Finally, you can calibrate your headphones to your ears.

[Neuro]

Sean Olive shares a tip:

This looks interesting. An inexpensive in-ear microphone to measure and calibrate your headphones.

As we all know, outer ear anatomy causes deviations of up to 20 dB!!! in the perceived frequency response when listening with headphones. The target curve established by Harman and Sean Olive is an "average" curve based on a large population. That curve is meaningless for the individual case due to significant inter-individual variations.

Now, affordable calibration equipment is finally available.

EarEqual — Personal acoustic measurement

EarEqual makes measurement hardware that calibrates headphones and earbuds for the ears you actually have — designed by an ear-canal acoustics researcher.

earequal.com

 

[whazzup]

I see now! That does sound like the future! Minidsp going to give us umik-3, in-ear measurement mics?
*Wait, how about in-ears then? So this 'solves' outer ear customisation only?

 

[solderdude]

Disregards the ear canal which is an important part of test fixtures.
You will measure about +10dB around 4-5kHz when recording sounds coming from in front of you and around +5-7dB for sounds coming from a headphone driver on the side of your head.
Still fun to experiment with.

alternative:

Sound Professionals Low noise in-ear Binaural/ASMR microphones – Made in USA SP-TFB-2

Low noise in-ear Binaural microphones, available in two different sensitivities Please select the desired microphone sensitivity, cable type, and any optional items below:

soundprofessionals.com

 

[Sean Olive]

Disregards the ear canal which is an important part of test fixtures.

alternative:

Sound Professionals Low noise in-ear Binaural/ASMR microphones – Made in USA SP-TFB-2

Low noise in-ear Binaural microphones, available in two different sensitivities Please select the desired microphone sensitivity, cable type, and any optional items below:

soundprofessionals.com

He claims to be working on an ear tip with an embedded microphone to replace the one on your ear buds.

But soon this may become a standard feature for TWS since they already include a feedback microphone for ANC that can be used to measure and predict the pressure at the ear drum. I believe Bose and perhaps a few others are already doing this.

Soon we will all be measuring our headphones in our ears and calibrating them to our personal downward-tilted DF HRTFs for perfect sound. Headphone reviewers like Amir and expensive industry standard test fixtures like B&K 5128 based on population averages will become redundant in the age of AI and personalization

 

[solderdude]

Have you experimented with blocked ear microphones and did the results match with that of HATS when removing the ear-canal portion ?

 

[Sean Olive]

Have you experimented with blocked ear microphones and did the results match with that of HATS when removing the ear-canal portion ?

The Hyperion project at Harman measured blocked canal measurements of HRTFs of +300 people and industry manikins as well as high res scans of their heads/ears to do statistics on various landmarks/features of ears/heads. The main purpose of that was to find a better population average of HRTFs than KEMAR and B&K 5128 and they found a subject which better represented the average of 300 people -- referred to as Archie (Archetype). They also found the quartiles and printed the heads and ears.

So the answer to your question is that the HRTFs of Archie don't match KEMAR's or B&K 5128's when the ear canal is removed. This is not a surprise when you look at the physical differences in their heads/ears. And you can hear the differences when you listen to music and test signals convolved with the different sets HRTFs. They produce differences in timbre, localization, image width/depth and externalization. Sadly, none of this was ever published and I understand the facility has been moth-balled.

Crutchfield did a nice story about the research and you can see pictures of the Hyperion Sphere and the different manikin heads that were produced,

https://www.crutchfield.com/S-DxrAFq0HpTW/learn/crutchfield-visits-harman.html

 

[Redacted]

So, it's automatic room correction for a very, very small room?

 

[solderdude]

The main purpose of that was to find a better population average of HRTFs than KEMAR and B&K 5128 and they found a subject which better represented the average of 300 people -- referred to as Archie (Archetype)

Thanks for the info.

How big was the delta between Archie and the various individual measurements ?

