Master Complaint Thread About Headphone Measurements

[Robbo99999]

Sorry about that. I was coming to this as well, Robbo. But got sidetracked by some other stuff.

The diffuse field plots of the 2018 Harman over-ear curve were produced essentially the same way as the sound power plot above in Equalizer APO's Configuration Editor.

View attachment 144603

You are absolutely correct that the bass on the Harman curve should rise above the baseline. This is true also on a diffuse field compensated plot, because the DF curve is essentially flat below about the 160-200 Hz range in the bass. Without the negative Preamp enabled, you can see above that the compensated plot in the Analysis Panel does indeed rise above the baseline of 0 dBFS.

Just to clarify what's going on above, the raw plot of the Harman curve is the first filter or item in the stack, plotted using a variable graphic EQ curve. The second curve below that is Oratory's new 2021 diffuse field curve, turned upside-down (inverted to the geeks). Both curves are plotted in high precision, using a large number of points or bands The third filter is a Preamp control, which is currently disabled above. (You can tell it's off because its "power button" on the left isn't highlighted in light blue like the other filters.)

The bass level on Oratory's new 2021 DF curve is slightly above 0 dB. So when it's flipped upside down, it adds a small amount of negative gain on the order of -0.3 to -0.4 dB. Which brings the peak bass level on the Harman curve's compensated plot down by just a small amount. Otherwise the peak level in the Harman curve's bass response would be closer to 6.1 dB above the baseline. You can see the difference though after the two curves were combined in the lower right of the Analysis Panel, where it says "Peak gain: 5.7 dB". The area that excurses above the 0 dBFS baseline is shown in red, btw, to warn you that audio in that area could be clipped, because it exceeds the unity gain level.

With the -8.00 dB Preamp filter enabled though, the whole curve falls somewhat below the 0 dBFS clip point, So the bass is no longer red, as shown below.

View attachment 144611

Since these curves and plots are mostly for my own informational/analytical purposes, and not for actual listening use (at least not in this particular form). And my primary concerns are the overall shape, tonal balance, and slope of the curve, what I'll generally do is drop the curve low enough so it's below the 0 dBFS clip point. And in the case of a diffuse field or sound power plot, I'll usually try to align or normalize the resulting response curve to one of the horizontal gridlines at around 1000 kHz or thereabouts, using the Preamp control. (If I'm in a hurry though, sometimes I may forget to do this.)

View attachment 144619

The choice of the normalization frequency is somewhat arbitrary though. And sometimes I may try to normalize other parts of the curve, such as the bass instead. It just sort of depends on what I'm focusing on at the moment.

Normalizing to a point somewhere around 1 kHz in the midrange though helps me to better assess how much of a dip there is in the curve at the crossover point of the treble and midrange at around 2 to 2.5 kHz. The level of that dip is particularly important to me because my ears seem overly sensitive in that range. Probably because alot of the pop recordings I listen to are mixed too hot there, because the in-room monitors used for mixing and mastering the content likely have a similar depression there in their dispersion and off-axis response.

It also helps me to gauge some of the differences in the levels between the midrange and bass. And also the midrange and the treble, particularly around 6 to 8 kHz in the low treble. Imo, the treble in that 6 to 8k range on Oratory's DF graphs probably shouldn't exceed the level in the mids at about 500 Hz to 1kHz by too much, or the headphones may begin to sound a bit too bright. (This is just a general rule of thumb though, which may not apply to all headphones or listeners.)

The Harman curve does not represent a perfectly neutral response to my ears btw. But I can go into some more detail on that another time. The main area that I have some issues with though is the lack of air in the upper treble. There is some unevenness in the upper mids as well imo, which leads me to believe that it's possibly also a bit too forward in some spots there. Perhaps around 1.5 to 2 kHz, and maybe also in some other area around 4k or so. This is still something I'm experimenting with though.

The dip in the Harman curve's DF response in the upper mids is positioned a little higher in frequency than a typical loudspeaker's midrange-tweeter crossover. Which tends to be more around the 2 to 2.5k band range (though it can technically occur almost anywhere from 1 to 3 kHz, depending on the speaker's design). The artificial pinna that Oratory uses complicates the response in the upper mids and treble though. So it's not quite an ideal match to the average sound power response of the good loudspeakers. So that's why I use various references, including the measured responses of headphones, to try to get a better handle on what should be going on in that area. I think Ora's pinna can produce both bumps and dips/notches around that whole area though, which may make his plots look slightly off, or uneven, or overly bright or depressed in some parts of that range.

I'll have some more thoughts on diffuse field, and on your other post above with the HD560S plots another day.

That is another headphone I've been looking at pretty closely though lately. Although I haven't tried them yet myself, the new polymer drivers and good bass extension supposedly give those HPs a some wider sounding soundstage than some of Sennheiser's other classic 6-series HPs. Some reviewers also claim that they have a pretty neutral sound. And it looks that way from the graphs I've seen so far. They look like they might be just a bit uneven in some spots in the upper mids and treble though. And maybe just a tad bright around 4.5k or so. Which should be easily correctable with an EQ.

Although the Senn 5 and 6 series headphones tend to have pretty good driver symmetry overall, I've noticed that some of the models in those lines will sometimes appear a little imbalanced in some spots in the upper mids or treble. And it looks like the HD560S may possibly fall into this category as well, due to a little dip in the right channel at around 2k. Based on this compensated Rtings graph...

https://www.rtings.com/headphones/1-5/graph#18492/7903

A feature like this is slightly harder (but certainly not impossible) to correct with an EQ, because it requires a somewhat different adjustment for the right and left channels.

I don't currently have any open back HPs (or Senns) though. And this looks like it could be one of the better options in the Sennheiser lineup in my general price range. So the HD560S is still very much on my radar. There's a local store where I can easily get one, which also helps.

The audio gear I'm currently using is listed in the link in my signature btw.

Thanks for the reply there. So by subtracting the Diffuse Field Target from the Harman Target, what is it that you are wanting to show/achieve, what does it prove? It's only showing the difference between the two targets, and you seem to be trying to relate that graph showing the difference directly to graphs of speaker sound power....I'm not really sure how that is possible to compare them, I'm not joining the dots there at all....it seems like apples to oranges. It seems that you're relating them simply because they look similar, but just because they look similar doesn't mean a thing....I'm missing your logical step of why/how you're comparing the two.

