Dan Clark Expanse Headphone Review

[Robbo99999]

I wonder how much of these differences in target curve preferences are actually a merely function of differences in the measurement rig used to create that target?
Let's say Harman created headphones that precisely matched the Harman-target measured on a Harman rig.
If you then took those same headphones and measured it on the Oratory rig, it would measure different. And Oratory would then issue EQ settings that supposedly "correct" the headphones so that they match the Harman target. But in reality, the headphones with 0 EQ are, by definition, are matching the Harman target.

Well, the Oratory & Amir rigs (GRAS) are effectively the same as the original Harman rig. The pinna is not exactly the same as apparently Harman used a custom made pinna, but the rigs that Oratory/Amir/Crinacle/Resolve use are all GRAS rigs that are the current best thing in terms of a legitimate "Harman Rig". I don't know how much actual difference you would see in measurements using the original Harman pinna vs the current one used on GRAS, I have a feeling that the main difference was just the flexibility of the pinna, the original Harman pinna was stiff, so I'm thinking the GRAS pinna used mimics the Harman one but is more flexible (increased pliability/malleability) and therefore more like a real human ear in terms of better headphone fitment but someone can correct me on that if they for sure know the differences.

 

[solderdude]

There are 12 different configuration possibilities for 45CA fixtures so not all 45CA are created equal

Amir uses the 45CA-10 which complies with IEC 60318-4.
This norm specifies a transfer impedance tolerance range between 100Hz and 10kHz. It provides a 'human-like' load up to 10kHz and allows a 5dB tolerance at 10kHz and 1.5dB between 100Hz and 2kHz.
However, this norm does not specify the tolerance of the frequency response accuracy (logical, as it can be all over the place above a few kHz) which is not the same as the transfer impedance.
The IEC 60318-4 norm only specifies dimensions/shape the pinna and ear canal have to comply to. The 'acoustical load' basically.

The configuration Amir uses is claimed by GRAS to provide a human-like acoustic load up to 20kHz within +/- 2.5dB.
It does not seem to comply to IEC 60318-7 which the B&K5128 does
That HATS has even more 'human-like' shape of the fake ear canal and thus provides an even more accurate 'acoustical load'.
It only specifies the acoustic load though not the accuracy of the frequency response.

Frequency response tolerances are specified elsewhere and merely state in at which frequency the measured response may (= not equal to actually does) deviate.
With diffuse field above 6kHz there is a 10dB tolerance and above 8kHz a 12dB tolerance which is not strange given the measurements of 1 single headphone in various positions of the HD600 below. It is easy to observe that between 100Hz and 1kHz the deviation is small but above a few kHz all bets are off.
Now... we have to realize that in practice narrow (high-Q) dips and peaks are not that audible (smoothed by the hearing) and the variations shown below are not representative to the ways one would wear this on the head. It does, however, show that measurements above a few kHz have to be taken with a grain of salt and thus also a 'highly smoothed' average target has to be taken merely as a 'guide-line' (for each different fixture) and a deviation from that 'target' line in a measurement does not mean it actually deviates as much ... in reality. It only means that specific headphone, on that specific fixture, on that specific 'seating' (or averaged seatings) shows a certain deviation from a very 'smoothed/averaged' drawn line.

'Exact' EQ or believing a single 'measurement trace' is 'exact science' and has to closely follow some 'target line' is equally silly as believing measurements say nothing.
But at least 'science' provides some 'handles' one can hang on to.
Exact EQ looks very nice when applied to a single measurement but IRL it is only accurate for that specific measurement but... at least it might be better than just winging it by ear.
Measurements matter but one has to know its limitations and above all those of headphone measurements.

This, including product variance, all explains why there are different 'squiggles' between measurements done by others, even with the same fixture.

To make this about DCA Expanse again... on all measurements done with 'reputable' measurement gear this headphone, without EQ and a good seal conforms pretty well to the 'Harman' target. How much it deviates and where it exactly deviates is unclear but deviations seem small over the known measurements so quite 'trustworthy' IMO.

