• 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!

HBK Headphone Measurement Talks from Head-Fi and Sean Olive

OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,390
Location
Seattle Area
Adding on, I consider the high frequency aspect of Harman curve as a target. And just like targets for room response, we always use high amount of filtering to get the overall tilt we prefer. We can easily get high resolution measurements in room at high frequencies but we don't want to. This is what it would look like:

f635d4_a85ac53b1d384bb3873a8d6c3581d042~mv2.png


And smoothed version for the purposes of target curve:

f635d4_a74fe0a0680345d2bfffb2e7b37b38d9~mv2.png


And regardless, preference definitely comes into the picture here. It would be next to impossible to create a "high resolution" target and then tell people that is right.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,390
Location
Seattle Area
Agree though that the response of anechoically flat loudspeaker in the bass is closer to the B&K target, Dr. Toole also questioned and tried some possible explanations why the preferred levels at both frequency extremes differed to the anechoic ones and said further research must be done:
I am not seeing how that is close to anything 5128 produces. Such data does not exist.

That aside, I have now equalized some 70 headphones in bass response to Harman target and performed controlled listening tests with or without such boost. It is absolutely correct thing to do. It creates a level of high fidelity that even exceeds that of most speaker systems due to easy with which headphones can produce sub-bass. And freedom they have from "room modes" which real listening spaces suffer from.

To enjoy the benefit of the above, you need to listen to right kind of music that has such extreme low response. I have collected a long list of them now and it is such a delight to play them through a headphone with such response. I think many people who are critical of this have not been in this situation (right music and right bass response).

In other words, we have good bit of practical research under out belt here.
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,747
Likes
16,182
I am not seeing how that is close to anything 5128 produces. Such data does not exist.
There is a misunderstanding there, I din't mean the B&K 5128 but was referring to the 1974 B&K loudspeaker target https://www.bksv.com/media/doc/17-197.pdf which @Mad_Economist had linked in my quote.

To enjoy the benefit of the above, you need to listen to right kind of music that has such extreme low response. I have collected a long list of them now and it is such a delight to play them through a headphone with such response. I think many people who are critical of this have not been in this situation (right music and right bass response).
I and quite few others I know who find the Harman target too bass heavy, actually are bassy music lovers and don't prefer it at modern better recordings, where on older more bass-shy ones it can work really well. In the end it depends of course on the listened music (audio's circle of confusion), playback level and habituation.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,390
Location
Seattle Area
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,390
Location
Seattle Area
I and quite few others I know who find the Harman target too bass heavy, actually are bassy music lovers and don't prefer it at modern better recordings, where on older more bass-shy ones it can work really well. In the end it depends of course on the listened music (audio's circle of confusion), playback level and habituation.
It is not about bass but sub-bass. Get some music with proper sub-bass in it and play it with a headphone flat and boosted. In flat speakers, it is barely perceptible. With boost, then it comes to its own.
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,747
Likes
16,182
Oh, sorry about that. :) I was wondering why you were saying that.
No problem :) , B&K is everywhere :D and I personally I find it impressive that I really like a target today that they researched 47 years ago! :cool:
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,390
Location
Seattle Area
No problem :) , B&K is everywhere :D and I personally I find it impressive that I really like a target today that they researched 47 years ago! :cool:
Indeed. I was once interested in a vibration sensor from B&K and sent email asking about it. Problem is, I was on BK Precision site, not B&K! I got an angry message back from their support people almost yelling at me, "we are NOT B&K!"

But yes, their work has survived and validated so many times since then.
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,747
Likes
16,182
It is not about bass but sub-bass. Get some music with proper sub-bass in it and play it with a headphone flat and boosted. In flat speakers, it is barely perceptible. With boost, then it comes to its own.
It still depends on the recording (and playback level), for example if it was mixed with a flat bass headphone or loudspeaker it will probably sound more neutral also with such a bass target as the same playback level, but could you maybe give a couple of example tracks so people test them with both targets? :)
With loudspeakers it depends also on the room acoustics, if you have a sub-bass absorbing room like I unfortunately have right now, then even a flat bass response at the listening position can sound too sub-bass heavy with quite many recordings... Its all unfortunately not simple, that's why also for example generic targets don't work for loudspeakers.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,390
Location
Seattle Area
It still depends on the recording (and playback level), for example if it was mixed with a flat bass headphone or loudspeaker it will probably sound more neutral also with such a bass target as the same playback level, but could you maybe give a couple of example tracks so people test them with both targets? :)
Sure, here is the start of my playlist:

1634199975243.png


And a few from the next page:

1634200025343.png


Fading Sun especially is revealing. It starts with deep bass that typical bookshelf speaker doesn't even bother to reproduce. With most headphones equalized, it is jaw droopingly good.
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,747
Likes
16,182
Sure, here is the start of my playlist:

View attachment 158961

And a few from the next page:

View attachment 158962

Fading Sun especially is revealing. It starts with deep bass that typical bookshelf speaker doesn't even bother to reproduce. With most headphones equalized, it is jaw droopingly good.
Thank you, will have a listen to it hopefully in the weekend.
 