Will Archie ever become a product for headphone measurements ?
Preferably affordable

 

[Sean Olive]

Disregards the ear canal which is an important part of test fixtures.
You will measure about +10dB around 4-5kHz when recording sounds coming from in front of you and around +5-7dB for sounds coming from a headphone driver on the side of your head.
Still fun to experiment with.

alternative:

Sound Professionals Low noise in-ear Binaural/ASMR microphones – Made in USA SP-TFB-2

Low noise in-ear Binaural microphones, available in two different sensitivities Please select the desired microphone sensitivity, cable type, and any optional items below:

soundprofessionals.com

Interesting you mention the importance of the ear canal. The person behind EarEqual in-ear microphone is Marko Hiipakka, an acoustics researcher and entrepreneur with a Ph.D. in Acoustics and Audio Signal Processing from Aalto University. I met him maybe 11 years ago at Aalta University (he was a PhD graduate doing a post-doc I believe) when I was an external examiner for a PhD student working on headphones. PhD defenses in Finland are like Persian weddings - a big ceremony followed by 2 days of celebration. The Finns certainly know how to party. and I missed my flight the next day.

Marko was working on a in-ear headphone that self-calibrated itself. There are AES papers that describe it. I received a prototype but could never manage to get it to work as expected and eventually the company folded. Maybe an idea before it's time?

" Before launching EarEqual, Marko spent nearly two decades specializing in personalized audio, ear-canal acoustics, and eardrum sound-pressure estimation. His notable previous work includes: [1]

• Hefio Oy: Marko was the original founder and CEO of Hefio Oy (which operated from 2014 to 2022). Hefio was an audio company focused on developing self-calibrating earphones that could automatically measure the unique characteristics of a user's ear canal to deliver mathematically optimized, personalized sound. [1] "
  
  I wish him the best of luck in this new adventure.

 

[Sean Olive]

Thanks for the info.

How big was the delta between Archie and the various individual measurements ?

Will Archie ever become a product for headphone measurements ?
Preferably affordable

I wish I had that information but not sure I could even tell you if I did for legal reasons.

As you expect the differences get larger above 1 kHz. I did some blind experiments where listeners compared stereo loudspeakers playing music in the Reference Room versus binaural reproductions of them reproduced through in-ear headphones using different sets of BRIRs: their own, Archie, KEMAR and B&K 5128. They rated each one based on similarity to the reference (stereo speakers) in terms of timbre, localization and externalization. They could switch quickly via a floor pedal. There was no head-tracking -- since it was a single set of BRIRs. In most cases, the personalized BRIR were rated highest. The KEMAR was rated the lowest. I would love to repeat those experiments again using more than 2 loudspeakers or something that is less focused on the frontal plane,

I don't think Harman would ever commercialize Archie.... but I could be wrong.

 

[solderdude]

How would the Hefio account for hearing loss ?

As my left ear is slightly different in response as my right ear (causes 13kHz tinnitus in my 'worse' left ear).
I can control/lower the tinnitus by listening to music using headphones corrected for what is missing compartively (particular boost in left ear) to almost becoming inaudible.
Of course real life sounds 'resets' this so need to listen to 'hearing loss corrected' music which I enjoy a lot.

I believe the boys (you know who I mean ) also have been playing with their 'partly open' miniature in-ear microphone for individual calibration and were looking into the influence of ear-canal influence.
They seem to have been getting good results with individual EQ (not surprisingly).

 

[Sean Olive]

Of course, none of this accounts for hearing loss, taste, and the degree of listening experience/training — all factors known to influence what we like but seldom reported in headphone reviews. These factors could explain more why a certain frequency response is preferred than the size and shape of our head and ears.

There is a debate whether the ear should be semi-open or blocked when measuring the headphone since it affects the acoustic losd on the headphone. Konstantin D. says this product is useless except for measuring a K1000 headphone or comparable free air equivalent headphone.

Dan Clark and I built a pair of these semi-open microphones that Konstantin recommends to measure the headphones used on listeners for the target curve research as a control —but that is the extent to which I used them. There were certainly measured differences among listeners.

 

[solderdude]

There is a debate whether the ear should be semi-open or blocked when measuring the headphone since it affects the acoustic load on the headphone.