About the HD560s, yes it's a good headphone, currently my joint favourite with the K702....they are the two of my headphones that do soundstage uniquely & accurately in comparison to my others headphones. K702 is an impressive immersive wide soundstage and has more detail & nuance around female vocals, and HD560s is narrower than the K702 but potentially more accurate spatially within that smaller soundstage. Bass is really good on HD560s, well defined, and soundstage on 560s is certainly wider and more integrated between the two cups than the HD600.....HD600 is terrible for soundstage in my experience, worst of my headphones, but it does everything else really well (apart from average bass). All my experience listed here is when EQ'd to Harman. I would definitely recommend you the HD560s over the HD600. Re the RTings HD560s review, I reckon they had an unlucky sample with poor channel matching as solderdude's diyaudioheaven website showed HD560s to have good channel matching, and if you look at the graph showing the 3 measurements that I posted earlier you can see that there is very close & strong agreement between all the measurements between 100-10000Hz and even above that, so I would expect that would mean that channel matching is also likely to be good.....having said that Oratory said channel matching was average on his sample he tested, but that seems to go against the strong agreement between all the 3 different measurements I showed.

 

[ADU]

Thanks for the reply there. So by subtracting the Diffuse Field Target from the Harman Target, what is it that you are wanting to show/achieve, what does it prove? It's only showing the difference between the two targets, and you seem to be trying to relate that graph showing the difference directly to graphs of speaker sound power....I'm not really sure how that is possible to compare them, I'm not joining the dots there at all....it seems like apples to oranges. It seems that you're relating them simply because they look similar, but just because they look similar doesn't mean a thing....I'm missing your logical step of why/how you're comparing the two.

I'm not sure I can answer all of the above yet, Robbo. Because some of this still involves a little speculation on my part. The similarities between the Harman curve and the DF responses of some headphones, and the sound power responses of some loudspeakers is one aspect of why this is important to me though. But its not the only aspect.

After looking at this subject from just about every direction, I've reached the conclusion that, for me personally, diffuse field is the better method of analyzing frequency response data for my needs, rather than using raw measurements. Both because of its more system-independent nature, and because of its correlation to the use and measurements of other transducers than just headphones. I don't think I'm the only one who feels this way btw. I believe there is in fact a whole community of headphone users and measurers, speaker designers, engineers, etc. for whom diffuse field is a bit like a second language. That does not mean I will not use both types of data though. Because, imo, the raw measurements can still have some value within certain confined contexts.

I think Tyll H probably came to a similar conclusion though on this, after his somewhat unsuccessful experiences in trying to adapt the Harman target to his data. And I think he eventually probably realized why the diffuse field measurements would likely have been alot more useful and valuable in this regard than the less-standardized independent of direction function that he chose for all of his compensated headphone plots. It seems that way to me from some of things he was saying not long before he decided to leave Inner Fidelity anyway. And the company eventually got dissolved or absorbed back into its parent company, Stereophile.

Until someone comes up with a better standard though, that can be applied more liberally than to just one specific closed type of measurement system, DF is the better way to go, imho. There are obviously some others out there though, for whom other factors and considerations may be more important, that may well have a different POV on all this. For some very good reasons. Including I suspect some monetary ones. And fwiw, I really have nothing against that. People need to make a living after all.

If, for example, you've tried some of the various EQ solutions that are out there, which are based on the Harman curve (as it seems you have done). And have found those results to be mostly to your liking, then most of what I'm talking about here will probably make no sense. And just seem like pointless mental masturbation. There are alot of folks that I've run across on forums like this who seem to feel this way. And for most of these folks, the "Harman curve" is all they need. As Steve Guttenberg expressed, some of those kinds of folks may not even give a rip about how their headphones measure! And wonder what all the fuss re the Harman curve is all about.

I've tried most of the Harman-based solutions for headphones though, and have not been as impressed or satisfied with the results as alot of other folks. So I'm still looking for some better solutions to the question of what represents a "neutral" and "accurate" response. Both in headphones and in loudspeakers. Because they really go sort of hand and in hand imo. And I think diffuse field is one of the potential common threads between the two... That's what all of the measurements seem to be telling me anyway.

So yes, there is a bit of conjecture and speculation involved here, in drawing the conclusion that the diffuse field response of headphones and the sound power of loudspeakers are related in some meaningful way. That conjecture is based on alot of very hard data and measurements though. And also on Harman's own measurements and subjective testing. So I think it's probably worth looking at.

The only other alternatives that I can see to this that could tell us more from a scientific standpoint would be either more double-blind subjective tests that would involve people listening to both headphones and speakers at the same time, and then giving their impressions on which headphones are the best matches to the speakers. Or in-ear measurements made from a variety of speakers in "typical" (whatever that is) semi-reflective home listening spaces.

This, in and of itself, involves alot of assumptions though, that many people still don't fully agree on. Like the notion that it's even possible to somehow replicate the sound or response of a pair of loudspeakers in a pair of headphones. And what are the best ways to go about that. Or the idea that "room sound" is also the best model for a headphone's sound. So more research is needed in these areas as well imo.

Most of the above involve time and money and wherewithal that I do not personally possess. So until there is a standard cross-platform method for making in-ear measurements of loudspeakers and comparing them to headphones, diffuse field and sound power is pretty much all I've got to work with, to try to find my own better answers to some of the above questions!

The cross-platform aspect of this is important imo. Because that makes it open to more than just one vendor and manufacturer. And to more users as well. Which measns more choices and options for people like me. So if the solution doesn't fit that basic criteria, then I'm much less interested in it. I will look at the raw GRAS measurements. But until you can put those into a context that I can apply more broadly (such as the diffuse field), they'll be of much less value and use to someone like me. Because I can't as easily compare the raw data to similar measurements made elswehere, on different systems. Like the new HBK 5128 rig, for example. To try to draw some more in-depth conclusions from it.

This is not just something that I've brought up here, btw. I have had some pretty long (and also continuing) conversations about this with some of the folks on other forums as well. And also with some others who do headphone measurements and reviews (like Rtings). And in other contexts as well. So its not as though I'm singling ASR out on this.