Mad_Economist

Addicted to Fun and Learning
Audio Company
Joined
Nov 29, 2017
Messages
543
Likes
1,618
I forgot I wrote this quote:

“While most loudspeaker manufacturers today aim to achieve a flat frequency response on-axis, headphone manufacturers seem to be aiming at a target response that is as variable and random as the weather.” ( I was thinking Canadian weather --not southern California weather.) I think CA Poldy said much the same thing many years earlier.
It's my second favourite quote of yours, after
Why? Your ERB is already getting quite large by those frequencies so those ups and downs are not heard. You can easily experiment with this by creating high Q filters and listening for effects. I have done this countless times and they simply are not audible.

That is on top of the problem of small variations of headphone on the fixture can vary these reflections and resulting vector responses.
Does this mean that you'd favour representing headphone measurements in the same band with smoothing equivalent to the ear's ERB?
No problem :) , B&K is everywhere :D and I personally I find it impressive that I really like a target today that they researched 47 years ago! :cool:
Every time I end up on a phone call with somebody from a niche and interesting audio firm, they seem to have some tie back to B&K. It's very central to the history of scientific audio.
 

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,584
Likes
1,086
<<<I see no appreciable differences between your preferred in-room loudspeaker curve or PIRL (and the current Harman over-ear headphone targets based to some extent on this), and the estimated in-room responses of some of the best extended speakers in the sub-bass, which also have a reasonably flat/neutral on-axis response...>>

I agree. And that was the point of showing the anechoic spins of the Revel vs the in-room and headphone target curves that you show in the next post. The bass roll-off of the speaker only means that a subwoofer or EQ needs to be applied.

Thank you for explaining this a bit better, Dr. Olive.

I enjoyed your recent discussion with Resolve and Mad_Economist on the Headphone Show as well.

For those who did not get a chance to participate in or hear HBK's recent measurement conference, btw, it looks like they have now started to post some of the presentations on their YouTube channel here...


I don't see Dr. Olive's discussion on Modeling And Predicting Listeners’ Headphone Preference Ratings, or Jude Mansilla's Audio Measurements As Consumer Content there yet. But these have been coming out a little at a time, for the last few days. So hopefully those other presentations will also be posted soon.

Some other presentations from the above conference that might also be worth a look and listen...

Erik Ziegler's - Keynote Presentation
Martin Alexander's - ElectroAcoustic Principles – The Complete Idiot’s Guide to Basic Electroacoustic Measurement
Mark Serridge's - Why do so many of us like the sound of analogue recordings on vinyl when the measurements look so bad?
Søren Jønsson's - Wideband impedance measurement technique in the human ear canal
Lars Birger Nielsen's - 5128 High Freq HATS and conformance to the new ITU-T P57/P58 standard
 
Last edited:

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,584
Likes
1,086
This is a bit OT, but I just want to quickly apologize to Robbo, Mad and thewas for not responding to a few of your recent posts. I'm in the middle of a rather large and difficult move at the moment. Which is going to make it difficult to stay on top of the things here.

I've definitely enjoyed some of the interesting debates we've had here though. And have appreciated some of the insights that others have had to share on this subject. And hope we can maybe continue a bit more with some of this in the future... And will try to add my 2c where I can, and think it might be of some help, when I'm able.

This is a really great community of people, with some very heavy-duty thinkers on this subject. And I do love and appreciate y'all for that. :)
 

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,584
Likes
1,086
Have not had the chance to watch/listen to these yet or the other HBK videos linked above (except fror Erik Ziegler's Keynote address), but here are the presentations by both Jude Mansilla of Head-Fi and Dr. Olive from HBK's virtual measurement conference. They were both posted by HBK earlier today.

 
Last edited:

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,584
Likes
1,086
Watched/listened to all of the above HBK conference videos. The first roughly 2/3 of Dr. Olive's talk on Modeling And Predicting Listeners’ Headphone Preference Ratings was mostly material that I'd heard before, which basically summarized alot of the research that led to the development of the Harman headphone target. So the more recent study involving the HBK 5128 system doesn't really begin until around 28:30.

