What Causes This Output from Headphones?

[KSTR]

Well, if the APx fan is as freakingkly loud as the one in my AP System Two, then this really might be a problem.
EDIT: I see Amit has put the APx in box but was still thinking about an isolation box for the phones at that time:

I will probably build a quiet box in the future for it but it makes the workflow much slower.

Slower workflow... well, that's engineering life, is all I can say.

 

[KSTR]

But the real pity (with all acoustic measuerments made here) is that the measured impulse responses are not made avaibable so all we have is plots and (IMO useless) magnitude-only FR data in text form (like if it were 1980).

 

[GaryH]

The 'poor results in the group delay department' seen on here are most often simply due a measurement set-up that allows open-back headphone measurements to be polluted by uncontrolled environmental noise (and possibly acoustic crosstalk due to using stereo test signals), which has nothing to do with the performance of the headphone. This is quite obvious when you notice that open-back headphone group delay measurements on here commonly exhibit 'mess' above 1 kHz, none more so than the least isolating headphones in the treble as measured by Rtings e.g. the (very) open-back HifiMan Arya, whereas closed-back headphones commonly don't have this mess in their group delay, e.g. one of the most passively isolating closed-back Rtings measured, the Drop + THX Panda. Note that Rtings' own group delay measurements of e.g. the Arya do not show any of this mess, because they actually control for this environmental noise by measuring within an acoustic isolation chamber. As noted by Jude in his review of the Dan Clark Expanse, those headphones are actually quite isolating for an 'open-back' design, which explains the minimal 'mess' in their group delay. Also note that this pollution of the measurements on here by noise is not confined only to the group delay measurements, but also shows up in the frequency response plots as 'fine grass' roughness in the same region above 1 kHz, which again has nothing to do with the acoustic output of the headphones. And even if it wasn't polluted by noise, the group delay plot would still not be fit for purpose, as it's excess group delay that matters for audibility.

 

[solderdude]

Acoustical crosstalk, sorry.

I don't think so either.

 

[solderdude]

but also shows up in the frequency response plots as 'fine grass' roughness in the same region above 1 kHz, which again has nothing to do with the acoustic output of the headphones.

What's your reasoning for that. I think it is something that is a property of the drivers themselves.
For several dB variations to be shown in FR at 94dB the 'interference' from outside noise would be huge. Not all large drivers exhibit this either.

 

[KSTR]

The 'poor results in the group delay department' seen on here are most often simply due a measurement set-up that allows open-back headphone measurements to be polluted by uncontrolled environmental noise (and possibly acoustic crosstalk due to using stereo test signals), which has nothing to do with the performance of the headphone.

If that is true (and I fear it is) then many of Amir's measurements are doomed as not only useless but actually severly compromised data, spreading fake news around the world, actually disadvantaging many manufacturers. Truly a pity if not an outright distaster.

 

[KSTR]

Note that Rtings' own group delay measurements of e.g. the Arya do not show any of this mess, because they actually control for this environmental noise by measuring within an acoustic isolation chamber.

They've also used proper smoothing, it seems, and a coarser, more approbriate time scale.

 

[Verig]

So, as I mentioned in my first comment in this thread, if it’s a characteristic of the device rather than a measurement issue, some may perceive it as a signature rather than a flaw. That’s why I believe it’s difficult to simply judge it as good or bad.

Interesting topic. I'm very partial to open back and definitely to open back planars.
I find them "smooth and diffuse" compared to dynamic drivers. Even if the FR looks a bit suspicious (like Arya) they really sound excellent and how I perceive the soundfield is markedly different. Everything is ever so slightly less accurate (or less punchy?) but the whole is enjoyably together and errors don't sound nearly as bad as I'd think. Oddly enough, not bad at all.

I'm a bit of a Hifiman fan but the same feel applies to other planars I've tried, which covers most of them.
Why? I really don't know so I'm going to grab a cup of coffee and hope you guys figure it out.

 

[solderdude]

If that is true (and I fear it is) then many of Amir's measurements are doomed as not only useless but actually severly compromised data, spreading fake news around the world, actually disadvantaging many manufacturers. Truly a pity if not an outright distaster.

You're talking about the GD aspect only or all of the measurements.
FR is:
seal related
positioning related
pad condition related
product variance related
target related
test fixture (configuration) related
SPL related
smoothing related
so how can any headphone measurement be anything else than 'compromised data' ?

The distortion measurements are:
SPL related
seal related
product variance related
pad related
background noise related

I think you are being unfair towards @amirm 's measurements.
I'm sure if you ask Amirm he is willing to provide the measurement data.