I supect that the folks at GRAS are enjoying the situation as it currently stands though. Because as long as people are locked into just one method of using, analyzing and doing measurements with their rig and the Harman curve for compensation, they pretty much have a corner on the headphone measurement market. And can sell there equipment to basically anyone with a few bucks that wants to start measuring and reviewing headphones. That's both the upside and also (imho) a potential future downside to what they're doing.

The GRAS system, and most others like it, are still based on some rather antiquated and inaccurate technology, that was developed decades ago, mostly for the design of hearing aids. So it is, imho, far from a perfect solution for doing this kind of thing. And there are all manner of potential problems with trying to accurately analyze the data from this type of technology for different kinds of transducers, in, on, and out of the ear. I'm sure that the folks at GRAS also know this, btw, and are currently developing some new and better systems to compete with HBK's 5128 rigs, with more anatomically accurate human geometry.

This video was prepared for HBK by Jude of HF. So it is in a technical sense, "marketing". And should be viewed in that context. But it explains a few of the differences between the older 711 systems which are currently in used by most of the noteworthy graphers in this field, and the 5128 system. And why they could potentially be of some importance...

Until Amir happened to mention it, I did not know (or had forgotten) that Harman had actually picked up this new system. If Harman continues its work though in the area of headphones, then I suspect that some or possibly all of that work could be done on the new HBK 5128 system. Which could present a bit of a problem down the road for some of the graphers and other reviewers who are currently using the GRAS systems. Because rather than being designed specifically for their rigs, the next "Harman curve" could well be designed instead for the new 5128 rigs.

For those with the $$ and wherewithal to afford a new 5128 rig, that's not likely to be a huge problem. It will still take some time, effort, and research though to get up and running reliably with a new system though. Which involves costs and investments that go beyond just the few bucks for a new mannequin, with some better ears and microphones in it.

Alot of folks simply won't be able to afford the new systems though. And may fall to the wayside as a result.... unless Harman also puts the new data into a format that's more compatible with other platforms, such as the diffuse field format.

To be clear, I am all in favor of efforts to also find and develop a new and better method of standardizing and translating this type of info across multiple platforms. I think that is very much within the realm of what's doable. Until this happens though, and there's a more standardized and reliable method for the developers of these types of rigs to include a "neutral speakers in a typical listening room" type of compensation curve, diffuse field measurements are probably the next best thing... If you don't simply want to adopt a closed system, like the GRAS/Harman curve system.

Since I've now rambled on again for several type-written pages on this subject, I'm going to stop there for now. And pick this up again when maybe I have a few more visual aids to offer, to maybe help with better explaining some of this. Hopefully some of the above has begun to give a little better picture though on where I, and probably also a few others, may be coming from on some of this.

 

[Robbo99999]

I'm not sure I can answer all of the above yet, Robbo. Because some of this still involves a little speculation on my part. The similarities between the Harman curve and the DF responses of some headphones, and the sound power responses of some loudspeakers is one aspect of why this is important to me though. But its not the only aspect.

After looking at this subject from just about every direction, I've reached the conclusion that, for me personally, diffuse field is the better method of analyzing frequency response data for my needs, rather than using raw measurements. Both because of its more system-independent nature, and because of its correlation to the use and measurements of other transducers than just headphones. I don't think I'm the only one who feels this way btw. I believe there is in fact a whole community of headphone users and measurers, speaker designers, engineers, etc. for whom diffuse field is a bit like a second language. That does not mean I will not use both types of data though. Because, imo, the raw measurements can still have some value within certain confined contexts.

I think Tyll H probably came to a similar conclusion though on this, after his somewhat unsuccessful experiences in trying to adapt the Harman target to his data. And I think he eventually probably realized why the diffuse field measurements would likely have been alot more useful and valuable in this regard than the less-standardized independent of direction function that he chose for all of his compensated headphone plots. It seems that way to me from some of things he was saying not long before he decided to leave Inner Fidelity anyway. And the company eventually got dissolved or absorbed back into its parent company, Stereophile.

Until someone comes up with a better standard though, that can be applied more liberally than to just one specific closed type of measurement system, DF is the better way to go, imho. There are obviously some others out there though, for whom other factors and considerations may be more important, that may well have a different POV on all this. For some very good reasons. Including I suspect some monetary ones. And fwiw, I really have nothing against that. People need to make a living after all.

If, for example, you've tried some of the various EQ solutions that are out there, which are based on the Harman curve (as it seems you have done). And have found those results to be mostly to your liking, then most of what I'm talking about here will probably make no sense. And just seem like pointless mental masturbation. There are alot of folks that I've run across on forums like this who seem to feel this way. And for most of these folks, the "Harman curve" is all they need. As Steve Guttenberg expressed, some of those kinds of folks may not even give a rip about how their headphones measure! And wonder what all the fuss re the Harman curve is all about.

I've tried most of the Harman-based solutions for headphones though, and have not been as impressed or satisfied with the results as alot of other folks. So I'm still looking for some better solutions to the question of what represents a "neutral" and "accurate" response. Both in headphones and in loudspeakers. Because they really go sort of hand and in hand imo. And I think diffuse field is one of the potential common threads between the two... That's what all of the measurements seem to be telling me anyway.

So yes, there is a bit of conjecture and speculation involved here, in drawing the conclusion that the diffuse field response of headphones and the sound power of loudspeakers are related in some meaningful way. That conjecture is based on alot of very hard data and measurements though. And also on Harman's own measurements and subjective testing. So I think it's probably worth looking at.

The only other alternatives that I can see to this that could tell us more from a scientific standpoint would be either more double-blind subjective tests that would involve people listening to both headphones and speakers at the same time, and then giving their impressions on which headphones are the best matches to the speakers. Or in-ear measurements made from a variety of speakers in "typical" (whatever that is) semi-reflective home listening spaces.

This, in and of itself, involves alot of assumptions though, that many people still don't fully agree on. Like the notion that it's even possible to somehow replicate the sound or response of a pair of loudspeakers in a pair of headphones. And what are the best ways to go about that. Or the idea that "room sound" is also the best model for a headphone's sound. So more research is needed in these areas as well imo.