The points in the video regarding the custom modified GRAS pinna, and its improved modeling of leakage effects were fairly important imo. And helped me to understand why Harman felt that some comparisons between the two systems would be valuable.

At ~34:25, Dr. Olive commented on some other discrepancies though between the two systems, which seemed to be somewhat specific to the DCA Stealth measurements, pointing out the fairly sizable dip between the main resonances at ~3k and 8k on Jude's 5128 plot of same. This is something Amir also commented on at the beginning of this topic, and circled for us in the graph below...

index.php


The dip on the 5128 Stealth plot looks a bit pronounced. But it is not that unusual to see dips like this in the headphone plots of some other measurement rigs than the GRAS. There are similarly recessed areas between the first two ear resonances in several of Harman's other 5128 measurements shown in the above video. (Though they aren't generally quite as obvious or pronounced as on Jude's Stealth plot.) The graph of the Sennheiser HD650, which is the third graph in the top row at around 36:15 in the above video, is one of the better examples of this. The blue curve on that graph is the 5128 measurement.

You can see the same type of feature on some the SoundGuys 5128 measurements as well, especially on their Audeze Mobius and Senn HD 6XX graphs.

Audeze Mobius
Sennheiser HD 6XX

The same type of dip can also be seen in the pre-Harman target Olive-Welti in-ear response curve, based on the sound of speakers in a room...

WdEe0CH.png


Which I believe came from this 2013 AES white paper...


Several years ago (around 2015?), I also computed the average response for some of the better-sounding headphones which were available at that time, based on raw FR meausrements on a now defunct Korean headphone review/measurement site, called GoldenEars. Which I believe was using a Head Acoustics system similar to the one that Tyll was also using at Inner Fidelity. And you can see a similar dip and series of resonances in the upper frequencies in the average response of those 21 headphone measurements as well.

It's a bit hard to see, but the FR ranges on these plots are the usual 20-20000 Hz. And the amplitude range is 50 dB. And the green curve in the center of the datapoints on this first graph is the average of all the samples.

points-jpg.922218


The peaks and valleys formed by this average response curve also appeared to me to follow a general pattern of concentric arcs.

Peaks:

peaks-jpg.922226


Valleys:

valleys-jpg.922250


Pattern of concentric arcs:

arcs-jpg.922242


And this became one of the features that I increasingly began to look for in the headphone measurements on these types of rigs.

Most of the 21 headphones in the above sampling were audiophile type headphones, btw, rather than consumer headphones. So as a group they were probably a bit on the brighter side.

A somewhat similar pattern of peaks and valleys could also be seen in many of the Rtings raw response graphs. Especially on plots of some of the better headphones. So I tried computing some averages from that data as well. I was still trying to follow the Harman target model at this time (more or less), so the resulting average curve had a similar rise in the upper midrange at around 1.5 to 2 kHz as that target response curve. (Which is something I've since decided that I'm probably not such a huge fan of.)

Here is one of those averaged Rtings curves compared to the 2013 Olive-Welti curve (top), and 2018 Harman over-ear target for the GRAS 45CA with custom pinna (bottom)...

3654746


Although this is only vaguely hinted at by the three rather tantalizing peaks at around 3, 8 and 16 kHz in the 5128's diffuse field compensation curve (purple curve below), the same type of dip and resonant structures in the treble can also be seen fairly plainly (imho) in Amir's raw 5128 graph of the Sennheiser HD650...

index.php


Based on the DCA Stealth and some of the other 5128 plots by the SoundGuys and Amir above, it appears as though at least some of the better headphone measurements on the 5128 system could end up following a similar pattern of dips and resonances in the higher frequencies, maybe putting it a bit closer to the response of the Head Acoustics system in that area on average, than the GRAS 45CA.

I think this difference between the GRAS and 5128 resonances was also briefly touched on in the recent Headphone Show livestream with Dr. Olive as well. (Though I can't remember if it was Mad_Economist, Resolve, or Dr. Olive who initially brought this up.)

The resonances on the Head Acoustics and similar systems (which appears to maybe also include the HBK 5128) seem to be fairly consistent with the resonant frequencies of the ear canal show here...

10239357.jpg


And the somewhat sizable dip between the first two resonances also seems to coincide with what alot of audiophiles refer to as the "sibilant range", which generally falls in or around the region between the first two resonances in the ear canal shown above. Substantial peaks in this area between the two ear canal resonances are generally disliked for their sibilant characteristics.

