• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Dan Clark Expanse Headphone Review

Rate this headphone:

  • 1. Poor (headless panther)

    Votes: 10 2.7%
  • 2. Not terrible (postman panther)

    Votes: 12 3.3%
  • 3. Fine (happy panther)

    Votes: 65 17.7%
  • 4. Great (golfing panther)

    Votes: 281 76.4%

  • Total voters
    368

Robbo99999

Master Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
6,874
Likes
6,672
Location
UK
That's exactly what Harman themselves did to "transfer" the Harman target to the new 5128 (see here).

It's also what they took considerable flak for from the audio science community as this method is quick and dirty, with less than optimal results.
You just can't really compensate for differences in acoustic impedance, as the response delta between GRAS and the 5128 will be different for each headphone.
Sure you can find a compensation curve that'll on average get you closer to compliance, but the individual error for each headphone will be quite considerable regardless.
The right thing for Harman to do would've been to remake the target from scratch.

As a result of Harman's shoddy work and their decision to keep their results under NDA, people like Resolve and oratory were encouraged to do better, and are now working on 5128 targets of their own.
Ok, that's very cool that Resolve & Oratory are working on 5128 targets of their own. Yeah, I understand why that's a quick & dirty method re what I suggested in my post and that Harman got flak for it re using GRAS Harman equalised headphones to transfer from GRAS to 5128 - I also don't think it's a particularly good idea for the reasons you mention but also because if you're going to take that approach then you may as well just stick with using the GRAS. I agree that it makes more sense to create a fresh target for the 5128. Praps @Resolve can think about getting his B&K into a good listening room with some Anechoic Flat speakers and create a target by measuring at the eardrum of the 5128 to sine sweeps from the anechoically flat speakers (in the same way that Harman did it for their GRAS in their studies). Then you could tweak bass & treble levels to taste for a few or many participants, then Bob's your Uncle you've got a Target Curve for the 5128, lol! It would be quite difficult and time consuming to sort out all the participants depending on how many you were aiming for, and it would require quite some effort & time to get the whole study setup. I'd say that kind of approach would be the best thing to do with the 5128 to create a target curve.
5128 ear + ear canal is closer to that of an (average) human being.
The acoustic impedance also differs so the results will differ.
This is why the 5128 would be a good choice. T
he more measurements are made public the more people can relate to the plots.

Note that the exact same headphone on different fixtures always will measure different and the exact same headphone will also be perceived differently by different individuals.
The goal of the 5128 is to get closer to the average human in measured response.

The thing is IF you are going to construct a target based on the 43AG you are basically calibrating to an older standard that isn't as anatomically correct and thus give incorrect measurements even though it was considered one of the 'correct' fixtures (but not the most 'correct'.

Resolve is on the correct path it seems but may not end up with a Harman target. While the Harman target may be preferred by many (the bass boost part) many find it a bit too much. The Harman target is impressive sounding and may well be correct for those listening at say 70dB average (all day listening level) but some like to enjoy music at a louder level for say 2 hours tops. For these folks the bass target proposed by Resolve may fit that better. Also the bass levels of 5128 are measured slightly lower so a Harman target on the 5128 will appear to have less 'bass boost' in the raw graphs anyway.
Yeah, I get your points @solderdude re listening levels & effect on bass & treble, and also your point about the quick hack I suggested in my previous post would mean you're just calibrating to an old standard (GRAS) like you mention, so it should instead be a Target Curve that is created "fresh", like I replied just now to StaticV3 in this post. I don't know how ambitious and how much time & effort Resolve has got to create his Target Curve for the 5128, it is interesting though.
 