Most of the above involve time and money and wherewithal that I do not personally possess. So until there is a standard cross-platform method for making in-ear measurements of loudspeakers and comparing them to headphones, diffuse field and sound power is pretty much all I've got to work with, to try to find my own better answers to some of the above questions!

The cross-platform aspect of this is important imo. Because that makes it open to more than just one vendor and manufacturer. And to more users as well. Which measns more choices and options for people like me. So if the solution doesn't fit that basic criteria, then I'm much less interested in it. I will look at the raw GRAS measurements. But until you can put those into a context that I can apply more broadly (such as the diffuse field), they'll be of much less value and use to someone like me. Because I can't as easily compare the raw data to similar measurements made elswehere, on different systems. Like the new HBK 5128 rig, for example. To try to draw some more in-depth conclusions from it.

This is not just something that I've brought up here, btw. I have had some pretty long (and also continuing) conversations about this with some of the folks on other forums as well. And also with some others who do headphone measurements and reviews (like Rtings). And in other contexts as well. So its not as though I'm singling ASR out on this.

I supect that the folks at GRAS are enjoying the situation as it currently stands though. Because as long as people are locked into just one method of using, analyzing and doing measurements with their rig and the Harman curve for compensation, they pretty much have a corner on the headphone measurement market. And can sell there equipment to basically anyone with a few bucks that wants to start measuring and reviewing headphones. That's both the upside and also (imho) a potential future downside to what they're doing.

The GRAS system, and most others like it, are still based on some rather antiquated and inaccurate technology, that was developed decades ago, mostly for the design of hearing aids. So it is, imho, far from a perfect solution for doing this kind of thing. And there are all manner of potential problems with trying to accurately analyze the data from this type of technology for different kinds of transducers, in, on, and out of the ear. I'm sure that the folks at GRAS also know this, btw, and are currently developing some new and better systems to compete with HBK's 5128 rigs, with more anatomically accurate human geometry.

This video was prepared for HBK by Jude of HF. So it is in a technical sense, "marketing". And should be viewed in that context. But it explains a few of the differences between the older 711 systems which are currently in used by most of the noteworthy graphers in this field, and the 5128 system. And why they could potentially be of some importance...

Until Amir happened to mention it, I did not know (or had forgotten) that Harman had actually picked up this new system. If Harman continues its work though in the area of headphones, then I suspect that some or possibly all of that work could be done on the new HBK 5128 system. Which could present a bit of a problem down the road for some of the graphers and other reviewers who are currently using the GRAS systems. Because rather than being designed specifically for their rigs, the next "Harman curve" could well be designed instead for the new 5128 rigs.

For those with the $$ and wherewithal to afford a new 5128 rig, that's not likely to be a huge problem. It will still take some time, effort, and research though to get up and running reliably with a new system though. Which involves costs and investments that go beyond just the few bucks for a new mannequin, with some better ears and microphones in it.

Alot of folks simply won't be able to afford the new systems though. And may fall to the wayside as a result.... unless Harman also puts the new data into a format that's more compatible with other platforms, such as the diffuse field format.

To be clear, I am all in favor of efforts to also find and develop a new and better method of standardizing and translating this type of info across multiple platforms. I think that is very much within the realm of what's doable. Until this happens though, and there's a more standardized and reliable method for the developers of these types of rigs to include a "stereo speakers in a typical room" type of compensation curve, diffuse field measurements are probably the next best thing... If you don't simply want to adopt a closed system, like the GRAS/Harman curve system.

Since I've now rambled on again for several type-written pages on this subject, I'm going to stop there for now. And pick this up again when maybe I have a few more visual aids to offer, to maybe help with better explaining some of this. Hopefully some of the above has begun to give a little better picture though on where I, and probably also a few others, may be coming from on some of this.

Hey ADU, thanks for that, nice post. Based on what you've said there and your searching for a Target Curve that matches a good subjective experience for you, I really think you should try the Impulcifier Project, it sounds like you have the intelligence & tenacity to do it: https://github.com/jaakkopasanen/Impulcifer
I would have tried it already if I wouldn't have gotten good results from a simple Harman EQ.....that Impulcifier Project is literally there to create a personalised target for you to emulate speakers in a room based on your own anatomy.....so a DIY personalised Harman Curve if you like (given that of course the Harman Curve attempts to do that but through an average anatomy whilst adding some preference studies to the curve too). Point is though, Impulcifier Project will create a unique curve for your own anatomy. If you do it then report back, because I'm interested in it....kinda wish I didn't get such pleasing results with the bog standard Harman Curve!

Regarding DF as a means of a standard & for comparing headphone measurements.....sorry man, I'm still not convinced, there's too many inaccuracies associated with those conversions as I outlined in previous posts (as well as one other poster). Try that Impulcifier & stop faffing around with DF!

You're right that if Harman adopts the 5128 as a new standard then that means that current reviewers and amazing contributors like Oratory are put a step back.....but that's life, if 5128 is indeed better then so be it, but I'm not convinced, ultimately it all boils down to the same line when it comes to headphone "simulation", the dummy is only as accurate as it's accuracy of portraying a convincing average of the human populace and then you've still got the obvious statistical human population variance vs the average (no matter how accurate the average is portrayed). To me the future of headphones is easy personalisation/incorporation of HRTF into headphone design through easy to use commercialised EQ apps.....but we're not there yet, Impulcifier Project is of that ilk yet is complicated. Try Impulcifier & stop faffing with that DF man!!

 

[Deleted member 16543]

I'm with Robbo, as I've stated earlier, in not being convinced that DF is such a converging measurement method in terms of stability and accuracy.
Speaking of impulcifier and the likes, here's probably the "grandfather" of this kind of methods (or at least one of the old wisemen).
http://www.davidgriesinger.com/

 

[amirm]

Until Amir happened to mention it, I did not know (or had forgotten) that Harman had actually picked up this new system. If Harman continues its work though in the area of headphones, then I suspect that some or possibly all of that work could be done on the new HBK 5128 system. Which could present a bit of a problem down the road for some of the graphers and other reviewers who are currently using the GRAS systems. Because rather than being designed specifically for their rigs, the next "Harman curve" could well be designed instead for the new 5128 rigs.

