• 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!

Is there any way to objectively measure headphone resolution?

Joined
Oct 31, 2020
Messages
27
Likes
16
I just realized something. As far as I know, the only sources indicating that some headphones have greater resolution / detail retrieval than others are subjective reviews. Oftentimes from the same people who would have you believe in benefits of multi-thousand $ cables, 192+ kHz audio and other such nonsense.

My question is... Perhaps, just perhaps, might it be possible that the differences in resolution between various headphones are also mostly imagined? The fact that headphone blind testing is complicated to say at least only adds to my concerns.

What objective proof do I have that something like a Stax SR009/S, an Abyss Phi TC or a Raal SR-1A have greater resolution than something like this: https://drop.com/buy/drop-hifiman-he4xx-planar-magnetic-headphones ?

Anything other then ''I compared them myself and heard a clear difference'', because as we know, in the world of audio, people swear to hear all kinds of differences which objective measurements prove to be imaginary.
 

BillG

Major Contributor
Joined
Sep 12, 2018
Messages
1,699
Likes
2,266
Location
Auckland, New Zealand
If you're referring to frequency response and distortion levels, these would be the fundamentals of such a measurement I suppose. The professionals here could provide a more comprehensive answer, though.
 

bobbooo

Major Contributor
Joined
Aug 30, 2019
Messages
1,479
Likes
2,079
Here's the best answer I've seen to this question, from reddit user Oratory1990, a professional acoustic engineer who measures headphones for a living:
"Detail" is the ability to correctly reproduce small timescale events. In order to do this, the system needs a high upper frequency extension, at least 20 kHz, and also (more importantly in my opinion) no big resonance peaks that would obscure them.
The tricky thing is that due to the resonance „storing“ energy in its frequency region and releasing it over a longer time, some people mistake this for "increased detail", when in fact it is not the music that has more detail now, it‘s a resonance of the headphone that obscures what is actually happening in the music.

What are we looking for? We are looking for smooth, extended treble reproduction without sharp resonance peaks.
So basically, resolution/detail, like the vast majority of a headphone's perceived characteristics that influence judged sound quality, can be determined by frequency response, as is supported by the scientific literature e.g. Dr Sean Olive's extensive research.
 
Last edited:

Feelas

Senior Member
Joined
Nov 20, 2020
Messages
390
Likes
315
Well, if we can go from impulse response to FR and back-forth, then (as far as I remember, tho I'm no major in the topic) both FR and IR contain complete system's response to the stimulus.

Consider looking at masking phenomena (the same ones used with lossy compression), relatively wide-spread high-Q peaks (of lower amplitude) could cause less masking than low-Q ones with higher, maybe? If quanitzation and artifacts can be put accounted for in lossy compression, then the same behaviour might occur in headphones, it seems that oratory is onto the relatively same way of thinking. I guess we could try comparing the performance of subjective "detail killers" alongside FR graphs and see for any correlation.

It's actually interesting whether there's any connection between FR smoothness and detail masking; yet that would require subjective testing, since "detail" cannot be easily quantified by a PC. If the a low-Q resonance is so overpowering in a certain freq-range it might enforce massive masking (due to big differences). Should look at lossy compression for how the artifacts are hidden, again.

Great topic, wondering about that a lot myself recently and needed a space to discuss.
 
Last edited:

Blujackaal

Active Member
Joined
Feb 7, 2020
Messages
221
Likes
102
The Hyped Stax L700mk2 and 009/S. Have really heavy roll of under 70Hz that makes the already forward treble sound brighter than normal, Which can fool people that it has fast bass/more res. I don't think being <0.2% at 100db guarantee clean sound, Grado headphones do the same thing like roll of under 70Hz and boosted treble it using a 40mm driver with high THD. The ER4SR can be 0.85% still sound clean from no masking effects since it FR is so smooth and it using a stiffer driver to stop bass overshoot(planar/estat can move the whole diaphragm).

Oftentimes from the same people who would have you believe in benefits of multi-thousand $ cables, 192+ kHz audio and other such nonsense.

The same reviewers have done reviews where they hype the fast bass of Stax L700, LCD2 but then when doing BA headphones they'll argue that fast decay is a issue?. Then defend a artifact that full range dynamic drivers suffer from?.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
teufel-fr-ok.png


good treble extension ... check.
no audible sense of 'resolution'... alas ... check.
Looking at the second condition... no peaks and dips... indeed peaks and dips in the treble.
In this case explains the lack of treble quality/air and resolution.

The question is if these peaks and dips could be 'compensated' properly (10dB high Q) and if that would increase SQ. Never tried, never will sent it back to the manufacturer. IMO it is best to 'start' with a headphone that needs the least amount of EQ and take it from there.
 

bloodshoteyed

Major Contributor
Joined
Jul 31, 2020
Messages
4,774
Likes
20,957
Location
n/a
That's the thing though... Couldn't two headphones have the same frequency response and yet different levels of ......

of anything, really
it's the same as a speaker playing in different rooms (though exchange the room with the shape/size of your head, ears, earcanal..)
 

