• 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!

"Bias" of some members towards headphone measurements?

This seems like a straw man argument, I don't know where you've got this view that we've thrown our hands up in despair and said "its all too subjective"? People have quantified the acoustic performance of headphones, you are just unhappy with the devices used and perhaps some of the methodology if I understand you correctly, in part because as outlined in my previous posts I think you oversimplify how different headphones and speakers are to necessitate this difference in methodology.
Except that current methods are not repeatable and/or reliable. I propose that measuring through a pinna and ear canal simulator is an error in the protocol. It will always be difficult to define target function while overlaying some averaged pinna and ear canal filters over the top of the metric. It does not appear to be a necessary part of the measurement. I have not found a method currently that would measure the acoustic responses without the pinna and ear canal HRTF's but yet retains the acoustic loading (acoustic impedance) of the pinna wave/guide effect.
 
So when you say "Reading a headphone metric should be as easy as reading a Klippel NFS set of data for a loudspeaker." this seems to me unscientific because it is an assumption nobody else is making, and which seems completely counterintutive. Due to how different headphones and loudspeakers are, I'd say having made that statement the burden of proof is on you to explain why the differences are not as consequential as previously thought.
What would be the reasoning behind why you think these two are so different? You are simply taking an impulse response or noise and using an FFT analyzing the acoustic signal and then comparing it to a reference. My proof is that no one seems to be able to look at a headphone metric and then say why it is better or worse. Whereas you can do that with loudspeaker metrics. Why do you think this is? Quantifying something is supposed to help in assessing what is doing and why it is doing it. If it does not accomplish that then it is of little use.
 
My proof is that no one seems to be able to look at a headphone metric and then say why it is better or worse.

They can't?

1737171176498.png


versus ...


1737171219854.png



Quantifying something is supposed to help in assessing what is doing and why it is doing it. If it does not accomplish that then it is of little use.

Quantification is attaching numerical identities to observations. In that sense, yes, it helps in assessing what something is doing ... or at least, how much it is doing. It is not intrinsic to answering the question. "Why?". That is for observers to deduce. For example, in the charts above, the resonances are more numerous in the HiFiMan phones, and they are also of greater amplitude. That, along with the deviations from target response, mark it as inferior to the Dan Clark set.

But quantification does not tell us WHY those resonances are so extreme, nor why there are more of them in one set rather than in the other.
 

Attachments

  • 1737171205884.png
    1737171205884.png
    48.7 KB · Views: 24
Last edited:
They can't?

View attachment 421964

versus ...


View attachment 421966




Quantification is attaching numerical identities to observations. In that sense, yes, it helps in assessing what something is doing ... or at least, how much it is doing. It is not intrinsic to answering the question. "Why?". That is for observers to deduce. For example, in the charts above, the resonances are more numerous in the HiFiMan phones, and they are also of greater amplitude. That, along with the deviations ffom target response, mark it as inferior to the Dan Clark set.

But quantification does not tell us WHY those resonances are so extreme, nor why there are more of them in one set rather than in the other.
Yes, one is a hamburger, the other is a sandwich!;) I like burgers. :cool: How can people be so confused?!? Or are they just being obtuse?
 
Yes, one is a hamburger, the other is a sandwich!;) I like burgers. :cool: How can people be so confused?!? Or are they just being obtuse?

Obtuse? Here at ASR?

Heaven forfend!!!!! :p :p :p

p.s. - I prefer braunschweiger with onions and hot mustard. Mm-m-m-m-m-m!
 
OK, OK, I am going to suspend my defense for now. But I am going to explore David Griesinger recent work on proximity and equal loudness simulations in order to emulate timbre balance in headphones and attempt to look at measuring the acoustic impedance of headphones and what impact they have on their overall frequency response. Remember all I want is for us to get reliable and repeatable headphone metrics that are objective only. This is a great discussion by the way.

 
OK, OK, I am going to suspend my defense for now. But I am going to explore David Griesinger recent work on proximity and equal loudness simulations in order to emulate timbre balance in headphones and attempt to look at measuring the acoustic impedance of headphones and what impact they have on their overall frequency response. Remember all I want is for us to get reliable and repeatable headphone metrics that are objective only. This is a great discussion by the way.

I just want to comment on the response here. This is a thoughtful, respectful, and open-minded reply to having your belief challenged. Typically, at this point, I see people either doubling down or resorting to ad hominem attacks.

That's all -please continue this interesting discussion. :)
 
Last edited:
That is at least a confounding factor.
Also room interaction is not taken into account so the end result will likely not mimic speaker in a room but that of very near field speakers.
It could still be an improvement over no EQ.
 
Last edited:
That is at least a confounding factor.
Also room interaction is not taken into account so the end result will likely not mimic speaker in a room but that of very near field speakers.
It could still be an improvement over no EQ.
He is attempting to explore one area of why headphones sound so different from listening to a loudspeaker in a room. So the experiment only covers part of what is happening. Timbre is a big part why something sounds the way it does. So he is isolating that aspect by attempting to mimic the octave to octave balance by having people equalize in third octave bands the sound to duplicate in the headphone what is happening when hearing the loudspeaker in a room. According to his work it only works when you do the DRP metric in each persons ears. Which makes sense. He is essentially doing one part of what the Smyth realizer does. It's more of a study of how we hear rather than a study of what a headphone should be doing. ( I wouldn't presume to guess what he was attempting to find out) But it's all useful information. Because of Sean Olives research we know (to some degree) a general window of where we want to end up on timbre balance. What we don't know is definitively why at this point.
 