Harman was one of the first customers of 5128 so it has been years and nothing has come about for it. They consider their headphone research completed and have moved on to spatial audio so I doubt anything will materialize.

I have told B&K that they need to formally contract Harman and see if they can get them to do this work as it is much needed. But I doubt that this is going to happen.

At the end, the fixture itself is not important. What is important is to have correlated target curve and listening test results which we have for Gras but not B&K 5128. As to Jude, there is criticism of his comparison of GRAS and B&K 5128. The few measurements he has published with 5128 have been of little value due to lack of target curve and any reference to judge the results. He can only use it to compare two products, not produce any measurements of a single headphone and speak to its audibility. In general, he uses measurement gear like women use jewelry. It is all for show to say "I can measure if I want" but apparently he never wants as such measurements despite being promised, rare come about. So don't look to him having a unit as any measure of usefulness. I would buy it in a heartbeat if I could get useful information out of it but it simply is not possible.

 

[Deleted member 16543]

At the end, the fixture itself is not important. What is important is to have correlated target curve and listening test results which we have for Gras but not B&K 5128.

I would say that the fixture IS very important in order to get as accurate and meaningful measurements as possible.
The fact that most of the work has been done on a 'lesser' type of rig doesn't take away the inherent merit of a more anatomically accurate rig.
But.. it is also true that if there's no research done with the better rig, than that that rig is pretty much worthless. At least until somebody rolls up their sleeves and does some more research with it.

I think a few of us were hopeful that this would be the case for ASR. Without any moral judgment whatsoever (since doing some research is easier said than done!), I think it's fair to say that the beaten path was instead taken as a strategy.

 

[amirm]

I would say that the fixture IS very important in order to get as accurate and meaningful measurements as possible.
The fact that most of the work has been done on a 'lesser' type of rig doesn't take away the inherent merit of a more anatomically accurate rig.
But.. it is also true that if there's no research done with the better rig, than that that rig is pretty much worthless. At least until somebody rolls up their sleeves and does some more research with it.

Well, the other rig is not "better" until such research is done. If correlation between listening tests with B&K measurements is lower than GRAS, then that is that. The fact that it costs more and they did more doesn't help that.

As an example, in the case of B&K 5128, it is darn near impossible to test IEMs with it. I could get any of them to stay in there and seal. As a result, the measurements are all over the place. With GRAS 45C, they fit like a glove! The choice of material makes a big difference in how much friction there is for the IEM tips to stay seated in them.

 

[ADU]

I would say that the fixture IS very important in order to get as accurate and meaningful measurements as possible.
The fact that most of the work has been done on a 'lesser' type of rig doesn't take away the inherent merit of a more anatomically accurate rig.
But.. it is also true that if there's no research done with the better rig, than that that rig is pretty much worthless. At least until somebody rolls up their sleeves and does some more research with it.

I think a few of us were hopeful that this would be the case for ASR. Without any moral judgment whatsoever (since doing some research is easier said than done!), I think it's fair to say that the beaten path was instead taken as a strategy.

Fwiw, I am not judging your decisions on this^ either, Amir. And I think you probably made the best choice for your and ASR's present needs, demands, and circumstances.

And I also understand the skepticism that many of you are expressing on alot of this, with regard to the diffuse field and whatnot. (This "ain't my first rodeo", as some of us midwesterners like to say.) If you feel any of this is getting out of hand though, that's ok. Because I'm just a newbie and guest here after all. And not really here to try to convince anyone to do something they don't want.

I hope some of you can keep just a little bit of an open mind on some of this though, so we can continue this conversation. And that you'll allow me to post a few more examples, to try to maybe better illustrate where I'm coming from on some of this.

So far, I haven't had much luck finding a DF compensation curve to use with the raw 45C measurement data here, so I may try to give GRAS a ring about that during the week. I'm assuming the data would probably have to come from the 45C coupler installated in some type mannequin though, to incorporate the head and torso effects on the diffuse sound field, as well as pinna and simulated ear canal effects. This is assuming GRAS has made such measurements.

As I mentioned previously though, I've found what I think is probably an appropriate DF compensation curve for the HBK 5128. So I can start doing some DF plots from that to compare to both the DF version of the Harman target, and Oratory's DF measurements, and also the ASR sound power spinorama plots. So that might at least be a starting place for some type of a neutral 5128 reference target.

I'm still sort of at the beginning stages of all though (even though this kind of thing has been in the back of my head for quite a long time).

 

[Robbo99999]

I would say that the fixture IS very important in order to get as accurate and meaningful measurements as possible.
The fact that most of the work has been done on a 'lesser' type of rig doesn't take away the inherent merit of a more anatomically accurate rig.
But.. it is also true that if there's no research done with the better rig, than that that rig is pretty much worthless. At least until somebody rolls up their sleeves and does some more research with it.

I think a few of us were hopeful that this would be the case for ASR. Without any moral judgment whatsoever (since doing some research is easier said than done!), I think it's fair to say that the beaten path was instead taken as a strategy.

Well, the other rig is not "better" until such research is done. If correlation between listening tests with B&K measurements is lower than GRAS, then that is that. The fact that it costs more and they did more doesn't help that.

As an example, in the case of B&K 5128, it is darn near impossible to test IEMs with it. I could get any of them to stay in there and seal. As a result, the measurements are all over the place. With GRAS 45C, they fit like a glove! The choice of material makes a big difference in how much friction there is for the IEM tips to stay seated in them.

Yep, 5128 doesn't mean anything till research is done on it to correlate to "speakers in a room" combined with "preference", and I believe both of those variables would need to be incorporated, exactly as Harman did the the research (for one you don't experience bass in the same way through speakers that you do in headphones so preference has to come into it on that aspect). GRAS & Harman is the best we got until something more effective & easy come our way......... @ADU you still got the Impulcifier to try, that's the next best thing to try vs Harman I reckon! (and maybe you need to add extra bass to that if not already accounted for....don't know enough about Impulcifier)

 

[ADU]

Harman was one of the first customers of 5128 so it has been years and nothing has come about for it. They consider their headphone research completed and have moved on to spatial audio so I doubt anything will materialize.

I have told B&K that they need to formally contract Harman and see if they can get them to do this work as it is much needed. But I doubt that this is going to happen.

