"Bias" of some members towards headphone measurements?

[solderdude]

Are loudspeakers measured with a HATS apparatus? Like in Klippel NFS...... (Not that I am aware of) . Is this wrong??? So for a objective evaluation a loudspeaker we don't want to use a HATS. But for headphones we do? Can you list the scientific reasons why an pinna acoustic filter is required for a headphone objective metric? I am oversimplifying but this still applies generally.

They are not measured using a HATS. HATS can be used for acoustic measurements.

It also is not desirable to measure speaker(s) using a HATS because of the errors that are introduced.

For headphones the circumstances are very, very different so in these cases HATS are more suited.
Measurements will be correct to the standard the fixture complies to.
Not all HATS are equal and they can be configured with different combinations of couplers (mic + ear canal) and pinnae.

A fake ear canal and pinna change the incoming sound waves. To obtain equal amplitude in the measurement it would have to be 'undone'.
Therein lies the problem as the needed correction is not always the same and depends on factors.

 

[Jaxx1138]

They are not measured using a HATS. HATS can be used for acoustic measurements.

It also is not desirable to measure speaker(s) using a HATS because of the errors that are introduced.

For headphones the circumstances are very, very different so in these cases HATS are more suited.
Measurements will be correct to the standard the fixture complies to.
Not all HATS are equal and they can be configured with different combinations of couplers (mic + ear canal) and pinnae.

A fake ear canal and pinna change the incoming sound waves. To obtain equal amplitude in the measurement it would have to be 'undone'.
Therein lies the problem as the needed correction is not always the same and depends on factors.

So for headphones there are no standards. But for loudspeakers it creates errors. This is not scientific in any way. We are doing it wrong. But the religious adherence to bad standards and sycophantic idol worship is taking us far away from achieving good sound. I say we return to scientific standards and utilize Newtonian physics to return to a reliable and repeatable standard. I worked with Sean Olive and Floyd Toole and I believe that first principle acoustics can get us back on track regarding the metrics of IEM's and headphones.

 

[solderdude]

For headphones one uses standards that are intended for acoustic measurements and modifies them a bit.
Harman did create a standard that was founded the way you proposed and a target was added (based on preference for a majority of listeners).
I am not really happy about it but sadly that comes closest to a documented standard (aside from the standards for FF and DF measurements.

It is as scientific as it gets and all there is now as far as standards and measurement methods that exist.

headphone measurements on HATS (that comply to a standard) are reliable and above all repeatable to the standard they are measured to.
That said, mounting on the fixtures, seal issues, tolerances in the standard, tolerances in DUTs and above all applied corrections and targets make headphone measurements the wild west.

I would say ... build a better fixture. Create a standard for it. Solve the problems mentioned above. Provide scientific evidence that the method is preferred. Get it certified. Produce (or have it produced) to the standard and then measure the crap out of everything.

Get someone to fund all that and make a profit.

 

[ZolaIII]

So for headphones there are no standards. But for loudspeakers it creates errors. This is not scientific in any way. We are doing it wrong. But the religious adherence to bad standards and sycophantic idol worship is taking us far away from achieving good sound. I say we return to scientific standards and utilize Newtonian physics to return to a reliable and repeatable standard. I worked with Sean Olive and Floyd Toole and I believe that first principle acoustics can get us back on track regarding the metrics of IEM's and headphones.

You want science? Dig in how rigorous are medical putting in use a cure standard's and how good need to be results to under 1% and well explained. Because it's directly tied to human health. Until treated the same and enforced you won't have such in audio. This will happen but too late when primary IEM's make soch hearing damage to majority of population in their highly productive age. Then money for research won't be a problem because social care cost will be much higher.
When Descartes put human back in the center of everything he didn't think you should become one dimensional numeric thing but a true mesure of all things. Social scientific research's are very real and important and you can explain subjective related appearance's psy and physically. Some things from audiology including data could be used with lots of consulting, I believe they also want to get able to do better measurements and improve their own methodology.

