Headphone preference targets vary with age and hearing loss

[Keith_W]

Preamble
All of us on ASR who follow headphone science are aware of the Harman headphone curve. Amir draws in the target curve for all his headphone and IEM measurements so that we can see how closely the device follows the curve. An excellent discussion on the Harman curve can be found here.

It is important to remember that the Harman curve is a preference curve - it was derived by applying the HRTF to an anechoic flat speaker response, and then allowing listeners to adjust the curve to their liking. Over the years, the Harman curve has been refined as shown above.

Sean Olive found that there are 3 groups of listeners which can be broadly divided into "Harman curve lovers", "less bass", and "more bass". The first group made up 2/3 of all listeners tested, with the remainder equally divided into "less bass" and "more bass".

It is also known that issues with the test procedure affect the repeatability of the result. Anything below 200-300Hz is subject to variables which may produce air leaks such as clamping pressure, wearing eyeglasses, improper seals, and so on. Anything above 2-3kHz is subject to individual HRTF - shape of ears and pinna, length of auditory canal, etc.

Miller and Downey 2023
I was recently made aware of a study performed by the above two authors published in AES in Oct 2023. The PDF can be downloaded for free from the AES website. The most interesting finding of the study was that there is a clear preference for more high frequency energy > 10kHz. In addition, individuals with hearing loss or older age preferred even more high frequency boost.

The authors subjected the study participants to a hearing test and collected data on hearing loss.

The Harman curve was then modified with several treble boosts, up to 27dB. The study participants then listened to music and the treble shelves were tested blind using the MUSHRA (Multiple Stimulus with Hidden Reference and Anchor) protocol.

Result

As can be seen, older individuals had a clear preference for treble boost, up to 12-27dB over the Harman curve. What is also interesting is that all test participants preferred more treble over the Harman curve.

Based on the above findings, the authors proposed a modification to the Harman curve, with a different curve profile based on the age of the listener.

So now we know that some listeners prefer less bass / more bass / neutral, and this new study shows that preference for treble boost > 10kHz varies according to age group. Some ASR members are dogmatic about Harman curve compliance - this may be a reminder to keep a more open mind.

 

[LGD_]

I am 75 with a pretty significant HF hearing loss - I wear hearing aids.

Experimenting with my EQ program (Electri-Q) in foobar, I came up with a curve that gives me excellent sound in my headphones (Sennheiser HD 650), with NO hearing aid.
This is my headphones curve:

The < 500Hz part is a copy of Amir's bass EQ, while the >500Hz is my hearing loss correction.
I use a similar curve in my second system (using speakers), both work great for me.

Obviously, this is an individual thing, I'm not suggesting this particular curve for anyone else.

 

[Curupira]

As can be seen, older individuals had a clear preference for treble boost, up to 12-27dB over the Harman curve. What is also interesting is that all test participants preferred more treble over the Harman curve.

The age group from 22-35 and 36-45 preferring a much more subtle decay on higher frequencies (>9khz), and the chaotic nature of those frequencies with dips and peaks, might indicate something that I always suspected when looking at the IEM industry (and even some OEs) and "community" target curves: Harman heavy smoothing may be the main culprit behind the aggressive slope present on their target.

For me personally, being 28, the slope always seemed off, too inclined and "muted" when considering treble naturalness.

Intriguing to find that all of the participants preferred MOAR treble overall though, even considering the high mids and lower treble, which is a common criticism of Harman coming from the new "meta" enjoyers. Were they tested with Over ear or in-ear headphones?

Edit: seems like they used IEMs

 

[Snarfie]

Preamble
All of us on ASR who follow headphone science are aware of the Harman headphone curve. Amir draws in the target curve for all his headphone and IEM measurements so that we can see how closely the device follows the curve. An excellent discussion on the Harman curve can be found here.

View attachment 426421

It is important to remember that the Harman curve is a preference curve - it was derived by applying the HRTF to an anechoic flat speaker response, and then allowing listeners to adjust the curve to their liking. Over the years, the Harman curve has been refined as shown above.