Last edited:

Resolve

Active Member
Reviewer
Joined
Jan 20, 2021
Messages
212
Likes
531
Ok, that's very cool that Resolve & Oratory are working on 5128 targets of their own. Yeah, I understand why that's a quick & dirty method re what I suggested in my post and that Harman got flak for it re using GRAS Harman equalised headphones to transfer from GRAS to 5128 - I also don't think it's a particularly good idea for the reasons you mention but also because if you're going to take that approach then you may as well just stick with using the GRAS. I agree that it makes more sense to create a fresh target for the 5128. Praps @Resolve can think about getting his B&K into a good listening room with some Anechoic Flat speakers and create a target by measuring at the eardrum of the 5128 to sine sweeps from the anechoically flat speakers (in the same way that Harman did it for their GRAS in their studies). Then you could tweak bass & treble levels to taste for a few or many participants, then Bob's your Uncle you've got a Target Curve for the 5128, lol! It would be quite difficult and time consuming to sort out all the participants depending on how many you were aiming for, and it would require quite some effort & time to get the whole study setup. I'd say that kind of approach would be the best thing to do with the 5128 to create a target curve.

It would certainly be interesting but I think you don't necessarily need to go to all that trouble to get a closer Harman target for the 5128. You just need to calculate Harman in-room from DF for the 5128, then apply the shelf filters from the research. But, our contention is that DF and Harman in-room are fairly close to begin with and DF doesn't carry the Harman conditions forward, like the specific Revel speakers and the Harman room. The other aspect is the shelf vs slope question, and while either method would potentially achieve the bass to treble delta that most people prefer, these two targets would be perceived differently. I can see a reason to do either one, but the slope fits better with a larger body of research, and there is even a paper from Dr. Olive (2013) indicating a preference for slope over a shelf. Regardless, we're not hard locked in to anything here, as if we find there are compelling reasons to change it up we can quite easily do that. But at the moment there's enough to justify our DF + slope approach over the calculated Harman in-room approach or even DF + shelf approach.
 

Robbo99999

Master Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
6,874
Likes
6,672
Location
UK
It would certainly be interesting but I think you don't necessarily need to go to all that trouble to get a closer Harman target for the 5128. You just need to calculate Harman in-room from DF for the 5128, then apply the shelf filters from the research. But, our contention is that DF and Harman in-room are fairly close to begin with and DF doesn't carry the Harman conditions forward, like the specific Revel speakers and the Harman room. The other aspect is the shelf vs slope question, and while either method would potentially achieve the bass to treble delta that most people prefer, these two targets would be perceived differently. I can see a reason to do either one, but the slope fits better with a larger body of research, and there is even a paper from Dr. Olive (2013) indicating a preference for slope over a shelf. Regardless, we're not hard locked in to anything here, as if we find there are compelling reasons to change it up we can quite easily do that. But at the moment there's enough to justify our DF + slope approach over the calculated Harman in-room approach or even DF + shelf approach.
Interesting, thanks for the response! ASR had the 5128 at the start of ASR's headphone journey, and I remember @Mad_Economist was able to calculate a Harman Target for 5128 using the reverberation characteristics of the Harman Listening Room, maybe he applied that to DF to calculate the Harman Target on 5128 - but apparently it didn't sound good, I guess Amir didn't think it sounded authentic, I don't know who else was involved in actually testing/listening to the results (as it wasn't publicly done I don't think). Ah, I remember, he used the Chris Struck Method to calculate it. Is that the other option you are talking about, the other option to your DF+slope? You haven't got a graph have you, that shows your DF+slope vs your calculated Harman on your 5128, that would be interesting to see the difference?

On the slope vs shelf topic, if I think about it now, I'm of the opinion that both would be used ideally - because slope is created as a natural by-product of anechoic flat speakers in the room, and then also a small amount of room gain in the bass is quite common when people listen to music, so potentially artists have taken some of that into account during their music making. The Harman Speaker Curve for example has both Slope & Low Shelf, firstly slope applied throughout the frequency range, and secondly Low Shelf boost added to the bass. But, I do appreciate that you reference that Sean Olive 2013 paper that says that slope is preferred over shelf in some instances - although I'm not personally familiar with that paper so can't comment on that. It will be interesting to see how this all pans out for you, and what you come up with, are you showing the evolution of what you're doing anywhere, or just doing it all behind the scenes. If you've got a series of articles or forum threads you're dedicating to your explorations I'd love to have a read of that, have you got some links?
 

markanini

Major Contributor
Joined
Feb 15, 2019
Messages
1,718
Likes
1,768
Location
Scania
This is false, because EQing to match the target there doesn't improve it to a sufficient degree. Also false because other headphones EQ'd to match our target on the B&K don't sound shouty. Moreover the default shouldn't be to assume these headphones would be the same in that region because A) different pinna and B) different head, and C) not a plate. But, we've also been over much of this ground already.

