Initial Results for Head Acoustics HATS Measurements at Harman

Photo Credit: Warren TenBrook
The measurement head positioned in the Harman listening room.

I'm up against a bit of a wall here as I'm going on vacation next week—John Grandberg will have a headphone amp review going up, so stay tuned—so I haven't been able to do all the number crunching I'd like. None the less, I'm far enough in to provide some cool graphs of the results so far.

Room Measurements

The above plot shows the measurement of the room with a calibrated microphone at the listening position 40" above the ground. The smoothed line in the background is the Harman preferred listening curve. I did the measurements with this EQ as I wanted to measure the preferred sound of speakers with the head. I do have this measurement data, and later will provide views of the results both with this curve and with it subtracted, which would model a flat response.

Stereo Measurements

In the above plot the light blue line is the average of all measurements made with both speakers running. The red plot is a moving average smoothed plot. I will be doing further number crunching on this data set to see if there's anything to be learned from angular changes.

Measurements with Left Speaker Only

The above plot shows the total averages of left and right ear measurements for the left speaker only. Notice the closer ear has the notch feature at 6-8kHz.


The above plot shows an average of all the azimuth measurements at the three different elevations. It seems the main effect of elevation change occurs in the 600Hz to 2kHz area.

MeasuringHeadAcoustics_Measurements_Graph_LeftSpeakerAzimuthAverages Here's the interesting one: The above plot shows the average of the three elevation measurements at the various azimuth (left to right) angles. Here you can see a definite trend that as the head gets closer to straight on to the speaker, the dip at 6-8kHz and subsequent rise in treble gets higher.

I'll have a lot more to say about these plots in a couple of weeks after I get a chance to really digest them. I've also had a chat with Arnaud—long-time headphone enthusiast and hard-core audio measurement geek—and he's going to do some curve fitting with tools I don't posses and should be able to aid in coming up with some useful smoothed curves. Perhaps to use to build a compensation curve for this head.

The tricky bit is figuring out what curve, or average of the curves above, should be used in compensating headphones. This is not a trivial question. For a start, on speakers you hear both speakers at each ear; with headphones you only hear the left channel in the left ear and the right channel in the right ear. Also, audio energy adds a bit differently in the room with both speakers going as opposed to just one. I think we're going to have to try a variety of curve compensations.

Once I've got some good prospective curves, I'll use them to compensate real headphone measurements so readers will have a chance to weigh in on which ones look most like the headphone sounds. Should be interesting.

I'd very much like to thank Sean Olive, Todd Welti, Henry Goldanski, Charles Sprinkle, Omid Khonsaripour, and any others who've help on the Harman team. This was a golden opportunity for InnerFidelity readers and myself...thank you! Also want to give a big shout out to Warren TenBrook who's been a big inspiration with his emails that stimulated this event and for showing up to add his comments and take pix. Thanks, mate.

Those of you who would like to download the spreadsheets from this DropBox folder.

Happy number crunching!

zobel's picture

Did the overall frequency balance suit your ears? A tad heavy in the bottom end maybe?

detlev24's picture

I am also a bit surprised by the sudden dip around 100 Hz (thought transition would be flatter). Maybe this is because of how we perceive sound intensity at different frequencies - our ears are very insensitive especially below 100 Hz - see:

Does the 'Harman preferred listening curve' just represent the preferences their test subjects showed, meaning which sound signature they liked more; or does it actually represent a/the frequency response close to what sounds natural (based on real-life experience) to human brains? Or both?


Tyll Hertsens's picture
It sounded REALLY good, but I'm no expert on speaker sound. The thing about bass is that in a room like that there's all sorts off visceral, impactive sense of the sound you don't get with headphones. I didn't have much time listening so I really can't say how it relates to bass on headphones. Also was their observation that people liked more bass level in a room than they did on headphones....counterintuitive, but that's what they found.

My gut tells me that there's too much bass in the measurements though, and will likely need about 2-3dB less.

ADU's picture

Also was their observation that people liked more bass level in a room than they did on headphones....counterintuitive, but that's what they found.

I think Harman essentially reversed themselves on this after the more extensive 2015 headphone study, and concluded that the bass and treble preferences were, in fact, the _same_ for both headphones and loudspeakers. That's why the bass (and treble) was revised upwards on the latest Harman headphone target shown here in blue...

