Resolve's B&K 5128 Headphone Target - you can try the EQ's.....

[amirm]

If two guys can create a target for a headphone fixture in little time with so little effort, we would have to conclude that Harman and crew didn't know what they were doing! Hobby work like this hugely trivializes the research needed to find acoustic measurements that have a prayer of correlating with listening tests.

I will repeat what I told B&K folks after my evaluation of 5128: the only way this fixture would become useful is if they commissioned Harman to repeat their testing and produce a new target for it. And that this was a requirement if they wanted the fixture to have much usefulness. Without it, it is not a solution. It is a component needing the other parts to work.

If the above was done and hence we had a comparable solution with 5128 that we have with GRAS 45, then there would be no fragmentation. Anything else will create noise and confusion.

 

[solderdude]

Well.. I don't know. We have no idea how much time went into this and how long this was already a hot topic of debate.
Well... I have some idea as I have been following some of their discussions/ideas for a while.

The Harman research was based on speaker preference in their own room and the headphone research was simulated where it is not with these hobbyists.
That research is done properly as far as I can see.

I do not share the opinion that it hugely trivializes the Harman research. I see it as learning from and continuing on that research. Perhaps Harman will repeat their research seeing they own one too, and come up with a target of their own one day.
That conclusion (whether the hobbyists can actually create a valid target) can only be drawn after they finalized their curve and if their results differ drastically from what Harman found. I suspect the bass boost won't be similar though.

The GRAS 45 compensation is also needed to obtain results. The B&K guys obviously are not really very interested in getting the headphones right but only went for FF and DF targets which was easy as this is a standard which they can calibrate to with great accuracy.

That what creates confusion is because they involved the headphone community in listening tests. On one hand this creates confusion on the other hand it can provide some valuable insight from the interested headphone community instead of experimenting on salesmen, musicians, school going kids, audiophiles.

I wonder what @Sean Olive thinks of the endeavors of these guys. They do seem to have a line between them and bounce ideas.

So that's why I am of the opinion to just wait and see what they come up with and how well (or not) it matches with current measurements.

 

[amirm]

That conclusion (whether the hobbyists can actually create a valid target) can only be drawn after they finalized their curve and if their results differ drastically from what Harman found.

??? How would they evaluate the efficacy of their target? With an Internet poll??? Any target needs to be tested across 50 to 100 headphones, and with large population study in controlled environment. This is not in the cards for any hobby level work like this.

The 5128 has already shown odd results as published by Harman. There is currently no explanation for shortfalls like this: https://www.audiosciencereview.com/...ment-talks-from-head-fi-and-sean-olive.27017/

There are a number of problems to be solved here and I am not seeing them remotely attacking them properly.

 

[solderdude]

I prefer to wait till those guys are done.
Yes, they do not have access to the body of work done by Harman. They do not have access to all the tools Harman has.
That does not mean something fruitful can not come out of this IMO.

The different 'dip' is obviously originated from the ear-gain differences (pinna + earcanal) and as long as the target is not set in stone who is to say what the 'corrected' curve would look like and above all how the majority of owners would prefer the sound with EQ generated from either test fixture.
Whether or not that ear-gain is representative for the average human ear (which does not really exist) or that the old, less human like, acoustic impedance is more accurate I really can't tell. I hope that I can be pleasantly surprised one of these days.

I have no direct contact nor am I involved in any way nor would that be wise.

And I agree with you... there are a number of issues to be solved but have no idea if they are attacking them properly. And yes, they are a bunch of amateurs (in the good sense of that word) toying around with a potentially good test fixture they just have to figure out.

 

[MayaTlab]

If two guys can create a target for a headphone fixture in little time with so little effort, we would have to conclude that Harman and crew didn't know what they were doing! Hobby work like this hugely trivializes the research needed to find acoustic measurements that have a prayer of correlating with listening tests.

This is just disingenuous. No one at Headphones.com, Crinacle or LMG would pretend that their current target(s) have been vetted to anywhere near the same degree as Harman's. This thread and others are just here precisely for that reason : start to gather feedback.

