Measuring HRTF for headphone use

[pkane]

In all fairness to Griesinger, in addition to using multiple (albeit not many) individuals as Mitchco noted, he makes reference to the work of Henrik Møller and Dorte Hammershøi, which includes a paper which could be construed as supporting his premises - note section 3.1 and figure 11.

Edit: apologies, I have two Hammershøi and Møller papers open and appear to have linked the wrong one. Corrected link impending.

Edit 2: Prior to correction, my AES link to Hammershøi & Møller directed to this moderately pertinent - but not including the chart I was trying to indicate - paper

Quote from the 1995 paper, conclusion:

The choice of reference point at the blocked ear canal entrance has proven rational [...] The variation between subjects of HRTFs measured at the blocked ear canal is smaller than at the same point with open ear canal. The directional characteristics of the sound transmission to the ear are thus best described by blocked ear canal HRTFs.

 

[Mad_Economist]

Quote from the 1995 paper, conclusion:
The choice of reference point at the blocked ear canal entrance has proven rational [...] The variation between subjects of HRTFs measured at the blocked ear canal is smaller than at the same point with open ear canal. The directional characteristics of the sound transmission to the ear are thus best described by blocked ear canal HRTFs.

Indeed, and I would agree - blocked canal has since been shown to be at the least perceptually accurate enough for very good subjective responses to blocked canal corrections. IMO it is validated as the default.

However, while I feel that the onus is on Griesinger to demonstrate the perceptual relevance of this, I will point out that it is the bolded section of this quote that Griesinger himself is referencing - he feels that the open canal's differences are indicative of some interaction which negatively impacts the timbre of HRTF corrections done with a blocked canal. I am highly skeptical, but I've been wrong before.

The conclusions for the 6 individuals are not based on in-ear measurements, but rather on a listening test trying to match volumes at different frequencies. It appears to be using a similar method to the one I was proposing for measuring without a microphone. What's not clear to me is how accurate or sensitive this method is, and if there might exist a significant variation between individuals in their ability to match volumes properly across frequencies.

Frequency response determination by subjective comparison with loudspeakers is an older approach to determining headphone frequency response - IIRC it was first established in IEC268-7. Memory serving, neither Hammershøi & Møller nor Theile were pleased with its consistency when tested, and I would say that it is not the mainstream approach, but I think there is some pertinence to that sort of comparison in some circumstances - if you measure the same response but perceive a timbre difference, it shows something's amiss, no?

Edit: Theile's complaints regarding the subjective loudness matching method are well covered in section 2 here.

 

[March Audio]

If it were a mature area I would ask for the canonical texts.

Such texts would cast light on background for inquiry. Because this is an application oriented crowd I wonder how our insight into the problem can help us design and find neutral headphones, cfr. @Floyd Toole ’s many practical insights into speakers.

You are off again. You need to stop asserting that this is an unknown unresearched unscientific area. See olive, as mentioned, or maybe this, the Audeze article etc etc
https://www.bksv.com/en/about/waves/WavesArticles/2014/inside-headphones

Or maybe some of your own searches on the subject. The same message keeps coming out. It's actually a simple problem, but difficult to solve, until we find a cheap and consistent way of measuring individuals ears.

 

[pkane]

Frequency response determination by subjective comparison with loudspeakers is an older approach to determining headphone frequency response - IIRC it was first established in IEC268-7. Memory serving, neither Hammershøi & Møller nor Theile were pleased with its consistency when tested, and I would say that it is not the mainstream approach, but I think there is some pertinence to that sort of comparison in some circumstances - if you measure the same response but perceive a timbre difference, it shows something's amiss, no?

Edit: Theile's complaints regarding the subjective loudness matching method are well covered in section 2 here.

The reason I was thinking of using this kind of test was because it incorporates my hearing idiosyncrasies into the measurement. Correcting for such a transfer function would be addressing errors in the transducer, the HRTF, and my hearing deviations from flat. Not too many things left to correct! But I can easily see how these tests can be inconsistent between individuals, and vary for a single test subject at different times of day, or after running a test or two.