Fwiw, this is a recent conversation with Todd Welti from May 21 of this year.

It mostly deals with loudspeakers and sub-woofers. But he also talks about his work on headphones towards the end as well. And mentions several times that he believes that there is alot more research to be done on this subject, that he's interested in doing. I've pre-cued the video to where the discussion on headphones begins, at around 55:30.

I'm not sure if Sean Olive has the same opinions on this. But thought that I heard intimations along some similar lines from him as well not too long ago. (Though I don't have any links I can post on that.)

Todd is obviously a very bright and well-informed guy. So the rest of the vid is also worth a listen. Much of it is related to the use of multiple subs, which is an area that I guess Todd knows alot about, and did some pioneering work in.

 

[ADU]

@ADU you still got the Impulcifier to try, that's the next best thing to try vs Harman I reckon! (and maybe you need to add extra bass to that if not already accounted for....don't know enough about Impulcifier)

Yes, bro. I got your message on that above, loud and clear!

And I'm not quiet finished tryin to bring you over to the diffuse-field side! (I'm kidding on that btw.) I agree that there are most definitely a few wrinkles to be ironed out on this though.

 

[Robbo99999]

Fwiw, this is a recent conversation with Todd Welti from May 21 of this year.

It mostly deals with loudspeakers and sub-woofers. But he also talks about his work on headphones towards the end as well. And mentions several times that he believes that there is alot more research to be done on this subject, that he's interested in doing. I've pre-cued the video to where the discussion on headphones begins, at around 55:30.

I'm not sure if Sean Olive has the same opinions on this. But thought that I heard intimations along some similar lines from him as well not too long ago. (Though I don't have any links I can post on that.)

Todd is obviously a very bright and well-informed guy. So the rest of the vid is also worth a listen. Much of it is related to the use of multiple subs, which is an area that I guess Todd knows alot about, and did some pioneering work in.

They shouldn't stop work on it, but if they're basing it purely on an average, which they have to do if they're not scanning an individual's HRTF then all they're doing is fine-tuning the average to be more representative.....yet still the population's deviation from the average will exist. Personally, in terms of creating a one-size-fits-all target curve, I think they're already there & there's nothing more to really create.....instead I think they need to start along the lines of easy personalisation based on an individuals anatomy (scanning, etc)......I think they're at the pinnacle of their current approach.

 

[ADU]

Fwiw, the individualizing of the measurements doesn't really concern me that much, because the DF compensation is supposed to remove that, and take the individual ear and rig-specific effects largely out of the equation... That's the way it's supposed to work anyway. In actual practice though, this process may not always work quite as smoothly as one would hope. And I may have a couple of examples of this that I can post.

I think there are probably some ways of dealing with some of the rig-to-rig variances with DF compensation though. My (rather crude) solution to that problem right now is simply to use the DF target curve(s) more as a guide for selecting some individual headphone measurements that are in the same ballpark tonally. And then trying to figure out some sort of an average based on the measurements of those HPs. It's a bit messy, and work-aroundy (like just about everything I'm doing along these lines). But I think there might at least be some potential there.

Whether this kind of thing would actually yield a noticeably or substantially better result for an average listener than simply doing a rough EQ to match the raw Harman target is debatable though.

 

[MayaTlab]

They shouldn't stop work on it, but if they're basing it purely on an average, which they have to do if they're not scanning an individual's HRTF then all they're doing is fine-tuning the average to be more representative.....yet still the population's deviation from the average will exist. Personally, in terms of creating a one-size-fits-all target curve, I think they're already there & there's nothing more to really create.....instead I think they need to start along the lines of easy personalisation based on an individuals anatomy (scanning, etc)......I think they're at the pinnacle of their current approach.

I'd like to learn more about headphones transfer function per se. Todd mentioned his experiments on leakage which is a part of it. The way I see it, if we can't design headphones which HPTF can be reliably predicted for each individual (and perhaps compensated for), whether it's because of sample variation, positional variation, pad compression, ear anatomy, etc., it might be difficult to convincingly implement an HRTF profile, or am I wrong ? I mean, the directional part of the HRTF would be correctly fed, but the constant part not - perhaps to the point of ruining the overall result (I can't imagine that a basal FR at the drum that's completely out of whack wouldn't have an effect on our capacity to properly interpret HRTF cues, even individualised) ?
I can only speak of my subjective experience here, but I never managed to coax anything in terms of front / back / up / down presentation with object based formats and generic HRTF profiles without EQing the basal FR of my headphones to a reasonably satisfying (and probably personal to some degree) point. And applying presets was not enough - particularly since most of the remaining differences were in the HRTF modulation area.

 

[ADU]

Well, the other rig is not "better" until such research is done. If correlation between listening tests with B&K measurements is lower than GRAS, then that is that. The fact that it costs more and they did more doesn't help that.

As an example, in the case of B&K 5128, it is darn near impossible to test IEMs with it. I could get any of them to stay in there and seal. As a result, the measurements are all over the place. With GRAS 45C, they fit like a glove! The choice of material makes a big difference in how much friction there is for the IEM tips to stay seated in them.

Both interesting points, Amir. My principal interest though would be in the over-ear 5128 measurements. Because I think they have the best potential for some degree of standardization. IEMs are obviously a bit more of a challenge in this regard.

Maybe HBK will find a way of sorting out the issues you mention above though. Since it seems that better IEM measurements were probably one of their goals for this new system, from Jude's remarks in the above video.

 

[ADU]

I'd like to learn more about headphones transfer function per se. Todd mentioned his experiments on leakage which is a part of it. The way I see it, if we can't design headphones which HPTF can be reliably predicted for each individual (and perhaps compensated for), whether it's because of sample variation, positional variation, pad compression, ear anatomy, etc., it might be difficult to convincingly implement an HRTF profile, or am I wrong ? I mean, the directional part of the HRTF would be correctly fed, but the constant part not - perhaps to the point of ruining the overall result ?

Welcome aboard, Maya!

 

[Robbo99999]

Fwiw, the individualizing of the measurements doesn't really concern me that much, because the DF compensation is supposed to remove that, and take the individual ear and rig-specific effects largely out of the equation... That's the way it's supposed to work anyway. In actual practice though, this process may not always work quite as smoothly as one would hope. And I may have a couple of examples of this that I can post.

