Headphone and Headset Measurement Seminar G.R.A.S.


Hoo boy, this was a dangerous presentation for me.

Headphone measurements are terrifically difficult when you have to perform them on an artificial head. The ear canals of traditional 711 coupler heads are cylindrical and the placement in-ear monitors can be remarkably hard. The ears themselves tend to be a bit stiff, and the design includes the intention to place hearing aids in position behind the ears. As a result, most measurement head ears stick out a bit too far from the head. This makes placement of on-ear headphones quite troublesome.

In addition, traditional couplers (the device that provides the ear canal and microphone) are plagued with problems of erroneous measurements above 10kHz. The simple geometrically shaped ear canal and acoustic properties of the couplers create a very strong 1/2 wavelength resonant peak around 13kHz. This peak can shift significantly as the ear canal length changes with IEM insertion. I suspect it also is quite sensitive to the varied geometry of on-ear and over-ear headphones. It's very common to hear headphone engineers say, "Measurements above 10kHz are useless."

The new G.R.A.S. KEMAR head and torso simulator is designed to address these issues with anatomically correct ears, and a newly designed coupler.


A look inside KEMAR's head.

The key features of this new measurement head design are:

  • A new backwards compatible coupler (matches the 711/IEC 60318-4 coupler physical dimensions and acoustic response to 10kHz) that extends accuracy (+/-2.2dB) to 20kHz; introduces less distortion into measurements; and lowers measurement noise floor 20dB!
  • Anatomically correct ear canal that delivers more repeatable measurements of IEMs.
  • Pinnae that are more anatomically correct in how close the rear part is to the head, and with flexibility more like a real ear. Both of which are claimed to improve measurement repeatability with on-ear headphones.

Again, I encourage you to log in and download this presentation as it is very enlightening, but a few pictures will be worth a thousand words as I make a few comments.

Many readers will be familiar with the controversy regarding the Sony MDR-Z1R and the disparity between my measurements and Jude's, much of it surrounding the significant difference of the measured peak at around 10kHz. I think much of the curfuffle can be written off with the understanding that measurements from differing systems really can't be compared. I continue to believe these headphones have too much energy in that area, but if you look at all the InnerFidelity measurements you can see a clear tendency for my head to show a peak in that area...and I think that's a measurement accuracy error.

Jude has one of the new KEMAR head and torso simulators with the anthropometric ears and high-resolution coupler from G.R.A.S., and at this point I'm convinced it's more accurate. I do have a hard time interpreting Jude's measurements as he does compensate the measurements and I haven't seen enough of them to develop an eye for interpreting them, but I'd bet a case of good Belgian beer that the raw data is more like what we hear at our ear drum than my measurements.


Seen at left is the older G.R.A.S. RA0045, a coupler designed to meet the IEC 60318-4 (former IEC 60711) specification. To the right is the new G.R.A.S. RA0401 High Resolution Ear Simulator. The new coupler is mechanically and electrically backwards compatible with the older couplers, and will deliver the same response below 10kHz.

The ear of a dummy head consists of two parts: The coupler, which simulates the acoustic termination impedance of the ear canal and contains the microphone to detect the acoustic signal at the ear drum reference point. And the Pinna, which includes all external ear geometry and entrance part of the ear canal, it then mates to the coupler. For a good hard look at the new coupler performance check out this whitepaper from G.R.A.S. From the whitepaper, here is a response comparison of the 711/IEC 60318-4 with the new G.R.A.S. RA0401 high resolution ear simulator coupler.


You can see the troublesome very high Q peak at 13kHz of the old coupler compared to the much lower Q, and more stable in amplitude, peak of the new G.R.A.S. coupler. The amplitude and Q of the peak in the older couplers is unspecified in the IEC specification, and products are allowed to vary in this region and still remain within the spec. The peak in the new coupler, however, is specified to provide a calibrated response to 20kHz (+/-2.2dB).