 

[mitchco]

@svart-hvitt In this section on The Science of Preferred Responses for Headphones and Loudspeakers, I summarised Sean Olive and teams research and had Sean review it for any errors. Scroll to the end of the section where there are several reference links. This link to one of Sean's presentations contains references to 14 AES papers on the subject area.

 

[pkane]

I tried some tests without a microphone. My method is a bit different than Griesinger's. It's basically a way to measure the lower threshold of hearing through the headphones. It may not be representative of HRTF at normal volume levels, but curious nevertheless.

I wrote a software program (Earful) that plays pure sine tone through the headphones at various frequencies. I can increase and decrease volume in increments of 0.5dB. It takes me through a logarithmic sweep in the 20-20KHz range, letting me find the lowest volume where I can still hear the tone at each frequency. Here's the result. My right ear, with HE560 headphones, no DSP applied, 50 points measured, two test runs averaged:

The level in the chart is relative: 0dB represents a 0dB level of the sine wave (-1 to 1 amplitude). I didn't measure the actual SPL at the headphone.

While it looks like I'm losing hearing (not surprising, I'm well over 50), this is a combination of my ear response and HE-560s. From my measurements HE560s are down about 12dB from 700Hz to 20Khz, which seems to follow the pattern in the above threshold curve. At least some of this is due to the headphones, maybe some due to the noisy environment (in the basement with AC units running), but yes, I'm losing some HF hearing, can't deny it

Maybe Earful can be a good tool for audiologists out there, or for the golden-ear audiophiles to prove their abilities

 

[dc655321]

Maybe Earful can be a good tool for audiologists out there

But, isn't this very nearly how audiologists determine hearing acuity already?

How are you concluding a hearing loss is indicated here?
This looks Fletcher-Munson (-ish) for a reasonably loud source (~90 dB SPL).

Nice work, BTW...

 

[pkane]

But, isn't this very nearly how audiologists determine hearing acuity already?

How are you concluding a hearing loss is indicated here?
This looks Fletcher-Munson (-ish) for a reasonably loud source (~90 dB SPL).

Nice work, BTW...

Thank you! Yes, I realized after I thought about this method that that's exactly how it's measured by an audiologist.

I'm not sure if there is or isn't hearing loss, but I expected a slightly less exponentially rising curve after about 7KHz. Maybe it was just wishful thinking

I'll add the Fletcher-Munson lower threshold equal-loudness curve to the software, to make it easier to compare. Probably should also add the option to subtract the FR measurement of the headphones, so that the 'real' hearing threshold can be displayed.

EDIT: Added ISO-226-2003 curve (red) next to my threshold measurement, for comparison.

 

[solderdude]

you could probably drop the red line about 10dB in level to get a better sense of the actual hearing loss.

I found that the bottom hearing limit and hearing at say 70dB or 80dB SPL have little to do with each other.
You cannot base any HRTF on it.
Measuring the ears is a good thing though.
Your brain calibrates itself to daily sounds so is why you generally don't notice some hearing loss unless it is sudden (ear clogged) or people start complaining about you not hearing anything.

When older and still being able to hear 20kHz chances are the level needs to be turned up a lot. In this case you might well hear distortion or sideband products and not the actual tone.

Audiologists usually only measure up to 8kHz as they are only interested in speech, not in fidelity of music.
They use calibrated headphones. The headphones themselves are characterized and a correction for that model is applied (which can be selected from a just few types).
Usually the plots they generate show a loss of response relative to average humans with good hearing (young people), not calibrated to 0dB SPL.

I once took my ear measurements (in a similar format but calibrated to actual SPL) to my GP and she had no idea what it meant.
Had to go to an audiologist to get my ears measured which generates numbers that she 'understood'.

HRTF for headphones is quite different when sounds are coming from the side of the head only (as with headphones) angling drivers does make a change so does the Concha a little but not nearly as much as the ear plots show.

There is an online 'hearing test' as well. ( http://newt.phys.unsw.edu.au/jw/hearing.html ). Should be used with known flat speakers or something like the HD600 or HD650.
Lower bass will be 'off' in that case though but is well documented.

