just infer DFHRTF from anechoic data...). These are simply saying "with a flat equalized speaker in an anechoic chamber, at this position relative to the HATS, the frequency response was this". A 0/0 (that is, frontally located and neither above or below the canal entrance) free field frequency response and a diffuse field frequency response are on page 2, figure 2.

In principle, our HATS and ear sims and suchlike are equivalent to population averages, and their HRTFs ostensibly match population averaged HRTFs, so we could also use averaged HRTF data from populations, and this is what some people do with 43AGs and other ear sims where placing the test device into a representative sound field won't actually show us a proper HRTF. At the end of the day, however, the compensation data is a frequency response reflecting the ear drum/ear simulator frequency response of whatever your "reference point" acoustic situation is - historically either a free or diffuse field, nowadays people like the Harman target, which is a bit of a mix of diffuse and free field (direct and reflected speaker sound), plus some preference filters.

Edit: In hopes of making what I'm saying clearer, have this visual aid taken from Audio-Precision's somewhat cursory but nonetheless quite good application note on headphone measurements:
View attachment 77120

Well you definitely covered the points I was raising, thanks for all of that well explained detail. So the HRTF is based on that specific mannequins head, so the questions we have to ask ourselves if we compare against other sites data:

1. Is the mannequin head different to say GRAS Oratory1990 and other sites (I'm thinking yes it's a different head).
2. What "simulation" have GRAS Oratory1990 & others used to determine the HRTF - eg the Freefield or Diffuse Field options you mentioned as possible configurations for your B&K.

Because the above 2 points are gonna affect the measurements greatly I would have thought as well it will affect the Target Curve for headphone EQ. Does that mean the B&K comes with different Target Curves associated with each of the HRTF simulations, in terms of for use to target headphone EQ's?

We'd also have to ask ourselves, which configuration file are we gonna use from that floppy disk, is it gonna be the Freefield or Diffuse Field (or any other options if there are some).....which one is most valid.

Am I right in thinking we can't really use the Harman Headphone Target Curve then with this B&K because that Harman Headphone Target Curve has a number of different assumptions baked into that frequency response curve than the B&K in relation to the HRTF: different mannequin head, perhaps different protocol for speaker positions (etc), it's also not Freefield nor Diffuse Field. If this is true, then perhaps it's only relevant to use the B&K in isolation to create headphone EQ's, with whatever target curves B&K provides? (EDIT: well you could add the Low Shelf Bass Boosts of the Headphone Harman Curve to whatever B&K provides I guess.)

EDIT #2: and if we do compare to other sites data we gotta ask why are we comparing & what do we want to learn about the B&K by doing so. Any comparisons would really only be halfway valid if they use the same Diffuse Field or same Free Field simulation, given that the mannequin head/ears would be different for a start. Any comparisons would just be to see how different they were, I'm not sure what other conclusions could be drawn.....although the value of learning how much different they are could give an indicator of how far off any measurement could be from an individual person out there in the population (but I feel studies have already been done to show how much variation there is in HRTF in populations). What it comes down to ultimately in terms of how valuable/worthwhile the B&K is would be the following question: how good do headphone EQ's from B&K sound in preference vs EQ's from other sites/mannequinns. The answer to that question would show whether B&K was a better average of someone's HRTF than the other averages out there associated with GRAS or other equipment - and that's a tricky thing to conclude in terms of gathering preferences from people & having enough people test it.

 

[Jimbob54]

Well you definitely covered the points I was raising, thanks for all of that well explained detail. So the HRTF is based on that specific mannequins head, so the questions we have to ask ourselves if we compare against other sites data:

1. Is the mannequin head different to say GRAS Oratory1990 and other sites (I'm thinking yes it's a different head).
2. What "simulation" have GRAS Oratory1990 & others used to determine the HRTF - eg the Freefield or Diffuse Field options you mentioned as possible configurations for your B&K.

Because the above 2 points are gonna affect the measurements greatly I would have thought as well it will affect the Target Curve for headphone EQ. Does that mean the B&K comes with different Target Curves associated with each of the HRTF simulations, in terms of for use to target headphone EQ's?

We'd also have to ask ourselves, which configuration file are we gonna use from that floppy disk, is it gonna be the Freefield or Diffuse Field (or any other options if there are some).....which one is most valid.