Now remember that these peaks can shift significantly with the effective ear canal length while measuring headphones. That means that it is impossible to make one compensation curve that will effectively subtract this feature as it moves around. Here are some plots of both couplers response with changing ear canal length.


Coupler response with changing ear canal length. Top is a standard 711/IEC 60318-4; bottom is the new G.R.A.S. high resolution coupler.

Well, if you're not going to be able to compensate this feature due to its movement, I sure as heck know which one I'd rather be measuring with. The new coupler should provide a significantly less ragged and peaky response above 10kHz than older couplers.



On the left you can see the older IEC spec ear with cylindrical ear canal, on the right is G.R.A.S. new anthropometric ear.

The new anthropometric pinna will also provide more stable and repeatable measurements for both IEMs and on-ear headphones. The ear on my measurement head is just like the ear on the left above. I always have a pair of hooked needle nose pliers at my measurement stations because ear tips will usually pull off and remain in the ear canal when I remove the IEM. That ear canal is a pain in the butt.

Also, inserting tips into the ear canal of my head is quite troublesome. The entry angle isn't right, and if the tip is slightly too large it will fold up and create a leak. I will very often have to try numerous tip to get the IEM to seal properly, often not the tips that come with the IEM, and often I'll have to use foam tips instead of silicon.


The plot to the left above is 20 insertions of an IEM in the old standard pinna; to the right is 20 insertions of an IEM into the new pinna.

While I'm not certain of the amount of attention paid to proper insertion and the training of the operator of the plots above, my experience tells me it's a totally believable result for insertion attempts in the absence of feedback to help guide the operator. I'm pretty sure my practice of monitoring a square wave while inserting IEMs prevents me from having such a wide variation in measurements, but there's no doubt I have to struggle often and refit the IEM numerous times to get a reliable response.

One thing not mentioned in the G.R.A.S. presentation and literature is that there is a small port on most IEMs to tune the bass. At times this port is close to the nozzle holding the ear tip. I have found on occasion that inserting the IEM to the proper depth of my head's ear canal causes the flange of the tip to cover and occlude this port, resulting in uncharacteristic excess bass to be measured. It took me a long time to figure that one out and I suspect I have numerous older measurements of IEMs that over report bass response.

It seems to me quite likely this new anthropometric ear will go a long way toward addressing both these issues. I did quickly insert an IEM into the head at the seminar and it did seem easier. When I talk to Jude at RMAF he told me inserting IEMs was much easier with the new ears.


On-ear headphones with new anthropometric ear (left) and old type ear (right).

The new ear is also softer and more representative of the flexibility of a human ear. The older IEC spec ear was designed to be able to have a hearing aid easily rest behind the ear, so it sticks out farther from the average human ear.


Response variation with 20 placements of on-ear headphone with the standard ear (left) and with the new anthropometric ear (right).

On-ear headphones may be even more difficult to deal with than IEMs when it comes to placement on the head for measurement. Sometimes I just can't make it work and I will put a rubber band around the headphones on the head and let it sit there for a few hours to get the pad to take on the shape better. Then I can take the rubber band off and will usually be able to get valid measurements. Thing is, all you have to do is look at my raw measurements of on-ear headphone and you can see that they vary in the bass much more than any other headphone type.

I didn't try placing on-ear headphones on the seminar head, but I did twiddle around with its ears and they were significantly softer than my measurement head ears. It certainly did feel like it would be much easier to measure on-ear headphones with these new ears.

Knowing that the new Listen Inc. gear will have me eventually remeasuring many headphones, I did approach the G.R.A.S. representatives to ask whether they might want to sponsor InnerFidelity's measurement program with a new head to use. Frankly, I wasn't looking for any commitments. Mostly I wanted to know if it was worthwhile having to worry my way through how compatible the measurements would be with my existing database. It makes me almost sick to my stomach to think about the close to 1000 measurements I've taken, and how risky and difficult the compatibility issue is for readers if I change heads. But it's also unsettling to think about how much error might be lurking in past and future measurements using my now less than ideal equipment. This new instrument from G.R.A.S. may make for a more reliable InnerFidelity headphone measurement data set into the future.

