Getting Ready to Measure My Head at Harman's Labs

Thursday I will be at Harman's research laboratories to measure what my head hears when in front of a very highly calibrated speaker system. It's an effort to drill into the measurements and, slowly but surely, come to a better understanding of what's going on.

If you want to catch up, it started with my Audeze LCD4 review. Then Audeze brought to my attention that these so-called calibrated heads measure differently one from the other. Pretty convincing stuff, so I started looking into what I might do to investigate the issue. Then Warren TenBrook stirred the pot with a very interesting email exchange with myself and Sean Olive. And that got me thinking: Since the premise Harman target response curve is that headphones should be tuned like good speakers in a good room, wouldn't it be a good idea to have a look at what my head, in particular (since it's the one used for the 800 or so measurements in the InnerFidelity headphone measurement database), hears when in front of a good speaker system? Since I have to go to L.A. for T.H.E. Show Newport next weekend, I asked Sean Olive if he would allow me to measure my head as part of the trip. He very graciously agreed.

And then I got to thinking, Geez, I'm not really set up to measure speakers with my head...I better do some cogitating and get this figured out in advance. MeasuringHeadAcoustics_DryRun_Photo_Scraps

The first step was to mount the head on a tripod...that turned out fairly easy. I won't bore you with details. The next bit is more complicated. I will be measuring the head's response at a number of angles and will then average all the responses to get some spatial averaging to smooth things out a bit.

I started with paper and cardboard models as you can see from the above photo. The result was the little gizmo you see below.


It allows me to set the azimuth and elevation fairly precisely. I'll be taking 15 measurements; 0, +/-10, and +/-20 degrees in azimuth, and 0 and +/-10 degrees in elevation. I'll be doing that with the speakers driven in stereo, but I'll also be repeating the measurements with only one channel driven in case there might be some differences worth note.

Here's the result of the dry run at my house with some pretty middle of the road speakers a friend built for me a long time ago, measuring at the 15 angles, averaging them all together, and then smoothing the data.


I was surprised it came out as well as it did. Big notch at 45Hz is probably a room mode. The peak at 3.5kHz from the concha bowl is right where it should be, but the run-up from about 600Hz to 2kHz is flat when it should be rising gently. This could be the speakers, but I've got a hunch that it actually may be from the fact that my head has no torso, and that rise is, in part, due to the presence of a torso. I suppose we'll know more after the measurements this Thursday.

Anyway, stay tuned, I'll post the results (and spreadsheets for all you math geeks) next week.

Josuah's picture

I am trying to duplicate your initial test setup but am having difficultly temporarily removing my head so I can attach it to a tripod. My first attempt resulted in a man in a white coat yelling at me for a good 30 minutes.

I'm looking forward to reading more about this and seeing the results!

Tyll Hertsens's picture
Science are careful!
Johan B's picture

I hope that the Harman studios have some KEF UniQ drivers or perhaps electrostatic speakers for some proper omnidirectional room response. Otherwise the mesuring head is only measuring the inadequacies of the speaker drivers

Tyll Hertsens's picture
Oh...this room has some serious speakers. Pix for sure in the coming week.
wktenbrook's picture

I agree that KEF uniQ drivers have some of the finest directivity in the business. Harman's room will likely feature JBL Professional and/or Revel speakers, which are equal or better than the KEF line in most respects. KEF may be a bit superior in controlled vertical directivity, but the Harman speakers' horizontal directivity is as good and they likely play reference levels with less distortion and low dynamic compression.

Electrostatic and planar speakers have their fans, but they are not omnidirectional. They have a lobe of sound toward the listener, another lobe directly rear, and nulls right, left and vertical. In my opinion this is irregular directivity, or at least unconventional, and would not be suitable for Tyll's tests.

MRC01's picture

Planar speakers having ribbon tweeters (like Magnepans) radiate sound much more evenly than electrostats, at least in the treble. Since the ribbon tweeter is narrower than the wavelength of sound it emits, it radiates sound evenly and you don't get the complex beaming patterns that electrostats have.

Sebas's picture

I can't wait. Can we get the raw data as you did with other measurements to see what goes on? I loved to do the exercise of processing big chunks of data on Excel and R!


johnjen's picture

This is great.
To be able to follow along with a form of calibration exercise…

It reminds me of that tv show with the tag line of…
"Don't you just love it when a plan comes together…"

I see this as a way for us to 'peer under the hood' so to speak.
And learn a bit more about measurement systems.

2x dbl thumbs up.


GNagus's picture

It took me a month to build these speakers you you!

--your friend

ednaz's picture

Always knew that I'd get better audio if I didn't have to drag around this meat scaffolding.