Edit: To be clear, I'm not saying the Expanse IS shouty necessarily for every head, merely that this indicates some potential variation of headphone behavior depending on the head (or rig) that's wearing it.
If you would share your data used to validate that point it would be useful. Until then GaryH conclusion isn't technically unreasonable, neither would the conclusion that the 5128 is less good at correlating 2-5kHz to real ears.
 

Resolve

Active Member
Reviewer
Joined
Jan 20, 2021
Messages
212
Likes
531
If you would share your data used to validate that point it would be useful. Until then GaryH conclusion isn't technically unreasonable, neither would the conclusion that the 5128 is less good at correlating 2-5kHz to real ears.

It's not less good, nor necessarily better. It's just different - as are real heads/ears. And yeah validating this point comes in the form of comping the same headphone FR to rig specific DF targets on each rig respectively, which show differences in those regions. We noticed this fairly early on with the effect of ear gain at 2khz when comparing between the two rigs. Those results will be made public in a few weeks most likely, since we're showing how headphone behavior varies between the two systems. But, if you're looking for further confirmation from other sources, Dr. Olive has also pointed out the same - that we can't expect comp'd results to rig specific targets to match - even if those targets use the same DF principle.
 

GaryH

Major Contributor
Joined
May 12, 2021
Messages
1,348
Likes
1,804
This is false, because EQing to match the target there doesn't improve it to a sufficient degree.
It does improve it then. Normalizing your 5128 measurement of the Expanse and target at 500 Hz (the industry standard, as you did with your GRAS measurement), actually puts the Expanse's response 1 dB below your target at 3 kHz (and not peaking above it anywhere at all in the mids/treble except above 10 kHz, which not even the loosest, most slippery definition of 'shouty' would cover). And yet after EQing this down even further by the 3 dB necessary to take it to the Harman target from your GRAS measurement you still find it shouty? Now 4 dB below your target at the ear gain peak? You need to first definitively rule out that either it's your (new, theoretical, scientifically-untested) target that's 'shouty', or your (I presume sighted/uncontrolled and so cognitive bias-prone) listening tests/reports are inaccurate/unreliable, or both, before even considering anything else.
Also false because other headphones EQ'd to match our target on the B&K don't sound shouty.
Which headphones? What are their 5128 and GRAS measurements? To how many listeners? Trained or untrained? Confirmed normal audiometric hearing? What's your listening test protocol and what controls did you implement? Blind, double-blind? Loudness matched, and to what standard/level (e.g. ITU-R BS 1770-4 with an absolute level of 85 dB slow, C-weighted equivalent diffuse field level as Harman used)? What program material and what are their spectra? How many trials per person (e.g. reseating the headphone on the head between them)? How exactly did you define 'shouty' and ensure everyone reports it consistently? Was it a binary shouty / not shouty question or scaled? If the latter what scale did you use for 'shoutiness'? What randomisation of e.g. program/headphone did you implement in order to control for order/learning biases? What are the statistics of the results (e.g. ANOVA)?
Moreover the default shouldn't be to assume these headphones would be the same in that region
The default should be to assume the tonality depicted by the delta between the Harman target (produced from our current best headphone science) and measurements of a headphone made on a GRAS system with the same acoustic impedance to that used to create said target, will likely be a good representation of its tonality to the majority, unless definitively demonstrated otherwise for the specific person/headphone in question using a methodology as or more scientifically robust and controlled as Harman used, which my above questions allude to.
this indicates some potential variation of headphone behavior depending on the head (or rig) that's wearing it
Without the methodological controls mentioned above, no it doesn't. It could be any number of things, including your 5128 target being at fault. And it's not just the Expanse, your target also seems off with the HD600/HD650 at the same ear gain peak region:

1e3d89a34a6000c40c4f29cf84ff31f74532acd2.jpeg


Compare with GRAS measurements (normalized at 300 Hz to match your non-standard normalization) where it's already ~2 dB above the Harman target around 3 kHz:

graph-32.png


The number of people who would want even more ear gain at 3 kHz from the HD600 (to match your target) are likely to be few and far between. With that it would be 4 dB above the Harman target, which of course unlike your target, has been scientifically tested to be the majority preference in controlled, blind listening studies of many users, trained and untrained listeners alike, producing solid data showing high correlation (86%) between closeness to their target and perceived sound quality (quantified by predicted calculated from the former, and actual given for the latter, preference ratings). And that's the key point in all of this. As Sean Olive has said:
Jude maintains that the 5128 is more accurate than the other test fixtures. In terms of representing the acoustics of average human ear canals that may well be true. But there still needs to be some way of interpreting headphone measurements made on the 5128 in terms of human perception of sound quality. In my view, it doesn't matter if the 5128 is more accurate, if there isn't any data on how the measurements correlate with listeners' sound quality ratings.
 
Last edited:

Resolve

Active Member
Reviewer
Joined
Jan 20, 2021
Messages
212
Likes
531
It does improve it then. Normalizing your 5128 measurement of the Expanse and target at 500 Hz (the industry standard, as you did with your GRAS measurement), actually puts the Expanse's response 1 dB below your target at 3 kHz (and not peaking above it anywhere at all in the mids/treble except above 10 kHz, which not even the loosest, most slippery definition of 'shouty' would cover). And yet after EQing this down even further by the 3 dB necessary to take it to the Harman target from your GRAS measurement you still find it shouty? Now 4 dB below your target at the ear gain peak? You need to first definitively rule out that either it's your (new, theoretical, proprietary, untested scientifically) target that's 'shouty', or your (I presume sighted and so cognitive bias-prone) listening tests/reports are inaccurate/unreliable, or both, before even considering anything else.

Which headphones? What are their 5128 and GRAS measurements? To how many listeners? Trained or untrained? Confirmed normal audiometric hearing? What's your listening test protocol and what controls did you implement? Blind, double-blind? Loudness matched, and to what standard/level (e.g. ITU-R BS 1770-4 with an absolute level of 85 dB slow, C-weighted equivalent diffuse field level as Harman used)? What program material and what are their spectra? How many trials per person (e.g. reseating the headphone on the head between them)? How exactly did you define 'shouty' and ensure everyone reports it consistently? Was it a binary shouty / not shouty question or scaled? If the latter what scale did you use for 'shoutiness'? What randomisation of e.g. program/headphone did you implement in order to control for order/learning biases? What are the statistics of the results (e.g. ANOVA)?

The default should be to assume the tonality depicted by the delta between the Harman target (produced from our current best headphone science) and measurements of a headphone made on a GRAS system with the same acoustic impedance to that used to create said target, will likely be a good representation of its tonality to the majority, unless definitively demonstrated otherwise for the specific person/headphone in question using a methodology as or more scientifically robust and controlled as Harman used, which my above questions allude to.

Without the methodological controls mentioned above, no it doesn't. It could be any number of things, including your 5128 target being at fault. And it's not just the Expanse, your target also seems off with the HD600/HD650 at the same ear gain peak region:

1e3d89a34a6000c40c4f29cf84ff31f74532acd2.jpeg


Compare with GRAS measurements (normalized at 300 Hz to match your non-standard normalization) where it's already ~2 dB above the Harman target around 3 kHz:

View attachment 256901

The number of people who would want even more ear gain at 3 kHz from the HD600 (to match your target) are likely to be few and far between. With that it would be 4 dB above the Harman target, which of course unlike your target, has been scientifically tested to be the majority preference in controlled, blind listening studies of many users, trained and untrained listeners alike, producing solid data showing high correlation (86%) between closeness to their target and perceived sound quality (quantified by predicted calculated from the former, and actual given for the latter preference ratings). And this is the key point in all of this. As Sean Olive has said:
It seems perfectly accurate with the hd600 and hd650, and HEk v2 which has less variation across rigs, which was used to EQ flat to the target and guess what... Not shouty. So yeah... Headphones behave differently on different heads, something we need to get used to collectively.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
So yeah... Headphones behave differently on different heads, something we need to get used to collectively.