 

[MayaTlab]

Well, the Oratory & Amir rigs (GRAS) are effectively the same as the original Harman rig. The pinna is not exactly the same as apparently Harman used a custom made pinna, but the rigs that Oratory/Amir/Crinacle/Resolve use are all GRAS rigs that are the current best thing in terms of a legitimate "Harman Rig". I don't know how much actual difference you would see in measurements using the original Harman pinna vs the current one used on GRAS, I have a feeling that the main difference was just the flexibility of the pinna, the original Harman pinna was stiff, so I'm thinking the GRAS pinna used mimics the Harman one but is more flexible (increased pliability/malleability) and therefore more like a real human ear in terms of better headphone fitment but someone can correct me on that if they for sure know the differences.

Spoiler

Harman's own pinnae (there are two of them) use a modified version of this one : https://www.grasacoustics.com/products/accessories/for-artificial-ears-and-mouths/product/430-kb0070
It was 3D scanned, the geometry was altered a bit for each version (less protruding pinna among other modifications), and then cast in Shore 00-35 silicone for the first one (same as current GRAS anthropomorphic pinna), and a softer silicone still for the second one.
Source : https://www.aes.org/e-lib/browse.cfm?elib=17699

Harman has compared their own pinna(e?) with GRAS' various pinnae, including the anthropomorphic one, but it hasn't been published AFAIK.

If you digitise the traces from various Harman's papers or Sean Olive's measurements and compare them to the various GRAS measurements available the difference can go from quite moderate (HD650 below 6kHz for example) :

To potentially more significant :

https://twitter.com/i/web/status/1368458551618805760

This one is interesting as we may presume that the same sample was tested by the same operator (not certain though) on the exact same fixture (minus the pinna difference), and both pinnae are of a soft type.
The poor visibility of the data prevents a clean digitisation but after quite a bit of Photoshop wizardry and perspective correction this can give a rough idea of the difference (please don't directly compare it to other graphs and the Harman target) :

It's probably a case by case problem depending on the exact model of headphones.

I don't think that the older KB0065 pinna is in use in most databases available anymore, but comparing it and the newer anthropomorphic pinna can be interesting to assess the importance of the pinna on the overall results ceteris paribus.

SoundStageSolo has published the difference for several headphones at some point in time : https://www.soundstagesolo.com/index.php/equipment/headphones/headphones-text-listing

I haven't bothered to digitise all of it but generally it looks like this :

i.e. a general trend line with noise around it.
Oratory1990 has performed a similar test and the trend line is quite similar (also with a good deal of variation around the trend for individual headphones).

TLDR : the difference between Harman's modified pinna and GRAS' anthropomorphic pinna isn't well documented at the moment, but I wouldn't worry too much about it for headphones such as the HD800 up to several kHz, and perhaps a bit more for other designs .

 

[Rthomas]

I have learned a lot from these discussions.

You cannot use compliance to a target and low distortion to declare anything the 'best headphone in the world'

For example for a lot of music I prefer Bob Katz target to the Harman Target (using LCD5)

My ranking of all the flagships I've heard (EQ allowed)

Tier 1. Audeze LCD-5 or Audeze CRBN
Tier 2. Audeze LCD4 or Stax SR009S
Tier 3. DCA Stealth/Expanse or Focal Utopia or Stax 007
Tier 4. HD800S or LCD3
Tier 5. DCA Ether 2 or Meze Empyrean

Never buy a flagship headphone based on a reviewer's reccomendation unless money is no object. You will lose $1000 to $2000 if you don't like it and want to sell it.

 

[Adi777]

My ranking of all the flagships I've heard (EQ allowed)

Tier 1. Audeze LCD-5 or Audeze CRBN
Tier 2. Audeze LCD4 or Stax SR009S
Tier 3. DCA Stealth/Expanse or Focal Utopia or Stax 007
Tier 4. HD800S or LCD3
Tier 5. DCA Ether 2 or Meze Empyrean

You hear a big difference between these all headphones?

 

[Rthomas]

You hear a big difference between these all headphones?

Even after EQing to the same target there is a clear difference to my ears.

