First Peak at Head Measurements at Harman

Sweet! Had a great day at Harman yesterday with Warren TenBrook (his conversation with Sean Olive here) and the Harman engineers. Two series of measurements were made, one with both speaker channels being driven, and one with only the left speaker driven. Fifteen measurements were made for each series at +/-20 degrees, +/-10 degrees, and 0 degrees azimuth at elevations +/-10 and 0 degrees.

It's going to take me a little while to massage the data for meaningful display, but I thought I'd pop up one preliminary plot for your enjoyment to hold you over while I spend the weekend at T.H.E. Show Newport.

The following plot is an overall average of all measurements with both speakers running. The room was equalized for Harman's preferred listening curve, so one might be safe to assume this plot is a reasonable representation of how a really good speaker set-up looks on my measurements head.


The thing that set me off on this journey is Audeze's post on how measurement heads may differ, especially above 3.5kHz. One of the things that stimulated their response was my contention that the raw response plots should show a smooth, ever-increasing fall in response after the peak at 3.5kHz, as specified by the published preliminary Harman target response as shown in the following plot.


Well, at least as far as measurement of speakers using my head goes, what I found—so far—is that a dip-like feature does exist between about 6kHz and 9kHz that does relate to Audeze's claims that upper-treble response may not be as simple as the Preliminary Harman target response.

Very interesting! Of course, the question is much more complex. This firs plot shown above is the total average of both speakers being driven, but in headphones we only hear one channel in each ear. Is it more reasonable to measure one speaker only and look at only the ear on the same side? Or should I measure one speaker only and add the left and right ear together? Or should I average the response of both speakers driven with that of only one speaker driven as music has both mono and difference components?

I know these question may be a little confusing, they're a bit confusing to me as well. I'm going to approach this on a step-by-step basis and see if we can't ferret out some clarity and understanding as we unpack the data. The next post on this subject will include a variety of views of the various data. Once this preliminary set of plots is available, I'm going to spend some time in email correspondance with both Harman and Audeze engineers to elicit ideas on how to interpret the data, and will report back once I think I've got some sort of grip on the thoughts and what they might mean.

Until then, we'll just have to be satisfied looking at plots and watching the meaning slowly unfold.

Off to T.H.E. Show!

detlev24's picture

Why would Harman tolerate such a big room mode in their listening environment? Does this show up on a classical measurement - which should follow their preferred listening curve - as well?

Lots of interesting stuff going on! I would also like to see how a, by Sonarworks calibrated headphone measures on your system. :)


mat's picture

That's the preliminary plot from Tyll's home speakers. It was included in his last blog post this past Monday.

Tyll Hertsens's picture
I just love how nothing gets past you guys. Thanks for that...seriously.
It's great to know people are coming along for the ride.
detlev24's picture

You are testing your readers constantly, aren't you? ;D

I really don't want to annoy you; but if possible, please perform a measurement of a Sonarworks calibrated headphone. I think this would fit perfectly into the discussion!

And again, thanks for your unique dedication.

Ghost_Pack's picture

I use Sonarworks almost religiously when listening or mastering, but I've always found it a little dark. I'd love to see how the final result of a Sonarworks calibrated headphone compares to the target curve you find at Harman, especialy above 1KHz.

Mad_Economist's picture

It would appear that the plot you have labeled as being the average of the Harman measurements ( is exactly identical to the average of the measurements of your own speakers ( Either your friend who built them was way ahead of his time, or I suspect that there may have been a mixup somewhere...

inventionlws's picture

It may be a wrong link I suspect...

RasherBilno's picture

Unless, based on NWAVGUY's description of your intellect, it was a very clever play on words that will escape 99.99% of readers, that is...

Tyll Hertsens's picture
Um...I'm going with, I planned that and acknowledge it might have gone over the head of many.

(You should have noticed by now how bad a speller I am; lysdexia and I are familiar acquaintances.)

Tyll Hertsens's picture
Okay...proof positive that you shouldn't be writing getting up at 4AM to get a post up. I've replaced the original plot in this post (which was from last week at my house) with the proper graph. Sorry about that.
Phoniac's picture

> (Note: The big dip at around 50Hz is probably due to a room mode.)
I don't see that...forgot to remove this text?

Tyll Hertsens's picture
I was at Audeze's HQ talking about the data with the engineers when Karthick (one of them) said, "Hey, that's not the plot you posted today." So I got their wi-fi password, logged on, and saw I screwed up. I changed the graph there and then. Just got back from dinner, going to do the edit now.
poleepkwa's picture

What is the height of the dummies ears from the ground? That picture make it look pretty high compared to normal seated height.