Am I right in thinking we can't really use the Harman Headphone Target Curve then with this B&K because that Harman Headphone Target Curve has a number of different assumptions baked into that frequency response curve than the B&K in relation to the HRTF: different mannequin head, perhaps different protocol for speaker positions (etc), it's also not Freefield nor Diffuse Field. If this is true, then perhaps it's only relevant to use the B&K in isolation to create headphone EQ's, with whatever target curves B&K provides? (EDIT: well you could add the Low Shelf Bass Boosts of the Headphone Harman Curve to whatever B&K provides I guess.)

I may be wrong, but the target curve is the target curve . End of.

The measurements are the measurements from whichever device is used.

The EQ settings are calculated to take the measured response as close to the curve as possible.

If the measurements are totally wayward, the EQ settings from in are way wrong too. But the target curve (whichever is chosen) does not get adjusted.

Showing all of my ignorance right there.

 

[Robbo99999]

I may be wrong, but the target curve is the target curve . End of.

The measurements are the measurements from whichever device is used.

The EQ settings are calculated to take the measured response as close to the curve as possible.

If the measurements are totally wayward, the EQ settings from in are way wrong too. But the target curve (whichever is chosen) does not get adjusted.

Showing all of my ignorance right there.

Ah, @Mad_Economist , so those HRTF files that came on your floppy disk (the different ones like Diffuse Field & Free Field) they would be the actual Target Curves then for headphone EQ? I suppose you could add Harman Headphone Low Shelf Bass Boost to those curves right, but what else could you do to those Target Curves legitimately? (When I say "legitimately", I suppose I'm talking about retaining an accurate Flat Speaker simulation......I suppose the only other thing you could do is add different degrees of downwards tilt to simulate different rooms, eg Harman have downward tilt associated with their listening room I think.)

EDIT: thinking about it, re "downward tilt" mentioned in previous paragraph, any downward tilt applied to the headphone target curve would have to weighted according to frequency because from the headphone target curves we can see some frequencies being attenuated more than others at arrival on eardrum, so any idea of applying a downward tilt would have to weighted differently at different points on the curve.

 

[solderdude]

Those correction curves (free field and diffuse field) are obtained with 2 different methods. The free field is done in an an-echoic room (as if in non reverberant outside conditions). That correction file is there to compensate for the HRTF of the HATS itself so that when a 'flat' response is coming at the HATS right from the front the plot that comes out of the system also is a flat line.
Very similar as to what comes out from a 'flat' microphone pointed at the same source at the same distance.
That's what the free field is there for... sounds coming from the front.

With headphones sound is coming from the side, totally different result so not usable.

The diffuse field correction is a correction done in a similar way but not with sounds coming from the front but from all around. So from the front, sides, rear... all around. This is yet another curve.
With headphones the sounds are only coming from the side so will also give the wrong results. But at least it is 'better' than freefield.

So both correction files (the opposite of the what the pinna and ear canal 'change' relative to an omnidirectional microphone) are not really suited when the goal is to get correlation between headphone and perceived sound.
But... it is a standard and standardised test rigs should (when presented with free and diffuse field signals) give similar end results.
There will be differences in what the mic capsule inside picks up though. Therefore each HATS will have its own correction files and you cannot really compare 'raw' plots (what the mic inside actually measured) because the used pinnae and ear canals are not the same.

The challenge for HATS users is to create (I assume HATS manufacturers by now will provide a correction file for this use case) an 'average' correction for headphones.
That file too will differ from HATS to HATS (for simplicity as one can fit it with various couplers and pinnae).
The problem is there is no 'reference' headphone like there are reference speakers. Large, small drivers, angled or not, pads etc. differ from headphone to headphone so that's the biggest issue.

So far the correction. Which is the crux here and starting point for another correction.
That would be the room correction. The reason for adding that to the correction is because speakers in a room that radiate a 'flat' response from their position(s) in the room will be changed due to room modes, reflections and distance (small effect) to the listener.
That 'change' in general will be a downward slope from higher frequencies and some 'boost' of lower frequencies.
Below a plot that shows (not exactly but ballpark) what 4 different 'room correction' plots look like

There are several standards that have been determined over the years. The Harman one (Olive-Welti) is based on preferences of listeners and changed a bit over the years as it is newer than the other ones.
That can easily be added to the 'corrected' signal of the microphone so that the resultant plot will be perceived as 'flat' (or should I say realistic) with the room correction.