Yep, industry standard just means the fixture conforms to a standard and can be checked to a standard and can conform to that standard within the specified tolerances of that standard.
Very handy for measurements and to compare measurements on copies of the same rigs/standard.
The 5128 is closer to a real 'average' human ear.

Glad you (and others) are looking into the relation between perceived sound and the measurements and try to come up with a standard that, likely, may deviate somewhat from that of a manufacturer who's goal is preference. Keep up the good work.

Too bad that the size and shape of their heads and pinnae all differ, as well as wearing habits, seal, pad compression (wear, head-width, hairs, glasses) all can mess up the tonal balance on human heads. Even when they happen to have similar ear canal shape and pinna shape there is still hearing condition (especially for older folks).

Measurements can be a very useful tool to get an idea of the frequency response and tonality but acoustic measurements aren't always as precise as most nerds believe/want them to be in relation to what's perceived. And the reason I do not believe in 'exact' computer generated EQ based on an average of, or specific measurement but believe just some nudges in the right direction for the worst offending deviations is good enough and probably even better than 'exact' EQ only really valid for that (averaged) measurement on that rig.
 

markanini

Major Contributor
Joined
Feb 15, 2019
Messages
1,718
Likes
1,768
Location
Scania
Putting aside that real ears vary, and the new rig might very well be better at predicting bass performance.

Would it be an issue if the new rig had less correlating mids with an average of real ears across multiple headphones? You could keep tweaking target curves at that point but to what end?
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
The Harman curve also wasn't born in one day and took some iterations.
The same will be true for this fixture but not based on Harman research (unless Harman or B&K comes up with one and turn it into a standard) but somewhat different but not invalid. It just probably would not be seen as a standard when it is not registered as being a new standard.

I think the challenge is to get a good correlation between measurements and perception. The acoustic impedance and pinna as well as the coupler are fully characterized already.
 
Last edited:

markanini

Major Contributor
Joined
Feb 15, 2019
Messages
1,718
Likes
1,768
Location
Scania
That's why I pose the question as a hypothetical. Because of the disparity of data at the moment. On the GRAS side tests have show a high correlation between user scores and predicted/calculated scores for the Harman target curve. On 5128 we don't have a universally accepted target yet to conduct a similar test with, or tests with proposed targets. It should fall to reason that predictability tests should be repeated with 5128 once we do. If we reach a point in the future where say multiple targets happen to show less predictability in 5128 units vs GRAS, do we revert to GRAS or keep using what becomes the Harman target of the 5128 despite it? A third alternative in that scenario would be prompting B&K for a hardware revision. That's what Sean Olive and his collages did with GRAS, only it was focused on outer ear anatomy and bass frequencies IIRC.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
First there needs to be more data before starting to walk alternative paths.
 

markanini

Major Contributor
Joined
Feb 15, 2019
Messages
1,718
Likes
1,768
Location
Scania
First there needs to be more data before starting to walk alternative paths.
Possibly less than before the GRAS/Harman relationship.
As far as the subjective blind tests go those were published by Sean Olive for a handful of headphones with redacted model names. Could be re-used in theory, as the reference dataset, If we identify the models by the graphs. The remaining work to produce a predictability comparison would involve making new measurements on the 5128 of the relevant headphone models and computing preference scores using the current model only substituting the target curve with the one proposed by Resovle. That would give us a lot to go on, without starting completely from scratch, and could be re-tested against any number of proposed target curves for the new rig.
 
Last edited:

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
Sure, this could also be done with all the headphones that already have been measured.
The problem I see is the variability of headphones on the head. The guys are probably on that already using headphones that do not change much with seal.

I'll just wait and see what the measurement buffs are coming up with time.
 