As far as I can tell, Harman is now using the same preferred in-room loudspeaker curve (or "PIRL") for both their reference listening room, and for the room correction on their latest headphone target above.

The term "PIRL" comes from the black Loudspeaker curve on page 38 of this PDF...

I'm pretty sure it's the same as the red curve on this graph...

See my comments/links in the other recent Harman-related articles/blogs here for more on the PIRL.

ADU's picture

I'm pretty sure it's the same as the red curve on this graph...

I should say the _pink_ (rather than red) curve on this graph. Namely, the smooth "SP Preferred 2knob" curve.

ADU's picture

I think the curves on this new plot are shifted a little too far to the left btw...

On all the other graphs above, the main resonance in the treble is closer to 3.5 kHz.

Arve's picture

Having tried the Harman target response for speakers in my own listening room, I've come to the conclusion that it's much too bass heavy, and it tends toward producing bass that is "one-note" in its quality, because there is usually way less musical information below 50 Hz than people think with most music.

It also needs to be understood that the -5 dB roll-off from 200 to 20000 Hz needs to be tailored individually to the room and listening position. In my case, since I have very little sidewall reflections, it just about tips the scale toward sounding too dark (with or without the insane bass bump)

wktenbrook's picture

As shown in the photos, Harman provided their flagship M2 for Tyll's measurements. They only sell the M2 monitor system with accompanying DSP and amps so it can be individually tuned for the room. The curve used for Tyll's measurements was basically Harman's standard Reference Room curve, but that curve may not translate to another room with significantly different acoustics.

Phoniac's picture

'Notice the closer ear has the notch feature at 6-8kHz.' I only see a peak. Where is there a notch, please? 'Dip 6-8 kHz'? Same.

Phoniac's picture

Oh wait - that 2 dB (!) notch? I might have got it...

tony's picture

I've been following S.Olive & Harmon for the last couple of years, now we have Tyll going deep into the Research World.

Collecting, correlating Data, reasoning thru concepts, measuring, resulting in the actual manufacturing devices that give us civilians a Symphony Orchestra in the palm of our hands, it seems a miracle!

We are seeing all this unfold in one life time.

I'm trying to imagine Ben Franklin sitting in Tyll's Montana Research Lab ( sitting in that Room containing all that Big Sound gear ) experiencing what those participants experienced, thinking that maybe a little Lightning Experiment ( 250 years ago ) started all this! Phew.

My Psychiatrists tell me the Human Brain is a Computer with an Operating System. Here ( today ) we have folks like Harmon and Tyll writing program Applications that will further enhance it's functionality.

I remember 70 years ago. We had little cartridge Microphones and cartridge speakers in our Western Electric House Phone. We had sound at the Movie Theaters from those Voice of the Theater EV Loudspeakers ( probably the best sound Quality we'd experience ) and we had Table Radios for Base Ball Games and we had 78 Records ( just a few ). We seemed happy enough!

But 33 Vinyl brought lovely sound quality within reach. That was only 50 years ago!

Tyll & friends are discovering and describing the Target ideal that manufacturers will use to create the next generation of devices that I'll be listening to.

Thank You Tyll & Harmon, I have a feeling that we all will be standing on your shoulders from now on.

Tony in Michigan

ADU's picture

Tyll wrote...


Here's the interesting one: The above plot shows the average of the three elevation measurements at the various azimuth (left to right) angles. Here you can see a definite trend that as the head gets closer to straight on to the speaker, the dip at 6-8kHz and subsequent rise in treble gets higher.

The "peaks and valleys" in the treble will definitely change with the direction of the sound source.

This PDF shows free field plots from 4 different angles for 3 different HATS systems (B&K Type 4128-C, G.R.A.S. KEMAR Type 45BM, and Head Acoustics HMS II.3)...

There are differences in the frequency response of all three simulator heads. But they each respond similarly as the direction of the sound changes.

When the heads are directly facing the source (0 degrees azimuth), there's alot of energy around 12-13 kHz in the treble.

When the head is turned so sound is coming more directly at the ear from the side (90 deg azimuth), the "notch" between 3 and 12-13 kHz disappears, and other resonances begin to appear at different frequencies in the treble, esp. around 9 kHz.