I think there are probably some ways of dealing with some of the rig-to-rig variances with DF compensation though. My (rather crude) solution to that problem right now is simply to use the DF target curve(s) more as a guide for selecting some individual headphone measurements that are in the same ballpark tonally. And then trying to figure out some sort of an average based on the measurements of those HPs. It's a bit messy, and work-aroundy (like just about everything I'm doing along these lines). But I think there might at least be some potential there.

Whether this kind of thing would actually yield a noticeably or substantially better result for an average listener than simply doing a rough EQ to match the raw Harman target is debatable though.

DF or any generalised target curve can never ever remove the anatomical differences between all of us that make our ideal target curve unique - it's exactly that, our own ideal target curve is entirely unique (for all intents & purposes............of course very similar looking individuals, eg identical twins will have almost identical HRTF's & hence identical target curves but that's the extent of it). The only way to get to true sound is to have an individualised target based on how a known signal (frequency sweep) is received at our own eardrum.......and yes the manner in which any given model of headphone would have to be EQ'd, it would be a different EQ for each model of headphone, exactly because it's one thing to have a personalised target, but it's another to predict a headphone's response to your own anatomy (HPTF)......if you really want to dive into it & perfect it, then you have to solve both of those questions.

@MayaTlab , yes, as alluded to in previous paragraph, they'd need to scan our HRTF and also take our HPTF into account (I suppose they would have already taken this into account as part of their scanning process, but really that would depend on the intricacies of both the inner & outer ear, and I'm not sure how they gain visibility on the inner ear canal with "scanning"). I'm not an acoustician, but there are challenges to be had here I think. I still think it's clear that progress can really only be made via individualisation to the individual (not an average as Harman currently does) based on their anatomy and their obvious good understanding of what makes a good speaker in a room (along with their trends in preference studies of headphone response that show a desire for more bass in headphones vs speakers beyond the "dummy in room measurement" which would also translate to a personalised measurement in terms of additional bass preference).

I do still think that their Harman Headphone Target is "good" and enjoyable, just I know it's not the pinnacle of what's possible.

I'd like to learn more about headphones transfer function per se. Todd mentioned his experiments on leakage which is a part of it. The way I see it, if we can't design headphones which HPTF can be reliably predicted for each individual (and perhaps compensated for), whether it's because of sample variation, positional variation, pad compression, ear anatomy, etc., it might be difficult to convincingly implement an HRTF profile, or am I wrong ? I mean, the directional part of the HRTF would be correctly fed, but the constant part not - perhaps to the point of ruining the overall result (I can't imagine that a basal FR at the drum that's completely out of whack wouldn't have an effect on our capacity to properly interpret HRTF cues, even individualised) ?
I can only speak of my subjective experience here, but I never managed to coax anything in terms of front / back / up / down presentation with object based formats and generic HRTF profiles without EQing the basal FR of my headphones to a reasonably satisfying (and probably personal to some degree) point. And applying presets was not enough - particularly since most of the remaining differences were in the HRTF modulation area.

 

[ADU]

DF or any generalised target curve can never ever remove the anatomical differences between all of us that make our ideal target curve unique - it's exactly that, our own ideal target curve is entirely unique (for all intents & purposes............of course very similar looking individuals, eg identical twins will have almost identical HRTF's & hence identical target curves but that's the extent of it). The only way to get to true sound is to have an individualised target based on how a known signal (frequency sweep) is received at our own eardrum.......and yes the manner in which any given model of headphone would have to be EQ'd, it would be a different EQ for each model of headphone, exactly because it's one thing to have a personalised target, but it's another to predict a headphone's response to your own anatomy (HPTF)......if you really want to dive into it & perfect it, then you have to solve both of those questions.

Howdy again, Robbo99999.

I am not as into to all the fancy new DSP and HRTF-related stuff for headphones that's out there right now as some other folks seem to be. So you (and Maya) may be much better informed in these areas than I am, Robbo. And feel free to correct me again, if you think I'm looking at any of this incorrectly, or through too rosey-colored glasses. That does not mean I have anything against such things though, or those who are into it. (Quite the contrary, in fact.)

My primary objective though in this is NOT to determine the best sound signature, or HpTF (if I'm understanding that term correctly) for my own particular ears. It's to determine, as well as can possibly be done within the limited means available to me, what a neutral, accurate HRTF-independent response curve should look like for a pair of over-ear headphones. That is, one which is NOT reliant on customization to an individual person's particular ears. And I believe that using DF compensation is one potential way to accomplish this (albeit a somewhat imperfect one).

That isn't the whole point of why I'm using DF compensation. But it is certainly one of them.

Alot of folks in the headphone community unfortunately seem to be under the false impression that there is only one diffuse field curve (like in The Highlander... THERE CAN BE ONLY ONE!!!!! Aaaaaahhh!) that you can use on any measurement system to get that groovy, bright "diffuse field sound", baby! But it dudn't quite work like that.

The diffuse field is not a "generalized target" like the Harman curve in this sense. It's a compensation or correction or calibration curve which is based on a common or reference stimulus. And not one particular curve... Which means that the DF compensation is different for every different HATS rig. And tailored specifically to each of their HRTFs.

It seems like you sort of get this, Robbo. But some of the objections you keep raising lead me to think that maybe you're not quite seeing the connection between this, and HRTF compensation or correction. As far as I'm concerned, DF compensation is HRTF compensation or correction! Because it's different for each individual rig.

You can certainly make the argument that it's poorly standardized. And not reliable or accurate because the DF measurements may not all be performed exactly the same way, and with the same type of equipment, on different rigs by different manufacturers. And that's something I'd probably accept and have to agree with to some extent. But arguing that it is somehow not a form of HRTF compensation or correction seems a bit of a stretch to me. (It may not be a particularly GOOD form of HRTF compensation!... But it is HRTF compensation in my book.)

Maybe I'm still missing or not fully grasping something in what you're saying though. And the confusion is more on my end on this. And if you think that's the case, then I'm asking you to please try to clarify it a little better, or point it out to me, if you can.