There can be some low level 'ticks' audible when an (inaudible) tone starts and stops though.

 

[pkane]

you could probably drop the red line about 10dB in level to get a better sense of the actual hearing loss.

I found that the bottom hearing limit and hearing at say 70dB or 80dB SPL have little to do with each other.
You cannot base any HRTF on it.
Measuring the ears is a good thing though.
Your brain calibrates itself to daily sounds so is why you generally don't notice some hearing loss unless it is sudden (ear clogged) or people start complaining about you not hearing anything.

When older and still being able to hear 20kHz chances are the level needs to be turned up a lot. In this case you might well hear distortion or sideband products and not the actual tone.

Audiologists usually only measure up to 8kHz as they are only interested in speech, not in fidelity of music.
They use calibrated headphones. The headphones themselves are characterized and a correction for that model is applied (which can be selected from a just few types).
Usually the plots they generate show a loss of response relative to average humans with good hearing (young people), not calibrated to 0dB SPL.

I once took my ear measurements (in a similar format but calibrated to actual SPL) to my GP and she had no idea what it meant.
Had to go to an audiologist to get my ears measured which generates numbers that she 'understood'.

HRTF for headphones is quite different when sounds are coming from the side of the head only (as with headphones) angling drivers does make a change so does the Concha a little but not nearly as much as the ear plots show.

There is an online 'hearing test' as well. ( http://newt.phys.unsw.edu.au/jw/hearing.html ). Should be used with known flat speakers or something like the HD600 or HD650.
Lower bass will be 'off' in that case though but is well documented.

There can be some low level 'ticks' audible when an (inaudible) tone starts and stops though.

All good points.

HRTF is still functional at the low threshold of hearing, but it's not clear if that's the dominant effect there -- most likely not. For this reason, I don't think Earful is a good way to measure HRTF, but I knew that going in. I'm still thinking about how to change this to measure unobstructed ear response at normal listening levels.

I was going to add the ability to use DSP to flatten out the output of Earful to correct for headphone frequency response. This should let me measure the pure ear response by removing any response variations in the headphone. Seems like a bit more work than I'm willing to do at the moment -- adding a proper convolution engine would take some effort.

Clicks occur when the output transitions from one waveform to another, suddenly. The same problem can occur when adjusting digital volume by a large amount, not just when switching frequencies. I've taken care of this by having a gradual transition between two tones using a smooth window function -- no clicks now.

I also found it hard to detect the tone at such a low level. Sometimes I thought the tone was still playing, but after muting/unmuting, I realized the tone continued to play regardless: it was all in my imagination. For this reason, I made the tone play interrupted, similar to a phone busy signal. One second on, one second off. This is much easier to hear and to detect. When I hear this, I know I'm not imagining it

 

[Shikamon]

I'm sooooo glad I found this thread online! I've been working at finding my own personal HRTF ever since I watched David Griesinger's video on YouTube. I have tried several attempts (even made my own little in ear measuring device) but seemed to getting barely close to sounding right.

My way of getting my HRTF is very much Griesinger's method, but I feel that I'm missing something. The playing tones back and forth kinda confuses me on whether higher frequencies are too loud or soft when compared to 500hz or so. Seeing your method pkane is something I haven't tried. Is it worth doing when you got your results?

 

[pkane]

I'm sooooo glad I found this thread online! I've been working at finding my own personal HRTF ever since I watched David Griesinger's video on YouTube. I have tried several attempts (even made my own little in ear measuring device) but seemed to getting barely close to sounding right.

My way of getting my HRTF is very much Griesinger's method, but I feel that I'm missing something. The playing tones back and forth kinda confuses me on whether higher frequencies are too loud or soft when compared to 500hz or so. Seeing your method pkane is something I haven't tried. Is it worth doing when you got your results?

Hi Shikamon,

What are you trying to achieve by measuring HRTF?

My method produces the lower threshold measurement that includes HRTF, any hearing deficiencies, and the frequency response of the headphones. So, yes, it produces a curve but perhaps not the one you are looking for.