This same sibilant frequency range also appears to be described by the ITU-R 468 noise weighting curve. Which is the black curve on the graph below that peaks in the low treble at around 6 kHz. Roughly in between the resonant peaks of the ear canal, and many of the measurements made on the Head Acoustics system (and it appears now also at least some of the measurements on the 5128 system)...

533px-Lindos3.svg.png
 
Last edited:

Robbo99999

Master Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
6,874
Likes
6,672
Location
UK
Watched/listened to all of the above HBK conference videos. The first roughly 2/3 of Dr. Olive's talk on Modeling And Predicting Listeners’ Headphone Preference Ratings was mostly material that I'd heard before, which basically summarized alot of the research that led to the development of the Harman headphone target. So the more recent study involving the HBK 5128 system doesn't really begin until around 28:30.

The points in the video regarding the custom modified GRAS pinna, and its improved modeling of leakage effects were fairly important imo. And helped me to understand why Harman felt that some comparisons between the two systems would be valuable.

At ~34:25, Dr. Olive commented on some other discrepancies though between the two systems, which seemed to be somewhat specific to the DCA Stealth measurements, pointing out the fairly sizable dip between the main resonances at ~3k and 8k on Jude's 5128 plot of same. This is something Amir also commented on at the beginning of this topic, and circled for us in the graph below...

index.php


The dip on the 5128 Stealth plot looks a bit pronounced. But it is not that unusual to see dips like this in the headphone plots of some other measurement rigs than the GRAS. There are similarly recessed areas between the first two ear resonances in several of Harman's other 5128 measurements shown in the above video. (Though they aren't generally quite as obvious or pronounced as on Jude's Stealth plot.) The graph of the Sennheiser HD650, which is the third graph in the top row at around 36:15 in the above video, is one of the better examples of this. The blue curve on that graph is the 5128 measurement.

You can see the same type of feature on some the SoundGuys 5128 measurements as well, especially on their Audeze Mobius and Senn HD 6XX graphs.

Audeze Mobius
Sennheiser HD 6XX

The same type of dip can also be seen in the pre-Harman target Olive-Welti in-ear response curve, based on the sound of speakers in a room...

WdEe0CH.png


Which I believe came from this 2013 AES white paper...


Several years ago (around 2015?), I also computed the average response for some of the better-sounding headphones which were available at that time, based on raw FR meausrements on a now defunct Korean headphone review/measurement site, called GoldenEars. Which I believe was using a Head Acoustics system similar to the one that Tyll was also using at Inner Fidelity. And you can see a similar dip and series of resonances in the upper frequencies in the average response of those 21 headphone measurements as well.

It's a bit hard to see, but the FR ranges on these plots are the usual 20-20000 Hz. And the amplitude range is 50 dB...

points-jpg.922218


The peaks and valleys formed by this average response curve also appeared to me to follow a general pattern of concentric arcs.

Peaks:

peaks-jpg.922226


Valleys:

valleys-jpg.922250


Pattern of concentric arcs:

arcs-jpg.922242


And this became one of the features that I increasingly began to look for in the headphone measurements on these types of rigs.

Most of the 21 headphones in the above sampling were audiophile type headphones, btw, rather than consumer headphones. So as a group they were probably a bit on the brighter side.

A somewhat similar patterns of peaks and valleys could also be seen in many of the Rtings raw response graphs. Especially on plots of some of the better headphones. So I tried computing some averages from that data as well. I was still trying to follow the Harman target model at this time (more or less), so the resulting average curve had a similar rise in the upper midrange at around 1.5 to 2 kHz as that target response curve. (Which is something I've since decided that I'm probably not such a huge fan of.)

Here is one of those averaged Rtings curves compared to the 2013 Olive-Welti curve (top) and 2018 Harman over-ear target (bottom)...

3654746


Although this is only vaguely hinted at by the three rather tantalizing peaks at around 3, 8 and 16 kHz in the 5128's diffuse field compensation curve, the same type of dip and resonant structures in the treble can also be seen fairly plainly (imho) in Amir's raw 5128 graph of the Sennheiser HD650...

index.php


Based on the DCA Stealth and some of the other 5128 plots by the SoundGuys and Amir above, it appears as though at least some of the better headphone measurements on the 5128 system could end up following a similar pattern of dips and resonances in the higher frequencies, maybe putting it a bit closer to the response of the Head Acoustics system in that area than the GRAS 45CA.

I think this difference between GRAS and 5128 resonances was also briefly touched on in the recent Headphone Show livestream with Dr. Olive as well. (Though I can't remember if it was Mad_Economist, Resolve, Dr. Olive who initially brought this up.)

