Measuring HRTF for headphone use

[pkane]

What we call cross-talk is a natural element of hearing individual sources, and one of the features enabling us to localise. Question is, if it’s always a benign mechanism, e.g. considering reproduction of composite content?

For those of us going to AES in NYC next month, David Griesinger will be playing excellent binaural recordings using ultra-nearfield monitoring with cross-talk cancellation; which is an elegant way of separating humanisation in recording from personalisation in reproduction. I suggest you go listen and make up your own mind.

As I'm not going to be at the AES, I'd love to hear some feedback from those who will be there. And what does ultra-nearfiled monitoring mean? Is it just sitting very close to the speakers? Or even closer, speakers sitting inside one's ears?

I've played around with crosstalk cancelation in a normal listening position speaker set up. This wasn't a huge improvement but noticeable, and took a few iterations with the folks producing these filters (@Home Audio Fidelity ) to get it just right. Thierry worked with me to troubleshoot and improve the convolution filters until they produced a noticeable improvement on well-recorded, acoustic music. IIRC, the amount of cross-talk cancellation produced by these filters at the listening position was not much more than 10%. But these were not individualized, other than being configured for the proper relative position of the speakers to the listening position.

 

[Shikamon]

Welp, just did a little measurement session for my Status CB1 headphones. Here are my results:

And then after correcting them:

After listening to them when corrected, it sounds pretty dull in some way. Was this the same case for you @pkane when you did yours based on your initial thread response? :0

 

[solderdude]

I don't think there really is 14dB peak at 6kHz at all.
At least these measurements suggests not so nor have I ever measured any peak being that high.

This is most likely a measurement error.

The 'result' may well be flat measuring but the actual result will not be.

 

[Shikamon]

@solderdude Yeah I'm been getting some odd measurements with these in ear mics. I feel they're not as flat as they're projected to be based on the product's frequency response. I did a comparison of the SP-TFB-2 with my Dayton iMM-6 (which I trust is flat with the given .cal file they provide). I held them a about over a foot away from my speaker and this is what i get:

That peak could have been due to this but I was told that calibrating mics for compensation of headphones shouldn't matter. I feel to disagree with that. Thoughts? :0

 

[solderdude]

To check speakers I would measure them outdoors, away from reflective surfaces, on axis and 1 meter away.

For measuring headphones I don't think it is wise to alter the frequency response, measure an altered response and correct on that.

 

[Shikamon]

@solderdude The measurements I've shown are meant to compare how the mic's differ. The two above are the left and right mic of the SP-TFB-2 and the lower is the Dayton iMM-6. :0

 

[solderdude]

I assume with the mics in your ear and same distance as the Dayton.

How do the mics compare in free-field pointed towards the speaker. I would assume they will be comparable.

 

[Shikamon]

@solderdude Yeah they're comparable. The graph shows just that. The in ear mics are held in free field in front of the speaker as well as the Dayton mic.

 

[pkane]

Welp, just did a little measurement session for my Status CB1 headphones. Here are my results:

View attachment 33859

And then after correcting them:

View attachment 33861

After listening to them when corrected, it sounds pretty dull in some way. Was this the same case for you @pkane when you did yours based on your initial thread response? :0

That's a large jump at 6kHz! Try correcting the response to a 10dB (20Hz-20kHz) declining line. See if this is a little better sounding. How do they sound without the correction?

Something is strange if you are getting nearly a 20dB bump at 6kHz...

 

[Shikamon]

@pkane yeah but the slope wont do much good if the in ear mics, when measure in free field, compare differently to my Dayton iMM-6. Check out my previous thread to see what I mean. I wonder if there is a way other than comparing mics that I can do to verify if my in ear mics have a flat response...

 

[pkane]

@pkane yeah but the slope wont do much good if the in ear mics, when measure in free field, compare differently to my Dayton iMM-6. I wonder if there is a way other than comparing mics that I can do to verify if my in ear mics have a flat response...

The simplest way is to equalize flat response from a speaker using a known flat-response mic. You can then try the inear mic in the same position as the measurement mic to see how different the response is. This is not perfect due to different dispersion pattern, but may be enough to detect large differences.

 

[Shikamon]

@solderdude Yeah I'm been getting some odd measurements with these in ear mics. I feel they're not as flat as they're projected to be based on the product's frequency response. I did a comparison of the SP-TFB-2 with my Dayton iMM-6 (which I trust is flat with the given .cal file they provide). I held them a about over a foot away from my speaker and this is what i get:

View attachment 33863

That peak could have been due to this but I was told that calibrating mics for compensation of headphones shouldn't matter. I feel to disagree with that. Thoughts? :0

@pkane In that case, I should calibrate the difference from my other measurement mic I have as I have more faith that Dayton took the time to calibrate the mic properly as no two same mics have the same exact frequency response curve. :0

 

[pkane]

@pkane In that case, I should calibrate the difference from my other measurement mic I have as I have more faith that Dayton took the time to calibrate the mic properly as no two same mics have the same exact frequency response curve. :0

Right. I’d trust a calibrated mic much more than an uncalibrated one.

