D
Deleted member 16543
Guest
Pass, but thanks.
-
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!
Master Complaint Thread About Headphone Measurements
- Thread starter amirm
- Start date
Pass, but thanks.
Perhaps, Robbo99999,
Since I'm a new here (but not new to the subject of headphone measurements), I thought it might help to go into a bit more detail on a couple of your inquiries, just to be sure that we're on a similar page. And to show I have some understanding of the more technical aspects and nuances of this subject (within certain limits). I apologize if some of my comments came across as condescending, or overly explanatory though. And will try to be more succinct, if there's anything else to add on any of the above.
I'm enjoying some of the different tangents in this discussion. But understand that it might make things a bit harder for some folks to follow.
A good point about the directionality of the HPTF source. The plot of the transfer functions for different parts of the ear, head and torso was mostly just to give a general idea which parts of the ear and HRTF might be included in the HPTFs for different types of headphones though. And which might not.
Since I'm a bit new to the concept though, I'm sure you'd know more about it than I do.
Last edited:
Not necessarily, I've just pieced together bits of understanding I've gained from visiting this site since beginning of 2020....I've got some scientific background so visualising difficult concepts or extrapolating from them isn't that alien to me, and I do like a visual graph or schematic! If you've got some theories and want to discuss it in detail with folks then start a new thread on any particular topic you want to discuss.
If you want to bail, I understand. I appreciated having the opportunity to try to answer a few of your questions though on some of this stuff.
Since you seemed somewhat curious before about how the diffuse field response of a headphone might be related to the sound power of a loudspeaker, I went looking for some quotes from some of the experts on this subject to maybe try to explain it better than I can. I'm pretty sure I heard either Dr. Olive or possibly Dr. Toole refer to sound power as the sound of a loudspeaker in a diffuse field, in one of the rather numerous videos discussing their research on either headphones or loudspeakers. But I'm still trying to find that. It sort of makes sense though. Because there's really nothing else that it could really be. (Because the direct/free-field, listening window, early reflections, room curve and directivity indices are pretty much all covered by the other curves in a speaker's spinorama plot.)
While I continue to look for the above quote though, here are a couple others I found that say pretty much the same type of thing. I've pre-cued these, so you don't have to try to sift through them to find the quotes. In this video (which is a little hard to see), Dr. Olive describes sound power as the sound in a "reverberant room that was highly reflective", which is the same type of space and sound field used for measuring a diffuse field.
https://en.wikipedia.org/wiki/Reverberation_room
https://en.wikipedia.org/wiki/Diffuse_field_acoustic_testing
I believe the "room curve", which is sometimes but not always included on spinorama plots, is actually a weighted combination of the speaker's direct sound (or early listening window, I can't remember which), the early reflections, and the sound power.
Last edited:
At 13:45 in this video, which is also already pre-cued, he talks about some of the tests that Harman did with different headphone reponse curves that Amir referred to earlier, including the diffuse field curves. And he describes a diffuse field as "a speaker in a very reflective, diffuse field, like a reverberation chamber". Which is similar to the description that he gave above for sound power.
Last edited:
The diffuse field source which is used for measuring the in-ear response of a HATS rig is a bit different though than the diffuse or sound power response of a loudspeaker. Because the former is spectrally flat, and designed to measure the spectral balance of the HATS rig at the DRP, to give you an idea how it distorts a spectrally flat diffuse sound field. While the diffuse or sound power response of a speaker represents its spectral balance in a highly reflective reverberant room, measured (out of the ear) at the listening position in the room.
So they are similar, but not precisely identical functions, which are really more like opposite sides of the same coin. And one gives the diffuse response of the sound source, at the listening position in a room. While the other gives the response to a spectrally flat diffuse source inside the ear, at the sound receptor.
In this section of the above video though, at 34:52 (again pre-cued), he sort of pulls the two threads above together, and suggests that after some additional subjective tests, their preferred target headphone response curve approximates the sound power of a well-designed speaker, with about 2 dB less bass and treble... So not a perfect match in this test, but pretty close.
This video is from 2014 though, and I believe in some of their subsequent testing they revised the bass on their target headphone response curve back up again. Which would likely have put it pretty close to the sound power response of a well-designed speaker.