The Audezes and the Focal are simply more fun to listen to.

The Stealth and the Stax are a softer more gentle presentation, not good for rock music imho.

The SR009S and HD800S are amazing for classical.

The LCD-5 and CRBN are the best all rounders so that's why I sold everything else.

The differences are not huge but at this level only the nit pickiest amongst us are buying and comparing these products.

At $4000+ I have the right to be hyper critical.

The price of 2 of these flagships nearly get you a SOTA speaker system like the Dutch & Dutch 8C or Genelec 8351 so I'm going to be hyper critical in my evaluations.

 

[Dazerdoreal]

When you say that they sound different even when tuned to the same target, do you take measurements from your personal phone (e.g. Sonarworks) or did you take the measurements from Oratory/Crinacle etc?

I mean, even if there is a high quality control on high end phones, no headphone does sound exactly the same. So I wonder if their difference really is something "above" Frequency Response or if small deviances from the measurements can explain most of the perceived differences.

For example, these are two measurements of the Hifiman Susvara by Crinacle. Based on which Susvara measurement you base your target on, you will get a noticeably different tuning.

 

[Robbo99999]

There are 12 different configuration possibilities for 45CA fixtures so not all 45CA are created equal

Amir uses the 45CA-10 which complies with IEC 60318-4.
This norm specifies a transfer impedance tolerance range between 100Hz and 10kHz. It provides a 'human-like' load up to 10kHz and allows a 5dB tolerance at 10kHz and 1.5dB between 100Hz and 2kHz.
However, this norm does not specify the tolerance of the frequency response accuracy (logical, as it can be all over the place above a few kHz) which is not the same as the transfer impedance.
The IEC 60318-4 norm only specifies dimensions/shape the pinna and ear canal have to comply to. The 'acoustical load' basically.

The configuration Amir uses is claimed by GRAS to provide a human-like acoustic load up to 20kHz within +/- 2.5dB.
It does not seem to comply to IEC 60318-7 which the B&K5128 does
That HATS has even more 'human-like' shape of the fake ear canal and thus provides an even more accurate 'acoustical load'.
It only specifies the acoustic load though not the accuracy of the frequency response.

Frequency response tolerances are specified elsewhere and merely state in at which frequency the measured response may (= not equal to actually does) deviate.
With diffuse field above 6kHz there is a 10dB tolerance and above 8kHz a 12dB tolerance which is not strange given the measurements of 1 single headphone in various positions of the HD600 below. It is easy to observe that between 100Hz and 1kHz the deviation is small but above a few kHz all bets are off.
Now... we have to realize that in practice narrow (high-Q) dips and peaks are not that audible (smoothed by the hearing) and the variations shown below are not representative to the ways one would wear this on the head. It does, however, show that measurements above a few kHz have to be taken with a grain of salt and thus also a 'highly smoothed' average target has to be taken merely as a 'guide-line' (for each different fixture) and a deviation from that 'target' line in a measurement does not mean it actually deviates as much ... in reality. It only means that specific headphone, on that specific fixture, on that specific 'seating' (or averaged seatings) shows a certain deviation from a very 'smoothed/averaged' drawn line.

'Exact' EQ or believing a single 'measurement trace' is 'exact science' and has to closely follow some 'target line' is equally silly as believing measurements say nothing.
But at least 'science' provides some 'handles' one can hang on to.
Exact EQ looks very nice when applied to a single measurement but IRL it is only accurate for that specific measurement but... at least it might be better than just winging it by ear.
Measurements matter but one has to know its limitations and above all those of headphone measurements.

This, including product variance, all explains why there are different 'squiggles' between measurements done by others, even with the same fixture.

To make this about DCA Expanse again... on all measurements done with 'reputable' measurement gear this headphone, without EQ and a good seal conforms pretty well to the 'Harman' target. How much it deviates and where it exactly deviates is unclear but deviations seem small over the known measurements so quite 'trustworthy' IMO.