That curve coming out of this system (coupler/HATS) = correction (diffuse field or specific for 'average' headphones) + 'room correction' based on the choice of the guy doing the measurements will result in a 'flat line plot' and is a target one can EQ on when played at reference level loudness.
Since Harman research has shown that people prefer the boosted bass (the reasons is another debate) an 'average' is established and when we take this more serious than other 'room corrections' then it makes sense for measurement folks/EQ tinkerers to take that as the most 'fitting' target.

 

[solderdude]

What is interesting for Amir to test is how the corrected results are compared to other HATS users that have documented the same headphone.
Of course there is tolerance and unit-to-unit variance of said headphones that can also show differences that do not come from the measurement rigs but from the headphones themselves. Unfortunately this can not be distinguished unless the same headphone or a whole bunch of them (Sonarworks for example) is used to get an accurate 'average' for that type of headphone. That should be done by all the other HATS users.

Fortunately the HD600 and HD650 are pretty constant over a larger production period, have little pinna action and have very little production spread.
From my P.O.V. using this headphone can be used as a 'reference' (just not a 'flat' one) to evaluate/compare results.
The other headphones are also interesting when they make use of pinna activation, have large or very small diameter drivers, angled ones etc.
These headphones are useful to compare how far off the 'corrections' of HATS are opposite each other.

The next question would be which one is the most 'correct' one. This requires listening to the headphones and comparing them with speakers (in a conditioned room by an experienced listener) That's where Amir comes in, in this particular test he is about to do.

Very curious about the results and if and how he will take this further.

 

[Jimbob54]

What is interesting for Amir to test is how the corrected results are compared to other HATS users that have documented the same headphone.
Of course there is tolerance and unit-to-unit variance of said headphones that can also show differences that do not come from the measurement rigs but from the headphones themselves. Unfortunately this can not be distinguished unless the same headphone or a whole bunch of them (Sonarworks for example) is used to get an accurate 'average' for that type of headphone. That should be done by all the other HATS users.

Fortunately the HD600 and HD650 are pretty constant over a larger production period, have little pinna action and have very little production spread.
From my P.O.V. using this headphone can be used as a 'reference' (just not a 'flat' one) to evaluate/compare results.
The other headphones are also interesting when they make use of pinna activation, have large or very small diameter drivers, angled ones etc.
These headphones are useful to compare how far off the 'corrections' of HATS are opposite each other.

The next question would be which one is the most 'correct' one. This requires listening to the headphones and comparing them with speakers (in a conditioned room by an experienced listener) That's where Amir comes in, in this particular test he is about to do.

Very curious about the results and if and how he will take this further.

Its going to be interesting alright. Much debate. Many viewpoints

 

[Doodski]

ASR has a fair amount of traffic wanting headphone advice. The resource of headphone tests done by @amirm will be popular. Have a lil faith in this people.

 

[Mad_Economist]

Ah, @Mad_Economist , so those HRTF files that came on your floppy disk (the different ones like Diffuse Field & Free Field) they would be the actual Target Curves then for headphone EQ? I suppose you could add Harman Headphone Low Shelf Bass Boost to those curves right, but what else could you do to those Target Curves legitimately? (When I say "legitimately", I suppose I'm talking about retaining an accurate Flat Speaker simulation......I suppose the only other thing you could do is add different degrees of downwards tilt to simulate different rooms, eg Harman have downward tilt associated with their listening room I think.)

EDIT: thinking about it, re "downward tilt" mentioned in previous paragraph, any downward tilt applied to the headphone target curve would have to weighted according to frequency because from the headphone target curves we can see some frequencies being attenuated more than others at arrival on eardrum, so any idea of applying a downward tilt would have to weighted differently at different points on the curve.

They can be used as target curves, but ultimately they are simply measurements of the eardrum response of the head under certain acoustic conditions. The 0 elevation, 0 azimuth free field HRTF, if applied as compensation like a mic calibration file, will result in a flat line from a frontally located sound source with a flat frequency response, in an anechoic chamber or other situation where there are no other incidences for sound.

Well you definitely covered the points I was raising, thanks for all of that well explained detail. So the HRTF is based on that specific mannequins head, so the questions we have to ask ourselves if we compare against other sites data:

1. Is the mannequin head different to say GRAS Oratory1990 and other sites (I'm thinking yes it's a different head).
2. What "simulation" have GRAS Oratory1990 & others used to determine the HRTF - eg the Freefield or Diffuse Field options you mentioned as possible configurations for your B&K.