Sean Olive found that there are 3 groups of listeners which can be broadly divided into "Harman curve lovers", "less bass", and "more bass". The first group made up 2/3 of all listeners tested, with the remainder equally divided into "less bass" and "more bass".

It is also known that issues with the test procedure affect the repeatability of the result. Anything below 200-300Hz is subject to variables which may produce air leaks such as clamping pressure, wearing eyeglasses, improper seals, and so on. Anything above 2-3kHz is subject to individual HRTF - shape of ears and pinna, length of auditory canal, etc.

Miller and Downey 2023
I was recently made aware of a study performed by the above two authors published in AES in Oct 2023. The PDF can be downloaded for free from the AES website. The most interesting finding of the study was that there is a clear preference for more high frequency energy > 10kHz. In addition, individuals with hearing loss or older age preferred even more high frequency boost.

View attachment 426422

The authors subjected the study participants to a hearing test and collected data on hearing loss.

View attachment 426423

The Harman curve was then modified with several treble boosts, up to 27dB. The study participants then listened to music and the treble shelves were tested blind using the MUSHRA (Multiple Stimulus with Hidden Reference and Anchor) protocol.

Result
View attachment 426424View attachment 426425

As can be seen, older individuals had a clear preference for treble boost, up to 12-27dB over the Harman curve. What is also interesting is that all test participants preferred more treble over the Harman curve.

Preamble
All of us on ASR who follow headphone science are aware of the Harman headphone curve. Amir draws in the target curve for all his headphone and IEM measurements so that we can see how closely the device follows the curve. An excellent discussion on the Harman curve can be found here.

View attachment 426421

It is important to remember that the Harman curve is a preference curve - it was derived by applying the HRTF to an anechoic flat speaker response, and then allowing listeners to adjust the curve to their liking. Over the years, the Harman curve has been refined as shown above.

Sean Olive found that there are 3 groups of listeners which can be broadly divided into "Harman curve lovers", "less bass", and "more bass". The first group made up 2/3 of all listeners tested, with the remainder equally divided into "less bass" and "more bass".

It is also known that issues with the test procedure affect the repeatability of the result. Anything below 200-300Hz is subject to variables which may produce air leaks such as clamping pressure, wearing eyeglasses, improper seals, and so on. Anything above 2-3kHz is subject to individual HRTF - shape of ears and pinna, length of auditory canal, etc.

Miller and Downey 2023
I was recently made aware of a study performed by the above two authors published in AES in Oct 2023. The PDF can be downloaded for free from the AES website. The most interesting finding of the study was that there is a clear preference for more high frequency energy > 10kHz. In addition, individuals with hearing loss or older age preferred even more high frequency boost.

View attachment 426422

The authors subjected the study participants to a hearing test and collected data on hearing loss.

View attachment 426423

The Harman curve was then modified with several treble boosts, up to 27dB. The study participants then listened to music and the treble shelves were tested blind using the MUSHRA (Multiple Stimulus with Hidden Reference and Anchor) protocol.

Result
View attachment 426424View attachment 426425

As can be seen, older individuals had a clear preference for treble boost, up to 12-27dB over the Harman curve. What is also interesting is that all test participants preferred more treble over the Harman curve.

View attachment 426426
Based on the above findings, the authors proposed a modification to the Harman curve, with a different curve profile based on the age of the listener.

So now we know that some listeners prefer less bass / more bass / neutral, and this new study shows that preference for treble boost > 10kHz varies according to age group. Some ASR members are dogmatic about Harman curve compliance - this may be a reminder to keep a more open mind.

View attachment 426426
Based on the above findings, the authors proposed a modification to the Harman curve, with a different curve profile based on the age of the listener.

So now we know that some listeners prefer less bass / more bass / neutral, and this new study shows that preference for treble boost > 10kHz varies according to age group. Some ASR members are dogmatic about Harman curve compliance - this may be a reminder to keep a more open mind.