Blujackaal

Active Member
Joined
Feb 7, 2020
Messages
221
Likes
102
teufel-fr-ok.png


good treble extension ... check.
no audible sense of 'resolution'... alas ... check.
Looking at the second condition... no peaks and dips... indeed peaks and dips in the treble.
In this case explains the lack of treble quality/air and resolution.

The question is if these peaks and dips could be 'compensated' properly (10dB high Q) and if that would increase SQ. Never tried, never will sent it back to the manufacturer. IMO it is best to 'start' with a headphone that needs the least amount of EQ and take it from there.

What headphone is that?.
 

Feelas

Senior Member
Joined
Nov 20, 2020
Messages
390
Likes
315
The ER4SR can be 0.85% still sound clean from no masking effects since it FR is so smooth and it using a stiffer driver to stop bass overshoot(planar/estat can move the whole diaphragm).

It might be a plain trade-off between overshoot (and more correct impulse response) vs undershooting (and having less correct impulse). Getting this "exact", as in "neither over nor under" might be impossible. These *might* be some cause of these strange HF artifacts in various TOTLs.

The same reviewers have done reviews where they hype the fast bass of Stax L700, LCD2 but then when doing BA headphones they'll argue that fast decay is a issue?. Then defend a artifact that full range dynamic drivers suffer from?.
Depends on what you're listening to, remember that e.g. bass guitar has many high harmonics and thus when producing full-bodied sound the instrument's speed rendering might not be considered as a whole and judged as "fast" (since high harmonics occur properly) and not properly evaluated in the lows. Presence region is much higher! Also, some music *might* be mastered with typical living-room reverb in mind, thus faster response could cause problems with any bass-related content, thus impacting point-source character, imaging and so forth.

The question is if these peaks and dips could be 'compensated' properly (10dB high Q) and if that would increase SQ. Never tried, never will sent it back to the manufacturer. IMO it is best to 'start' with a headphone that needs the least amount of EQ and take it from there.

Since they could have a resonant character, smoothing a response by compensating dips might cause additional THD+N issues due to new resonances happening, doing it the passive way by bringing down could indeed regulate masking - yet it could also contribute to worse resolution, if the big peak cause distortion masking - it seems that we'd require a complex model to get down to even the basics. Don't trust me on that, though. It is just merely interesting, considering i.e. what was brought up by your K612 measurements w/ being coarse despite being EQ'd down. I believe that's what I encountered on HD-681s EQ'd to Harman (for pure experiments) and it completely put me off that it's so harsh, what I've never heard before w/o EQ. Could you try boosting the dips w/ K612 (or anything else you have on hand with non-smooth treble) programatically after having the filter for the peaks applied? Targeting a smooth treble character on measurement. Just to see whether getting the treble objectively "smoother" (either with boosting or downboosting) gets us anywhere further into the better SQ and if the impact is negligible or not.
 
Last edited:

Dreyfus

Active Member
Joined
Feb 7, 2019
Messages
247
Likes
296
Location
Germany
I think the explination is most likely the composition of the treble response where narrow dips could decrease the perception of detail whilst peaks could increase the saturation and sharpness. Such resonances, being related to the driver and cavity design, can possibly even contribute to the spatial perception which is mainly prescribed by the spectral maps of your specific PRTF / HRTF.

The trouble is the fact that standardized measurement systems are not capable of rendering the exact performance that is perceived by the human listener. Just looking the treble response, a lot of experts consider the accuracy of ear simulators to be only valid until roughly 4-5 kHz in practice. That is not only due to the individual ear geometry (pinna and ear canal size and shape) but also the fact that you are entering a frequency band whose wave lengths start to fit into the cavity where a lot of constructive and descrutive resonances can occur. The errors introduced by those parameters cannot be sufficiently predicted by measuring with an average ear whose response is also overlapping with the self-resonances of the coupler behind it.

I doubt that you can accurately capture and describe what resolution actually is if you only rely on measurements done with systems described above. In fact, that is not even how most of the (higher performance) manufacturers voice their products. They mostly treat it more like a guideline to confirm what they are actually hearing.
 
Last edited:

Feelas

Senior Member
Joined
Nov 20, 2020
Messages
390
Likes
315
I think the explination is most likely the composition of the treble response where narrow dips could decrease the perception of detail whilst peaks could increase the saturation and sharpness. Such resonances, being related to the driver and cavity design, can possibly even contribute to the spatial perception which is mainly prescribed by the spectral maps of your specific PRTF / HRTF.

Good observation, but what do you mean exactly by "saturation and sharpness"? Despite being said in pretty visual terms, I cannot really translate this to audible analogue and am pretty interested what that'd mean. Whilst we're speaking about treble - isn't it possible that balanced response could be perceived from "correct" (as in "non-oversaturated" in amount) SPL of treble, despite having a "wrong" detail and tembral character?

