• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Measuring HRTF for headphone use

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
1,264
Likes
1,784
Location
North-East
#1
In a recent thread, I asked if anyone could recommend a simple way to measure HRFT. Rather than to try to correct for some average or modeled HRTF, I would like to have a more direct measure of my personal one. The plan was to then apply DSP to correct for it, in combination with headphones (and possibly speakers).

My initial thought was to try to run listening tests to determine the lowest dB level at which I can still hear each frequency in a frequency sweep. This sounds like it might actually produce a transfer function, combining headphone FR with my HRTF and also with my hearing acuity. A bit of work and possibly not very high degree of accuracy.

Meanwhile, I found a simpler and more direct way to measure HRTF that involves in-ear microphones. These microphones get inserted behind the pinnae in the ear so I can then use headphones over the ears to play a sweep to measure FR directly. I found in-ear microphones meant for binaural recording from Sound Professionals, part # SP-TFB-2. These are under $100, but if you don't have a recording interface that can provide 10-12v phantom power (48v will fry these!) you'll also need their battery power supply. They do sell an XLR version of these that work with 48v, but these cost more.

The mics are not calibrated, but their reported FR is pretty flat:



I found that the position of the microphone in the ear makes a difference to the measured response. Pushing it in as far as I could, I found the results fairly repeatable with +/- 1dB accuracy over the audible range.

Using REW log sweep into HE-560 headphones in combination with the binaural mics, here are the results for my ears:

hrft.png


This is different from the recommended Harman correction curve, which is not surprising as this represents a combination of HE-560 FR and my HRTF. Right is red, Left is blue. You can also see there are some differences between my left and right ears.

I then created parametric EQ settings to flatten the response (I'll spend more time in the future to fine-tune, but this was close enough for government work!) The corrected version, again measured with the same setup, is shown below. I didn't try to correct anything above 6KHz.

hrtf-corrected.png


What does it sound like? Very, very nice! :) This is an obvious improvement over what I've been previously able to achieve with an average Harman HRTF correction curve, a simple flat curve, or a 10dB tilted one (20-20KHz). I have a few binaural recordings, and used those for testing.

Flat has never sounded good to me. Not over speakers, and not over headphones. Until now, the 10dB declining curve was my goto correction curve for both.

Interestingly, flat curve sounds great now, when it's measured from the inside of my ears! I also wonder if these measurements can be used to predict which headphone's natural FR would be a better match for my ears, if say I can't or don't want to do DSP. I'll have to subtract FR of HE-560s from the measured curve to determine my actual HRTF.

Comments?
 
Last edited:

mindbomb

Active Member
Joined
Nov 11, 2017
Messages
284
Likes
147
#2
I thought you were going to go in a different direction. What I would try to do, is measure a binaural impulse response from each of your speakers, and then convolve it with headphones, and see how close you can get to recreating that experience.
 

Grave

Senior Member
Joined
Jun 30, 2018
Messages
382
Likes
178
#3
I was confused when I first read this post, but I see what you did now. If you could post the EQ setting you used as well as the result when compensating for the FR of your headphones, then that would be interesting. I am not sure that all of this effort would be worth it when you can just listen. I do not see the problem with "trusting your ears" when it comes to headphones since they are hard to measure and the perception of headphone frequency response seems to vary per person.
 

mitchco

Senior Member
Joined
May 24, 2016
Messages
394
Likes
1,143
#4
Cool! I used the same in ear mics and measurement approach for NAD Viso HP50:

NAD VISO HP50 onesixth octave smoothing.jpg


That's with no eq. Note this approach is using a blocking measurement method in the canal, not measured at the ear drum, which is what Harman uses. and target curve derived from... The main issue with this blocked meatus approach of measuring the headphone is it doesn’t include the acoustic impedance of the ear drum versus the headphone.

On a related note, perhaps this would be an experiment worthy of trying:

 

Mad_Economist

Active Member
Joined
Nov 29, 2017
Messages
103
Likes
154
#6
Unfortunately this technique is not suitable for measurements of HRTF. For proper measurement of blocked-ear HRTF, the microphones must be flush with the canal entrance. A microphone which sits in the concha, as the Sound Professionals mics do, can yield extremely misleading results. Figure 2 of this excerpt shows the variation that even slight insertion depth irregularities can cause with a blocked canal measurement - now imagine what will happen when you go from this to this.

I suppose there's additionally Griesinger's contention (since he's been brought up in thread already) that blocked canal measurements are fundamentally flawed vs. open ear, but honestly, I'm of the opinion that the body of scholarship supports blocked canal working - equalization based on blocked canal measurements (including, if I understand their premises correctly, Smyth's Realizer products) seems to be consistently and strongly well-received. Still, worth a ponder.

It is unfortunate that there isn't an easily available option for consistent and accurate in-ear measurements. In addition to more ideal headphone equalization, they would be a boon to DIYers.
 