Has this actually been measured? I mean the difference between what actually is being emitted from a driver (over-ear/on-ear) if that really changes between blocked and open ear canal ?
I can see the acoustic load differing slightly around the resonance frequency band of the ear canal at best but reckon this is too small to affect the actual membrane movements.

In impedance measurements alone one can see a clear difference in acoustic load between free air and on a fixture and probably also between proper and compromised seal.
This would be difficult to measure I reckon.

Konstantin D. says this product is useless except for measuring a K1000 headphone or comparable free air equivalent headphone.

I assume you mean the test microphone in the OP ?
In the past I have experimented with microphone capsules directly in the ear canal and got poor results.

The guys did report good results with the (half-open) miniature mic in the hearing entrance though.
Alas I did not see a follow-up but don't really follow it so there might have been.
Indeed just not for creating an average 'correction' for a certain headphone model but with such a microphone and clear instructions and compensation generating SW I can see good results are likely to be had.

I also suspect that this would have to be done for each different headphone model and fit/positioning will still be a thing.

A bit of a chicken-egg situation where one would have to use a reference in order to get the right correction and the true reference is lacking so would have to be derived from known measurements.

IMO there are plenty of headphones that sound good and enjoyable and good enough that going through a lot of trouble is probably not worth it other than for science.

 

[Marko_EarEqual]

Hi all, Marko here, the person behind EarEqual.

Sean, thank you for the kind words and for remembering that defense. And yes, you got a Hefio prototype that did not behave, and I learned a lot from where that design fell short.

Let me take the ear canal point head on, because solderdude and Konstantin are both raising something real and I do not want to wave it away.

The Plug does a blocked canal measurement. It captures the Thévenin source pressure Ps that the headphone produces. This approach assumes a free air equivalent (FEC) load, so Konstantin is right that the strict case is a FEC type headphone. Konstantin raised exactly this on my LinkedIn post a few days ago, and I will give the same honest answer here.

A closed back headphone is not FEC, so the radiation impedance seen looking outward from the canal is not the same as the loudspeaker reference. Equalising Ps then does not give a perfect match in theory, the way it would for a fully FEC headphone. That residual is real, but how big it is depends on the headphone, it is not black and white, and in my view it is small next to what you gain from a robust and repeatable individual measurement. solderdude, on your point in the other thread, I agree the change in mechanical load on an over ear driver itself is small.

The ideal chain needs two things: FEC headphones, and a high acoustic impedance at the measurement point. The first is a property of the headphone, nothing the device can do about it. The second is exactly what the plug is built to maximise. A stiff, high impedance body, while the ear canal side stays soft, hollow and elastic for comfort and sealing, gets much closer to the open circuit (true Thévenin) condition than a foam plug with the mic sitting at, or outside the entrance.

It is worth being clear that this is a trade off. An open canal method avoids the FEC issue, but it pays for it in repeatability, above all in mic location stability, which is the hardest thing to control. The blocked approach accepts the FEC caveat in exchange for a controlled, repeatable measurement point, and the plug is a feasible, readily available way to get that.

On what it is for: this is individual calibration. The target can be your own reference loudspeakers measured the same way, or a generic target defined at the blocked ear canal entrance. The point is that it is your own ears in the loop, which is the part a population fixture can never capture for you.

So to be clear, this is not a lab reference and it will not replace a 5128. It is an affordable way to measure your own headphone response and correct it toward a target you choose.

 

[solderdude]

Thanks for your answer.

The idea looks solid but there are a few caveats here.
The most obvious one is the reference that will be used that the correction will be based upon.

The reference being 'the owner's speakers in the owner's room' and I suspect in stereo and not gated.
That reference thus will not be a 'true reference' but will just be how the speakers in that room behave on the listening position.
This can go in all directions when not measured using the right techniques.

While technically one may get to make the tonal balance of the headphone closer to how the owner's speaker in a room 'measures' it is still missing crossfeed and one might be getting dips and peaks that aren't really wanted.

That does not mean that the mic is useless but in practice will be limited and results are highly dependent on the owner's speakers in that room at the preferred listening position.

The ear canal does not matter much here (assuming the mic is at the exact position of the ear canal entrance) as that is not in play and thus does not need correction.