 

[RandomEar]

So for headphones there are no standards. But for loudspeakers it creates errors. This is not scientific in any way. We are doing it wrong. But the religious adherence to bad standards and sycophantic idol worship is taking us far away from achieving good sound. I say we return to scientific standards and utilize Newtonian physics to return to a reliable and repeatable standard. I worked with Sean Olive and Floyd Toole and I believe that first principle acoustics can get us back on track regarding the metrics of IEM's and headphones.

I think we need to take a step back and look at this another way: What does the transfer function driver (speaker driver or headphone/IEM driver) -> eardrum look like?

For speakers, it can be separated into two parts: Speaker -> ear & ear -> eardrum. You can argue where exactly "ear" starts and ends, but for the sake of this argument, we can just assume an arbitrary point as long as it is the same for every subject. If different people listen to the same speaker in the same room at the same head position, the first transfer function (TF) speaker -> ear will be the same for everybody. The second TF (ear -> eardrum) will be different, but it is consistent for these people in the sense that they always have this TF when listening to anything in their life - music, wind noise, cars outside.

With headphones, you can't separate the TF in this way, because the driver position (distance to eardrum and angle to pinna opening) is different for each headphone and each person. In addition, you will also have interactions like direct reflections driver -> pinna -> ear cushion -> ear canal which simply do not exist or are below the threshold of hearing for speakers in a room. Other interactions can include such things like squished or malformed tips on IEMs due to mechanical constraints of the ear. This is all very complex, which is why headphones and IEMs can sound very different to different people even if they measure close: The TF of each headphone/IEM is different for each person using it. If you measure just the headphone or just the IEM "clean" in a free space, all these interactions, which have a major influence on the sound, will vanish and be ignored. A HATS allows you to capture these interactions, which are effectively baked into the TF and are different for each person.

When measuring a speaker with the Klippel NFS, the NFS software will calculate a baseline "clean" result without the TF of the room the measurement was taken in. This removes the first part of the TF (speaker -> ear, or in this case micophone) mentioned above. Results can then be presented only for the speaker without any TF, but there is also the option to generate the estimated in-room response: This applies the TF of a "standard room" to the clean results of the speaker to make it comparable. That is what a HATS is: It applies the TF of a "standard head" to the clean response of the headphone or IEM.

All our rooms are different, but an estimated in room response is still valuable, because it makes speaker measurements comparable. All our heads and ears are different, but a heapdone or IEM measurement on a HATS is still valuable, because it makes headphone measurements comparable.

I think you could argue that also measuring and publishing "clean" headphone/IEM measurements without a HATS could be valuable. But I don't think you can argue that publishing measurements using a HATS isn't valuable.

 

[Talisman]

Do you have any measurements to support this statement?

It seems absurd to read this question on this site.
Do you want proof that not everything that is measured is detectable acoustically?
Listen to two decent DACs but with different measurements, like the last one in the green row and the first one in the blue row.
They have very different measurements, it is practically impossible for you to hear a difference.

 

[InfiniteJester]

Measuring according to a standard is one thing.
Measurements can be done correctly and incorrectly and have a certain accuracy.
Addition of target curves is another thing.
Proper assesment of measurements is another thing. This might require expert knowledge.
Then there is the discrepancy between measurements and the person that is listening.

Harman is fine. It is based on valid research and the majority (so not everyone) seems to prefer it.
This requires measurements according to standards. Those measurements can be done correctly and incorrectly.

We can have yet another standard but don't see how that would help or give progress.

I would bet that the problem with the Harman target comes from too little listening time before judging.

Harman sounds very good in short bursts, but tiresome in the medium and long term.

It probably sounds best for a single song, but not for listening to three albums in a row, and especially not for hundreds of hours of mixing.

Has this effect ever been rigorously studied? I.e. how your preferences change with the duration of a listening session.

I think it is likely that one would prefer less "exciting" tunings over time.

 

[Miguelón]

I would bet that the problem with the Harman target comes from too little listening time before judging.

Harman sounds very good in short bursts, but tiresome in the medium and long term.

It probably sounds best for a single song, but not for listening to three albums in a row, and especially not for hundreds of hours of mixing.

Has this effect ever been rigorously studied? I.e. how your preferences change with the duration of a listening session.

I think it is likely that one would prefer less "exciting" tunings over time.

I usually get fatigued when long headphones listening, whatever the target curve.