So they are similar, but not precisely identical functions, which are really more like opposite sides of the same coin. And one gives the diffuse response of the sound source, at the listening position in a room. While the other gives the response to a spectrally flat diffuse source inside the ear, at the sound receptor.
In this section of the above video though, at 34:52 (again pre-cued), he sort of pulls the two threads above together, and suggests that after some additional subjective tests, their preferred target headphone response curve approximates the sound power of a well-designed speaker, with about 2 dB less bass and treble... So not a perfect match in this test, but pretty close.
This video is from 2014 though, and I believe in some of their subsequent testing they revised the bass on their target headphone response curve back up again. Which would likely have put it pretty close to the sound power response of a well-designed speaker.
Last edited:
We can test how close a graph of the diffuse field response of the Harman curve is to plots of the sound power responses of some well-designed loudspeakers by simply inverting each of the loudspeakers' sound power curves. And then using that as a correction or compensation curve for the diffuse field version of the Harman curve. Which is what I've done in the four graphs below.
An exact match between the Harman target and speaker would be represented by a flat line in the Analysis Panel at the bottom of each of the images. Most are in the general ballpark, except in the very low and high frequencies. Some of the differences would be a bit more noticeable though if the graphs were expanded or scaled a bit more in the vertical axis.
DUTCH DUTCH 8C:
INFINITY PRELUDE MTS:
MESANOVIC RTM10:
GENELEC 1032A:
This is one of the other reasons why I think it would be good to see both the raw and diffuse field measurements of the headphones that are reviewed here.
An exact match between the Harman target and speaker would be represented by a flat line in the Analysis Panel at the bottom of each of the images. Most are in the general ballpark, except in the very low and high frequencies. Some of the differences would be a bit more noticeable though if the graphs were expanded or scaled a bit more in the vertical axis.
DUTCH DUTCH 8C:
INFINITY PRELUDE MTS:
MESANOVIC RTM10:
GENELEC 1032A:
This is one of the other reasons why I think it would be good to see both the raw and diffuse field measurements of the headphones that are reviewed here.
Last edited:
This is also just my opinion. So you can take it or leave it. But I think the reason Harman's early headphone tests with diffuse field curves probably failed is because they were calibrating their headphones to match the in-ear response of a spectrally flat diffuse sound field compensation curve, which is well to the brighter side.
What they probably should have done was try EQ-ing the diffuse field responses of their headphones to match the diffuse/sound power responses of some of their loudspeakers. If they had done this, then I think they might've gotten a much better result. (Or so it seems from all of the graphs above.) And maybe we'd have a more convenient DF target response curve, that was designed to work with more devices than just the GRAS system they were using for their raw headphone measurements.
Their testing with different raw response curves for headphones (based largely on the in-ear measurements of speakers in a room, and also some subjective testing) has helped to demonstrate the validity of the above diffuse field / sound power relationship though. So the tests were largely successful from that standpoint.
What they probably should have done was try EQ-ing the diffuse field responses of their headphones to match the diffuse/sound power responses of some of their loudspeakers. If they had done this, then I think they might've gotten a much better result. (Or so it seems from all of the graphs above.) And maybe we'd have a more convenient DF target response curve, that was designed to work with more devices than just the GRAS system they were using for their raw headphone measurements.
Their testing with different raw response curves for headphones (based largely on the in-ear measurements of speakers in a room, and also some subjective testing) has helped to demonstrate the validity of the above diffuse field / sound power relationship though. So the tests were largely successful from that standpoint.
Last edited:
@ADU , it's not about "bailing" and I'm not obliged to reply to all your posts anyway regardless, you're not taking the hint, if you want to start a new thread discussing your diffuse field theories then do so....I don't really see this thread as being a chalkboard for your detailed musings on the subject. (I've not looked at all your new 5 long posts in succession, I'm a bit concerned about the quirky persistence in light....)
I got your hint, Robbo.
I don't really want to piss you and some of the other members of this community off though in my first week on the forum. So maybe I'll quit while I'm ahead on some of this... Unless someone else has some other questions on anything I've posted above that is within the scope of this topic.