Also, as it's already been mentioned, Harman's research, which isn't just about spewing out targets and predictive models, is being used by them to provide some guidance. I'd prefer that for now they'd stick to the more empirical findings in terms of preferential tilts / shelves and modifications to the baseline used and make sure that the coupling insensitive headphones (but very specifically not the others) don't produce an error curve that is too far off from Harman's up to around 5kHz, but that's just me.

Any target needs to be tested across 50 to 100 headphones,

You mean evaluated against a cohort of 50 to 100 headphones ? Does that then mean that Harman's research is bogus (it never compared the target to more than 30 alternatives) ? Now that's above my understanding but I believe that there are plenty of statistical means to evaluate which sample size is needed to arrive at good enough results.

Actually I don't think that you need to compare a target to a cohort of headphones to validate it, you just need to compare it to alternative curves, and these alternative curves can be randomly generated with varying amounts of deviation from the target, which with a certain listening test method would have been quite an interesting endeavour to pursue. In a certain sense Harman's articles on preferred levels of bass and trebles are closer to that spirit.

and with large population study in controlled environment. This is not in the cards for any hobby level work like this.

Hopefully larger companies such as LMG will have the resources to do proper listening tests. Just like you I think that this is essential to provide some degree of validation for any target devised on test fixtures incompatible with Harman's research, which is also why for now I'd like to see the error curve of coupling insensitive headphones remain similar to Harman's, unlike what we see with Headphones.com's current 8dB tilt approach.

The 5128 has already shown odd results as published by Harman. There is currently no explanation for shortfalls like this: https://www.audiosciencereview.com/...ment-talks-from-head-fi-and-sean-olive.27017/

I don't see why these results are odd given that the Stealth has already been demonstrated as being stupidly sensitive to coupling issues, including in that range (well throughout the FR spectrum anyway) - and in extenso more susceptible to variance in operators' practices as well. There could be other factors at play but the former already is enough to explain that.

The Stealth is a perfect example of a pair of headphones which shouldn't be used to produce a transfer function between test rigs and for which, if the error curve from any target devised for the 5128, differs from the one obtained on Harman compatible rigs, it should be seen as totally fine and not indicative of a fault on behalf of the fixture or the target, but rather plausibly the headphones'.

 

[markanini]

One could also say your personal expectation of the sound is biased by the ranking (as a form of a self fulling prophecy).

It would only be accurate if you were handed a bunch of headphones, ranked them by ear and then compared the rankings.
Bias works both ways.
Whether the HD600 is ranked higher than the HD650 or vice versa is mostly a preference thing or slope thing.

I think those guys are doing a pretty good job with creating the target for 5128 as I see it right now. No idea about IEM's which would seem could be even more valid.
As far as I know those guys did not create a ranking list (yet). The target as they have now seems much more promising than the overly smoothed and obviously incorrect above 6kHz that is in use now.
Once they are close enough to the final target (which I hope will not be revised and revised again over the years) you could enter that in the ranking algo and get a ranking list.
Only when that has been done you can compare rankings and most likely that list will not change much, perhaps shuffle the ranking within a few (pointless resolution) points.

We'll see. I would not discredit those guys that soon and comment on the way they go about things until they are ready to 'publish' their target. So far so good IMO.

That's why I started with my personal subjective rankings before comparing to Harman scores. Sighted as far as model so not perfect I admit.

I don't think it's necessarily discrediting anyone to encourage pursuing a level of validation trough blind testing of multiple listeners.

 

[solderdude]

Indeed, so for that reason I see no reason to discredit the attempts made by resolve and friends as that is exactly what they are doing.
I say let them finish what they started and only criticize when they finished what they are doing.
Ofcourse the assistance they asked from interested people can only help once they weed out the less helpful guidance they were looking for. The latter is the most difficult part.
I would not dismiss the 5128 simply because the manufacturer nor Harman created their target and the fact that the measurements results differ from all other fixtures which may well be an indicator for better technical performance in specific areas. Once the target curve has been created only then we can say how 'accurate' the results are.