Griesinger’s two tone method is something I’m going to experiment with but I can see how a low frequency reference can be hard to use with a high frequency test tone. I have a few ideas on how to improve on this test, but not enough time to test them all

What I found worked well for me to correct for headphone FR and pinna interference (part of HRTF) is to use in-ear microphones with headphones on top. Correcting for this measured curve produced the best sounding audio through headphones that I’ve heard so far.

 

[Shikamon]

Hey pkane! I plan on try to find my HRTF based on what my ears hear when listening to a anechoically flattened speaker in my room. Getting the data of my speaker response in a quasi anechoic way is a bit tough when doing it in my small bedroom. I tried doing it outside too, which helped, but the environment is hard to control (noise and whatnot). But I think I would like to get my in ear mic device working again as that makes it a bit easier for me in doing Griesinger's method, which at the end sounds really good when done properly.

 

[pkane]

Hey pkane! I plan on try to find my HRTF based on what my ears hear when listening to a anechoically flattened speaker in my room. Getting the data of my speaker response in a quasi anechoic way is a bit tough when doing it in my small bedroom. I tried doing it outside too, which helped, but the environment is hard to control (noise and whatnot). But I think I would like to get my in ear mic device working again as that makes it a bit easier for me in doing Griesinger's method, which at the end sounds really good when done properly.

I made an attempt to compensate for pinna/HRTF and room+speaker response using in-ear mics. Unlike with the headphones, the result was disappointing. It was a quick test, so it’s possible I messed up something in the process.

 

[Shikamon]

I made an attempt to compensate for pinna/HRTF and room+speaker response using in-ear mics. Unlike with the headphones, the result was disappointing. It was a quick test, so it’s possible I messed up something in the process.

Ah I see... It's tough but I feel that I'm getting close to what I like coming from my headphones. Let me know how you fair with Griesinger's method. Hope you find a better work around than I can. lol

 

[solderdude]

Below a few brainfarts from me.

Let's suppose one could determine the measured frequency response at the ear entrance of one's ear.
So flush with the entrance and not a few mm in front of it.
Assuming the used speaker is 'flat' at 1 m in anechoic room and you could accurately measure the FR using noise while averaging measurements moving the head slightly around at 1 m distance in free field (no noises around you).
You would also be able to do the same but at 1m to your right and would be able to subtract those two responses and get the difference in FR.
Then you could also do this at a few meters away and with 2 speakers in a typical listening position a few meters apart and follow up with 1 speaker on your side again... as many meters away and get that HRTF.
These won't be the same, even when speakers and mics are perfect.
Which HRTF to choose ?

Let's assume you found the difference in FR, most likely a peak like that from Concha gain and Pinna as well as Tragus influence but without the ear canal gain as shown in plots by folks who already done this.
Let's assume there are differences found.. What would one do with this info and is the measurement or research by others valid ?

Forget the above .... Let's assume you found your FR differences and EQ'ed the speaker to the side of your ear with that compensation.
Will the tonal balance be the same in that case when switched between a front and side speaker or would phase differences still throw dirt in the mix ?

Now let's assume the FR did match exactly and you would use a (perfect) headphone with the same EQ would the tonal balance be the same as that from the speaker x meters away ?

I fear not... and will explain.
The speaker at a distance could be considered as a point source. audio-waves entering your ear will come from a specific axis and 'excite' the Concha in a certain way.
Now imagine the point source would be a wall of speakers a few meters high and wide all set up in a straight plane (not circular)
Would you still measure the same frequency response ?
The answer is no as phase shifts due to speed of sound and Concha activation at different angles result in different FR.

Well ... your headphone driver is the equivalent of that AND enclosed in a small confined space as well.

Now it is time to incorporate the following variances when it comes to headphones.
Driver diameter, driver-ear distance, driver position in the baffle, modal differences across the membrane/cone differing per frequency (planar/dynamic), porting, resonances (and other driver properties), pad bounce, amount of air trapped inside, leakage of pads (intentional or not), driver angle, position of the headphone over the ears.
All these differences will result in a different Concha activation and Pinna properties.
Not to mention the wildly varying frequency response (no headphone will ever be as flat as a speaker at 1m in free field).
So the derived frequency compensation is worthless when using headphones, especially between different models.