(I'm pretty sure that HD600 graph you're showing is more than one unit to put it in perspective as it seemed that you were referring that to be just one unit, it's all his measurements of all his HD600 units he's tested, he sent me that same graph one time when I asked him why the peaks around about 5 & 6kHz had been smoothed in his latest published measurements - this was around a year or more ago, 7th Dec 2020 he sent me that or maybe he posted it in a thread when I asked him about it, one of the two).

But to respond generally to you points, I don't think any of that excludes the validity of EQ'ing accurately to the Harman Curve as a starting point, for the reasons I mentioned in earlier posts, I won't keep dragging it up as the discussion has to end somewhere.

 

[Robbo99999]

Spoiler

Harman's own pinnae (there are two of them) use a modified version of this one : https://www.grasacoustics.com/products/accessories/for-artificial-ears-and-mouths/product/430-kb0070
It was 3D scanned, the geometry was altered a bit for each version (less protruding pinna among other modifications), and then cast in Shore 00-35 silicone for the first one (same as current GRAS anthropomorphic pinna), and a softer silicone still for the second one.
Source : https://www.aes.org/e-lib/browse.cfm?elib=17699

Harman has compared their own pinna(e?) with GRAS' various pinnae, including the anthropomorphic one, but it hasn't been published AFAIK.

If you digitise the traces from various Harman's papers or Sean Olive's measurements and compare them to the various GRAS measurements available the difference can go from quite moderate (HD650 below 6kHz for example) :
View attachment 238587

To potentially more significant :

https://twitter.com/i/web/status/1368458551618805760

This one is interesting as we may presume that the same sample was tested by the same operator (not certain though) on the exact same fixture (minus the pinna difference), and both pinnae are of a soft type.
The poor visibility of the data prevents a clean digitisation but after quite a bit of Photoshop wizardry and perspective correction this can give a rough idea of the difference (please don't directly compare it to other graphs and the Harman target) :
View attachment 238596
It's probably a case by case problem depending on the exact model of headphones.

I don't think that the older KB0065 pinna is in use in most databases available anymore, but comparing it and the newer anthropomorphic pinna can be interesting to assess the importance of the pinna on the overall results ceteris paribus.

SoundStageSolo has published the difference for several headphones at some point in time : https://www.soundstagesolo.com/index.php/equipment/headphones/headphones-text-listing

I haven't bothered to digitise all of it but generally it looks like this :
View attachment 238591
i.e. a general trend line with noise around it.
Oratory1990 has performed a similar test and the trend line is quite similar (also with a good deal of variation around the trend for individual headphones).

TLDR : the difference between Harman's modified pinna and GRAS' anthropomorphic pinna isn't well documented at the moment, but I wouldn't worry too much about it for headphones such as the HD800 up to several kHz, and perhaps a bit more for other designs .

Thanks, interesting. Do all of Sean's published measurements (that he posts in Twitter "recently") use one of Harman's customised pinna's, or does he sometimes use the GRAS KB5000/5001 anthropometric pinnae?

It looks like for the HD650 graph that the various GRAS KB5000/5001 anthropometric pinnae measurements average out to be around what Sean's measurement is showing as I eyeball it, below 6kHz (apart from the bass measurements which are likely more influenced by measurement protocol and decisions on which measurements to discount before publishing the final measurement - Oratory shows best case bass for instance). But yes, from what you're saying we can't be totally sure just how much Harmans modified pinna varies from the GRAS KB5000/5001 anthropometric pinnae, it doesn't seem disastrous though.

 

[MayaTlab]

Thanks, interesting. Do all of Sean's published measurements (that he posts in Twitter "recently") use one of Harman's customised pinna's, or does he sometimes use the GRAS KB5000/5001 anthropometric pinnae?

Unknown to me. Harman's research hasn't always been consistent in that regard either, "Factors that Influence Listeners’ Preferred Bass and Treble Balance in Headphones" was published after "Improved Measurement of Leakage Effects for Circum-aural and Supra-aural Headphones" (where the custom pinnae were introduced) and yet still used the KB0071 (but also used a pair of HD518 as replicator headphones).