The biggest concerns in headphone measurements will always be positioning, deviation from the owners actual ears from the standard and seal which can cause much higher deviations than the differences between measurement standards. So even with 5128 it will merely be indicative so will be the derived ratings.

That's why I started with my personal subjective rankings before comparing to Harman scores. Sighted as far as model so not perfect I admit.

Did you arrive at the exact same preference order or a similar-ish one in broad lines ?

 

[Deleted member 16543]

??? How would they evaluate the efficacy of their target? With an Internet poll??? Any target needs to be tested across 50 to 100 headphones, and with large population study in controlled environment. This is not in the cards for any hobby level work like this.

The 5128 has already shown odd results as published by Harman. There is currently no explanation for shortfalls like this: https://www.audiosciencereview.com/...ment-talks-from-head-fi-and-sean-olive.27017/

There are a number of problems to be solved here and I am not seeing them remotely attacking them properly.

One of the reasons why I posted that curve from 10 years ago is exactly the "problem" highlighted in your graph.
My measurement shows something similar too (with all the due disclaimers of my rig being less than anatomically accurate at the time I took the measurements, and the measurement procedure non being ideal, etc...).
It's not a problem at all. It shows a compliance to the expected HRTF.

 

[Robbo99999]

This is just disingenuous. No one at Headphones.com, Crinacle or LMG would pretend that their current target(s) have been vetted to anywhere near the same degree as Harman's. This thread and others are just here precisely for that reason : start to gather feedback.

Also, as it's already been mentioned, Harman's research, which isn't just about spewing out targets and predictive models, is being used by them to provide some guidance. I'd prefer that for now they'd stick to the more empirical findings in terms of preferential tilts / shelves and modifications to the baseline used and make sure that the coupling insensitive headphones (but very specifically not the others) don't produce an error curve that is too far off from Harman's up to around 5kHz, but that's just me.



You mean evaluated against a cohort of 50 to 100 headphones ? Does that then mean that Harman's research is bogus (it never compared the target to more than 30 alternatives) ? Now that's above my understanding but I believe that there are plenty of statistical means to evaluate which sample size is needed to arrive at good enough results.

Actually I don't think that you need to compare a target to a cohort of headphones to validate it, you just need to compare it to alternative curves, and these alternative curves can be randomly generated with varying amounts of deviation from the target, which with a certain listening test method would have been quite an interesting endeavour to pursue. In a certain sense Harman's articles on preferred levels of bass and trebles are closer to that spirit.



Hopefully larger companies such as LMG will have the resources to do proper listening tests. Just like you I think that this is essential to provide some degree of validation for any target devised on test fixtures incompatible with Harman's research, which is also why for now I'd like to see the error curve of coupling insensitive headphones remain similar to Harman's, unlike what we see with Headphones.com's current 8dB tilt approach.



I don't see why these results are odd given that the Stealth has already been demonstrated as being stupidly sensitive to coupling issues, including in that range (well throughout the FR spectrum anyway) - and in extenso more susceptible to variance in operators' practices as well. There could be other factors at play but the former already is enough to explain that.

The Stealth is a perfect example of a pair of headphones which shouldn't be used to produce a transfer function between test rigs and for which, if the error curve from any target devised for the 5128, differs from the one obtained on Harman compatible rigs, it should be seen as totally fine and not indicative of a fault on behalf of the fixture or the target, but rather plausibly the headphones'.

Just a quick question, can you expand on what you mean about the bit of your post that I put in bold? I don't fully understand what you mean there I don't think. "Coupling Insensitive", I think you mean headphones that don't change their measured frequency response much between people when measured with blocked ear canal mics, or do you instead mean not much variation with different spatial reseats on a measurement rig (eg GRAS)? And what do you mean by the error curve remaining similar to Harman's?