The resonances on the Head Acoustics and similar systems (which appears to maybe also include the HBK 5128) seems to be fairly consistent with the resonant frequencies of the ear canal show below...

10239357.jpg


And the somewhat sizable dip between the first two resonances also seems to coincide with what alot of audiophiles refer to as the "sibilant range", which generally falls in or around the region between the first two resonances in the ear canal shown above. Substantial peaks in this area between the two ear canal resonances are generally disliked for their sibilant characteristics.

This same sibilant frequency range also appears to be described by the ITU-R 486 noise weighting curve. Which is the black curve on the graph below that peaks in the low treble at around 6 kHz. Roughly in between the resonant peaks of the ear canal and many of the measurements made on the Head Acoustics system (and it appears now also at least some of the measurements on the 5128 system)...

533px-Lindos3.svg.png
We know measurements are different on different rigs, so what's the main point you're making? (just to be clear, because I wasn't sure).
 

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,584
Likes
1,086
We know measurements are different on different rigs, so what's the main point you're making? (just to be clear, because I wasn't sure).

Howdy, Robbo. Since both Amir and Dr. Olive seemed to be slightly confused by the dip between 3 and 8k on the 5128 DCA Stealth plot, I thought I'd try to show and explain why that's not such an unusual feature to see on some other rigs than the GRAS systems that they're perhaps a bit more used to.

It also took me some time to figure out why the treble looked and behaved the way it did on the GRAS measurements, with the rather large notches at 9 and 15k, for example. Since I was more used to seeing measurements like the ones above from GoldenEars, Rtings and Inner Fidelity. So I suppose I sort of understand some of their confusion on this, comin at it from the opposite direction.
 

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,584
Likes
1,086
Since there is a plethora of in-room speaker info available now, I was also going to suggest that maybe you try some different curves than Harman's PIRL, with the in-ear measurement of their flattened speakers. If you like the sound of that "flat" in-ear curve combined with their PIRL curve though, then trying other stuff may be just a waste of time. There's a wealth of estimated in-room spinorama measurements here though, fwiw...


Usually they're included on separate graphs from the main spinorama data. Maybe because they can be so similar to the shape and slope of the early reflections. (?)

For the 5128 though, my preference is still to use the 5128 DF curve with an average sound power response curve. Which works very similarly to the kind of thing you're doin above. What you are doing with the flat in-ear response curve and PIRL (or slopes based on in-room response curves) is for all intents and purposes the same as what I'm doing with the DF and sound power curves. The main difference is that the DF curves don't seem maybe quite as reliable and useful for this on the GRAS as on the 5128.

The 5128 DF curve is potentially also missing some important FR details though in the treble, in between the resonant peals at about 3, 8 and 16k. So it is not an entirely reliable system either imo. Aside from that fairly minor shortcoming though, the 5128 DF curve seems pretty accurate. And looks like it could still be pretty useful for generally approximating the positions and amplitudes of the peaks in the treble on that system,... when combined with an appropriate sound power correction curve. And it may also be useful for identifying headphones that have a more neutral response on that system... Which could eventually be used to calculate an even more precise average 5128 target response curve... maybe.

Having some actual in-ear measurements of well-extended, neutral loudspeakers in a room would still be my preference though for the 5128 rig. You use whatya got though. :)
 
Last edited:

ADU

Major Contributor
Joined
Jul 29, 2021
Messages
1,584
Likes
1,086
DIY-ers that do their own in-ear measurements inside their ears, or using a mannikin could also potential use the flattened in-room speaker approach that Harman used to help define its target headphone curve. If you have some decent, fairly well-extended speakers (or maybe a sub) that you could EQ to a totally flat in-room response in your home, then that probably wouldn't be so hard to do.

You would probably just measure the response to that flattened out sound in your own ears. And then combine or correct that in-ear response curve using an appropriate slope or in-room response curve of your own choice, either from the spinorama in-room data in the above link, or from the Harman research, or perhaps your own measurements, to better approximate the more tilted and warmer sound of speakers in a room. And then you'd probably have somethin that's reasonable close to a Harman in-ear headphone target for your own ears. (If it's good enough for Harman, then it's probably good enough for you too!)

This is probably not quite as easy to do as it sounds though. And I'm pretty sure I'm not the first person to think of doing something like this, other than Harman. So there may already be some DIYers out there who've tried an approach like this, and might have some more and better insights on it than I would.
 
Last edited:
Top Bottom