Rtings is one of the few (if not the only one I am aware off) that measures and publishes the effect a standard fake Pinna has on the measurements.
He measures each headphone without a Pinna and with a Pinna and compares that to a 'reference' he determined.
This shows no headphone reacts the same when all other variables remain the same (same dummy head, same headphone, same position).
Look at the section he calls: soundstage and compare different headphones.
You can watch his video that touches on this subject amongst other things and check out his measurements.
He also has a section explaining his measurements.

Then you can come to a conclusion that a single HRTF measurement is no measure for any kind of correction and that such EQ will not lead to better SQ.
Instead, when you want to EQ anyway I suggest to just set the EQ so that the headphones will sound to your preference (which also comes in play) at the SPL you prefer to listen to and just enjoy the music.
95% of all headphones are really crappy when it comes to frequency response flatness anyway. Sharp dips and peaks cannot be corrected perfectly anyway.

 

[pkane]

Then you could also do this at a few meters away and with 2 speakers in a typical listening position a few meters apart and follow up with 1 speaker on your side again... as many meters away and get that HRTF.
These won't be the same, even when speakers and mics are perfect.
Which HRTF to choose ?

Are you saying that frequency response of HRTF changes with the angle of incidence of the sound wave? Can you provide any references that demonstrate this?

Forget the above .... Let's assume you found your FR differences and EQ'ed the speaker to the side of your ear with that compensation.
Will the tonal balance be the same in that case when switched between a front and side speaker or would phase differences still throw dirt in the mix ?

Phase is a separate issue, IMO. Phase does not alter the frequency, so anything you do to correct for FR errors will not suddenly become useless if phase is not right. Certainly errors in phase can destroy the sense of direction and possibly interfere with the 'soundstage' perception, so they need to be accounted for, but does this somehow negate the benefit of correcting for FR errors?

Now let's assume the FR did match exactly and you would use a (perfect) headphone with the same EQ would the tonal balance be the same as that from the speaker x meters away ?

I fear not... and will explain.

Maybe not, but Harman testing does indicate that subjective preference for both, speakers and headphones is for a flat-line response, with a bit of a tilt from low to high frequencies. My own testing with speakers and headphones shows the same. While I don't get the same sense of space with EQ'ed headphones as I do with speakers, the tonal balance is just as accurate.

The answer is no as phase shifts due to speed of sound and Concha activation at different angles result in different FR.

Again, please explain how a phase shift results in a different frequency response?

Driver diameter, driver-ear distance, driver position in the baffle, modal differences across the membrane/cone differing per frequency (planar/dynamic), porting, resonances (and other driver properties), pad bounce, amount of air trapped inside, leakage of pads (intentional or not), driver angle, position of the headphone over the ears.
All these differences will result in a different Concha activation and Pinna properties.

There is definitely some variation with position. And headphone/ear is a complex non-linear system, I grant you that. But does that mean that no correction can or should be undertaken? That HRTF and headphone FR errors cannot be corrected for at all? I don't believe that's true.

Not to mention the wildly varying frequency response (no headphone will ever be as flat as a speaker at 1m in free field).
So the derived frequency compensation is worthless when using headphones, especially between different models.

I don't see how the complexity of headphone frequency response leads to the conclusion that it can't be as flat as a speaker. FR can be corrected. Phase response can also be corrected through DSP, as long as it's some measurable function. If you are saying that headphones can't simulate the multiple incidence angles that a speaker can, then yes. But a speaker in an anechoic chamber will have the same problem. The multiple incidence angles in a normal room come from the reflections in the room, and therefore are not part of the recorded audio. They are a distortion added by the room.

I do agree that no single frequency compensation will work using different headphones. None will probably work for different listeners with the same headphones. This is why I want to be able to measure my own HRTF in the first place!