GD, however, can still give indications about technical performance just like SINAD can be somewhat indicative.
Simply ignore the 'grass' between 1 and 4kHz or Amir could be persuaded to smooth the plots. That, however, would be very inconsistent and he would have to revisit all previous measurements if done correctly.

I think headphone measurements are the odd-one out compared to Klippel and the standard AP555X measurements of electronics.
Sure... reporting could be different and more elaborate (as more testing) etc but that is true for all measurements I have seen from others as well.

I'm sure to many engineers Amir's measurements are far from complete but for most readers going through an entire test report spewed out from the AP555X (like Schitt posts on their measurements) is not really going to inform the layman.
ASR is not a forum for test engineers but for consumers with an interest in audio gear and would like to see some basic plots they (think) they can understand.

 

[LIΟN]

Interesting topic. I'm very partial to open back and definitely to open back planars.
I find them "smooth and diffuse" compared to dynamic drivers. Even if the FR looks a bit suspicious (like Arya) they really sound excellent and how I perceive the soundfield is markedly different. Everything is ever so slightly less accurate (or less punchy?) but the whole is enjoyably together and errors don't sound nearly as bad as I'd think. Oddly enough, not bad at all.

I'm a bit of a Hifiman fan but the same feel applies to other planars I've tried, which covers most of them.
Why? I really don't know so I'm going to grab a cup of coffee and hope you guys figure it out.

As I mentioned in my post, the impulse response and, furthermore, the excess phase should also be checked. Additionally, as others have pointed out, it could be due to other factors as well.
And whether it is audible or preferable is another discussion altogether.
So, if someone likes it, they are free to enjoy it, and if they think it's bad, they are free to think so. Personally, I don’t have any strong opinions about it either way.

 

[Verig]

As I mentioned in my post, the impulse response and, furthermore, the excess phase should also be checked. Additionally, as others have pointed out, it could be due to other factors as well.
And whether it is audible or preferable is another discussion altogether.
So, if someone likes it, they are free to enjoy it, and if they think it's bad, they are free to think so. Personally, I don’t have any strong opinions about it either way.

By posting my preference I simply meant that for me there is a notable audible difference. And that difference does not seem to correlate with measurements that well so I'd very much like to know where it comes from. I can always roam free within matters of taste but the technical aspect is very interesting - especially is there something in the phase / group delay or nor.

 

[solderdude]

Also seem grille related a little

HD58X (note the scales used to better show the differences) taken on FP without pinna/ear canal.

stock (so grille with foam against the grille):

foam removed (so grille only behind the driver)

and below with the grille removed (so nothing behind the driver).

note: when the raw plot would have been done on a fixture with ear-gain the area between 1kHz and 5kHz would have been exaggerated by 10-15dB and who knows what that could have meant to the GD.

The tricky part was to not move the headphone too much during these tests.

Below Edition XX:

stock:

notice the crap between 500Hz ans 1.5kHz.

Below with damping felt in front and back of the driver:

some influence above 6kHz, nothing to write home about. Below 1.5kHz quite a difference in resonances.

Effect on FR of this modification:

Sounds a little better too, certainly with the 8kHz peak lowered.

 

[Verig]

The problem is that if ASR measurements don't have measurement error (or representation error) then it may be that I happen to perceive higher group delay as more naturally cohesive, spacious and preferred. That would be great news for me. But as others don't show similar results but smooth over them then I'm back to square one (in this regard at least).
I know this is a technical discussion but I'd very much like to be able to benefit from the results in a form of a graph I know how to read in relation to what I hear.

 

[solderdude]

I'd very much like to be able to benefit from the results in a form of a graph I know how to read in relation to what I hear

If only that exists... closest thing to it is FR measurements compensated to a target... but there too there are too many variables (see post #49).

 

[LIΟN]

The problem is that if ASR measurements don't have measurement error (or representation error) then it may be that I happen to perceive higher group delay as more naturally cohesive, spacious and preferred. That would be great news for me. But as others don't show similar results but smooth over them then I'm back to square one (in this regard at least).
I know this is a technical discussion but I'd very much like to be able to benefit from the results in a form of a graph I know how to read in relation to what I hear.

As another user mentioned, there are already too many variables to consider in measurements.
Therefore, it may be more suitable for you to equalize the frequency response of each device through in-ear measurements and compare specific devices or impulse responses.
Of course, the response of a device can be perceived as a characteristic depending on personal preference, but minimizing such variables as much as possible would make it easier to evaluate.
Here, sighted bias and fit also play a role as variables.
For example, last year, I remember testing an Apple Earpod that had been equalized based on my personalized measurements and applying the impulse response of the HD800s.
Yes, it sounded quite similar to the HD800s. But something still felt missing. So, I placed the unplugged HD800s on my head, and suddenly, it felt like I was truly experiencing the HD800s.