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
1,264
Likes
1,784
Location
North-East
#7
I thought you were going to go in a different direction. What I would try to do, is measure a binaural impulse response from each of your speakers, and then convolve it with headphones, and see how close you can get to recreating that experience.
That’s a possibility. Recreating impulse response for my speakers/room through the headphones sounds like a good experiment.
 

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
1,264
Likes
1,784
Location
North-East
#8
I was confused when I first read this post, but I see what you did now. If you could post the EQ setting you used as well as the result when compensating for the FR of your headphones, then that would be interesting. I am not sure that all of this effort would be worth it when you can just listen. I do not see the problem with "trusting your ears" when it comes to headphones since they are hard to measure and the perception of headphone frequency response seems to vary per person.
Trusting your ears is one way to do this, but testing out every possible headphones out there to find the goldilocks one that fits just right seems like a fool's errand. Considering HRTF is based on ear/head physiology and shape, everyone is different and what headphones will work with my noggin is a crap shoot. Do you think it's better to keep trying different different headphones if I can EQ the ones I already have to perform better than anything else?

What would the HE-560 EQ settings tell you? I do have the measurements for it and will subtract it from what I measured in my ears to see what the approximate HRTF really looks like.
 

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
1,264
Likes
1,784
Location
North-East
#9
Cool! I used the same in ear mics and measurement approach for NAD Viso HP50:

View attachment 14599

That's with no eq. Note this approach is using a blocking measurement method in the canal, not measured at the ear drum, which is what Harman uses. and target curve derived from... The main issue with this blocked meatus approach of measuring the headphone is it doesn’t include the acoustic impedance of the ear drum versus the headphone.

On a related note, perhaps this would be an experiment worthy of trying:

Yes, interesting! Seems to be a similar shaped response to what I measured, despite using different headphones. I was concerned about positioning of the mics as well. After trying various positions, I found I could reproduce the measurement fairly accurately if they were pushed in as far as they could go into the ear canal (without hurting the ear, obviously :) ).

Between initial measurements and EQ sessions, I took the mics off and then reinserted them, and then checked that I could still measure the same raw response before trying to correct. As I said, the differences were +/-1dB at some frequencies, but nearly an exact match up to about 1kHz.

The closed ear canal is a problem, as that removes some of the resonances that would otherwise be present in the ear. I'm planning on a separate test that will not involve microphones at all. My hope is that the general shape of what I measure in a listening test will be close enough to the microphone-based measured response, but time will tell.

On a related note, perhaps this would be an experiment worthy of trying:

Yes, I've seen this video before. The listening experiment I want to try is similar to this approach.
 

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
1,264
Likes
1,784
Location
North-East
#10
My only concern here is the positioning of the mics will change the response. Noting how different your left and right are. How do you ensure a consistent position?
Yes, definitely a concern. I found some variation due to the mic position, until I pushed the mics as far in as they could go into the ear canal. This produced a much more repeatable response curve, even after taking the mics off and putting them back in multiple times. Variation was within +/-1 dB range in places where the curve has the large excursions, but much less over most of the spectrum. In the end, I averaged 4 runs of measurements for each ear before trying to correct anything.
 

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
1,264
Likes
1,784
Location
North-East
#11
Unfortunately this technique is not suitable for measurements of HRTF. For proper measurement of blocked-ear HRTF, the microphones must be flush with the canal entrance. A microphone which sits in the concha, as the Sound Professionals mics do, can yield extremely misleading results. Figure 2 of this excerpt shows the variation that even slight insertion depth irregularities can cause with a blocked canal measurement - now imagine what will happen when you go from this to this.

I suppose there's additionally Griesinger's contention (since he's been brought up in thread already) that blocked canal measurements are fundamentally flawed vs. open ear, but honestly, I'm of the opinion that the body of scholarship supports blocked canal working - equalization based on blocked canal measurements (including, if I understand their premises correctly, Smyth's Realizer products) seems to be consistently and strongly well-received. Still, worth a ponder.

It is unfortunate that there isn't an easily available option for consistent and accurate in-ear measurements. In addition to more ideal headphone equalization, they would be a boon to DIYers.
I can't say if the measurements are accurate (at least not until I try to confirm these with another type of test that doesn't involve these mics). I spent some time trying to get a position for the mics that produced consistent results, and I think I succeeded. Just like one might do for room EQ, I took four measurements for each ear after removing and reinserting the mics, and then averaged them. I'm sure I can get repeatable measurements with these mics, but the potential changes due to blocking the ear canal remain.
 

Mad_Economist

Active Member
Joined
Nov 29, 2017
Messages
103
Likes
154
#12
I can't say if the measurements are accurate (at least not until I try to confirm these with another type of test that doesn't involve these mics). I spent some time trying to get a position for the mics that produced consistent results, and I think I succeeded. Just like one might do for room EQ, I took four measurements for each ear after removing and reinserting the mics, and then averaged them. I'm sure I can get repeatable measurements with these mics, but the potential changes due to blocking the ear canal remain.
Indeed, with a fixture in the outer ear, placement becomes a lot more consistent - I made a small cast of part of my concha and mounted a little electret in it once, and the results had slightly greater consistency than my own Head and Torso Simulator (turns out it's a lot easier to keep headphones in the same spot on your own head than a dummy). The trouble was, the measurements were useless in the HRTF band. I'd measure peaks and nulls which were not there when measured on the HATS (or subjectively there to my ear) or just outright incorrect tendencies for the FR. Useful for measuring low-frequency behavior on human heads (a point of substantial variation between headphones), but not for much else unfortunately.