Rtings is one of the few (if not the only one I am aware off) that measures and publishes the effect a standard fake Pinna has on the measurements.
He measures each headphone without a Pinna and with a Pinna and compares that to a 'reference' he determined.
This shows no headphone reacts the same when all other variables remain the same (same dummy head, same headphone, same position).
Look at the section he calls: soundstage and compare different headphones.
You can watch his video that touches on this subject amongst other things and check out his measurements.
He also has a section explaining his measurements.

Then you can come to a conclusion that a single HRTF measurement is no measure for any kind of correction and that such EQ will not lead to better SQ.

I agree that there's no standard HRTF correction curve. That each headphone will interact differently with different HRTFs. That's why I want my own, personal measurements so I can correct the playback to my set of ears and my specific headphones.

Instead, when you want to EQ anyway I suggest to just set the EQ so that the headphones will sound to your preference (which also comes in play) at the SPL you prefer to listen to and just enjoy the music.
95% of all headphones are really crappy when it comes to frequency response flatness anyway. Sharp dips and peaks cannot be corrected perfectly anyway.

Ultimately, that's what I do. But I prefer to start with something a little more determined, such as a reference curve that gets me close to the desired state and then make minor tweaks. I'd rather not start by arbitrarily changing frequency and phase response to achieve a better sound. In my experience, this leads down a rabbit hole where you get improvements on some content, and much worse performance on others.

 

[solderdude]

Are you saying that frequency response of HRTF changes with the angle of incidence of the sound wave? Can you provide any references that demonstrate this?

Freefield vs diffuse field does not yield the same response.
Or be your own reference, Set up a speaker in free field. Put noise on it and rotate your head.

Phase is a separate issue, IMO. Phase does not alter the frequency, so anything you do to correct for FR errors will not suddenly become useless if phase is not right. Certainly errors in phase can destroy the sense of direction and possibly interfere with the 'soundstage' perception, so they need to be accounted for, but does this somehow negate the benefit of correcting for FR errors?

With 2 speakers spaced apart or one big membrane you do get phase shifts which will alter certain frequencies depending on the wavelength (combfilter) so phase can alter the FR in those cases which is what I was pointing at.

Maybe not, but Harman testing does indicate that subjective preference for both, speakers and headphones is for a flat-line response, with a bit of a tilt from low to high frequencies. My own testing with speakers and headphones shows the same. While I don't get the same sense of space with EQ'ed headphones as I do with speakers, the tonal balance is just as accurate.

Yes, I fully agree and is not the point I am trying to get across.
I am saying that speakers (being point sources at a distance) are not the same as headphone drivers which produce a non uniform wavefront (more uniform in planars) and don't act like point sources.
I am also putting forward that speakers in a conditioned room or anechoic will show a much flatter response than headphones can.

There is definitely some variation with position. And headphone/ear is a complex non-linear system, I grant you that. But does that mean that no correction can or should be undertaken? That HRTF and headphone FR errors cannot be corrected for at all? I don't believe that's true.

I fully agree that FR response should be 'corrected' if possible. I may even be the biggest proponent of this.
If one wants one can ALSO introduce a tilt to the warmer side if ones taste desires such.
Also a big proponent of this.

What I am trying to get across is that the difference in a freefield speaker experiment to determine the tonal change differences with speakers from the front and side will yield a curve. You could use that but is inaccurate as is clearly shown by measurements from Sam (Rtings)

I don't see how the complexity of headphone frequency response leads to the conclusion that it can't be as flat as a speaker. FR can be corrected. Phase response can also be corrected through DSP, as long as it's some measurable function. If you are saying that headphones can't simulate the multiple incidence angles that a speaker can, then yes. But a speaker in an anechoic chamber will have the same problem. The multiple incidence angles in a normal room come from the reflections in the room, and therefore are not part of the recorded audio. They are a distortion added by the room.

Of course you can correct with measurements. It's what I do. But WHAT are you correcting for. You are correcting opposite a measurement and a total system. The quality of the measurement will determine the correction. This stands and falls with correction/compensation applied and quality of measurements.
It's how Sonarworks 'works'. Since their correction is based on a HATS, custom calibration of their own and based on multiple measurements over various units (or your personal headphone when sent in or bought) it will be more tonal balanced than when not corrected.
Your HRTF may differ somewhat from theirs.