 

[Heinrich]

... in relation to what I hear.

.... But something still felt missing. So, I placed the unplugged HD800s on my head, and suddenly, it felt like I was truly experiencing the HD800s.

The two comments above make clear that we don't just hear, but we feel--the unplugged (!) hd800 makes the actually active IEM to "sound" like the former.

The 'fine grass' in the group delay measurements first appeared to me as measurement error in that the GD is a derivative (in sense of calculus, or a high pass if you will) like delta phi / delta freq which may promote simple rounding errors or noise. The 'grass' is very dense, isn't it? The engineering approach to measurement often dismisses an investigation on measurement tolerances or other more systematic errors, what a natural scientist would put to limelight more often.

All investigations on GD in regard to human hearing were permormed with a single step in GD, way more mild than what we see here.

Simply put, the interpretations given in the discussion may be a tad exaggerated. Especially when considering my first point, on how we don't just hear, but feel. If you feel to make science out of it, I would rather suggest to call it a day.

 

[LIΟN]

The two comments above make clear that we don't just hear, but we feel--the unplugged (!) hd800 makes the actually active IEM to "sound" like the former.

The 'fine grass' in the group delay measurements first appeared to me as measurement error in that the GD is a derivative (in sense of calculus, or a high pass if you will) like delta phi / delta freq which may promote simple rounding errors or noise. The 'grass' is very dense, isn't it? The engineering approach to measurement often dismisses an investigation on measurement tolerances or other more systematic errors, what a natural scientist would put to limelight more often.

All investigations on GD in regard to human hearing were permormed with a single step in GD, way more mild than what we see here.

Simply put, the interpretations given in the discussion may be a tad exaggerated. Especially when considering my first point, on how we don't just hear, but feel. If you feel to make science out of it, I would rather suggest to call it a day.

You are misunderstanding my comment. Simply wearing the HD800s did not make everything sound like the HD800s.
I applied the "personalized" real impulse of each device, based on measurements, as if it were a kind of BRIR—essentially applying an impulse response from the earpod to the HD800s. That is already science.
My brain was merely looking for a certain cue, and that cue happened to be the sensation of the large ear cups covering my ears when I wear the HD800s.
And I highly value measurements and science, and I also enjoy the process of scientifically analyzing subjective reviews.
Some people often say that there are things that cannot be explained by measurements, but in many cases, the issue lies with the measurement itself or with how the measurement is interpreted.

 

[Heinrich]

You are misunderstanding my comment.

Yes and no, my summary was a bit shortened. You filter the IEM to the hd800, again put quite abbreviated, but then something was missing that came up when the feel of the headphone was there on your head.

That was my point, it is not "what we hear" but a more broad experience, taking input from all senses.

Sorry for the confusion.

In regard to science versus engineering, focused on the measurement part, the aim and so the how to are very different. A scientist would have discussed all the (discrete) mathematics of the GD readings in every detail. An engineer trusts for good reason the output of his tools if not otherwise prescribed.

This I mentioned not to critisize, but to bring expectations into perspective. Somebody wanted to get a presentation of measurements that served him directly with an intuitive understanding of what they tell. WIth such an approach we would be on the verge to leave engineering territory and march in direction of competitive advertizing..

 

[LIΟN]

Yes and no, my summary was a bit shortened. You filter the IEM to the hd800, again put quite abbreviated, but then something was missing that came up when the feel of the headphone was there on your head.

That was my point, it is not "what we hear" but a more broad experience, taking input from all senses.

In the specific case I experienced, that was the case, but others may or may not feel the same way—I don’t know about that.
However, to determine whether group delay (or various measurement artifacts?) affects the audible range and preference, one would need to match the magnitude response to their own ears and then compare the effects of excess phase or GD. -If someone says that the sound feels richer or clearer, we first need to separate whether that is due to the effect of excess phase or differences in frequency response.

 

[Heinrich]

... audible range and preference, one would need to match the magnitude response ...

Sure, we could throw full science force on this audio problem. I'm not working on the field of audio myself, still I know how science works in other fields. Think of writing a reliable paper discussing investigations other can reproduce. Otherwise it would be just anecdotal. While anecdotal observation is more often than not a wothwhile trigger for said science, it cannot be the result.

Realistically we cannot expect someone on this board to take the effort. It wonders me if anyone is willing to falsify it was not just noise, and I have none bad feelings about that either.