See the discussion on the usage of the word 'timbre'

It might be more prudent to use the word 'tonal balance' instead of the word 'timbre' for transducers as timbre entails more than just tonal balance.

He is not only attempting to apply HRTF but at the same time compensate for hearing loss and the the FR variations of the headphone in question for that combination of headphone (including product variance) + personal perception.

Where Harman excludes personal HRTF and builds in a preference for the majority of headphone users (of all walks of life) and not individual.
So for a measurement of a specific copy on a specific position (or average of positions) on a specific standard fixture and adds an averaged preferred tonal balance.
It disregards product variances, it disregards positioning and seal and disregards personal preference and uses a standard to estimate tonal balance and basically is science oriented. One can create an EQ profile to end up with that average on that standard fixture in those conditions.

An attempt that tries roughly the same namely to end up with tonal corrections for a used headphone using very different methods and most likely different EQ's as well.

What would be interesting (scientifically speaking) is to compare the derived EQ's obtained using both methods and see how they differ. Simply because for both methods the end goal is to end up with an EQ that corrects the tonal balance of headphones.
 
He is not only attempting to apply HRTF but at the same time compensate for hearing loss and the the FR variations of the headphone in question for that combination of headphone (including product variance) + personal perception.
I am not so sure about the hearing loss part in individuals . He asked his subjects to match the loudness of each third octave band perceptually. This would not change anything as far as hearing response functions individually. As what they heard from the loudspeaker would include any hearing loss and this this perceptual function would transfer over to the adjustments made for the headphone adjustments. The rest I am in full agreement with. ( I should have referred to it as spectral balance rather than timbre thanks for the correction)
 
Last edited:
What would be interesting (scientifically speaking) is to compare the derived EQ's obtained using both methods and see how they differ. Simply because for both methods the end goal is to end up with an EQ that corrects the tonal balance of headphones.
The part of the data I would like to see, is the correlation between listeners (or lack thereof) . If they are very similar, then we can make this usable in correcting headphones, if they are all different, then it only tells us information about individual HRTF's but not about what headphones should sound like. To me it wouldn't make much sense if they were extremely different. As everyone is hearing the same thing, even though they all have different HRTF's, it should come out as the same differences (headphone response plus or minus what they heard out of the loudspeaker).
 
I think some ASR users tend to overrate the meaningfulness of measurable data when it comes to headphones.
100%. My favorite is when folks eyeball 2 similar FR curves and confidently tell you which one will sound better.

a) Most of us agree that the frequency response is the most important parameter.
The real question is how much of the variation in perceived SQ can be explained by variation in FR? It's not 100%, not even close.

I have quite a few IEMs and Over-Ears, and when I tune them to Harman, they all sound different - some quite significantly - as they interact with my ear in a different way (also HRTF, hair, glasses etc.). Soundstage also seems a bit random.
Yep.

b) The Harman Target is a very helpful standard, but it is not the perfect target for everyone.
Agree. For starters, the methodology used by Harman to measure and report the Harman target is different from the methodology used by everyone else to measure headphones. The differences are significant and not easily correctible. So when you see FR measurements where Headphone A is closely following the Harman target curve, you actually don't know if it is - because the two plots (target and measured) were obtained using different measurement protocols (the only exception is when Harman provides the measurements using its own rig, such as in the tweets by Olive).
 
100%. My favorite is when folks eyeball 2 similar FR curves and confidently tell you which one will sound better.


The real question is how much of the variation in perceived SQ can be explained by variation in FR? It's not 100%, not even close.


Yep.


Agree. For starters, the methodology used by Harman to measure and report the Harman target is different from the methodology used by everyone else to measure headphones. The differences are significant and not easily correctible. So when you see FR measurements where Headphone A is closely following the Harman target curve, you actually don't know if it is - because the two plots (target and measured) were obtained using different measurement protocols (the only exception is when Harman provides the measurements using its own rig, such as in the tweets by Olive).
It does get messy of one starts digging. So many targets and measuring methods.

1737521219402.png
 
It does get messy of one starts digging. So many targets and measuring methods.

FWIW one reason I started giving more credit to subjective evaluation: Most IEMs and Harman 2019 represent a bit of an audiophile showroom tuning where the bass and treble are slightly emphasized. This doesn't work for me that listens for an extended time daily. I'll rather have clearer-smoother highs together with a bit less bass rumble. Even though I hate to say it, there's no universal target that knows your listening habits, and what you need.
 
FWIW one reason I started giving more credit to subjective evaluation: Most IEMs and Harman 2019 represent a bit of an audiophile showroom tuning where the bass and treble are slightly emphasized. This doesn't work for me that listens for an extended time daily. I'll rather have clearer-smoother highs together with a bit less bass rumble. Even though I hate to say it, there's no universal target that knows your listening habits, and what you need.
An HRTF would not be described as having boosted lower treble response. That's because it is a transfer function of the acoustic filter of the pinna. Hence it is natural for humans to hear this function when they listen. The bass and treble are most likely not emphasized at all in the Harman transfer function preference target. It is simply the transfer function (a fairly low resolution one) of the overall spectral balance of a good (flat free field) loudspeaker with good directivity in a listening environment that has a relatively "good" reflective and absorptive coefficients and a good RT (under 300ms). The pinna effect is well established. But the head shading effect (low pass shelf at 2K,ILD) , the Shroeder Frequency (transition frequency or room gain) , and the tactile sensation of "feeling" the bass, are all part of the transfer function and are often overlooked. It is most likely that we are skipping first principle acoustics and not really understanding what these curvy lines are really are telling us.
HRTF.png
 
Last edited:
Back
Top Bottom