 

[gzost]

The reference is a problem. For the Symth Realizer the German distributor offers measurements done in a 5.1 or 9.4.1 acoustically optimized studio. Pricey, but makes sense.

The home-made alternative would be a nearfield listening setup with neutral speakers. If the PostCOVID brainfog ever leaves me sufficiently for me to be able to make my way through the Smyth Realizer (original version) measurement process then I'll do that with my small Genelec (8010A + sub) system.

 

[Marko_EarEqual]

Thanks for your answer.

The idea looks solid but there are a few caveats here.
The most obvious one is the reference that will be used that the correction will be based upon.

The reference being 'the owner's speakers in the owner's room' and I suspect in stereo and not gated.
That reference thus will not be a 'true reference' but will just be how the speakers in that room behave on the listening position.
This can go in all directions when not measured using the right techniques.

While technically one may get to make the tonal balance of the headphone closer to how the owner's speaker in a room 'measures' it is still missing crossfeed and one might be getting dips and peaks that aren't really wanted.

That does not mean that the mic is useless but in practice will be limited and results are highly dependent on the owner's speakers in that room at the preferred listening position.

The ear canal does not matter much here (assuming the mic is at the exact position of the ear canal entrance) as that is not in play and thus does not need correction.

Thanks, fair points. Your caveats about the speaker in room reference are valid for that case. It was one example though, not the only possible reference, the target can just as well be a generic curve defined at the blocked canal entrance, or anything else the user prefers. At this stage I deliberately leave the choice of target to the user. The Plug is aimed at pros and measurement literate users, and they know what target serves their purpose. I will address the target question properly in the future.

 

[Marko_EarEqual]

The reference is a problem. For the Symth Realizer the German distributor offers measurements done in a 5.1 or 9.4.1 acoustically optimized studio. Pricey, but makes sense.

The home-made alternative would be a nearfield listening setup with neutral speakers. If the PostCOVID brainfog ever leaves me sufficiently for me to be able to make my way through the Smyth Realizer (original version) measurement process then I'll do that with my small Genelec (8010A + sub) system.

Good examples, and the Smyth Realiser parallel is a fair one. You are right that the quality of the reference is what matters, and both paths you mention work: a proper studio measurement if someone has access to one, or a careful nearfield setup with neutral speakers at home. That home nearfield route is a very reasonable option for people who want an individual speaker based reference without studio access. As with the target, I leave the choice of reference to the user, since the right approach depends on their setup and how far they want to take it.

 

[kemmler3D]

This is definitely interesting and I appreciate all the discussion and explanation so far. Noob question here: I understand the need for open headphones to create a valid measurement with this mic. Does this also mean the resulting correction is only valid for those headphones, or could it also be adapted to closed back cans?

Specifically: I have a set of Focal Elex that might work for the measurement but I'd also want to correct my Audeze closed-backs.

 

[olieb]

The Plug is aimed at pros and measurement literate users, and they know what target serves their purpose.

Ok, so this is more for experimenting and and tinkering.
I use a 6mm PUI mic capsule for similar experiments.

https://eu.mouser.com/ProductDetail/PUI-Audio/POM-2739L-HD3-LW100-R?qs=tlsG%2FOw5FFhhIHeRWgWyNA%3D%3D

Up to now I use silicone putty (the kind used for ear protection) as means to position it in the ear canal.

https://www.amazon.com/dp/B07VB9GYFD?ref_=ppx_hzsearch_conn_dt_b_fed_asin_title_1

This is a blocked situation of course. But simple and cheap.

I will try to measure with open canal too, for the advantages mentioned by @Marko_EarEqual .
I found this solution to place the capsule in the ear canal while keeping the canal semi-open.
I cut a silicone ear tip in half and then cut holes in the (half-)spherical part.
Will see how good that works. Placement should be reasonably repeatable and if one measures speaker-room and headphone in one go, small differences will cancel.

EDIT:
I used the 3-terminal solution from here
https://diymics.com/how-to-use-elec...balanced-input-that-provides-phantom-power-pp
For circuits using 48V phantom power you might look here.

https://groups.io/g/MicBuilders/attachment/30006/1/SimpleP48