 

[MayaTlab]

Just a quick question, can you expand on what you mean about the bit of your post that I put in bold? I don't fully understand what you mean there I don't think. "Coupling Insensitive", I think you mean headphones that don't change their measured frequency response much between people when measured with blocked ear canal mics,

Headphones like the HD800, Focal open HPs, some Hifiman open planars, would land in that category.

Blocked ear canal entrance measurements are a bit tricky to interpret properly past around 1kHz and very much so past around 5kHz (when presented the way Harman recently did it, in other contexts they can provide useful data higher), but probably the ones I'd prefer to see for now. Other types of in situ measurements have advantages but might also be trickier to perform well.

But in the context of that post I'd also exclude headphones that are consistent across individuals but don't follow the transfer curve between test rigs and real individuals other consistent, coupling insensitive headphones would follow (the DCA Noir could possibly be an example I believe). These headphones might actually be competently designed if that deviation is deliberate and ensures that they effectively deliver a good sound quality on most people's heads... or not.

or do you instead mean not much variation with different spatial reseats on a measurement rig (eg GRAS)?

I see spatial reseat consistency on a measurement rig, alongside pad compression and controlled leakage tests, as very useful proxies. It's not unreasonable to presume that a pair of headphones that has low positional variation, is insensitive to controlled leakage tests, and which SPL varies linearly with pad compression across a large part of the spectrum, is likely to also be consistent across individuals. The reverse however (ex : a pair of headphones that is very sensitive to leakage), might not be true (presuming that this pair will vary across individuals) as good design can mitigate these issues.

Since in situ measurements on a cohort of listeners are difficult to do, these proxy measurements might be the most practical way to get an idea of whether or not a pair of headphones has a chance to be insensitive to coupling issues.

And what do you mean by the error curve remaining similar to Harman's?

Since the OE HT has been so extensively tested via listening tests, and since for coupling insensitive headphones like, let's say, the HD800, we know that, once equalised to a target, this is what the test subjects actually experienced to a reasonable degree of exactitude, at least up to 4-5kHz, until comparable listening tests are performed with equalisation to other targets based on other test systems (ex : DF HRTF + tilt on 5128), I'd personally prefer to stick to what we know works. So if the error curve was very different for such headphones in that range, I'd have some apprehension about its relevance unless it's proven to be at least equally preferred.

Obviously we know that individuals vary in their preference to some degree around the HT, but you get the idea.

Headphones sensitive to coupling issues (in particular singling out the ones inconsistent across individuals here) may have a markedly different error curve, but that's most likely the fault of the headphones, not the target nor the HATS.

 

[Robbo99999]

Headphones like the HD800, Focal open HPs, some Hifiman open planars, would land in that category.

Blocked ear canal entrance measurements are a bit tricky to interpret properly past around 1kHz and very much so past around 5kHz (when presented the way Harman recently did it, in other contexts they can provide useful data higher), but probably the ones I'd prefer to see for now. Other types of in situ measurements have advantages but might also be trickier to perform well.

But in the context of that post I'd also exclude headphones that are consistent across individuals but don't follow the transfer curve between test rigs and real individuals other consistent, coupling insensitive headphones would follow (the DCA Noir could possibly be an example I believe). These headphones might actually be competently designed if that deviation is deliberate and ensures that they effectively deliver a good sound quality on most people's heads... or not.



I see spatial reseat consistency on a measurement rig, alongside pad compression and controlled leakage tests, as very useful proxies. It's not unreasonable to presume that a pair of headphones that has low positional variation, is insensitive to controlled leakage tests, and which SPL varies linearly with pad compression across a large part of the spectrum, is likely to also be consistent across individuals. The reverse however (ex : a pair of headphones that is very sensitive to leakage), might not be true (presuming that this pair will vary across individuals) as good design can mitigate these issues.

Since in situ measurements on a cohort of listeners are difficult to do, these proxy measurements might be the most practical way to get an idea of whether or not a pair of headphones has a chance to be insensitive to coupling issues.



