Measuring HRTF for headphone use

[pkane]

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.

The fact that heads are shaped differently, ears are shaped and positioned differently, resonances within one's head are different, all lead me to believe that HRTFs will also differ. Just by how much, I don't know, but that was the whole reason to try to measure this for myself. There are also some published studies that seem to address this specific issue: customizing headphone response to the specific qualities of the listener in order to achieve better, more realistic playback.

 

[andreasmaaan]

The fact that heads are shaped differently, ears are shaped and positioned differently, resonances within one's head are different, all lead me to believe that HRTFs will also differ. Just by how much, I don't know, but that was the whole reason to try to measure this for myself. There are also some published studies that seem to address this specific issue: customizing headphone response to the specific qualities of the listener in order to achieve better, more realistic playback.

Would be interested to have a look at these studies if you have some links handy?

 

[pkane]

Would be interested to have a look at these studies if you have some links handy?

Here are a few:

Headphone-To-Ear Transfer Function Estimation
Summary:

The estimation results show that HpTFs vary considerably with headphones and ear canals, which suggests that individualized compensations for HpTFs are necessary for headphones to reproduce desired sounds for different listeners.

Assisted Listening for headphones and hearing aids: Signal Processing Techniques
Relevant Extract:

• Human pinna is found to be as idiosyncratic as the fingerprint
• HRTFs are highly individual and differs substantially from one subject to the other
• For perfect 3D audio playback, individualized recordings/HRTFs and individualized headphone equalization are required

HRTF Personalization

Individualized head related transfer functions (HRTFs) are needed for accurate rendering of spatial audio, which is important in many applications.

Interesting: https://www.researchgate.net/public...naural_headphone_reproduction_in_web_browsers

 

[andreasmaaan]

Here are a few:

Headphone-To-Ear Transfer Function Estimation
Summary:


Assisted Listening for headphones and hearing aids: Signal Processing Techniques
Relevant Extract:



HRTF Personalization



Interesting: https://www.researchgate.net/public...naural_headphone_reproduction_in_web_browsers

Thanks

 

[RayDunzl]

I'm confused yet again:

Why is HRTF so important (as measured by being a topic of conversational criticism) when podering headphones - which, it seems to me, fire straight down yer ear hole... especially the IEM sort...

And all but ignored when talking about speakers, where, by my guesstimation, it would actually come into play. Just cup your hands around your ears to modify it (and the perceived response).

Secondarily, whatever you have is what you're used to.

If presented with "flat" pink noise, and two people (even if with wildly differing ear shapes) agree "Yup, that sounds flat to me!", should we look at their ear shapes and say to one (or both) "No! It doesn't! You're wrong! You need to set the EQ like this!"

 

[maverickronin]

Why is HRTF so important (as measured by being a topic of conversational criticism) when podering headphones - which, it seems to me, fire straight down yer ear hole... especially the IEM sort...

And all but ignored when talking about speakers, where, by my guesstimation, it would actually come into play.

Listening to anything more that a couple feet away adds adds any individual's HRTF naturally. It's because headphones more or less fire straight into one's ear that they need to compensate.

 

[Shikamon]

Yeah. Which in all honest, EQing headphones by ear is really tough once you get around the high frequency range (~ 2khz and up). :/

 

[solderdude]

The fact that heads are shaped differently, ears are shaped and positioned differently, resonances within one's head are different, all lead me to believe that HRTFs will also differ. Just by how much, I don't know, but that was the whole reason to try to measure this for myself. There are also some published studies that seem to address this specific issue: customizing headphone response to the specific qualities of the listener in order to achieve better, more realistic playback.

I understand the reasoning. The fact that heads and positions are shaped differently may well cause slight variations to HRTF when listening to speakers in front of you resulting in (perhaps) a slightly different Pinna and Concha gain. The Ear canal gain is nothing to worry about in this case and only relevant when you measure with one as that has to be compensated for.
A bit like measuring an amplifier or DAC output through a filter and then remove the filter effects afterwards.
The headshape may be an issue with getting good seal which is a problem with most closed headphones.

For on-ear headphones only the Concha Gain (+ Tragus) effect is there leading to yet another compensation also because there is less room for a mic and the shape of the Pinna can severely change the bass response through seal issues.