A properly canal mounted mic, on the other hand, is harder to keep consistent, at least for the DIY solution that tends to be used of sticking them in an IEM eartip or earplug. Thus I've usually considered in-ear measurements to be a tradeoff of risk of inaccuracy and risk of inconsistency, although it should certainly be possible to make a mounting piece which fits more consistently in the canal (and would, in my opinion, be the best way to go if you want to do on-head measurements).

I probably wouldn't worry too much about Griesinger's commentary on canal impacts for the moment, honestly - the project of attempted personalized EQ has enough challenges in terms of getting proper measurements of the headphones and defining target response without introducing hypothetical problems as well :p
 

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
1,264
Likes
1,784
Location
North-East
#13
I probably wouldn't worry too much about Griesinger's commentary on canal impacts for the moment, honestly - the project of attempted personalized EQ has enough challenges in terms of getting proper measurements of the headphones and defining target response without introducing hypothetical problems as well :p
Reading through Griesinger's presentation, I had a feeling that it was a rushed job trying to prove a point without much evidence. The impetus for all of it appears to be a single test subject (himself) not being able to adjust the FR response properly with IEM type mics. It's not at all clear to me if there was something different about his EQ technique, his mics, mic placement, or his ear physiology. Yet he makes far-reaching generalizations based on this single test result.

I'm certainly willing to consider that there might be some differences between HRTF measured just inside concha at the entrance to the ear canal compared to what may be measured right at the ear drum. But, I'd be a little more worried about sticking a DIY device that far into my ear to run this test! ;)
 

svart-hvitt

Major Contributor
Joined
Aug 31, 2017
Messages
2,375
Likes
1,155
#14
SCIENTIFIC REFERENCES?

This is the 14th comment in this thread on measuring headphones and HRTF, but I wonder:

Is it too soon to ask for some peer reviewed references, say from JAES or equivalent, on this very interesting but confusing topic?
 

mitchco

Senior Member
Joined
May 24, 2016
Messages
394
Likes
1,143
#15
Not suggesting Griesinger is the end all beat all, but he does bring up some interesting points in this much larger downloadable presentation. Why it is much larger is due to the audio samples that one can hear for themselves demonstrating how canal differences are quire different from a test panel of 6 individuals. See slides 33 to 39. It is rather ear opening :) PS. keep the volume down initially!
 

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
1,264
Likes
1,784
Location
North-East
#16
SCIENTIFIC REFERENCES?

This is the 14th comment in this thread on measuring headphones and HRTF, but I wonder:

Is it too soon to ask for some peer reviewed references, say from JAES or equivalent, on this very interesting but confusing topic?
Which part, that HRTF exists, or that it can be measured?
 

svart-hvitt

Major Contributor
Joined
Aug 31, 2017
Messages
2,375
Likes
1,155
#17
Which part, that HRTF exists, or that it can be measured?
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.
 

Mad_Economist

Active Member
Joined
Nov 29, 2017
Messages
103
Likes
154
#18
Reading through Griesinger's presentation, I had a feeling that it was a rushed job trying to prove a point without much evidence. The impetus for all of it appears to be a single test subject (himself) not being able to adjust the FR response properly with IEM type mics. It's not at all clear to me if there was something different about his EQ technique, his mics, mic placement, or his ear physiology. Yet he makes far-reaching generalizations based on this single test result.
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
 
Last edited:

Mad_Economist

Active Member
Joined
Nov 29, 2017
Messages
103
Likes
154
#19
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.
Toole himself did some investigation of headphones long ago, although for canonical texts I would argue that the works of the aforementioned Hammershøi & Møller and of Günther Theile are the most significant, as well as of course the modern work of Sean Olive and his collaborators at Harman.

Edit: Olive has a collection of headphone-centric AES papers which serve as a good introduction to the recent history and current state of the field.
 
Last edited:

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
1,264
Likes
1,784
Location
North-East
#20
Not suggesting Griesinger is the end all beat all, but he does bring up some interesting points in this much larger downloadable presentation. Why it is much larger is due to the audio samples that one can hear for themselves demonstrating how canal differences are quire different from a test panel of 6 individuals. See slides 33 to 39. It is rather ear opening :) PS. keep the volume down initially!
Interesting, thanks for the link! 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. Certainly the response charts shown in the presentation appear to vary a whole lot from individual to individual. Only two out of six charts appearing exactly the same. I'm not sure if that's just a copy/paste error (or Ben and David are the same individual :) ) but they have exactly the same responses. No differences whatsoever.

The cited paper by Hammershøi and Møller appears to be much more supportive of using blocked ear canal measurements than Griesinger, at least on first reading.
 
Last edited:
Top Bottom