There was a bit more to my explanation of why I think DF measurements would be a useful addition here. But it can easily wait for some other time and place if its too upsetting for you and others here right now. For those with a bit more knowledge and past experience on some of this acoustic measurement business though, most of this (including the diffuse field) is pretty basic stuff.
And while I may be drawing a couple of my own conclusions here and there, from some of the facts and measurements, and other info above, in the absence of a more well-defined study that I could cite (which may or may not exist), there isn't really anything in what I've said so far which is that far out, or beyond the pale imho... At least no more so than some of the other ideas and opinions that some other folks seem to be offering here.
I hope you'll reconsider your decision not to look at some of the above material though. Because some of it is pretty dang interesting imho!... And thank you also again for some your questions and comments, Robbo. Because they were much appreciated!
Last edited:
(discuss it somewhere else in this forum, night night)
D
Deleted member 16543
Guest
For what it's worth, I see where you're coming from. I also think that if a thread called 'complaints about something' isn't the right place to talk about different ways of doing that something.. Then it might as well be called 'complacency about something'.
However.. DF obtained curves are not that great for the purpose, in my opinion.
I touched on a few reasons why a couple posts ago.
But you do you, I do me and ASR does ASR.
I enjoy your posts.. Long and, in my opinion, slightly mislead as they may be.
There's a lot worse on this site.
OP
Thank you for the reply, Amir.
I think I'd have to see and understand the context in which these assessments were made to better understand what was meant by this. From looking at the spinorama plots of dozens of different speakers in Pierre's database though, I can tell you that it's the sound power curves which tend to vary the most of the 6 or 7 curves on a spinorama plot. And to be the least predictable in their behavior. So perhaps this is why they also have the lowest correlation to listener preferences, and are the least reliable for predicting that of all of the curves.
The only speakers whose sound power curves I really pay attention to though are the ones which also have a flat direct response. Because, generally speaking, those are the ones which are likely to have the most neutral (and satisfying) sound. And generally speaking, these kinds of speakers tend to have a somewhat more well behaved off-axis or sound power response as well, which is a bit more linear than some other speakers.
There are some exceptions to this though. And one that I happened to run across (and can still remember) was the Focal Twin6 Be. Which has a pretty flat direct sound (except for a small dip at around 2k). But some fairly big peaks and valleys in both its sound power, and directivity. So much so, that I felt I could not include something like this in my samplings...
https://pierreaubert.github.io/spinorama/Focal Twin6 Be/ErinsAudioCorner/index_eac-vertical.html
https://pierreaubert.github.io/spinorama/Focal Twin6 Be/ErinsAudioCorner/index_eac-horizontal.html
(This probably isn't the sort of speaker most people would use for mastering or listening to music anyway though, because it was designed primarily as a center channel. So there were multiple reasons for omitting it.)
There can still be alot of variation in the sound power responses of speakers with a flat direct response though, even if you leave out some of the more extreme examples, like the one above. Enough that I've sort of roughly divided alot them into three major groups (which are linear, V-shaped, and L-shaped). And since I'm mostly interested in what an "average" speaker with a flat direct response sounds like, I think I have to look at, and also include some speakers from all these different types in my samplings.
The point though is that I'm not just looking at the "random" sound power responses of speakers across the board (which may be what the Harman study was looking more at). And only interested in the sound power behavior of speakers which could realistically be thought of as having a "neutral response". And those would be the ones with a flat direct response, and generally also a pretty well-behaved or fairly linear sound power response... So there is some sifting of the wheat from the chaff in this.
If that doesn't fully answer the first part of your question, please say so, and I'll try to clarify it a bit further, if I can.
-----------------------------------------------
Re your second question, there are actually several reasons why I think that the average sound power response of a neutral loudspeaker could potentially make a good (but not necessarily the best) target for the response of a neutral headphone. Some of these reasons should be fairly intuitive, like for example its similarity in shape to the DF response of the Harman target and also the DF responses of headphones with a similarly neutral-ish response. While others are perhaps somewhat less intuitive, and have more to do with what a sound power curve actually represents (which is the diffuse response of a pair of speakers in reverberant room).