 

[Robbo99999]

Unknown to me. Harman's research hasn't always been consistent in that regard either, "Factors that Influence Listeners’ Preferred Bass and Treble Balance in Headphones" was published after "Improved Measurement of Leakage Effects for Circum-aural and Supra-aural Headphones" (where the custom pinnae were introduced) and yet still used the KB0071 (but also used a pair of HD518 as replicator headphones).

I see what you mean there, but you're talking about their published papers, I was just asking about Sean's apparently informal measurements of random headphones that he posts on his twitter, following is an example of a K702 he measured recently for example:

I don't suppose you know if that's the GRAS KB5000/5001 anthropometric pinnae or one of their custom pinnas? (As a probably unrelated aside it is a very strange measurement of a K702 though inasmuch that it doesn't have the two treble peaks that K702 normally have and it has no roll off in the bass!, but Sean did theorise that they may have changed the K702 design slightly in the latest produced units (his was like hot off the press from the factory)). I also see there's no 10kHz notch, that's weird! Actually, looking closer the notch is there but it's not very deep, but his y-axis is covering 70dB rather than the 50dB we're used to seeing, so it makes the notch look smaller.

 

[solderdude]

EQ'ing to the Harman Curve as a starting point

As a starting point... yes... expecting that overly smoothed and averaged plot to be valid for all headphones and expecting it to be accurate when you EQ to that curve... definitely no.

 

[Robbo99999]

As a starting point... yes... expecting that overly smoothed and averaged plot to be valid for all headphones and expecting it to be accurate when you EQ to that curve... definitely no.

Yes, I think & in my own experience too there's scope for different models of headphone to sound somewhat different when EQ'd to the same curve, (& beyond just the obvious unit to unit variation issue theoretically). (So there's the unit to unit variation issue to cope with as well for headphone users seeking out published EQ's). So yeah, I'll certainly agree that it's often just a starting point, it was this way with my HD560s for example - I do my own EQ using Oratory's measurements to the Harman Curve (accurately EQ'd to it for the most part), but then found adding a linear -1.25dB slope change from 20Hz through to 20kHz to the whole headphone on top of the Harman EQ got the best out of it for me.

 

[preload]

Well, the Oratory & Amir rigs (GRAS) are effectively the same as the original Harman rig. The pinna is not exactly the same as apparently Harman used a custom made pinna, but the rigs that Oratory/Amir/Crinacle/Resolve use are all GRAS rigs that are the current best thing in terms of a legitimate "Harman Rig".

As you know, the entire measurement apparatus includes both the GRAS and the pinna. While many other measurement gurus may be using the same GRAS unit, the fact that they are not using the Harman pinna is a big deal. It's not the same.

I don't know how much actual difference you would see in measurements using the original Harman pinna vs the current one used on GRAS, I have a feeling that the main difference was just the flexibility of the pinna, the original Harman pinna was stiff, so I'm thinking the GRAS pinna used mimics the Harman one but is more flexible (increased pliability/malleability) and therefore more like a real human ear in terms of better headphone fitment but someone can correct me on that if they for sure know the differences.

Here is a chart that shows the differences in measurements obtained on the SAME GRAS unit, except with different pinnae attached (blue, green, and yellow). Ignore red.

As you can see, the measurement differences are significant in both amplitude and bandwidth (all the way to 500Hz). The Harman target curve is referenced to measurements made using the Harman rig, which includes the Harman pinna. If you're not measuring using a Harman pinna, the chart gives an idea of the possible error that has been introduced. It's not pretty.

(And if you're not using a GRAS unit, the errors are likely even worse, as we saw a couple of years ago when Olive and Jude measured the same set of headphones...)

Ref: Olive S, Welti T, and Khonsaripour O. "Perception & Measurement of Headphone Sound Quality" slide deck, dated October 16, 2017.

 

[preload]

Here's another example of what I mean about how the same headphones can measure differently across rigs - which implies that unless you're looking at a Harman measurement, you don't really know how that headphone performs relative to the Harman target curve (which is referenced to Harman's rig only).

This is the AKG K371 Pro measured by Harman. (taken from a Sean Olive twitter post)

This is Amir's measurement on his rig. (Also, note the left/right differences above 3khz. Yikes.)