Because the above 2 points are gonna affect the measurements greatly I would have thought as well it will affect the Target Curve for headphone EQ. Does that mean the B&K comes with different Target Curves associated with each of the HRTF simulations, in terms of for use to target headphone EQ's?

We'd also have to ask ourselves, which configuration file are we gonna use from that floppy disk, is it gonna be the Freefield or Diffuse Field (or any other options if there are some).....which one is most valid.

Am I right in thinking we can't really use the Harman Headphone Target Curve then with this B&K because that Harman Headphone Target Curve has a number of different assumptions baked into that frequency response curve than the B&K in relation to the HRTF: different mannequin head, perhaps different protocol for speaker positions (etc), it's also not Freefield nor Diffuse Field. If this is true, then perhaps it's only relevant to use the B&K in isolation to create headphone EQ's, with whatever target curves B&K provides? (EDIT: well you could add the Low Shelf Bass Boosts of the Headphone Harman Curve to whatever B&K provides I guess.)

EDIT #2: and if we do compare to other sites data we gotta ask why are we comparing & what do we want to learn about the B&K by doing so. Any comparisons would really only be halfway valid if they use the same Diffuse Field or same Free Field simulation, given that the mannequin head/ears would be different for a start. Any comparisons would just be to see how different they were, I'm not sure what other conclusions could be drawn.....although the value of learning how much different they are could give an indicator of how far off any measurement could be from an individual person out there in the population (but I feel studies have already been done to show how much variation there is in HRTF in populations). What it comes down to ultimately in terms of how valuable/worthwhile the B&K is would be the following question: how good do headphone EQ's from B&K sound in preference vs EQ's from other sites/mannequinns. The answer to that question would show whether B&K was a better average of someone's HRTF than the other averages out there associated with GRAS or other equipment - and that's a tricky thing to conclude in terms of gathering preferences from people & having enough people test it.

1: All standards-compliant (the pertinent ones being ITU-T P57/58, IEC60318-4/7) mannequins and ear simulators have transfer functions that fall within certain tolerance bands given in the pertinent standards. Independent measurements of mannequin HRTF show that the commercially available models generally conform to these specifications, although Head Acoustics occasionally pushes the limits a bit. Thus while the transfer functions of the mannequins or ear simulators may differ somewhat - and model and unit-specific compensation data is preferable for this reason - there are limits to the range of variation. You cannot have an apples-to-apples comparison of uncompensated KEMAR data to uncompensated 5128 data, but the same data compensated to each unit's diffuse field HRTF should be comparable, and without compensation it should still be "close".

2: There are two targets in broad use for compensated headphone response: The first is diffuse field (drawn, as said, either from population averages, or the specific mannequin in question's HRTF in a diffuse field), which is a well defined sound field that should be possible to approximate (for example via the free field summation method, see Hammershoi & Moller 1995) by pretty much anyone willing to put in the work with any anthropomorphic system (and which most manufacturers will supply compensation data for).

The second is the Harman target, which is based on a set of in-room measurements within Harman's listening room, conducted with a 43AG set into a mannequin head, adjusted by a pair of shelf filters that a large population of testers were allowed to adjust to their preferences. The "baseline" of the Harman response is highly smoothed due to its origins as in-room unwindowed measurement, so the 43AG/KEMAR pinna specific HRTF elements it contains are fairly smoothed down, but applying it directly to data from another HATS would be improper. I generally champion the premise of either directly applying the Harman shelf filters to the DF-HRTF of the HATS in question, or if one feels particularly motivated, technically something even closer could be synthesized via Chris Struck's target methodology given that the acoustic parameters of the Harman room and the directivity of the Revels used in the Olive-Welti 2013 paper are known.

Broadly, if you want something that's got a well established statistical correlation with subjective perception, a GRAS unit and the direct use of the Harman target is the best validated approach - this would be the closest analogy to the speaker preference calculations already featured on this site. However, as said, Sean may be able to provide compensation data for the 5128, as he has had one for a while, and I see little reason to believe that the statistical robustness of the GRAS derived results would be absent with the 5128 (bar the aforementioned in-ear issues).