Generally speaking though, the only kinds of frequency response measurement information which is somewhat widely and readily available for a headphone is either its raw in-ear response, or its diffuse field response. And imo, the best way to compare the responses of a headphone and a loudspeaker in a room would actually be to use the raw in-ear measured responses of both types of transducers. There isn't really a huge wealth of info available on the loudspeaker side of this approach though, outside of some of the Harman studies. So the next best (and really the only other) option is to use the diffuse field response of a headphone.
This makes a good second option, imo, because there's a wealth of diffuse response information available on the speaker side of the equation (thanks in no small part to websites like this), that can also be used for comparison to the diffuse field info on headphones (which is currently available through graphing tools like Oratory's). And that diffuse info on the speaker's side would be the sound power curves in spinorama plots.
The other reason why I think sound power is a potentially good model for a neutral headphone's response is a bit more abstract, and harder to explain. But it sort of goes back to the reasons why the diffuse field response and compensation curves of a HATS measurement rig makes such a poor model for a neutral headphone's frequency response. And this is because the DF compensation curves for these rigs lack all of the essential timbral and tonal balance information that you get (in abundance) from a speaker's sound power curve.
I've already sort of explained a bit why this is. So I won't go into alot more detail on it again here. But it is, imho, primarily because the DF response of a HATS rig is based on a spectrally flat diffuse sound field, rather than on the more darkly-tilted response of a pair of speakers in a room.
The speaker's sound power response curve is what (potentially) fills in alot of the vital missing information between those two things. And it manages to do that because it contains both the reflective information of a reverberant room, and also the timbral info of a loudspeaker within that space, which is based to a big extent on the speaker's dispersion or directivity characteristics. The sound power directivity index of a loudspeaker is, in fact, derived primarily from a speaker's sound power response. And also either its listening window info (which is what Harman's spinorama uses) or its direct response. So that's how we know the sound power contains those dispersion or directivity characteristics.
And that is exactly the kind of info that's needed to bend or tilt the response of a measurement rig to a spectrally flat diffuse sound field down more in the treble and up more in the bass, to better match the response of neutral loudspeakers in room...
I give you, and Robbo99999 my word that I won't commandeer this topic for my own personal projects along these lines. Because that's not my goal here. But just to give you and maybe a couple others a quick idea of how someone might put some of the above info to use, here are several different raw in-ear response curves based on different loudspeakers, that were derived from this type of approach, by simply combining their sound power responses with the diffuse field compensation curve shown above. Which could potentially be used as some kind of a basis for a headphone's raw frequency response...
The horizontal gridlines on the graphs represent steps of 5 dBs. And most of these are pretty similar in shape and response to the raw plots of the 4 headphones on the previous graph above, and also to Harman's 2018 over-ear headphone target response curve.
The speakers are in order from top to bottom, the Dutch Dutch 8C, Infinity Prelude MTS, Mesanovic RTM10, Genelec 1032A, and Infinity Intermezzo 4.1T, btw. And as mentioned above, they were created by simply combining the speaker's sound power curves with Ora's diffuse field curve in Equalizer APO's Configuration Editor, as shown below...
There are often some other important details in the treble and upper mids though, which are generally related more to the resonant qualities or characteristics of a measurement rig, that will frequently be missing from both the rig's DF curve and the sound power speaker data (and also Harman's target response curve, for that matter). Which is why plots like the ones above, or some combination thereof, would still only be useful as a general guide for a neutral headphone's response. And why some of those other important details in the treble still have to be filled-in by simply looking at some of the measurements of actual headphones on the rig.
Raw in-ear measurements made from neutral loudspeakers in a room (using the same rig that's used for the headphone measurements) could also potentially help to fill in some of those details in the higher frequency ranges as well though. Ideally this would be done with a mannequin to also incorporate the head and torso effects with the room effects.
Last thing I'll mention here (since this has already gone way longer that I originally intended) is that there could possibly be a certain type or shape to a sound power response that results in a better (or worse) model for a headphone's diffuse response. So that's one other thing I'd also be looking for in alot of the data above. Maybe some users might prefer a more V-shaped sound in the upper mids than some of the plots above, that's more reminiscent of the response of some HFM headphones, for example...