I'm using the Harman curve to. With Wavelet
i'am able to change linear the curve such that it creates specific for me full transparency it depends how the music is masterd. It's a clever functionality i did not found before in any dsp software specific for headphones an it is for free including the database. Im 65 last hearing test (first one i ever did ) was well above average for my age.

https://imgur.com/Qa89SX5

 

[DVDdoug]

Some ASR members are dogmatic about Harman curve compliance - this may be a reminder to keep a more open mind.

I think most of us understand that it's an averaged preference and that the experiments were done with people who have "normal" hearing.

It wouldn't be easy to build a boost into passive headphones and EQ is probably a better solution than building different headphones for different age groups. ...Although different headphones are marketed to different market segments.

 

[Keith_W]

BTW if you do not feel like reading the original paper, you might prefer to read the Sound Guys article. Because both authors were working for Knowles, this proposed modification to the Harman target is known as the "Knowles curve".

BTW +27dB of treble is A LOT OF TREBLE. I would say a ridiculous amount of treble. It would be enough to drive your headphones into massive amounts of distortion unless they were specifically tuned to put out that much treble. I am in the 46-55 age bracket - according to the Knowles curve, I should be boosting my treble by 20dB. I just tried it, and it's unlistenable. Unfortunately, I can't tell you why it's unlistenable - maybe my IEM's are distorting. Maybe my IEM's already start off with the "correct" amount of treble for my age group and I am pushing them beyond. I don't have an IEM measurement fixture, so I don't know.

The other application is whether your home audio should have a treble tilt. Ever since I started to DSP my speakers, I have always given it a treble tilt based on my subjective preference. It seems that I like my treble slightly hotter than the Harman curve. Now I know why - I always thought that it was a personal thing.

 

[Galliardist]

To me, this just seems wrong. As my hearing has gone downhill in the last five years, the change in my treble preference is... none. And I don't see the point. If I hear live music, that is now how it sounds to me, so I'm pretty sure it should be the same in audio.

I think this is one experiment that needs repeating by a different team.

 

[Yuhasz01]

The key variable is not your age , but the ability to hear, or not, frequencies. Some of my friends have upper frequency problems at 55, others have hearing fine at 75.
Issue is to get your hearing tested by professional health people, to know where you stand regarding any hearing loss. Then insure you have audio equipment that can be adjusted with PEQ controls according to preferences.
All the published preference curves are just very, very general guidelines.

 

[swamps]

I've only skimmed the article but this massive treble boost past 10kHz seems weird like an instrumentation error. I prefer a bit more than Harman but nothing that extreme.

It seems like they used a non-commercial BA design for the listening tests which shows a massive roll-off past 10kHz, assuming it is accurate of course.
I'd choose (or make) an IEM with a lot of well extended treble (say up to 17kHz) and EQ it down to Harman from that.

 

[Yorkshire Mouth]

I note people’s comments about having hearing loss but not liking lots of extra treble. I think I can help a little.

I’m 59, and in a slightly different position. I’ve had a ‘treble issue’ in my left ear since my 20s. I finally got it checked a few years ago, and my right ear has no treble loss at all. The whole range is down a few dB, but only the amount they expected for my age. My left ear is the same as my right until the upper end, then it drops off - sorry, I don’t have the exact figures handy.

Anyway, where am I going. Well, when I listen on headphones or IEMs I never feel I have a treble loss in my left ear. Why? Well, I think you just get used to it.

With speakers it’s different, because you hear almost as much of the left speaker in your right ear as in your left year, and I’m sure your brain just compensates.

I’m currently about to have my ears re-tested, and I’m playing around with EQ on my WiiM Ultra.

 

[Yorkshire Mouth]

BTW if you do not feel like reading the original paper, you might prefer to read the Sound Guys article. Because both authors were working for Knowles, this proposed modification to the Harman target is known as the "Knowles curve".

BTW +27dB of treble is A LOT OF TREBLE. I would say a ridiculous amount of treble. It would be enough to drive your headphones into massive amounts of distortion unless they were specifically tuned to put out that much treble. I am in the 46-55 age bracket - according to the Knowles curve, I should be boosting my treble by 20dB. I just tried it, and it's unlistenable. Unfortunately, I can't tell you why it's unlistenable - maybe my IEM's are distorting. Maybe my IEM's already start off with the "correct" amount of treble for my age group and I am pushing them beyond. I don't have an IEM measurement fixture, so I don't know.