Tyll Hertsens's picture
It was 40". The room had been calibrated prior to my arrival for that height.
poleepkwa's picture

Ok, that is then at ear height according to most. It just looked higher because of the perspective. Looking forward to the details from this measurements and also how they did the speaker measurements, or is that irrelevant to the curve for the dummy head?

Dan Wiggins's picture

Acoustic measurement systems - all of them = will measure different. Microphones are not all equivalent, heads/fixtures are not all equivalent, processes are not always the same. In the case of headphones, even the way a person will orient and place the headphone will not be consistent.

Thus data collected from one measurement system cannot be directly compared to another WITHOUT first doing a calibration between the two. You need to have a common datapoint or reference for both systems where you can identify the differences (with linear and nonlinear differences accounted for - it's more than just frequency response). Then, and only then, can you draw conclusions across measurements of both systems.

Otherwise, with a single system, you've collected data that is highly valuable for the relative differences between all measurements made with that system - but you do not know how that set of data relates to another set of data collected by a different system.

Calibration of systems against each other is crucial if you're going to compare/contrast the results of those systems. And it needs to look at more than just raw sensitivity or frequency response, as both of those can (and often do) change with SPL, THD levels will change, noise floors are different, etc.

tony's picture

Big ported Bass Reflex with a wave guided Midrange/Tweeter. Hmm, looks like a JBL but is it?

The Room seems huge but maybe it's just right for critical working. I own that exact same rug but that is as close as I'd ever get.

Tony in Michigan

HammerSandwich's picture


Tyll, did you get to listen to them for a bit? Impressions?

wktenbrook's picture

At the suggestion of our hosts, Tyll and I listened to some brief selections after the measurements were done. This was my first opportunity to hear the M2, although I know them by reputation. In Tyll's earlier post I suggested them as a leading reference speaker for the headphone curve measurements, and there they were! In fact, one of the M2 designers, Mr. Charles Sprinkle, helped with the speaker set up and calibration for Tyll's tests. When Todd Welti and Sean Olive did earlier tests I believe the JBL Pro 6332 and Revel F208 were used. Now we have three top sets of speakers measured by two HATS, and that's encouraging.

My impression was a neutral tonal balance (almost boring, and I mean that as a compliment) combined with huge image width and depth, and ability to play at a live music level with no strain. Harman's reference room is carefully constructed for good acoustics and that certainly helps. It's about 500 sq. ft. so there is space for the speaker to breathe. I listened at the prime position where the speakers had been calibrated, where Tyll's head had been positioned, and where Harman had conducted their head measurements.

HammerSandwich's picture

Thanks, Warren. Seems that the M2s lived up to their reputation!

zobel's picture

From 400 Hz up, pretty near the same! You didn't have a subwoofer, and your smaller speakers were further away from the back wall, resulting in the reduced bass in your living room. By your measurements at least,those are very nice speakers your friend built for you, and combined with a sub, and improved room placement will get you very close to the Harmon sound, but with less maximum SPL, probably.

tony's picture


Thanks for figuring it out.

These are large speakers, the room itself must be gigantic to make them look so diminutive. Sweetwater is selling them for $20,000 per pair. They weight 130 lbs each. Plenty of dynamic range. Phew.

I wonder if Wilson's $250,000 speakers are as good? or 10 times better.

Tony in Michigan

detlev24's picture

Hi tony,

you cannot compare the M2 to something like the Wilson's. Target audience is different! The M2 is amongst the most neutral and accurate (see anechoic measurements) full-range loudspeakers on the market and is meant to be a working tool. So, if one strives for natural and accurate music re-/production, the Wilson's and most other heavily overpriced consumer loudspeakers do not even get close to the M2. Certainly, Harman/JBL Professional is not the only manufacturer capable of building exceptionally good loudspeakers; see Genelec for a further reference.

Unfortunately, the M2 are not the best loudspeakers for home use. The reason is not, that you need a big room (you do not!) - but to make them shine over their whole dynamic range, the recommended Crown amplifiers (those incl. DSP) are essential. They require a separate room, since cooling fans are very loud.^^ You can use other amplifiers, but then need a BSS Soundweb London DSP, since M2 presets have to be loaded for crossover etc.

[Bass reflex ports on the front, to enable in-wall mounting, see ]


tony's picture

I was being naughty with my satire. I was in the business of selling consumer Loudspeakers ( a long time ago ) and I'm a big fan of Genelec stuff and small loudspeakers like the ProAc Tablets and the Halbreths Tyll owns.
I can agree that the M2 system is a "Tool of the Trade".
I won't be surprised to see a Japanese audiophile having them in his home-based "Jazz Club".
Sennheiser HD series have replaced my Big Music system for a tiny fraction of the cost, I feel "liberated" to criticize the super-expensive gear I once tried to sell to my Golden Eared Audiophile customer base.

On another note:
I just had a brief meeting with one of our long time, Major Audio Reps. Ray Wright. Ray closed his Rep business in 2002, he worked in the Consumer Audio from way back in the day.
Ray has lost his hearing, his ears no longer function, to the point of pain and agony. The medical people drilled a hole in the left side of his skull and implanted an Electronic device to by-pass his ear drums. The new system is horrible. He now cautions everyone about high SPLs ( signal pressure levels ) that were always part of the Sales and enjoyment of home & Pro Audio. Ray is younger than I, his lovely wife tells his sad story. Ray's hearing problem is apparently a common peril for those working in Audio. I'm sharing this story as a request from Ray and his Wife as a caution to all.

Tony in Michigan

thefitz's picture

What's interesting is that head-fi seems to loathe the EL-8 exclusively due to your measurements of it, and that giant 6kHz notch. That seems to be an area your dummy head has a notch in by itself. Not saying that the EL-8 doesn't have a 6kHz notch, but the hate it receives might be amplified due to the dummy head and the EL-8 itself working together to scoop out as much of 6kHz as it can.

It also explains why there isn't this giant 7kHz ice-pick for your HD700 plots.

immersifi's picture

The mannequin head microphone shown is capable of a variety of equalizations ranging from Free Field (FF), Diffuse Field (DF), Independent-of-Direction (ID) and user-defined EQ's.

In contrast, using a pressure mic or even free-field microphone (i.e. Bruel & Kjaer type 4189 or equivalent) is close to flat over most of the audible spectrum, so the spectra observed from the mannequin's ear responses will not be similar to what a set of free-field mics would generate unless the mannequin EQ was set to FF. However, a true free-field exists in very specific cases only, and in the case of the mannequin shown, I seem to recall that FF implies a) anechoic conditions, and b) the source located directly ahead and at a distance of approximately 1m, neither of which are FF equalization would also be incorrect from an analysis standpoint.

I would argue that the ID equalization is likely the best choice given the room pictured, but for analysis, one should theoretically de-convolve (i.e. correct) the equalization that was imparted during data acquisition so that the equalization is not a part of the time domain signal when wanting to compare the data to what would achieve with free-field microphones. I didn't see any mention of that having been done, but I may have completely missed that part, and if that was indeed done, well then "never mind" as they say. BTW, you can find more information on the EQ curves at this link:


immersifi's picture

Sorry, I wrote: "However, a true free-field exists in very specific cases only, and in the case of the mannequin shown, I seem to recall that FF implies a) anechoic conditions, and b) the source located directly ahead and at a distance of approximately 1m, neither of which are FF equalization would also be incorrect from an analysis standpoint.".

What I should have written is as follows (corrections [appear in brackets]:

"However, a true free-field exists in very specific cases only, and in the case of the mannequin shown, I seem to recall that FF [equalization in that mannequin type] implies a) anechoic conditions, and b) the source located directly ahead and at a distance of approximately [10 ft (3m)], neither of which are FF equalization would also be incorrect from an analysis standpoint."

Sorry about that...

ADU's picture

Not sure I get your point, immersifi. The Stereo Speaker Response graph above is _uncompensated_. So you're seeing the _raw_ in-ear frequency response of the loudspeaker before any FF, DF, or IoD correction has been applied to the data.

The data above is still preliminary, but what we're trying to compare here are the raw in-ear measurements of the Inner Fidelity, and Harman head and torso simulators to see if there are any obvious differences between the two measuring systems.

In the article above, Tyll says that "the room was EQ'd for Harman's preferred listening curve". I assume that means the 2013 "Preferred In-Room Loudspeaker" response (or "PIRL", as I call it), which is the same room correction used in the latest 2015 Harman Headphone Target.

For reference, the raw _out-of-ear_ frequency response of that preferred listening curve should look something like the _black_ Loudspeaker curve on page 38 of this PDF...

The raw _in-ear_ frequency response of the Harman PIRL looks like the blue curve on this graph, when measured with Harman's head and torso simulator...

So this is really what we should be using for comparison with Tyll's raw in-ear Stereo Speaker Response plot above.

There are some fairly obvious differences between Tyll's plot and the blue curve above, especially in treble. Tyll's data may still need to be massaged a bit though, to insure that we're actually comparing "apples to apples".