DIFFUSE FIELD RESPONSE OF 5 HIFIMAN HEADPHONES:
Shutting up now (right after I respond to sax512's other post above).
Last edited:
Thank you, sax512.
I read up a little on the gating issue you mentioned before btw. If I understand the theory behind that, I guess it's to help isolate the response of speaker from its surroundings, when taking measurements. But is effected to some degree by wavelength, so it's not as practical for the lower frequencies. Maybe I'm not understanding it correctly though. And need to continue reading more about the subject.
I assume this is mostly for people doing their own measurements though, under less than ideal conditions. As opposed to measurements in an anechoic chamber, where room reflections are less of an issue. And I assume that the gating would also have to be adjusted differently if you also want to pick up the early and late reflections, or steady-state response in a room.
I do get a bit carried away with my enthusiasm on some of these subjects sometimes, btw. So I apologize if that scared anyone. And will try to be a little more level-headed in the future.
Last edited:
D
Deleted member 16543
Guest
It depends. The software that I use gates in accordance to psychoacoustics, therefore it takes into account early reflections and also standing waves in the lows (the gating varies with frequency).
Something like what Amir uses is tailored to gate right after the direct wave has passed, so that the reflections are not included in the response. It's a way to make quasi-anechoic measurements in a non-anechoic room.
It's still gating, but the philosophy behind these two types of gating is very different.
One looks at the end result and the balance at the listening position as a human would perceive it in the room where the measurement is taken, the other focuses on the behavior of the speaker itself, without much concern for the space in which it is placed (in fact, it actively erases those contributions from the response calculation).
What Amis uses also calculates the speaker's directivity, so that it can indirectly serve as an indication of how a speaker might sound in a typical room. But you probably already knew that.
I understand a bit about the basic principals behind the Klippel system that Amir uses. But not all of the specifics. What you're saying seems to make sense though, sax512. Including about the different uses for different types of measurements (and Amir's quasi-anechoics). And I suppose I can also imagine how differences in gating might potentially complicate the comparisons between two different sets of measurements, made on two different systems.
With regard to the DF measurements though (which I think is why this came up), I think I'm goin to have to put my faith in the geeks (no offense intended) and mathematicians that they pretty much knew what they were doin when they put alot of this stuff together. So that a DF measurement made with one type of system or transducer, would be the rough equivalent in terms of time domain effects and so forth as a DF measurement made on another system or transducer. And they are roughly comparable.
Perhaps that's a little naive on my part, since I don't really understand all of the complex underlying math on this stuff, and am mostly looking at the end results. But it seems to work... Quite well in fact. And since most of the key data points seem to be lining up so well, I think I have to give the folks who designed and developed this stuff largely the benefit of the doubt. I'm probably puttin Robbo to sleep again though. So better quit there.
With regard to the DF measurements though (which I think is why this came up), I think I'm goin to have to put my faith in the geeks (no offense intended) and mathematicians that they pretty much knew what they were doin when they put alot of this stuff together. So that a DF measurement made with one type of system or transducer, would be the rough equivalent in terms of time domain effects and so forth as a DF measurement made on another system or transducer. And they are roughly comparable.
Perhaps that's a little naive on my part, since I don't really understand all of the complex underlying math on this stuff, and am mostly looking at the end results. But it seems to work... Quite well in fact. And since most of the key data points seem to be lining up so well, I think I have to give the folks who designed and developed this stuff largely the benefit of the doubt. I'm probably puttin Robbo to sleep again though. So better quit there.
Last edited:
OP
- Thread Starter
- #658
Klippel NFS only uses gating at higher frequencies (set to about 1 kHz in my configuration). Below that, it is actually able to separate the direct sound of the speaker versus reflected ones. As such, it produces completely anechoic response at full resolution.
OP
- Thread Starter
- #659
So to be absolutely clear, measurement system I use is NOT quasi-anechoic. It truly is anechoic. Quasi-anechoic refers to gated high frequencies stitched manually with near-field measurements. It is an approximation with inherent error (unless you spend time compensating for it).
Got it. I won't make that mistake again.
Similar threads