In ears are another problem as you cannot measure that yourself and things like seal and insertion depth as well as ear canal shape/diameter/length all change the peak of the ear canal resonance.

I understand your desire to want to know this. Where I put my question marks is the references to which your measurements are held. When making a measurement it has to be calibrated which isn't the case. The mic plot on its box is not the real reference (most likely). I mean looking at your own results it appears as though the Concha gain which one would expect measures the other way around. There is no gain in that part of the FR but rather a drop off in treble. When you compensate for this the tonal balance will change.
It will sound brighter which most perceive as more detailed and 'better' unless overdone. So it seems you are not compensating for HRTF but simply add treble based on an incorrect measurement is the way I see it.

I understand the search for better SQ in headphones... I have been working on this for many years myself. I would like to encourage you to keep at it and preferably post results and discuss the experiments.

 

[solderdude]

Listening to anything more that a couple feet away adds adds any individual's HRTF naturally. It's because headphones more or less fire straight into one's ear that they need to compensate.

I will show what Maverickronin means here:

And basically is what you are saying and attempt to compensate for.

When looking at real ear pictures sometimes the tragus prevents firing directly into the ear effectively blocking higher frequencies a bit more.

As you can see this picture is from a G.R.A.S. Pinna and taken from the viewpoint of a headphone driver but so is the picture above.
Part of the sound is obstructed by the Tragus but most of it fires directly into the ear where sounds from the front bounce from the Concha into the ear and get boosted in a part of the FR because of it.

I get what you want to achieve is speaker alike reproduction which means the peak you get with speakers needs to be removed again to have the tonal balance of speakers. Most headphone manufacturers already do this as is evident by most headphones having a dip in the 2-5kHz range.
None get that perfectly right because acoustics is very difficult in such a small space.

You would want to know how much difference the concha gain makes in YOUR specific case. In order to correct this you need to measure with a high accuracy and correct reference. That's the issue here. Not the theory as to why one would want to.
When you get the reference of your method sorted out you will get closer. Get the reference wrong and you may be further away.

The question is how much will the speaker FR deviate from known references when freefield measured and how does this relate to your speakers at home in a living room which is what you will most likely compare to.

 

[Shikamon]

The question is how much will the speaker FR deviate from known references when freefield measured and how does this relate to your speakers at home in a living room which is what you will most likely compare to.

Hmmm... So would that mean even if I eq my speakers anechoic response to flat, it still won't give me the reference I need to find my HRTF in my own room? :0

 

[solderdude]

Correct.. a speaker that measures 'flat' in anechoic conditions will not sound/measure 'flat' in any (untreated) living room.

So when you were to be able to EQ a headphone 'flat' to mimic a speaker in anechoic conditions the headphone will sound different from the same speakers in the room.
Could be 'better' could be 'worse' depending on the room acoustics, speakers used, listening & speaker positions and personal preferences.

On top of that most popular studio albums (with the exception of audiophile labels) won't sound more 'realistic' when EQ'ed to 'flat in anechoic' conditions as most albums are produced to sound good on speakers in living rooms AND on small speakers and they should also sound good on car stereos and lately they have been asked to make the same album sound decent of the little speakers on phones.
That's where the majority of music is used. On the go, in a car and using phone speakers these days.
Audiophiles are just a very small subcategory of listeners.

Also, even when headphones are EQ'ed perfectly to (your) HRTF the sound will still be inside the head and won't pop forward.
This is what some HRTF seekers hope to accomplish. You can't fool the brain into 3D with just EQ, more trickery is needed.

Still EQ is essential to make the majority of headphones sound 'realistic' as in not 'colored' when audiophile music that is well recorded is played through them.

 

[Shikamon]

Sigh... So I guess I was doing it wrong then. I was focused in EQing for anechoic and not just EQing for the room and reflections. At this point, I feel like giving up this charade cause it's like going forward only makes me walk backwards.

 

[solderdude]

The final result of EQ-ing also depends which headphone you EQ.
Have EQ'ed a lot of headphones and even when the tonal balance is closely the same they still sound different.
Some headphones simply never sound good regardless how well their tonal balance measures.
A headphone that needs a lot of corrections (high and sharp peaks and dips) won't sound as good as a headphone that is already good by itself and merely has some small corrections.