Love to hear your comments on that one.

jim in cheyenne's picture

This is great news! Yeah it's time to update the equipment.

How to bridge between the old measurements and the new? You will have better ideas than I, but I would think a half dozen or so well chosen headphones, from modest to 'all-out' with carful discussion of the differences would suffice. Of couse all of us hope than half dozen includes our favorite....

Good luck, keep up the good work!

jim in cheyenne's picture

This is great news! Yeah it's time to update the equipment.

How to bridge between the old measurements and the new? You will have better ideas than I, but I would think a half dozen or so well chosen headphones, from modest to 'all-out' with carful discussion of the differences would suffice. Of couse all of us hope than half dozen includes our favorite....

Good luck, keep up the good work!

Oops, I tried to register, but Listen wouln'd accept it
Is ther some 'company' we are supposed to use?

JMB's picture

I quite appreciate the progress in Grass design. Our ear canals and pinnae are quite variable (even between one owns left and right ones) so exactly which would be the one to choose? Also the materials used to make artificial ears/heads are quite different to our tissues (especially regarding damping and resonance behavior). So there will be an ever changing improvement for more realistic ears (upgrade for ever like we do our toys).
I still have a more basic question about how measurements at the eardrum are relevant for headphones but not for loudspeakers. I understand that IEMs don not use our outer ear and only a part of the ear canal but that does not apply to circumaural headphones which leave the function of the ear canal and at least partially the outer ear intact as would listening to loudspeaker in free air. Basically this is a question about what is the right compensation curve.

coastman25's picture

Surely, the obvious candidates for revaluation using the new testing methods and equipment would be your current wall of fame headphones. However before going there perhaps look at models were the test results did not match your listening experience and see if that is still the case with the Listen Inc system. Then some visa a versa ie re-test some you are familiar with I know for sure your listening experience matches your current test results.

GTABC's picture

That would make the most sense to start.

I'm really interested in the subjectively poor headphones with good measurements(under the old methodology) along with subjectively good headphones with bad measurements (under the old methodology).

Gosh it would be nice if new measurement equipment/ techniques changed measurements in such a way that they would more often be aligned with subjective impressions.

I've often heard that headphones that measure poorly rarely sound good but that there are many examples of headphones that measure well but sound awful.

100VoltTube's picture

I would also love to see some more measurements of those types of headphones. In addition, it might also be interesting to see more measurements of headphones where the measurements tell one story about the sound signature and the listening tells a different one. For example, the beyer T70

100VoltTube's picture

It might be interesting to look at Fourier transforms of the headphones' output as a function of input frequency. A sort of 3D Fourier transform, if you will (or just a couple of normal Fourier plots). It might also be interesting to see how the distortion components change during the decay of a tone. Like a CSD plot, but the decay of a single sine wave.

Journeyman's picture

I do hope they help you out. It's great advertising in the long run given the fact you are a reference for many headphone enthusiasts.
I also think the headphone community would be very grateful.
Tyll I do hope you measure some Beyerdynamic Headphones models again. :-D I know you don't really like them but the DT880 is a reference for many people.

MRC01's picture

What a treat. It's great to see this level of research which will improve SOTA for everyone. Thanks for sharing the experience. As for what kind of new measurements to include, I suggest CSD plots.

zobel's picture

I imagine that the new distortion / noise measurements may be the biggest improvement in conveying perceived sound quality. Those, along with the waterfall plot, and improved SPL/freq. graph, there has to be be a much improved set of measurements that will go much further in describing the sonic features of all HP measured.

Thanks for keeping your tests as relevant as possible Tyll. I like your idea of covering the 'worthy' cans and letting the unwashed hoard go as is.

thu hien's picture
bobusn's picture

No excuses! Thanks for your dedication, Tyll!