Too funny. But I'm concerned that it may be correct... THEN what?

tony's picture

I suspect you'll pick-up a measure of "Earned Confidence" from working at the Harmon facility or B&O or GM Technical Center or Ford Labs or any of many other places like Phonak.

Bob Katz exudes confidence in his writings, well, he own's Standards that he created, he earned his confidence, in my book.

I think it's brave of you to venture into the Lab, a place where 2+2=4 !, and it speaks volumes about Harmon's appraisal of you, to invite you. I've never heard of any Journalist being allowed into an Auto Industry R&D facility ( for any reason, much less to do technical work ).

If you are allowed to write about all this it'll be a Must-read.

Tony in Michigan

ps. I'll buy the next Harmon product you review

Bennyboy's picture

Are these Harman people the same ones that made the utterly craptastic in ear monitors I bought a while ago, thinking they must know what they were doing? Jesus was I robbed.

Hi-Reality's picture

..I've got a hunch that it actually may be from the fact that my head has no torso..

Tyll, I'm thinking a factor more important than the presence of a torso: wouldn't it be a more realistic/genuine measurement if that generalized (Neumann-like) dummy head would be an exact and 1-to-1 scale dummy head of your actual head (i.e. head-size, ear and facial shapes)?

Regards, Babak
Hi-Reality Project

Hi-Reality's picture

That way, the 'Speaker Response Test Measured by Head Acoustics' could be an individual HRTF.

MRC01's picture

It would be great to see the in-room frequency response measured at the listener position with a standard microphone, to compare against the mic inside your head. Correcting both for mic response, of course. The key would be not the shape of either curve, but the difference between them.

ADU's picture

In-ear and out-of-ear measurements with different amounts of room reflectivity would also have been interesting, because the in-ear resonances at around 9 and 15 kHz seem to be more related to diffuse or indirect sounds coming from the sides, rather than direct sounds from the front of the listener.

Bass management might also be important for accurate in-ear readings in the lower frequencies, based on some of the new plots here...

Post 176 compares the new Harman headphone target's room correction curve (referred to as "PIRL") to the trained, untrained, etc. listener preferences in Floyd Toole's recent AES paper discussed here...

And post 177 compares the Harman target's room correction to some loudspeaker measurements in a small venue.

McLelland's picture

This maybe off-topic, but with diminished hearing above 8kHz in both ears at age 69 and with event-specific damage to my left ear causing a notch between 2-6kHz, I've been wondering how I might work with different tools to vary the channel output to my HPs to compensate for some of these deficits. Tune my ears, so to speak, instead of the room,. In regards to this, is there a way to adjust the mics in your head system so their capabilities mirror your own audiological frequency responses?

MRC01's picture

Philosophical question: do you want music on your headphones to sound like what you would hear during a live event - including your unique response curve? If so, flat response (psychoacoustically speaking) is best for you just like for any other audiophile. Or do you want it to change your unique response to "normal"? Then the music won't sound like live music anymore, but it might sound like it would have when you were younger.

If the latter, I suggest using a digital signal processor between your source and your DAC or preamp. I use a Behringer DEQ2496. It can manipulate frequency response and dynamics, and it can apply different processing to the L and R channels. I'm sure there are other options, that's just an example.

McLelland's picture

I am thinking in terms of the frequency range deficits in my hearing as equivalent to the recent unauthorized, permanent installation of an EQ into my hearing with significant presets I don't like. I'm in a new normal audio environment. I've used a flat response to listen to music in the past and I am now thinking my ears can no longer hear that flat response because of the deficits. Is that a reasonable way to consider this?

I'm not even sure how much I am affected by what I am missing regarding subjective enjoyment of listening, but am curious about that. I'd like to see if I can discern effects of those hearing deficits and establish parameters that correct them to whatever degree is possible for an improved subjective audio experience,particularly in terms of tonal balance and my experience of the sound stage. It will be OK if I can't because I am continuing to enjoy my listening as is.

The DEQ2496 looks like an excellent test bed for experimenting. I assume its described 24-bit/96kHz A/D and D/A converters won't affect the performance of my DAC, a Gungnir multibit. Yes/No?

MRC01's picture

If you put the DEQ2496 in the pure digital path: between your source and your DAC (this is how mine is set up, using optical in & out) then its DA & AD converters will never be used. In this mode it is a pure digital signal processor and introduces no measurable distortion. I ran some test signals through it with some simple bench tests for intermodulation & harmonic distortion and could not see ANY distortion AT ALL. Not even below -100 dB. That's as expected for a pure digital signal processor with no AD or DA conversion.

The DEQ2496 also has DA & AD converters so it can listen to, or drive, analog components. I don't use it in this mode, so I can't say whether its DA & AD converters are any good. It sounds like you don't need to use this mode either.

McLelland's picture

That is very helpful. I am using optical connections and that is exactly what I need to know. McL