If we're really speaking about completely unaccurate response in >4-5kHz range then I'd say that it's absolutely pointless to bother discussing in that case, since then we'd be completely in the dark if using actual measurements. Yet, I believe on GRAS we're measuring the summarical (so we can omit predicting the actual resonances happening, since we're measuring them in anyways) response and it's discussable whether this should even get mentioned. It would need to be proven that the difference between couplers (or even heads, in a large-scale test) is measurably high as to be regarded as important in measuring treble response. Maybe the difference is small enough, since skin and bones might not really be that different between people and the resonant character is mostly kept? As for the canal resonance - isn't it what the Nuraloop & friends account into the measurements?
 

Sgt. Ear Ache

Major Contributor
Joined
Jun 18, 2019
Messages
1,894
Likes
4,150
Location
Winnipeg Canada
I'd say it's 75% balanced freq response/low distortion, 25% recording quality. I suspect that any 2 headphones that have reasonably close freq response, low distortion/noise, compared at the same volume with the same recording would be pretty much indistinguishable from one another in terms of detail retrieval. In a blind test that is, as hard as that is to perform.
 

wasnotwasnotwas

Senior Member
Joined
Nov 11, 2020
Messages
329
Likes
372
I'd say it's 75% balanced freq response/low distortion, 25% recording quality. I suspect that any 2 headphones that have reasonably close freq response, low distortion/noise, compared at the same volume with the same recording would be pretty much indistinguishable from one another in terms of detail retrieval. In a blind test that is, as hard as that is to perform.

If 2 headphones can take the same pads and some after market headband padding/ cover I reckon you could have a decent stab at a truly blind test. Assuming weight was similar and both EQ on same rig to same target of course.
 

Feelas

Senior Member
Joined
Nov 20, 2020
Messages
390
Likes
315
If 2 headphones can take the same pads and some after market headband padding/ cover I reckon you could have a decent stab at a truly blind test. Assuming weight was similar and both EQ on same rig to same target of course.
See https://seanolive.blogspot.com/2016/04/a-virtual-headphone-listening-test.html

It seems that Harman found out that until 8-10kHz the difference is consistent enough to take it into account and after that there's no point EQ-ing. What brings us further (and is maybe further explained, I didn't geek out Olive's blog like there is no tomorrow) is also this:
Finally, it's important to acknowledge what the virtual headphone method doesn't capture: 1) non-minimum phase effects (mostly occurring at higher frequencies) and 2) non-linear distortions that are level-dependent.

I believe this might also contribute to the "needing the least amount of EQ" dilemma, since we want stuff that suffers least from that.

Citing Olive in comments (despite being dated 2016): "The ideal headphone for virtualization has an extended flat smooth response (20 Hz to 25 kHz), no distortion, no leakage or controlled leakage, and a consistent response across all listeners. If you find such a headphone let me know :)"

They used HD518 for their EQ rig.
 
Last edited:

Blujackaal

Active Member
Joined
Feb 7, 2020
Messages
221
Likes
102
Depends on what you're listening to, remember that e.g. bass guitar has many high harmonics and thus when producing full-bodied sound the instrument's speed rendering might not be considered as a whole and judged as "fast" (since high harmonics occur properly) and not properly evaluated in the lows. Presence region is much higher! Also, some music *might* be mastered with typical living-room reverb in mind, thus faster response could cause problems with any bass-related content, thus impacting point-source character, imaging and so forth.

TOTL measuring rigs show the ER4 are nearly a flat line to 17.5KHz, The ER4 needs a 3.5db(2.8 Q) to fix it 7k dip that gives it muted treble. Many that sound suffer from poor res seem to have -5 ~ -12db above 6.5khz and +4.5db bass boost under 180Hz. Since they're no masking/overshoot it could explain why ER4/LCD still sound clean when the bass is boosted by +8db. The Utopia only Low THD because it 40mm driver look like something I'd see in a Speaker, But the HD560S using a 38mm that low THD and can handle more bass because it seems both don't have excursion issues and can rest quick from <100Hz notes.



It might be a plain trade-off between overshoot (and more correct impulse response) vs undershooting (and having less correct impulse). Getting this "exact", as in "neither over nor under" might be impossible. These *might* be some cause of these strange HF artifacts in various TOTLs.

That because many TOTL's are just free profits the KSE1200 the only TOTL in my view that free from the artifacts & issues. The Stax needs a blu tack mod, The ribbon HP driver THD done here shows it should've been a IEM or closed back.
 

Sgt. Ear Ache

Major Contributor
Joined
Jun 18, 2019
Messages
1,894
Likes
4,150
Location
Winnipeg Canada
If 2 headphones can take the same pads and some after market headband padding/ cover I reckon you could have a decent stab at a truly blind test. Assuming weight was similar and both EQ on same rig to same target of course.

you could also do something like what Oluv's Gadgets does - get really good recordings of the output of different headphones and then A/B and try and pick the "higher detail-retrieval" phones...
 
Top Bottom