I do agree that no single frequency compensation will work using different headphones. None will probably work for different listeners with the same headphones. This is why I want to be able to measure my own HRTF in the first place!

I am fully convinced the actual soundwaves reaching the eardrum will be more similar than you may think. I think the biggest difference between how people perceive the same headphones are:
A: Taste/preference (think basshead vs treblehead)
B: Seal (don't underestimate the effect with closed headphones
C: used recordings / genre

When you figured out your own HRTF I suspect you will find it won't be much different than the ones others already figured out. A slight change in frequency peak (perhaps) and amplitude (maybe) but there won't be huge differences that can make the sound go from acceptable to heavenly.

I agree that there's no standard HRTF correction curve. That each headphone will interact differently with different HRTFs. That's why I want my own, personal measurements so I can correct the playback to my set of ears and my specific headphones.

I suspect your HRTF won't differ much from those of others. You will have to correct each headphone differently.

Ultimately, that's what I do. But I prefer to start with something a little more determined, such as a reference curve that gets me close to the desired state and then make minor tweaks. I'd rather not start by arbitrarily changing frequency and phase response to achieve a better sound. In my experience, this leads down a rabbit hole where you get improvements on some content, and much worse performance on others.

A noble goal. The point I am trying to get across is that your references must be good.
Seeing a 'straight line' from a microphone (spec sheet or folder) does not mean it performs that way unless it is a calibrated plot of that particular mic used under the same circumstances. That mic thus should be flush with the entrance of the ear canal and not in a different position for instance.
The same goes for the reference of the speakers you want to use. Would have to measure/confirm in an anechoic room or outside in free field (quiet surrounding) for the baseline and assume this is correct.
Then when using a headphone would need to subtract the FR of the raw headphone. That would need to be measured to be obtained.
Here again simply using a 'flat' mic acc. to specs won't provide the correct response as a mic in a small sealed chamber does not measure the same as in freefield.
That's at least 3 uncertain reference points to obtain that which ultimately will vary just a little bit (when done perfectly with good references) with the research already done.

By the way I am all for measuring and referencing. Also for applying corrections. Just pointing out some (possible) pitfalls when trying to get accurate results where you can actually depend on.
I encourage you to measure and above all try to improve the sound.
What I found was it is more important to know your own limits (in hearing abilities) and understand the differences in taste.

What makes you believe that your HRTF is significantly different (that warrants different correction) than say the research done below with only the combined 3 and 4 traces ?

of course this does not go past 10kHz and there certainly is valuable info above 10kHz in music.
It sort-of coincides with Rtings findings.

 

[andreasmaaan]

Maybe not, but Harman testing does indicate that subjective preference for both, speakers and headphones is for a flat-line response, with a bit of a tilt from low to high frequencies. My own testing with speakers and headphones shows the same. While I don't get the same sense of space with EQ'ed headphones as I do with speakers, the tonal balance is just as accurate.

Just to be clear, in respect of speakers, Harman testing indicates that a perfectly flat on-axis response is most preferred.

The low-to-high tilt is not about the on-axis response, but about the in-room response, which includes both on-axis (direct) sound and off-axis (reflected) sound.

Harman is not suggesting that speakers should have a downward-tilting on-axis response, but rather that they should have a flat-axis response with a downward-tilting power response.

 

[pkane]

Just to be clear, in respect of speakers, Harman testing indicates that a perfectly flat on-axis response is most preferred.

The low-to-high tilt is not about the on-axis response, but about the in-room response, which includes both on-axis (direct) sound and off-axis (reflected) sound.

Harman is not suggesting that speakers should have a downward-tilting on-axis response, but rather that they should have a flat-axis response with a downward-tilting power response.

That's correct. The tilted curve is what sounds best at a normal listening position in a typical room based on these studies. That's the target curve to correct to at the listening position, not at the speaker or in an anechoic chamber.