The other application is whether your home audio should have a treble tilt. Ever since I started to DSP my speakers, I have always given it a treble tilt based on my subjective preference. It seems that I like my treble slightly hotter than the Harman curve. Now I know why - I always thought that it was a personal thing.

I’m not so sure about distortion.

Amir tests speakers 86dB and 96dB, occasionally 76dB, so that’s up to 20dB difference. And whilst we see low end distortion taking off the louder it gets, the high end usually stays steady.

For headphones the HD650s tested at 96dB have very low distortion at the top end at 96dB.

If you think how much more a driver has to move to create all that low frequency pressure, it should be no surprise that, any headphone that can handle that, isn’t going to get too crazy at the top end.

 

[Yorkshire Mouth]

To me, this just seems wrong. As my hearing has gone downhill in the last five years, the change in my treble preference is... none. And I don't see the point. If I hear live music, that is now how it sounds to me, so I'm pretty sure it should be the same in audio.

I think this is one experiment that needs repeating by a different team.

I’m a firm believer in test and test again.

But I’m also wary of a knee-jerk reaction that says “Those results are a surprise, therefore the test must have been flawed”.

But yes, more testing is always good.

 

[ZolaIII]

@Keith_W at which SPL program (mids) where those (both) tests conducted? Is Miller and Downey done the same as original Harman with two self filter jiggles on disposal, I see graph for high self. You can add what ever you like in DSP, driver won't follow it if it can't. Actually nothing of this is real progress, you can partially compensate up to cuple dB in highs and even for this much min phase FIR is advised and still angle tilting works surprisingly good. For the bass self filter (105 Hz Q 0.71) simply does the job to driver capabilities as limitation. Future more hearing loss doesn't have to be same with age groups nor permanently same (Tinnitus for example). If we can get to bottom of healthy near perfect hearing wide huge demographic patern that it's representive in every way to shell, shape, cavity and chenel model classification it would be a huge step forward. From there to audiologist and tailored design to cope with more extreme capabilities for people in need for hearing aid.

 

[Keith_W]

I am not sure what you are asking @ZolaIII. If you want to know how I put in the high shelf in my own informal testing, I simply dialled it into my PEQ in JRiver. Very quick and dirty.

 

[ZolaIII]

I asked at which loudness and assuming it's fixed those tests were conducted if you knew of course. Simply just that and for bass preference. To a little less dirty I whose referring to how to struck highs better with EQ-APO and bake FIR of it which you then put in convolver and use a bass self filter of course if you want.

 

[m8o]

As a person pushing 62 who has listened to speakers and headphones way too loud, and after an incident a few years ago where I actually fell asleep listening to IEMs way too loud now has fairly significant tinnitus in his left ear, no debate or refute of the thesis. I just can't get past one thing however:

Can you shed some light on how the material you are basing this summary on defines Q? What is the formula for Q in how it relates to bandwidth that they are using? Thanx

The reference is repeated in other images. There is no equalizer I own that has much of a bandwidth effect beyond a few thousand Hz above or below the reported center frequency of 16K Hz when the EQ's Q is 4. Let alone the significant lift illustrated below 10K Hz reaching down as far as 2K Hz at the highest boosts used in the study.

 

[Keith_W]

Can you shed some light on how the material you are basing this summary on defines Q? What is the formula for Q in how it relates to bandwidth that they are using? Thanx

The reference is repeated in other images. There is no equalizer I own that has much of a bandwidth effect beyond a few thousand Hz above or below the reported center frequency of 16K Hz when the EQ's Q is 4. Let alone the significant lift illustrated below 10K Hz reaching down as far as 2K Hz at the highest boosts used in the study.

Please read the paper from p.711 onwards for discussion of listener preference for different Q's. I can not give you any information which is not already in the paper because I had no hand in authoring it.

I don't know what equaliser you are using, but it is pretty trivial to create a curve like that with DSP software. Here is one I threw together in a few seconds: