• 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?

I’m a little bit lost in the headphone discussion: until now I believed that imitating an optimal speaker in a moderate reflective room was a good measurement to headphones, with some modifications attempting to cover averaged bone conduction. Please correct me if I’m wrong.

But due to the individual deviations on headphones preferences compared to more compact agreement on speaker preferences, is difficult to content a huge group of listeners so various types of headphones tuning cover better the listener’s taste.

But the late doesn’t cancel the need of measurements, at least objective tests combined by preference tests should help to detect clusters of frequency curves and build some variety of headphones to satisfy customers.
 
- If a headphone does exactly hit the Harman target that doesnt mean that it will sound perfect to you (or that there is something wrong with your ear if it does not sound perfect to you).
Got the same issiue did found the Wavelet app that let you seemingly modify linear the Harman curve (or any other target curve) for personal maximum transparency. Free software have a try works for me fine.

 
Last edited:
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?
By that stating the above, you are actually implying that klippel measurements of speakers is pointless tor the consumer while being helpful to the speaker designer. If that is what you are saying, I agree 100%.
 
When it comes to headphone measurements and the Harman curve, some members of the ASR community should remember that there are at least three layers of uncertainty:

1. Headphone measurements themselves are uncertain. The measurement depends on the test fixture, technique, etc. It is known that headphone measurements have poor repeatability, which is why some headphone test sites measure the headphones several times and average the results.
2. Headphone measurements on a test fixture do not translate to individual heads.
3. The Harman curve is itself a preference curve, meaning that it has a normal distribution and standard deviations. Some individuals fall outside the normal distribution.

What the measurements do, and what the Harman curve does, is to predict the preference of the headphone under test for the population as a whole. It is correct to say that most people would prefer the sound of such-and-such headphone if it has been correctly tested. It is wrong to blame individuals for not liking what we consider a well-measuring headphone with perfect Harman compliance.

When I see ASR members criticise others for not preferring the Harman target (which is itself preference-based) I roll my eyes a little. That is putting the cart before the horse. It is someone's preference, and who knows why they have that preference? They may have hearing loss. They may have heads and ears which are a little unusual. They may just like funny sound! And as JJ says - you can not criticise someone's preference.
 
I AM SAYING WE NEED TO MEASURE IT CORRECTLY
What is correctly... what do you propose ?

Problems with headphones vs speakers:

Speakers are heard with both ears at the same time.
Headphones have one channel per side (lets not get into the 3-4 wire discussion either)

With speakers there is also tactile feel of the body which adds to the entire 'music perception'.
Headphones lack this so just targeting the exact same EDRP is silly as the 'body input' is ignored in this case.

Speakers in rooms have room modes altering the FR in certain ways, usually more bass.
headphones can have seal issues resulting in less bass.
This means that IF we choose the same 'target' at EDRP and the speakers are not used in a 'perfect' room with 'ideal acoustics' at the right distance (between speakers and listener) and angle and headphones do not have perfect seal, pads are a bit worn or whatever what is the same EDRP character worth ?

Speaker positioning in the room and listening distance in real rooms will differ from what Klippel will tell you. Unless one measures at the exact listening spot and corrects the most obvious errors one will always use the sound of those (expensive ?) speakers in their room as reference and their headphones (with or without EQ based on whatever 'perfect' EQ they assume they are using) and the sound will still differ.
Headphone positioning and seal affects the frequency response in many cases will not be the same as was achieved with a copy of the same headphone on some 'ideal' fixture either.

Then there is the unavoidable product variance (which can be much higher than with speakers) and (silent) product changes, pad wear or replacement pads differing between the tested product and a sold copy, interaction between headphone driver and test fixture (will differ from real ears).

Let alone the substantial difference between the angle sounds enter the ear canal between speakers (in front of you from a distance) and headphones (from the sides and cm away in a 'sealed' condition) and as ears differ and HRTF differs in both cases from 'fixtures' what would a 'better' fixture add ?

Then add tonal balance issues in the recording (circle of confusion) and even the most perfect headphone/speaker might still not produce 'pleasurable' sound.

So... whatever 'error free' measurements and 'perfect fixture' and perfect matching EDRP of 'ideal speakers in ideal rooms' one can think of with 'ideal coupling' of a 'reference' headphone one can obtain there will always be errors.

So the questions would be .... What should a test fixture look like (dimensions, materials used). How to ensure proper positioning on the fixture that can be the same as on one's head. Will there be a correct 'correction' suitable (works correctly) for all IEM's, ear buds, on ear and over ear for every type of headphone design ?
I mean... driver size, type, angle, positioning on the baffle, driver-ear distance will differ.

Nah... one can scream and demand for better measurements and better corrections and better targets but the relation between measured response and perceived sound will always be lacking. There is no ideal speaker in an ideal room and no perfect coupling reference headphone to base any 'perfect standard' on.
Compromises have to be made.
 
By that stating the above, you are actually implying that klippel measurements of speakers is pointless tor the consumer while being helpful to the speaker designer. If that is what you are saying, I agree 100%.

I haven't read the entire thread, as I probably should before responding, but what is common among both speakers, headphones, and IEMs is that the linearity of the preference curves that some manufacturers aim for is based on studies and tests conducted with various ears and preferences. These studies indicate a general preference for these curves across different transducer systems.
That doesn't imply that you or I should or do prefer these responses; however, research and studies indicate that the majority of people do.

I doesn't need to be more complicated than that, I believe.
 
When it comes to headphone measurements and the Harman curve, some members of the ASR community should remember that there are at least three layers of uncertainty:

1. Headphone measurements themselves are uncertain. The measurement depends on the test fixture, technique, etc. It is known that headphone measurements have poor repeatability, which is why some headphone test sites measure the headphones several times and average the results.
2. Headphone measurements on a test fixture do not translate to individual heads.
3. The Harman curve is itself a preference curve, meaning that it has a normal distribution and standard deviations. Some individuals fall outside the normal distribution.

What the measurements do, and what the Harman curve does, is to predict the preference of the headphone under test for the population as a whole. It is correct to say that most people would prefer the sound of such-and-such headphone if it has been correctly tested. It is wrong to blame individuals for not liking what we consider a well-measuring headphone with perfect Harman compliance.

When I see ASR members criticise others for not preferring the Harman target (which is itself preference-based) I roll my eyes a little. That is putting the cart before the horse. It is someone's preference, and who knows why they have that preference? They may have hearing loss. They may have heads and ears which are a little unusual. They may just like funny sound! And as JJ says - you can not criticise someone's preference.
As i describe here above that Wavelet could linear modify a target curve that you like till you found a personal maximale transparency for your liking is that not a practical solution?. I Agree that a correct measurment is favorble but regarding each individual head is different not practical i guess.
 
Ufff @Snarfie play with Auto-EQ especially tilt and adjust slope max to enclosure similar to speakers (12 dB closed, 18~24 open) and smoothing the response. Wavelet use simple txt header in a tap manner fixed point GEQ which is both too little in resolution (at bottom) and not precise enough in highs and future more not a minimum phase all leading to a little bit more loosely bass and not exactly hitting or influencing resonance peek's especially in highs very good. Afterwards try the same in convolution FIR minimum phases baked form (with frequency resolution pushed low to 6~7 Hz). Listen carefully and observe the difference. With such and simple paper DIY mood it's possible to address the highs peek's in 10 KHz+ range almost completely with almost any over ear's, not sure what can be done additionally with similar effect for IEMs, costume passive filler fur sure but not for anyone considering self work needed.
The flaw of AutoEq and similar is when it tries to make them do what they possibly couldn't in the first place or at least not being happy to (introduce THD by pushing what it can't or would not be happy to) so you cut that down to 0 dB (to initial) editing filter directly (txt file in case of Wavelet) or implement high pass (Q 0.71) to natural roll-off order.
Use same HAT's for measured/targeted please!
 
What is correctly... what do you propose ?

Problems with headphones vs speakers:

Speakers are heard with both ears at the same time.
Headphones have one channel per side (lets not get into the 3-4 wire discussion either)

With speakers there is also tactile feel of the body which adds to the entire 'music perception'.
Headphones lack this so just targeting the exact same EDRP is silly as the 'body input' is ignored in this case.

Speakers in rooms have room modes altering the FR in certain ways, usually more bass.
headphones can have seal issues resulting in less bass.
This means that IF we choose the same 'target' at EDRP and the speakers are not used in a 'perfect' room with 'ideal acoustics' at the right distance (between speakers and listener) and angle and headphones do not have perfect seal, pads are a bit worn or whatever what is the same EDRP character worth ?

Speaker positioning in the room and listening distance in real rooms will differ from what Klippel will tell you. Unless one measures at the exact listening spot and corrects the most obvious errors one will always use the sound of those (expensive ?) speakers in their room as reference and their headphones (with or without EQ based on whatever 'perfect' EQ they assume they are using) and the sound will still differ.
Headphone positioning and seal affects the frequency response in many cases will not be the same as was achieved with a copy of the same headphone on some 'ideal' fixture either.

Then there is the unavoidable product variance (which can be much higher than with speakers) and (silent) product changes, pad wear or replacement pads differing between the tested product and a sold copy, interaction between headphone driver and test fixture (will differ from real ears).

Let alone the substantial difference between the angle sounds enter the ear canal between speakers (in front of you from a distance) and headphones (from the sides and cm away in a 'sealed' condition) and as ears differ and HRTF differs in both cases from 'fixtures' what would a 'better' fixture add ?

Then add tonal balance issues in the recording (circle of confusion) and even the most perfect headphone/speaker might still not produce 'pleasurable' sound.

So... whatever 'error free' measurements and 'perfect fixture' and perfect matching EDRP of 'ideal speakers in ideal rooms' one can think of with 'ideal coupling' of a 'reference' headphone one can obtain there will always be errors.

So the questions would be .... What should a test fixture look like (dimensions, materials used). How to ensure proper positioning on the fixture that can be the same as on one's head. Will there be a correct 'correction' suitable (works correctly) for all IEM's, ear buds, on ear and over ear for every type of headphone design ?
I mean... driver size, type, angle, positioning on the baffle, driver-ear distance will differ.

Nah... one can scream and demand for better measurements and better corrections and better targets but the relation between measured response and perceived sound will always be lacking. There is no ideal speaker in an ideal room and no perfect coupling reference headphone to base any 'perfect standard' on.
Compromises have to be made.
I asked this question earlier in the other thread and this was the answer:

I am not certain at this point. But I believe one that measures the headphone and not how humans hear is imperative to good objective metrics. I think that a some sort of a flat plane for a headphone (not that simplified though) and a standard coupler for a IEM's(modified from the current standards). I do not know the answer without the resources to conduct all the experiments needed to complete the research.
 
The question remains... what 'good metrics' and which 'near perfect' test fixture as well as the actual test conditions Jaxx is referring to and what those might be and who should do that research with what (to be developed ?) fixtures, compensations and target(s).

In any case it is clear that he is uncertain and should be because there is no single objective good metric for headphone measurements nor will/can there be because of the nature of the measurement method(s).

One can always believe, think, hope, suspect, demand of course but facts remain facts.
Headphone measurements are and always will be indicative at best, regardless how science-y it appears to be.
 
That is the problem here. Noticing that the current method is failing does not mean that I have all the solutions. I need to conduct research in order to resolve much of this. But it is easy to see our current methods are not working effectively.
What do you mean "the current method"? The flat plane coupler you proposed and which we were describing isn't the current method. I'm interested to know what research you think needs to be done, as at the end of the day you'd still need to derive a target curve for this coupler. I still don't see how you'd do this without placing it in an ideal room with good speakers, which is exactly what's been done before. Only this time you'd have a HRTF unlike anyone with ears and ear canals, but you'd have to prove measurements made with this accurately translate to how real humans hear the headphones.

I feel you're striving for a perfection which doesn't exist due to all the issues raised above by @solderdude. I think you're also undermining how good the current method is, albeit not perfect.

OK, thanks for your time. Even though I believe you are intentionally misunderstanding the obvious.
I think you're seeing bad faith where there isn't any, or being very pedantic with wording. I think the point was that the measurement of an individual's HRTF is objective in that, using im-ear microphones in the manner used by Smyth, the result is a an objective fact of how sound arrives in that person's ear. Of course this is "subjective" in the sense each person is a subject and the HRTF is different for each subject, but that's just being pedantic with wording and shows agreement for the point being made anyway.

I think it's a good argument and goes back to my point about perfection. If you are correct then it should be possible to make a pair of headphones which, by applying a generic impulse response for speakers in a room (generic in the sense it's not for a specific HRTF), sound the same for everyone (the same in the sense that it sounds like sitting in that room). The fact Smyth exists, Creative had a go with S-Xfi, to name just 2, demonstrates it's not this simple because of the way headphone fundamentally interact with each person's anatomy differently.
 
@solderdude. I think you're also undermining how good the current method is, albeit not perfect.
Which method do you mean ?
All standard measurement rigs produce different results and are not accurate above several kHz. They are pretty accurate between 100Hz and 6kHz which arguably is the most important part of the music spectrum.
This is clearly verifiable and requires a sizable investment.
 
I suppose when individual HRTF varies as much as it does, and is only crudely average-approximated by the HATS system, it could be worth exploring an alternative system that takes that variability factor out of it. Ultimately though, will this path actually make graph interpretation more easy, consistent and meaningful? Or what's the prospective outcome?

All I know is that nerds on the internet take measurements too seriously and think they have all the information needed to fully evaluate headphones. This is empirically a wrong viewpoint of course given that individual HRTF variations affect FR substantially, but it ends up ignored even when pointed out to those that read graphs dogmatically.
 
Last edited:
Ufff @Snarfie play with Auto-EQ especially tilt and adjust slope max to enclosure similar to speakers (12 dB closed, 18~24 open) and smoothing the response. Wavelet use simple txt header in a tap manner fixed point GEQ which is both too little in resolution (at bottom) and not precise enough in highs and future more not a minimum phase all leading to a little bit more loosely bass and not exactly hitting or influencing resonance peek's especially in highs very good. Afterwards try the same in convolution FIR minimum phases baked form (with frequency resolution pushed low to 6~7 Hz). Listen carefully and observe the difference. With such and simple paper DIY mood it's possible to address the highs peek's in 10 KHz+ range almost completely with almost any over ear's, not sure what can be done additionally with similar effect for IEMs, costume passive filler fur sure but not for anyone considering self work needed.
The flaw of AutoEq and similar is when it tries to make them do what they possibly couldn't in the first place or at least not being happy to (introduce THD by pushing what it can't or would not be happy to) so you cut that down to 0 dB (to initial) editing filter directly (txt file in case of Wavelet) or implement high pass (Q 0.71) to natural roll-off order.
Use same HAT's for measured/targeted please!
Sorry i lost you already in the first sence but that is my flaw.
My Experience with this Wavelet slider is that i could fine tune my personal amount of correction linear which works for me better than the fixed corrected measuring. Basicly without the corrected measuring i can't listen to the headphone. For that matter the corrected measuring does a great job. The personal amount is the icing on the cake atleast for me subjectively. For sure it is not perfect their is more to consider but for a practical way of using (in this case jaakkopasanen Harman curve) measurements, Wavelet (an my limited knoledge :facepalm:) is the best i can do.

1000001439.jpg
 
Last edited:
@Snarfie https://autoeq.app/
I literally told you what to do and how to adjust knobs other than to play with tilt one and adjust bass and treble accordingly. It bakes you EQ for Wavelet (1.31 KB) or many other things which you download and impot with Wavelet. And yes those are the same EQ's but you can future adjust it more this way (that it sticks better to a same target or to your personal correction for it) along with editing it by hand (in text editor) if need be (with such bass cannon's there isn't).
 
@Snarfie https://autoeq.app/
I literally told you what to do and how to adjust knobs other than to play with tilt one and adjust bass and treble accordingly. It bakes you EQ for Wavelet (1.31 KB) or many other things which you download and impot with Wavelet. And yes those are the same EQ's but you can future adjust it more this way (that it sticks better to a same target or to your personal correction for it) along with editing it by hand (in text editor) if need be (with such bass cannon's there isn't).
Ah ok did find my headphone on autoeq.app will have a try thanks.
 
Last edited:
Back
Top Bottom