Don't let me discourage you though. Just pointing out some soar points when attempting this.

 

[andreasmaaan]

Sigh... So I guess I was doing it wrong then. I was focused in EQing for anechoic and not just EQing for the room and reflections. At this point, I feel like giving up this charade cause it's like going forward only makes me walk backwards.

Which speakers are you using for this?

 

[Shikamon]

Which speakers are you using for this?

I'm using an old Magnavox Speakers that my grandparents had lying around. Was able to get it working by soldering them to a typical 3.5 mm male jack. I tried to do some anechoically measurements outside in my backyard and away from boundaries. Got a pretty smooth response compared to what I have coming from my small bedroom.

And I understand what you're saying solderdude. It's best to understand and learn the truth about find our own HRTF like Harman is trying to do as we speak.

And seeing that my method seems a bit flawed, I might have to go back to square one and try to EQ my speakers while in a room and see if I can get a better grasp at measuring with my Dayton imm-6 mic and do the rest of Greisinger's method again when I get home....

 

[RayDunzl]

Listening to anything more that a couple feet away adds adds any individual's HRTF naturally. It's because headphones more or less fire straight into one's ear that they need to compensate.

Oh.

So I suppose I was thinking backwards.

Headphone EQ needs to add HRTF, is the idea.

Carry on.

 

[Shikamon]

So, if I were to set up my speakers to measure inside my bedroom (3.53 X 2.72 X 2.44 m in volume), how should I position my speakers from the wall boundaries for the best listening? If I could get the curve of what my ears hear at an enjoyable listening experience with my Magnavox Speakers, then I might be able to get my SHP 9500s headphones close to it. :0

 

[solderdude]

Oh.

So I suppose I was thinking backwards.

Headphone EQ needs to add HRTF, is the idea.

Carry on.

Basically an over ear needs FR correction + a correction for Pinna + Concha (not ear canal) if you want to mimic the FR of speakers in front of you.
It only needs FR correction if you don't want / feel the need for 'speaker alike' reproduction in free field.
It needs FR correction + some bass boost when you want to emulate speaker in a room and don't feel the need to correct for the Concha gain/Pinna
It needs FR correction + some bass boost + Concha correction (partly HRTF) when you want speaker in a room sound.

For on ears the Pinna does not need 'correction' only the Concha + FR + bass boost (if desired)

For in ears it is more difficult as FR is dependent on insertion. There still should be HRTF correction when you want to mimic speakers in a room.

Most headphone manufacturers do attempt to correct for Pinna/concha gain already as is clearly shown in measurements without Pinna where there is a dip between 2kHz and 7kHz. When measured with a Pinna that dip is often just a bit less deep, not present or even boosted a little.

I suspect differences between individual HRTF are much smaller than the differences in taste/preference and variances in FR response between headphones and is a non issue.

Of course there is nothing wrong with striving for 'perfection' and correct even the last dB when done accurately which I suspect is not possible.
Also IME correcting exactly for any measured FR response with headphones is not really desirable as one trusts the measurement explicitly.
Just look at the many many attempts by professionals and amateurs of the same headphone and see how differences of well over 10dB are shown in these plots.
Mission impossible.
My P.O.V. thus is correct for coarse errors in FR (wide peaks and dips, often only partially) found in all plots which usually leads to better tonal accuracy.

 

[solderdude]

So, if I were to set up my speakers to measure inside my bedroom (3.53 X 2.72 X 2.44 m in volume), how should I position my speakers from the wall boundaries for the best listening? If I could get the curve of what my ears hear at an enjoyable listening experience with my Magnavox Speakers, then I might be able to get my SHP 9500s headphones close to it. :0

I would suggest trying to apply EQ by hand.

Set the speakers up how you prefer its sound.

My SHP9500 measured (without Pinna) and sounded like this to me with recessed upper mids/lower treble:

I got it to measure (and sound) like this:

But.... sold it as even with the FR corrected and the tonal balance was correct it did not sound as good as others I had + I did not like the pads being glued on and not replaceable.
Without the peak around 4kHz it sounded boring and dull so this headphone obviously needed a little boost there.

 

[Shikamon]

I see... So how did you get to the second curve? Was it all done by ear after calibrating your speakers to flat in a room and then adding it to your headphones?