GaryH

Major Contributor
Joined
May 12, 2021
Messages
1,348
Likes
1,804
It seems perfectly accurate with the hd600 and hd650
So you're honestly saying you prefer a response 4 dB above the Harman target around 3 kHz (where our ears are most sensitive)? I very much doubt many others would. And I don't seem to remember you ever saying Harman has too little energy in that region before coming up with this new 'perfectly accurate' target.
and HEk v2 which has less variation across rigs, which was used to EQ flat to the target and guess what... Not shouty. So yeah... Headphones behave differently on different heads, something we need to get used to collectively.
I'll ask one more time then give up and conclude it's a no from your reluctance to give an answer. Were the individual subjective listening tests reporting a headphone with / without EQ to be 'accurate' / 'shouty' / 'not shouty' etc., from which you're coming to sweeping conclusions about the objective variability of certain headphones on-head (while outright rejecting the possibility your target is at fault), controlled tests e.g. at the very least blind and accurately loudness-matched?
 
Last edited:

markanini

Major Contributor
Joined
Feb 15, 2019
Messages
1,718
Likes
1,768
Location
Scania
The problem I see is the variability of headphones on the head. The guys are probably on that already using headphones that do not change much with seal.
I agree, even if it's a given. I'm more concerned specifically with the question of the new rig representing a better average of real ears or not. I'm frankly a bit shocked by the certainty of early adopters, considering the data I've seen. The least one could do is initially focus on reproducing predictability tests.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
To bring the thread back on track ...do you think the measurement or target of 5128 measurements of the DCA Stealth are incorrect or only the proposed target is incorrect ?
 
Last edited:

Resolve

Active Member
Reviewer
Joined
Jan 20, 2021
Messages
212
Likes
531
So you're honestly saying you prefer a response 4 dB above the Harman target around 3 kHz (where our ears are most sensitive)? I very much doubt many others would. And I don't seem to remember you ever saying Harman has too little energy in that region before coming up with this new 'perfectly accurate' target.

I'll ask one more time then give up and conclude it's a no from your reluctance to give an answer. Were the individual subjective listening tests reporting a headphone with / without EQ to be 'accurate' / 'shouty' / 'not shouty' etc., from which you're coming to sweeping conclusions about the objective variability of certain headphones on-head (while outright rejecting the possibility your target is at fault), controlled tests e.g. at the very least blind and accurately loudness-matched?
I did not do a blind test, but it was level matched. The thing is, this much of a difference was immediately recognizable, to the point that it wasn't even close - between both headphones, where level at 3khz was measured the same. And... I wasn't the only one to have this same experience between the two. Moreover, you should know by now that it's not uniquely about level in one region of the FR, but rather the overall balance. You can't just point at 3khz on our target without considering where the rest of the treble is. Now, I'm not saying it's perfect there for everyone, merely that matching headphones to the target there doesn't always yield the same perceptual experience... Because... And this is the last time I'm going to be replying to you... Headphones behave differently on different heads/rigs.
 
Last edited:

Resolve

Active Member
Reviewer
Joined
Jan 20, 2021
Messages
212
Likes
531
To bring the thread back on track ...do you think the measurement or target of 5128 measurements of the DCA Stealth are incorrect or only the proposed target is incorrect ?
What I'd love for folks to try is do some EQ to the target with some of the new measurements we're putting out. Obviously we can't account for unit variation, and I don't recommend necessarily EQing to match any target past a certain point (because it's not your head/ears), but it would be good to get a read on how others perceive it.
 

isostasy

Senior Member
Joined
Oct 19, 2022
Messages
354
Likes
628
Figured you guys might find this one interesting. We're doing measurements on the B&K 5128 now, so here's the Expanse FR relative to DF + 8dB slope. The target is still being tested out, but I thought this might be some useful data for folks. What's interesting is that perceptually, the results so far have been highly predictive for most headphones, but this one seems to be a notable exception for a few of us.

View attachment 256351

View attachment 256352

Is the Expanse an exception because it sounds worse than expected despite conformity above 400mHz, or better than expected despite increased bass below 200Hz? Sorry if I've missed something, I read through the last few pages and could only see some discussions about whether it is or isn't 'shouty'.
 
Top Bottom