The resonance at 9 kHz becomes even more well-defined when the sound is coming from the _back_ of the head (180 deg azimuth). And there are also pretty well-defined resonances at 15 kHz & 20 kHz on the B&K model, and at around 13-14 kHz & 20 kHz on the Head Acoustics model.

These are the same type of treble resonances that show up fairly routinely on both the Inner Fidelity and Golden Ears plots of the better headphones...

Average of 21 good GE headphones:

I assume they're showing up more vividly on the headphone plots because the acoustics, angle and diffusion of the sound are different than the loudspeakers in the Harman reference room.

The resonances at 9 and 13-15 kHz might show up more on the in-ear loudspeaker plots though if the Harman reference room was less "dead", and reflected more sound from the side and back walls; or if the simulator was seated in a chair closer to a back wall more like a normal listener, rather than positioned in the middle of the room.

Seth195208's picture

Compared to a wide despersion tweeter based design, the waveguide minimizes room interaction and defeats, to a certain degree, the purpose of the measurement. Hopefully, Sean Olive could comment on this.

ADU's picture

Any idea which type of tweeter is more common in home or audio mastering setups?

Seth195208's picture

are much less common in these setups. A waveguide is used to match the dispersion characteristics of the cone driver at the crossover point, which in this case (10" cone at around 2.5khz)is quite narrow compared to a normal loudspeaker.

wktenbrook's picture

I think there's a misunderstanding. Waveguides don't minimize room interaction unless it's by design. Waveguides typically encourage well-controlled dispersion so early reflections from the room match the direct sound. This is the philosophy behind the M2 used for the measurements.

It's tough to find a conventional speaker that *doesn't* have a waveguide.

Seth195208's picture

It's actually a 15" dynamic driver with even narrower dispersion at the crossover point. No matter how you slice it, the early wall reflections will measure nothing like the direct sound. That's just plain physics.
I should have clarified in the first comment that I was referring to deep horn type waveguides, not the ultra shallow type as on the Revel Salon2 and Studio2, which would have been ideal for these measurements.

germay0653's picture

The Harmon FR target curve is very similar to what many of the participants in autosound competitions strive for when SQ is judged. They also focus on the following: Tonal Accuracy, Sound Stage, Imaging, Linearity and Absence of Noise.

I know I'm stating the obvious but headphones do not have the same room/cabin/environment interactions that speakers do, i.e. reflections or room modes. With headphones you're typically only hearing the environmental interactions present in the studio/venue where the recording was made.

It's interesting to see the response curves using the "heads" in a somewhat controlled environment and hopefully that will allow correlations to be made between the two.

DC2Light's picture

Your tests aside a moment, I don't get Harmon's own in room goal with their speaker set up. What ever happen to a flat response or as close to a square wave in/ out concept? (This idea carries over well for headphone calibration) Their curve looks like a giant loudness button. I could never mix through that. I comprehend it to an extent in headphones but not speakers in a room. Sounding good is arbitrary and accuracy is something else.

wktenbrook's picture

Sean Olive replied to my earlier post: "...We agree that a flat in-room response for a loudspeaker will sound too bright and thin..."

The proper design objective is flat speaker response on-axis only in an anechoic chamber, with controlled directivity and sound power off-axis. Then, when the speaker is placed in a good listening room, natural room gain will tilt the response toward somewhat more warm bass and less treble. This is philosophy of the M2 system used for Tyll's measurements. If you try to fight the natural tendency of the speaker and EQ it flat at a mid-field or far-field listening seat, it will be bright and bass-shy.

The Harman speaker curve posted by Tyll is for the M2 in a ~500 sq ft room, listening about 12 ft away. If you are mixing near-field and hearing primarily direct on-axis sound, the curve won't be correct.

DC2Light's picture

A 10 db tilt is not a small amount. I would live with what others might call "thin" but of course I would have my room in much better shape then the average living room. I very seldom EQ and if I do I'm cutting with passives. As far as this test I would still go for "0" across because he was concerned with his own calibration I thought.

Jim Tavegia's picture

is what I have in my main room at my listening position. When I first measured it I was concerned that the HF was too soft, but after years of listening I have found it to be very accurate, or I've just become so used to it that I've come to prefer it.

I can tell you that headphones that have severe dips from 2khz out to 10khz just don't do it for me and I am dealing with an expensive pair (to me) right now that are very HF deficient.

Variations of + or - 3db can be lived with but -10db or more cannot.