Since the OE HT has been so extensively tested via listening tests, and since for coupling insensitive headphones like, let's say, the HD800, we know that, once equalised to a target, this is what the test subjects actually experienced to a reasonable degree of exactitude, at least up to 4-5kHz, until comparable listening tests are performed with equalisation to other targets based on other test systems (ex : DF HRTF + tilt on 5128), I'd personally prefer to stick to what we know works. So if the error curve was very different for such headphones in that range, I'd have some apprehension about its relevance unless it's proven to be at least equally preferred.

Obviously we know that individuals vary in their preference to some degree around the HT, but you get the idea.

Headphones sensitive to coupling issues (in particular singling out the ones inconsistent across individuals here) may have a markedly different error curve, but that's most likely the fault of the headphones, not the target nor the HATS.

Cool, I think I know what you're getting at now, you're saying that in order to do listening testing on headphones that have been equalised to a Target Curve (eg the B&K 5128 in Resolve's project) then it helps massively if you limit it to headphones that don't vary much with spatial reseating on a measurement rig and also don't vary much between people (with both of those things often going hand-in-hand) because this way you can more likely assure that people are experiencing the intended Frequency Response of the Target Curve - it makes the listening tests more accurate. I'd agree with that.

The only variable that doesn't explore is the consistency of the measurement rig between different headphone models - perhaps if you took 100 headphones and measured them all on say GRAS and then B&K, you might then find that there was overall more measured variance within each headphone population recorded on either the GRAS or B&K. My initial thinking is that the measurement rig that had the least measured headphone variance within the headphone population would be the rig that is likely to give you the best results for measuring headphones for our purposes - I mean you don't want a measurement rig that actually makes headphones perform more inconsistently, that's my point.

 

[solderdude]

you're saying that in order to do listening testing on headphones that have been equalised to a Target Curve (eg the B&K 5128 in Resolve's project) then it helps massively if you limit it to headphones that don't vary much with spatial reseating on a measurement rig and also don't vary much between people (with both of those things often going hand-in-hand) because this way you can more likely assure that people are experiencing the intended Frequency Response of the Target Curve - it makes the listening tests more accurate.

Well summarized... and should be done with fresh pads and not 'weird' positioning on the head and preferably not using glasses or hair between the pads and skin.

Some (mostly open) headphones are really seal-insensitive so make good test candidates for this particular type of testing. Preference is still an issue though.
Closed headphones can be all over the place. The Stealth, unfortunately, is one of those closed headphones.

 

[amirm]

This is just disingenuous. No one at Headphones.com, Crinacle or LMG would pretend that their current target(s) have been vetted to anywhere near the same degree as Harman's. This thread and others are just here precisely for that reason : start to gather feedback.

Did you not read what you quoted from me? I said *if* they managed to do that (create a successful target), it would mean the work is quite trivial compared to what Harman went through. Prediction was made that they could be successful and I am giving my opinion of their chances of success.

 

[amirm]

Also, as it's already been mentioned, Harman's research, which isn't just about spewing out targets and predictive models, is being used by them to provide some guidance. I'd prefer that for now they'd stick to the more empirical findings in terms of preferential tilts / shelves and modifications to the baseline used and make sure that the coupling insensitive headphones (but very specifically not the others) don't produce an error curve that is too far off from Harman's up to around 5kHz, but that's just me.

It is a complete misunderstanding to think there is one preferred tilt for room response as to then try to mimic that for headphones. Every room needs to employ a target curve which fits the listener preference as driven by what they listen to. Music product has no standard. No agreed upon frequency response or tonality. No way then you can try to lock onto a preferred bass vs treble response. Hence the reason Harman curves were kind of a moving target here.

This is why I always adjust the EQ to taste and don't mind at all, indeed I encourage people, to do the same when it comes to overall target curve.

No way can we achieve high precision in any of this without a standard in production.

We can solve the standardization by just agreeing on something, anything. And that something might as well be the Harman target curve with a much more reasonably priced rig like GRAS 45C than screwing around with 5128 with DIY type efforts not based on true understanding of the problem at hand.