Slightly less fatigued when using open headphones, can this have something to do with pressure on ear?

 

[InfiniteJester]

I usually get fatigued when long headphones listening, whatever the target curve.

Slightly less fatigued when using open headphones, can this have something to do with pressure on ear?

I don't know. Have you tried to substantially lower the ear-gain region?

 

[HalSF]

Harman sounds very good in short bursts, but tiresome in the medium and long term.

Do you have a source or citation for this, or is it a personal opinion?

Sincere question — I’m interested in knowing more about this, assuming it’s not just one person’s experience.

 

[Jaxx1138]

For headphones one uses standards that are intended for acoustic measurements and modifies them a bit.
Harman did create a standard that was founded the way you proposed and a target was added (based on preference for a majority of listeners).
I am not really happy about it but sadly that comes closest to a documented standard (aside from the standards for FF and DF measurements.

It is as scientific as it gets and all there is now as far as standards and measurement methods that exist.

headphone measurements on HATS (that comply to a standard) are reliable and above all repeatable to the standard they are measured to.
That said, mounting on the fixtures, seal issues, tolerances in the standard, tolerances in DUTs and above all applied corrections and targets make headphone measurements the wild west.

I would say ... build a better fixture. Create a standard for it. Solve the problems mentioned above. Provide scientific evidence that the method is preferred. Get it certified. Produce (or have it produced) to the standard and then measure the crap out of everything.

Get someone to fund all that and make a profit.

That would be great, but there are more politics on the AES committee standards, along with all the others (IEEE, ISO, EIA) .... as well as requiring funding, credentials(the kind that brings fame and noteriaty) and political clout it's an uphill battle. I know, I am some middle aged guy complaining, but I just want others to think about it. I argued many points about audio reproduction while at Harman, I won some and I lost some. The headphone research had just gotten started when I left. I am certain we can back engineer this from our current standard to what would be considered reference. The pinna gain is pretty straight forward, there are some tests that need to resolve the Schroeder frequency/room gain and the fact that in a room we are listening to something between free field and diffuse field, all of which contributes to a final transform that convolves to what would be truly neutral.

 

[solderdude]

to what would be truly neutral

The question remains... truly neutral to what and whom ?
What if 'truly neutral' is not preferred by the many ?
Would that require redefining 'neutral' or would the majority of people just have to make do just because 'certain science' and some 'test fixture' tells us we should.

The answers to many of these questions are already clear from the substantial body of work that is out there.
Why reinvent the wheel for the umpteenth time ?

 

[Jaxx1138]

I think we need to take a step back and look at this another way: What does the transfer function driver (speaker driver or headphone/IEM driver) -> eardrum look like?

For speakers, it can be separated into two parts: Speaker -> ear & ear -> eardrum. You can argue where exactly "ear" starts and ends, but for the sake of this argument, we can just assume an arbitrary point as long as it is the same for every subject. If different people listen to the same speaker in the same room at the same head position, the first transfer function (TF) speaker -> ear will be the same for everybody. The second TF (ear -> eardrum) will be different, but it is consistent for these people in the sense that they always have this TF when listening to anything in their life - music, wind noise, cars outside.

With headphones, you can't separate the TF in this way, because the driver position (distance to eardrum and angle to pinna opening) is different for each headphone and each person. In addition, you will also have interactions like direct reflections driver -> pinna -> ear cushion -> ear canal which simply do not exist or are below the threshold of hearing for speakers in a room. Other interactions can include such things like squished or malformed tips on IEMs due to mechanical constraints of the ear. This is all very complex, which is why headphones and IEMs can sound very different to different people even if they measure close: The TF of each headphone/IEM is different for each person using it. If you measure just the headphone or just the IEM "clean" in a free space, all these interactions, which have a major influence on the sound, will vanish and be ignored. A HATS allows you to capture these interactions, which are effectively baked into the TF and are different for each person.

When measuring a speaker with the Klippel NFS, the NFS software will calculate a baseline "clean" result without the TF of the room the measurement was taken in. This removes the first part of the TF (speaker -> ear, or in this case micophone) mentioned above. Results can then be presented only for the speaker without any TF, but there is also the option to generate the estimated in-room response: This applies the TF of a "standard room" to the clean results of the speaker to make it comparable. That is what a HATS is: It applies the TF of a "standard head" to the clean response of the headphone or IEM.