This is Oratory's measurement:

 

[MayaTlab]

Here is a chart that shows the differences in measurements obtained on the SAME GRAS unit, except with different pinnae attached (blue, green, and yellow). Ignore red.

As you can see, the measurement differences are significant in both amplitude and bandwidth (all the way to 500Hz). The Harman target curve is referenced to measurements made using the Harman rig, which includes the Harman pinna. If you're not measuring using a Harman pinna, the chart gives an idea of the possible error that has been introduced. It's not pretty.

Most current reviewers using a GRAS setup now use the anthropomorphic pinna which is softer than the IEC pinna used in that article though .

Personally I am not comfortable making blanket statements on that issue either on the side of "not an issue" or "big deal" until we see more published data. When I digitise some of the measurements from Harman's articles where the use of Harman's own pinna is explicit and align them with other GRAS setup, the difference is moderate until several kHz for some headphones I can recognise. And for some others, it isn't. In all cases other variables than the pinna may introduce or not a deviation of the average coupling for real humans vs. fixture from the average transfer curve (perhaps geometry around the pinna for example).

 

[preload]

Most current reviewers using a GRAS setup now use the anthropomorphic pinna which is softer than the IEC pinna used in that article though .

Great, but that doesn't solve the problem of understanding a headphone's FR deviation from the Harman target curve, which was based on their pinna. It solves OTHER problems, like problems with seal, and getting consistent measurements to compare to other non-Harman measurements, but not the problem of trying to learn a headphone's FR relative to Harman.

Personally I am not comfortable making blanket statements on that issue either on the side of "not an issue" or "big deal" until we see more published data.

That's fair. I have a different perspective - in the presence of reliable data (from Harman) that establishes that different pinna can cause concerning amounts of variation in FR measurements, I would like to see evidence that the currently used pinna are comparable to the ones used by Harman FIRST, before assuming that the currently used pinna will yield comparable measurements to Harman's own pinna. Again, this is from the perspective of determining a headphone's FR deviation from the Harman target.

I guess it's the old adage of: "prove to me that something is unsafe before we stop using it," rather than "prove to me that something is safe before we will use it." The former philosophy is what caused the Challenger disaster in 1986. I choose the latter.

When I digitise some of the measurements from Harman's articles where the use of Harman's own pinna is explicit and align them with other GRAS setup, the difference is moderate until several kHz for some headphones I can recognise. And for some others, it isn't. In all cases other variables than the pinna may introduce or not a deviation of the average coupling for real humans vs. fixture from the average transfer curve (perhaps geometry around the pinna for example).

See my post above that demonstrates differences in measurements from a recent headphone. It's not pretty.

 

[Jimbob54]

I guess it's the old adage of: "prove to me that something is unsafe before we stop using it," rather than "prove to me that something is safe before we will use it." The former philosophy is what caused the Challenger disaster in 1986. I choose the latter.

In relation to headphone frequency response? Probably the most apt analogy .

 

[preload]

In relation to headphone frequency response? Probably the most apt analogy .

OK, fine. I'll save it for the threads on not grounding the chassis of audio equipment.

The data are there showing that different pinnae lead to different measurements. There's even the akg 371 meausrments I've provided showing that the current pinnae used by oratory and Amir still don't produce a measurement that is negligibly close to the Harman measurement. So the data are there. But the human factor here is how to interpret it.

 

[Rthomas]

When you say that they sound different even when tuned to the same target, do you take measurements from your personal phone (e.g. Sonarworks) or did you take the measurements from Oratory/Crinacle etc?

I mean, even if there is a high quality control on high end phones, no headphone does sound exactly the same. So I wonder if their difference really is something "above" Frequency Response or if small deviances from the measurements can explain most of the perceived differences.

For example, these are two measurements of the Hifiman Susvara by Crinacle. Based on which Susvara measurement you base your target on, you will get a noticeably different tuning.

sorry for the late reply

I use Oratory's measurements. Sending each headphone to him for individual measurements would be better but so far I've only sent my LCD-4.