 

[Shikamon]

Just did a second measurement with my in ear mic's calibrated. Seems to look a bit better than before:

And after apply compensation EQ:

Pretty interesting with the compensated measure curve as it has a warm tilt to it. And I compensated to a flat line too which I think looks better than the uncalibrated measurements.

Right now I'm listening to some tunes and they sound pretty nice, at least in sounding flat with a warm tilt. Is that what I should be hearing for my HRTF? :0

 

[DrCWO]

I got an old STAX SR-X MK3 headphone but not the PRO type.
Did anyone of you guys know how big the difference in frequency response is compared to the PRO type?
I can find measurements of the PRO but non of the non-Pro.

Best DrCWO

 

[thewas]

I got an old STAX SR-X MK3 headphone but not the PRO type.
Did anyone of you guys know how big the difference in frequency response is compared to the PRO type?
I can find measurements of the PRO but non of the non-Pro.

Here you find measurements of both, look very similar:

Source: https://www.inexxon.com/stax-übersicht-history/1970-1980-kopfhörer-headphones/sr-x-mk-3/

Source: https://www.inexxon.com/stax-übersicht-history/1980-1990-kopfhörer-headphones/sr-x-mk-3-pro/

 

[bennybbbx]

i have read about htrf, but simple tests let me doubt it. when hear music with speakers. Wy can bend the ear from back to front so that it is in front of the ear entrance. can hear that FR of hear change alot, but distance location work same as before. ?
When put a closed headphone with very much damping on ears that output no signal and hear with speaker. high freqs are mostly gone but it is full possible to locate distance of sound that come thru the headphones much in FR change from the speaker. there is also tell that the body reflections is important for distance location. but when put hands or foam before body also did not change distance location. so i do tests which change hrtf results alot, but distance location stay same.

I have here https://www.audiosciencereview.com/...ck-of-2024-contest.59615/page-22#post-2325412 written about realphones and a example song on youtube upload from David gilmore which work good. this bring on this song the vocal out of the head in front for me. there are not many songs that can do this. I also have do a own test song and i notice it depend mostly on reverb settings FR is not important. but what reverb i need i did not know to reproduce it correct. I get same results with diffrent headphones. hans zimmer song time Live in prague or brass songs live in concert halls recordet also good with out of head location. in my realphones guitar example the guitar is in front and the highhat. but kick and snare drum sound very near head. in hans zimmer song bass and drums are out of head.

did you get with your hrtf measure results better out of head hear results as this linked testsongs ?. only when say the truth can things make better and always verify theorie with examples, dont believe.

 

[bennybbbx]

I read in german wikipedia about distance hearing, this fit to my testing results and experience. The reverb delay and level and reverb to direct level is the most important thing. I translate the german to english

Entfernungshören – Wikipedia

de.wikipedia.org

Direct sound component: The sound pressure level of direct sound decreases linearly with distance, while that of diffuse sound remains almost constant in a reflective environment, so that this ratio changes significantly with the distance from the sound source. This is why, together with visual perception, it is one of the most important characteristics of distance hearing.

Initial time gap: In the case of distant sound sources, the first strong reflections hardly have a longer path to the listener than the direct wave. They therefore reach the listener almost simultaneously, whereas with a nearby sound source there is a clear initial time gap due to the different detours. Their great importance for the spatial staggering of the sound field is often neglected in sound productions.

Translated with DeepL.com (free version)

I did not find anything about that in english because i did not know how the english word is. translate in google "Entfernungshören" to english to distance hearing did not find something science text about that.

so the good news for out of head hearing as with speakers, there need no customize hrtf. My guess is hrtf is mabye important to detect audio from behind location. because when i change in realphones the hrtf settings or in some locations, i hear that it seem the sound come from back

I have upload a example find on youtube that is record with a stereo camera and some instruments are play . it sound not in head . important is stereo. because when switch to mono it sound in head. and left and right channel should be same loud. so if it sound strange try correct balance so it sound good. analg stereo pots have high wear and tear and differ left and right

so there need room or cathedral impulses for reverb that are able to make studio records sound real and good out of head . then mixing engineers can use this impules and create great songs that sound on headphones and speakers great. I have in the past develop reverb algorithm with delay allpass, feedback 90 delay lines until it sound as a nice reverb. such working many do, but such reverbs did not contain usefull information the brain need to bring the sound out of head and a good depth of field.

The guys that create reverb impulses do also a simular way. they use more than 2 microphones, place it so that it sound nice. then happen same problem a nice reverb sound without usefull information for brain to create depth of field. realphones can with a correct reverb then create the additional information that brain need to create depth on headphones. video it is with realphones for headphones do. can other hear that sound come not out of head too in this video ?

EDIT: I notice now out of head sound work with mono audio input track too, only realphones need output stereo signal