 

[amirm]

I don't see why these results are odd given that the Stealth has already been demonstrated as being stupidly sensitive to coupling issues, including in that range (well throughout the FR spectrum anyway) - and in extenso more susceptible to variance in operators' practices as well. There could be other factors at play but the former already is enough to explain that.

I don't know that at all. But you are accepting one of the failings of 5128 that I mentioned at the outset: it is difficult to fit a headphone or even IEM on it. Consistency as a result is poor with it making it the wrong tool for the job when we already have too many variables.

I am sure Jude prides himself in knowing how to fit a headphone on the fixture yet we have you saying that he must have done that wrong. If he can get it wrong, so can other reviewers (I suffered from that as I noted).

 

[amirm]

The Stealth is a perfect example of a pair of headphones which shouldn't be used to produce a transfer function between test rigs and for which, if the error curve from any target devised for the 5128, differs from the one obtained on Harman compatible rigs, it should be seen as totally fine and not indicative of a fault on behalf of the fixture or the target, but rather plausibly the headphones'.

You have no data to back that. The testing and comparison could very well be pointing to a design flaw in 5128. Remember, I measured nearly identical results to Dan Clark proving the repeatability and reliability of using GRAS 45. Fixture is switched by skilled operators and results were clearly odd. This should point the finger at the fixture first.

But let's say you are right. How would anyone know their headphone is not a good fit for the 5128? They wouldn't of course especially when you have no idea what is really going on here.

 

[solderdude]

Music product has no standard. No agreed upon frequency response or tonality. No way then you can try to lock onto a preferred bass vs treble response.

This (the recording side with poor practices) seems to be the biggest problem and a real issue for music reproduction both with speakers and headphones.
With speakers the room, placement and directivity+FR of speakers are confounding issues, with headphones ear-gain and seal are confounding issues. For both situations taste also comes into play.
When one has a good reference and is an experienced listener this can help.

Viva la plain old tone control. Having the old style tone control helps greatly with quite a lot of recordings to make them more enjoyable.

 

[Deleted member 16543]

Actually, a balanced response for speakers, as measured by a simple omni mic at the listening position, is something that has already been extensively researched.
While it's true that you'll allways have the extremely dull or shrill records, the function of a target curve is to put you "in the middle", where the instances of both of those are minimized.
One of the targets from the above mentioned research looks like the measurement of my speakers I shared a couple posts ago.

Speakers measurement rig and target has been researched and standardized to a degree that is probably not even possible to achieve for headphones, due to the variability and sensitivity of measurement results.
Therefore, starting from a well balanced pair of speakers at the listening position and positioning a mic that resembles our anatomy at that position, to measure what our eardrums sense when we're physically there, and using that as a target for headphones, is a very valid strategy. Provided the headphones are measured with the same binaural microphone that the speakers were measured with.

 

[amirm]

Headphones like the HD800, Focal open HPs, some Hifiman open planars, would land in that category.

Not in my book. Here is what I said about Hifiman HE400: "The large cups made an easy job of mounting them on my fixture and getting good measurements on first try." or the Arya: "Fitment on the fixture was very good/easy due to large cups." Or the HE-560: "The large cups made an easy job of mounting them on my fixture and getting good measurements on first try." Or Focal Utopia: "Mounting the headphone on the text fixture was easy requiring almost no manipulation to get proper measurements."

Large cups make it easier to measure than small ones. Now, if you switched to 5128, yes, you are going to experience more problems due to curvature of the dummy head. And overall shape of it which put more restrictions on how you can fit the headphone on it. That was one of my major issues when I evaluated the 5128. I simply could not get measurements that correlated with what was out there.

 

[solderdude]

Therefore, starting from a well balanced pair of speakers at the listening position and positioning a mic that resembles our anatomy at that position, to measure what our eardrums sense when we're physically there, and using that as a target for headphones, is a very valid strategy.

Would that need to be gated or need room effects need to be included and if so which room and what angle and speaker distance ?
Also there is still the ongoing 'missing a few 3dB' question that still hasn't been fully researched which can skew a possible target.