There are some other potential methods of performing this type of HRTF compensation, btw, which Jaakko apparently used to adapt the Harman target to several different sets of measurements made by different graphers for his AutoEQ project. And I believe this was actually done by comparing the measurements of the same set of headphones (I don't know how many though) on the different rigs. And then computing an average correction curve for each rig, based on the differences in the measurements between that rig, and the GRAS rigs used by Harman, Crin and Oratory.

This is a technique I also experimented with briefly, back when Tyll was trying to adapt the Harman curve to his raw Inner Fidelity measurements. I didn't have a whole lot of time to dedicate to plotting back then though, so it was extremely slow going. And when Jaakko posted his AutoEQ results, it made most of my efforts along a similar lines largely irrelevant. And I don't think I ever got to the point where I shared any of my results on this with anyone at IF. (I think a few us of may have discussed it as one possible solution though. And perhaps that's where the AutoEQ project began? It is very possible that Jaakko is the one that first brought this idea up though, because he was working on something along those lines already... I honestly don't remember though.)

It employs basically the same principles as diffuse field HRTF compensation though. Namely, using a common sound stimulus (a specific model of headphone in this case, instead of a diffuse sound field) to compare, and then calculate the differences in response between two different measurement rigs. And then to use that difference curve to adapt a target curve developed on one rig for use by another.

It may not work 100% perfectly all the time (as Jaakko's AutoEQ project has demonstrated), possibly due to unit variation, differences in measurement technique, seal, and so forth. But, in theory, the principal is sound. And is also something that could potentially be used to adapt a reference response curve, like the Harman target, for use with measurements made on other systems. Including possibly also the kind of in-ear measurement systems that Maya is using.

A slight drawback to using a headphone or headphones for this reference stimulus is that it is perhaps not quite as easy to relate back to something a bit more concrete, like the sound power response of a loudspeaker, as DF compensation seems to be... Unless you're willing to accept on faith and evidence that something like the Harman curve, or a headphone with a similar response such as the AKG K371, is an appropriate and sufficient model for something like that... Which I regrettably do not.

There are some headphones (or combinations of them) that I think might get pretty close to this though. And one of the advantages of an approach like this is that the stimulus being used as a reference for adapting other measurements is a bit closer in character and response to the other stimuli or sound sources (headphones, presumably) that you're intending to measure and correct with this info, than the rig's DF curve would be. (I assume that this is probably one of reasons that Jaakko preferred this approach to using DF or FF compensation for his AutoEQ project. Though perhaps he just couldn't find the DF/FF sound field data for all the different measurement systems in his project.)

If you do believe that the Harman curve (or something similar, such as the K371) is the last word, or at least a sufficient model for a "room sound" (or whatever sound it is that you're personally going for), then calibrating your measurements to that target by comparing various other headphone measurements made on your own personal rig or in-ear setup to the measurements of the same headphones on a GRAS rig similar to Harman's may be enough to get the job done.

I apologize, btw, if this has again been too much to take in, in just one post. But it seems as though some here may already have been down some similar roads. And already formed some other opinions on this. And since we seem to moving a bit more out of the kiddy pool, and more into the adult swim, I figured maybe some of this was ok to throw out there. If I'm somehow way off base though on this, then please feel free to say so. Or tell me to go jump in a lake!

I don't buy into the idea that the response of an over-ear headphone has to be specifically tailored to each person's individual HRTF to sound acceptable or accurate though. Because it just doesn't hold that much water in my opinion. (Hmm... lots of aquatic references today, must be the Olympics or somethin.) And I think perhaps your experience with using the Harman target for compensation, Robbo, seems to demonstrate that to some degree... At least maybe for you.

It's also possible that this all boils down to just questions of degree though, and determining which differences in these things actually make a difference to a person's audio enjoyment. Those are probably also some valid things to consider as well imo.

Just to be clear though, I'm not really saying that individual HRTFs should simply be ignored, or do not matter. What I'm sayin is that I believe that diffuse field HRTF compensation, and calibration to a common stimulus by this method (or possibly some other similar method using several pairs of well-designed & well-fitting headphones), will probably be adequate for my purposes for the time being. And enough to factor out the differences between different rigs or HRTFs, so that they're not a significant impediment to determining a general neutral response for a pair of circumaural headphones.

This is just my 2c on the subject though. And your mileage may certainly vary on this. (Another pretty long post folks, so my apologies for that.)

 

[ADU]

I have looked at the research that Maya has posted elsewhere, btw, that explored this idea of using the differences between multiple headphones to compute a correction curve for different individuals. And while I think there could be some slight problems with it. I wasn't really convinced by the study that the differences in the computed correction curves for some of the headphones were substantial or consistent enough in their irregularity to be of much consequence. Perhaps this was the sort of thing you were referring to with regard to HpTF?

Again, I also tried using this approach myself at one point, a number of years ago, to compare and compute the correction curves for several different headphones measured by both Harman and Tyll (or possibly some other measurement site at the time?). And while I did notice some fairly obvious variations in the correction curves between the different headphones on the two systems. I did not personally deem them to be significant enough to invalidate the approach. The whole point of using multiple stimuli or headphones like this was to improve the accuracy of the correction curve. And based on most of the plots I was doing along these lines, it seemed like the approach would've worked pretty ok.

One thing that also needs to be taken into consideration on this is the users' ability to "self-correct" the sound of different headphones on their own head, if they sound off, by shifting the headphones around a bit, or adjusting the seal, headband, and so forth. Which is something that's probably not so easily take into account in such a study.

If you are dealing with a more troublesome headphone design though, then determining an accurate correction or translation curve from something like that will likely be much more challenging. I believe Maya has also commented on this, and determined that some open headphones are probably more reliable for this type of thing, because they produce more consistent results between usages in her measurements. (Hopefully she'll correct me though, if that's not right.)

Here is the link to the study I was referring to above, and some of the relevant plots from it (courtesy of Maya's post in another topic).

https://www.aes.org/e-lib/browse.cfm?elib=16877

 

[amirm]

I'm not sure if Sean Olive has the same opinions on this.

I don't know about Sean's opinion but what I expressed came directly from him. Research needs to be justified in the context of their corporate interest so I appreciate that the company wants to get coverage in other areas than pure headphone tonality. Note that they have also moved on from speaker research as well even though there are open questions there (directivity, distortion, etc.).