All our rooms are different, but an estimated in room response is still valuable, because it makes speaker measurements comparable. All our heads and ears are different, but a heapdone or IEM measurement on a HATS is still valuable, because it makes headphone measurements comparable.

I think you could argue that also measuring and publishing "clean" headphone/IEM measurements without a HATS could be valuable. But I don't think you can argue that publishing measurements using a HATS isn't valuable.

It's not that HATS that isn't valuable. It's the misapplication of why it is used. The short of it is that we are using one method to evaluate loudspeakers and then using another method for evaluating headphones. Then we are attempting to unravel the convolved transforms so we can interpret it. We do not take into consideration human HRTF's when measuring the acoustic parameters of a loudspeaker (because it is not relevant objectively, it is a 'reference'). Now when measuring a headphone we somehow believe that these human acoustic filters are now necessary to evaluate an objective metric of the audio reproduction of a headphone? I believe this is an error. Now let's not throw the baby out with the bathwater. The data is still useful for research purposes. GRAS, B&K and many others make wonderful products useful for many kinds of work and research. But we need to reevaluate the metrics for headphones so that we can arrive at something ,( not unlike free field response curves for loudspeakers) that we can look at and have it be immediately meaningful as to whether the DUT(headphone) is good or bad.

 

[RandomEar]

It's not that HATS that isn't valuable. It's the misapplication of why it is used. The short of it is that we are using one method to evaluate loudspeakers and then using another method for evaluating headphones. Then we are attempting to unravel the convolved transforms so we can interpret it. We do not take into consideration human HRTF's when measuring the acoustic parameters of a loudspeaker (because it is not relevant objectively, it is a 'reference'). Now when measuring a headphone we somehow believe that these human acoustic filters are now necessary to evaluate an objective metric of the audio reproduction of a headphone? I believe this is an error. Now let's not throw the baby out with the bathwater. The data is still useful for research purposes. GRAS, B&K and many others make wonderful products useful for many kinds of work and research. But we need to reevaluate the metrics for headphones so that we can arrive at something ,( not unlike free field response curves for loudspeakers) that we can look at and have it be immediately meaningful as to whether the DUT(headphone) is good or bad.

I explained my understanding of the reasoning behind it above. Apart from the "standard room" and "standard head" logic, I'd like to point to one aspect again: Our HRTF is the same, whether we listen to speaker A, speaker B, or grandma talking - things that are "far away" from our ears. It is not the same for headphones: Each headphone has a pretty distinct HRTF on each head, because headphones and IEMs are extremely close to our ears and interact directly with them. In consequence, arguing that HRTF's are ignored when measuring speakers isn't an argument to also ignore them when measuring headphones. The circumstances are not comparable. Distance matters.

 

[Jaxx1138]

I explained my understanding of the reasoning behind it above. Apart from the "standard room" and "standard head" logic, I'd like to point to one aspect again: Our HRTF is the same, whether we listen to speaker A, speaker B, or grandma talking - things that are "far away" from our ears. It is not the same for headphones: Each headphone has a pretty distinct HRTF on each head, because headphones and IEMs are extremely close to our ears and interact directly with them. In consequence, arguing that HRTF's are ignored when measuring speakers isn't an argument to also ignore them when measuring headphones. The circumstances are not comparable. Distance matters.

Headphones do not have HRTF's. HRTF's are a biological human function. Distance does not matter for the OBJECTIVE part. I am not certain you understand objective vs subjective. HRTF's are subjective and metrics of audio reproduction devices are objective. Can you explain what the change in frequency response should be if the transducer is closer too or farther away from a persons ear? Perhaps an AES paper on the subject? The circumstances are not that different. Further , I have seen no evidence that including a simulated pinna and ear canal has assisted us in any way in developing a more neutral headphone. Perhaps you could provide some information that I may be missing? My hypothesis is that we need a more objective metric for headphones. And including a pinna and ear canal muddles that goal. As far as the distance goes, I do not see this as a hindrance but I am of course assuming we will be measuring an distances similar to what we do now in order to maintain the proper acoustic impedance load for sealed headphones.