ADU's picture

"Insure" should have been "ensure" in my post above.

immersifi's picture

Ah, thanks for that.

When I skimmed the article I thought that the response from the mannequin head's conditioned signals (i.e. EQ having been applied by the mannequin's companion electronics) had been plotted without the EQ selected having been complemented. It sounds like you're suggesting that probe mics were used within the two mannequins' ears. In this case there would definitely be differences between different mannequins owing to differences in dimensions. Plus, measurements at the ear canal can be tricky in terms of repeatability which more or less necessitates multiple measurements and taking the average thereof. After that, the various EQ options will then affect the data - it might be wise (assuming the inner fidelity head and the Harman head are different) to place them both in a homologated reverb chamber (something suitable for ASTM C-423 for example) and compare their diffuse field EQ responses (averaging several locations for good measure). No, it's not directly applicable to the in-situ comparisons in the listening room, but at least you would have a feel for how much alike they are under a defined acoustical boundary condition with DF equalization invoked for each - I would think they would be quite similar.

Ah well...I'm rambling...I'll have to go back and re-read the article in any case.

Good stuff...

ADU's picture

There are a lot of pieces in this puzzle. And they're not always fully spelled out. So some confusion and/or befuddlement is certainly understandable. :)

This isn't just an exercise in plotting btw. The ultimate objective is to try to define a new and (hopefully) better target curve for Inner Fidelity's compensated headphone graphs... Something to replace the current Independent of Direction plots.

Tyll and the folks at Harman both feel that this new target headphone response curve should resemble the response of an anechoically flat speaker in a good listening space. So that's why we're comparing their in-ear measurements of loudspeakers in Harman's reference listening room, rather than in an anechoic chamber.

ADU's picture

I still have some questions about whether Harman's 2013 Preferred In-Room Loudspeaker curve is the best model for a neutral room (and headphone) response btw, some of which are discussed in my two most recent posts here...

I'll be doing more plots like the ones in the above link, including a comparison of the Harman PIRL to the JBL M2 in larger venues.

tony's picture

Hate is far too strong a word and incorrect.

I bought these same EL-8 phones from a Meet Audition, they played nicely on a Valhalla 2 well tubed-up with Russian super glass. Impulse buying got me, again. I should've bought the Valhalla 2 with Russian tubes.

At home, the Audeze EL-8s were a significant step down from the Sennheisers I already own.

I don't think anyone "Hates" any of this headphone stuff. Some stuff just seems better value for money.

As much as I felt "Buyer's remorse" from the Audeze purchase, I wouldn't "talk it down". I didn't write a "bad" review either. Plenty of folks seem to love the darn things, maybe I got a sample that doesn't favor my Asgard 2. Bob Katz seems to love his Audeze as do a great many others.

As far as Tyll, up in Montana, he's got some serious Gold Standard stuff he's used to. Pretty darn hard to impress guys like him or Bob Katz ( for that matter ).

Still, Audeze, Stax, Sennheiser and Hifiman are the Heavy Hitters in headphones, even their lesser stuff sounds darn good for a civilian like me. I'm still a dynamic lover though, I even sold my Stax gear for Sennheiser. Trying some Audeze further convinced me that I'm a dynamic driver guy.

I'll give-up on Tyll if Hate starts creeping into his writings.

Tony in Michigan

HiFiArt's picture

It's funny. I once bought three pairs of speakers from a headphone store, because I liked the sound (KEF X300A, Teac S300-Neo, Mirage OMD-15). Of these, only the KEF I was disappointed in, because it was the only one I listened to, actively.

I've found that I only enjoy the sound of speakers, or people sometimes, with the sound of headphones, playing, in the background.

I hypothesize (different music playing), it's equivalent to dither (noise) hypothetically increasing the apparent playback ... resolution. ;)

Although the Sennheiser (HD 800) did nothing for me, although the Audeze LCD-3 was only 80% of the Stax 007, I resolve to buy (and use a bit), both headphones. :)

Ah, HiFiMan. I wish I had the 500. I have the 300. The 1000, I wish the diaphragm (it's so good) was ... used for! speakers! :)

DonGateley's picture

One more bid to get you to add a right click ability to graphs that would allow download of the same info as linear phase sample files.

Once set up its a nasty filing job that I'd volunteer for.

Magic can be done and is being done with DSP but to interface this site to that world of praxis such IR's would be oh, so helpful.

DonGateley's picture

Aargh. I shouldn't limit that to linear phase if full phase data is also available.