• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Will the headphone industry ever agree on a design target curve? - Presentation by Sean Olive in November 2022

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,755
Likes
16,195
The presentation was recently held at https://danishsoundcluster.dk/danish-sound-day-2022-a-great-networking-day/ and can be downloaded at below link:


The examined main topics are:
  • Is there a universal headphone target curve that most people prefer?
  • What factors influence listener preferences?
  • Some challenges/considerations in designing and testing headphones that satisfy listeners’ sound quality preferences
The conclusions are:
  • Most listeners (64%) prefer a headphone target based on an accurate loudspeaker calibrated in a semi-reflective listening room w. 2 smaller segments preferring slight adjustments to the bass and treble
  • Personalization can improve headphone sound & spatial quality to accommodate differences in taste, hearing, listening experience and ear shape/size acoustics
  • Closed headphones produce inconsistent bass across listeners vs. open back desings
  • Different headphone test fixtures produce measurements that diverge below 200 Hz and above 2 kHz and here they don’t accurately represent average measurements made on humans.
  • While there is a strong argument for a better industry standard headphone target, loudspeaker history tells us it is unlikely to occur.

Another interesting paper from Sonarworks that is mentioned in that presentation can be found at below link:

 

MayaTlab

Addicted to Fun and Learning
Joined
Aug 15, 2020
Messages
924
Likes
1,512
Interesting presentation. A few observations and speculations :

Someone at Harman possibly messed up the individual traces for the variance between listeners. For the open headphones, one trace in B belongs to A (red arrow), and traces of B have been mixed up with the C and D traces (easy to spot because of the high-Q notches characteristic of B's response). For the closed backs, one trace each for F and H have been interchanged (red arrows).

Screenshot 2022-11-17 at 11.32.16 copy.png
Screenshot 2022-11-17 at 11.32.04 copy.png


It's also not impossible that the standard deviation graph has been polluted by these errors.

It would be interesting to know whether or not the individual traces marked by the green arrow were from the same person, possibly during the same session. They all seem to sharply drop at a specific point between 2 and 3 kHz, in my experience with both closed and open ear canal entrance microphones that could come from the blocked ear canal entrance mics failing to effectively close the canal. Or not !

Knowing that ear simulators don't perfectly capture the average of real human seatings is not uninteresting, but I'd find it more interesting to know which one comes closest to having a constant transfer function across headphones with the average of the human seatings (in other words, which one most accurately captured the relative difference between headphones) - as this is a condition for evaluating headphones according to their adherence to a target.

Speculations on my end on which headphones were part of that study, based on a combination of several tweets from Sean Olive, comparing ear simulator measurements with these blocked ear canal entrance measurements after applying a generic transfer fonction from DRP to blocked ear canal entrance, personally knowing how some of these measure with such method, and looking at some specific notches / peaks / dips across the spectrum that are invariant and stable across various measurement methods for some of these headphones :

The ones I'm somewhat confident in :
Much more speculative :
  • "C" was meant to be the Arya (Link 1), and "D" the HD800(S?), but the Sundara's traces mixed up with them makes it harder to decipher.
  • "F" could possibly be a pair of N700NC(M2?) measured passively.
  • "H" could be the Dan Clark Audio Stealth, based on these tweets (Link 1 Link 2) and a superposition of the graphs after scaling their aspect ratio :
Screenshot 2022-11-17 at 13.51.27.png


Most of these headphones were present in that tweet.

Could "G" then be the JBL Quantum One, run passively ? Would make sense if Harman wants to compare their data to the data users can obtain themselves when performing blocked ear canal entrance measurements with them (In the box you get a blocked ear canal entrance mic).

While this wasn't published, hopefully Harman also took the opportunity to measure the ANC headphone(s?) with their feedback system engaged to see how effective it is to sort out the inconsistencies at lower frequencies and whether or not the difference between the modes was consistent between ear simulators and the average of the human listeners. I would also have loved to see a pair of Bose in there (QC35 for example).

Also, just as interesting would be to assess whether or not the variations past 4-5kHz track more or less what would be most desired by each individual (For example, some headphones possibly being more adept than others at naturally "exciting" the users' pinnae in a way that brings the delivered FR closer to what would happen at their DRP if they took the place of the mannequin in Harman's reference room and / or compared to an individualised target based on each listener's HRTF measurements), or are instead for the most part random and most likely a nuisance then.
 

Thomas_A

Major Contributor
Forum Donor
Joined
Jun 20, 2019
Messages
3,422
Likes
2,407
Location
Sweden
Interesting. My own simple measurements seem to agree with respect to open and closed headphones. Except the Bose QC25 which has consistent bass across subjects.