 

[kemmler3D]

The pinna gain is pretty straight forward,

My impression is that this is actually really tricky, though? Or maybe not the pinna per se, but the entire HRTF, canal resonances, and all of it?

I remember a Shure engineer telling me they tried to characterize pinnae / ear canals and gave up after testing some dozens of them. Apparently coming up with a systematic characterization of individual ears for voicing headphones was / is hopeless.

I could believe that headphone neutrality could probably be standardized in a substantial way - except for the listener, who is a random variable that has a direct and material effect on the final sound.

So as far as that goes I think personal ear/head measurements will need to be a thing to reliably achieve neutral headphone sound.

 

[Miguelón]

I don't know. Have you tried to substantially lower the ear-gain region?

As soon as I learn what is the ear-gain region, I will try

 

[Jaxx1138]

My impression is that this is actually really tricky, though? Or maybe not the pinna per se, but the entire HRTF, canal resonances, and all of it?

I remember a Shure engineer telling me they tried to characterize pinnae / ear canals and gave up after testing some dozens of them. Apparently coming up with a systematic characterization of individual ears for voicing headphones was / is hopeless.

I could believe that headphone neutrality could probably be standardized in a substantial way - except for the listener, who is a random variable that has a direct and material effect on the final sound.

So as far as that goes I think personal ear/head measurements will need to be a thing to reliably achieve neutral headphone sound.

No, you have it backwards. How WE hear has nothing to doing with with the objective performance of the headphone. That is the point. We do not tune speakers by how individuals hear. We should not be doing it with headphones.

 

[RandomEar]

Headphones do not have HRTF's. HRTF's are a biological human function.

Splitting hairs. I should have used "transfer function" in the generalized way to describe the path from the transducer to the ear drum.

Distance does not matter for the OBJECTIVE part. I am not certain you understand objective vs subjective. HRTF's are subjective and metrics of audio reproduction devices are objective.

HRTFs are a result of human anatomy and therefore individual to a person, but measurable and therefore objective. Am I missing something, here?

Can you explain what the change in frequency response should be if the transducer is closer too or farther away from a persons ear? Perhaps an AES paper on the subject? The circumstances are not that different.

As explained, the interaction between the headphone as a whole (including pads and housing) and the pinna and ear canal change the sound that arrives at your ear drum. I never said distance has any influence on the transducer. A quick google search turns up papers on pinna interactions like this one, to give just one random example. And a distance of maybe 10 mm is pretty different from 1000-3000 mm. I mean, that's three orders of magnitude - if that isn't significant, what is?

Further , I have seen no evidence that including a simulated pinna and ear canal has assisted us in any way in developing a more neutral headphone. Perhaps you could provide some information that I may be missing?

I'm not claiming that it did, therefore I don't see why I would provide evidence for this. I simply explained why I recognize that heaving a realistic "standard head" seems like a good comparison tool to me.

My hypothesis is that we need a more objective metric for headphones. And including a pinna and ear canal muddles that goal. As far as the distance goes, I do not see this as a hindrance but I am of course assuming we will be measuring an distances similar to what we do now in order to maintain the proper acoustic impedance load for sealed headphones.

Our current measurements are pretty objective, but you can certainly argue that a specific HATS or pinna model isn't the right one or that there's not enough evidence to back up one geometry or another. In the end, the choice of one specific HATS geometry is subjective. Fair enough. I do think, however, that designing a headphone without ever measuring it on real or simulated heads is a bad idea.

 

[kemmler3D]

We do not tune speakers by how individuals hear. We should not be doing it with headphones.

Right, but speakers don't bypass any of the head/body/pinnae and headphones / IEMs do, so we can say that the Harman curve (or a similar tuning) achieves 90-95%, but the remainder needs to be personalized to work.

To put it another way, the sound that reaches your eardrum and the sound that reaches my eardrum from a given pair of IEMs is probably nearly the same. But, taking a neutral speaker as the reference for neutrality... the sound should actually be different for each of us. Because our HRTFs are different and are bypassed by the IEM, that individual difference needs to be accounted for to achieve "the same" sound. Right?