 

MayaTlab

Addicted to Fun and Learning
Joined
Aug 15, 2020
Messages
924
Likes
1,512
Except the Bose QC25 which has consistent bass across subjects.

That would be because of its feedback mechanism. What's interesting with your test in regards to the Bose is how much it varies in the 1-4kHz range. I wish that Harman had tested one of their design because I believe that some aspects of their acoustic design could be fundamentally different from a pair of headphones like the ML5909.
 

markanini

Major Contributor
Joined
Feb 15, 2019
Messages
1,719
Likes
1,768
Location
Scania
The choice of title might lead you to think that there's a large variability of FRs or some reluctance in adopting the Harman Target. On the contrary there some evidence showng that the Target has been pursued by the industry long before Harman published it: https://github.com/jaakkopasanen/AutoEq/blob/master/results/RANKING.md A list of headphones sorted by preference score according to Harmans model. Above a score of 85 it lists more headphones developed independently of Harman, and before 2013 when the first target has published, than not.

I don't know, maybe Harmans marketing team wants consumers to think the target is an exclussive feature. But it's not.
 

Sean Olive

Senior Member
Audio Luminary
Technical Expert
Joined
Jul 31, 2019
Messages
329
Likes
3,004
The choice of title might lead you to think that there's a large variability of FRs or some reluctance in adopting the Harman Target. On the contrary there some evidence showng that the Target has been pursued by the industry long before Harman published it: https://github.com/jaakkopasanen/AutoEq/blob/master/results/RANKING.md A list of headphones sorted by preference score according to Harmans model. Above a score of 85 it lists more headphones developed independently of Harman, and before 2013 when the first target has published, than not.

I don't know, maybe Harmans marketing team wants consumers to think the target is an exclussive feature. But it's not.
I'm not claiming someone didn't design a headphone tuned to the Harman Target Curve before we published our papers -- only that we were the first company to conduct, and most importantly, publish research that provides scientific validation (ie. controlled listening tests with many listeners) for it.

The same cannot be claimed for the current ITU-R/IEC target curve (e.g. diffuse-field), which does not appear to be widely practiced.

If you look at the headphone measurement surveys done by us, Breebaart, and Soundworks cited in the presentation, there is clearly no widely agreed-upon industry standard target -- hence the title of the presentation.

The graph below from Breebaart's study shows the average response of 283 headphones. They deviate quite far from the Harman Target, indicating to me that there is some reluctance to apply it. if you compare this to surveys of professional and consumer loudspeakers I would bet that there is less deviation from a flat=on-axis response.

If you have evidence that suggests otherwise, I would be greatly appreciative if you share it.

1.4984044.figures.online.f3.jpg


 
Last edited:

markanini

Major Contributor
Joined
Feb 15, 2019
Messages
1,719
Likes
1,768
Location
Scania
I'm not claiming someone didn't design a headphone tuned to the Harman Target Curve before we published our papers -- only that we were the first company to conduct, and most importantly, publish research that provides scientific validation (ie. controlled listening tests with many listeners) for it.

The same cannot be claimed for the current ITU-R/IEC target curve (e.g. diffuse-field), which does not appear to be widely practiced.

If you look at the headphone measurement surveys done by us, Breebaart, and Soundworks cited in the presentation, there is clearly no widely agreed-upon industry standard target -- hence the title of the presentation.

The graph below from Breebaart's study shows the average response of 283 headphones. They deviate quite far from the Harman Target, indicating to me that there is some reluctance to apply it. if you compare this to surveys of professional and consumer loudspeakers I would bet that there is less deviation from a flat=on-axis response.

If you have evidence that suggests otherwise, I would be greatly appreciative if you share it.

1.4984044.figures.online.f3.jpg


The link with consolidated headphone measurements shows that manufacturers have been ignoring ITU-R/IEC standards for a long time. For higher end models very likely tuned to in-house preference curves which haven't been published.

What makes you work different is immensely important. (publicly available, large sample size across cultures, as universal as a target curve gets-no small feat thanks to solid methodology)

At the same time we live in a world of manipulation and the facts that make it special also gives ammunition to critics that wish to misunderstand the utility of the target. That's why I'm raising concern about stressing the uniqueness angle. EDIT: A common take from such bad actors claims that Harman wants the industry to adopt a V-shaped response with excessive bass. This is of course false as shown by other manufacturers having landed on similar target curves through independent R&D cycles for decades.

Basically the Harman target resembles the old ex that your uncles wife has kept in touch with in secret. The uncle being the ITU-R/IEC standard.
 
Last edited:
Top Bottom