-
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!
Harman curve without ear canal? HD650, DT150_100P, DT770 Pro 250 Ohm, Bose QC25 measured
- Thread starter Thomas_A
- Start date
They'll have to be at least flush with the ear canal entrance to get reasonably useful results past 3kHz or so in my experience - so not in the concha
.If your goal is to perform in-ear measurements and directly compare them with the Harman target minus ear canal effect, I can see a few potential problems here. My understanding of these notions is incomplete, so don't hesitate to correct me here.
From that article (which you can find without paywall) : https://www.researchgate.net/public...mmission_From_Sound_Sources_Close_to_the_Ears
From what I understand what you see here is the difference between measurements done at the blocked ear canal entrance (EEP ?) of the ear simulator and measurements done with the ear simulator with its own microphones (at DRP ?).
You see a peak centred around the ear simulator's canal resonance that drops back to zero at around 5-6kHz (let's for now not bother with the data above 7kHz, that's a whole another kettle of fish). This matches pretty well this (idealised ?) representation of the overall gain we experience, that you've probably already seen :
The thing though, this figure only shows the transfer function between ear canal entrance and DRP for that particular fixture (which I believe isn't the G.R.A.S. used by Harman's research). So you probably shouldn't directly use that figure's data and apply it to the Harman target to get Harman minus ear canal effect. The "gist" of it, however, ie the peak centred at the ear canal resonance, that drops back to zero at 5-6kHz, might be a useful visual guide to follow.
Besides, your own ears are not similar to Harman's ear simulator. If your own head had been in the place of Harman's mannequin in their reference room, the same methodology Harman used to determine their target would have resulted in a more or less different target, particularly at higher frequencies. So you wouldn't be able to directly compare them.
Finally, you have to make sure that your in-ear mics are properly compensated. What properly means here, I have no idea ! I occasionally compensate my own in-ear mics against a UMIK-1 with a speaker in very near-field-ish conditions. Even when limiting myself to frequencies below 800Hz or so there still are moderate differences with Harman's target. Is it because my UMIK-1 isn't perfectly compensated ? Is it because my compensation methodology is flawed ? No idea.
So, what you should rather strive to achieve is not to directly compare your own in-ear measurements with the Harman target minus ear canal effect, but rather first and foremost try to determine what the Harman target would measure like on your own head, with your own microphones, and then only compare your individual HPs measurements with this "individualised Harman target".
A few things you could do :
- below 800Hz or so : buy a pair of ANC headphones with a solid feedback mechanism, no volume dependent EQ, and tight tolerances. These will always strive to deliver the exact same FR in the range where the feedback mechanism operates regardless of most coupling issues (illustration of that phenomenon here). This makes them very good "translators" between ear simulators and your own in-ear mics.
- above 800Hz or so, this is a lot trickier. You might want to read that post I wrote a little while ago on the subject. You may buy a whole bunch of headphones (preferably ones with known good properties to translate well from ear simulators to real heads, ex large open dynamics such as the HD650), that you equalise to Harman, either via presets, or better, by sending your own samples to people owning a Harman compliant fixture (ex : Oratory). Then measure these on your own head, and average the results. This should give you a rough idea of what the Harman target should measure like on your own head, with your own mics. Emphasis here on "rough" : the average of the headphones equalised according to Oratory's presets is not a perfect match for the average of the Harman headphones I've also measured, and its shape in the ear canal gain region is... not what I'd expect (if I were to apply the "gist" of Hammershøi and Møller's figure above). So I doubt that it's a perfect representation of what the Harman target would measure like on my own head. Note, in addition to these difficulties, that blocked ear canal microphones may show some moderate inaccuracies above 2kHz or so and that above 7kHz they may get more significant.
You may PM Mad_Economist about in-ear measurements, he knows a lot more than I do on these matters and unlike me actually has solid theoretical knowledge
.
OP
- Thread Starter
- #3
Thanks, will look into it. I would primarily experiment with different heads and ears to see the amount of variation, as well as looking at relative differences between headphones. It would however be nice with at least an approximate reference.
You mean that you intend to perform in-ear measurements on several real humans ? If you think you can safely feel confident in the data you acquire (particularly above 800-1kHz), I'd be quite interested in looking at it
!Measuring how headphones relatively differ across different test subjects (ie, the "different difference") seems like a more practical thing to do indeed.
IMO that's opening a whole another can of worms.
If you want to perform blocked ear canal entrance measurements, one thing you could do is measure a GRAS fixture with the KB5000 pinna with your blocked ear canal mics, and compare that with your test subjects' results.
That won't give you the Harman target minus the ear canal effect, but could provide some interesting comparative data, particularly for headphones which are sensitive to coupling issues.
OP
- Thread Starter
- #5
You mean that you intend to perform in-ear measurements on several real humans ? If you think you can safely feel confident in the data you acquire (particularly above 800-1kHz), I'd be quite interested in looking at the data
Measuring how headphones relatively differ across different test subjects (ie, the "different difference") seems like a more practical thing to do indeed.
IMO that's opening a whole another can of worms.
If you want to perform blocked ear canal entrance measurements, one thing you could do is measure a GRAS fixture with the KB5000 pinna with your blocked ear canal mics, and compare that with your test subjects' results.
That could provide some interesting data, particularly for headphones which are sensitive to coupling issues.
Yes, my plan is to experiment with at at least three human subjects with clearly different ear and head size, but I may include more. If I find something interesting and worthwhile, I'll post it. Three HPs; HD650, Bose QC25, DT150/DT100 pads in plan. Matched microphones.
Good idea to have a HD650 as a baseline, and I'm very interested in how the QC25 in particular will compare across different test subjects !
So I really recommend trying to perform blocked ear canal entrance measurements (while acknowledging that they may have moderate inaccuracies in terms of the relative differences between headphones in the 2-7kHz range and higher above), and avoid anything that protrudes in the concha. You won't get accurate results past 3-4kHz or so if they do : https://audiosciencereview.com/foru...ted-harman-oe-curve-at-home.28130/post-975888
Sidenote : I don't know if that is the case for the QC25, but the QC35/QC45/700 have a volume dependent EQ that changes the FR according to the internal volume they're set at. So pick a volume and stick to it, otherwise some of the differences you'll see will come from that volume dependent EQ, not the subjects' anatomy.
OP
- Thread Starter
- #7
May I ask why you want to this?
solderdude
Grand Contributor
The only thing this experiment will teach someone is that all pinnae are slightly different and react differently to different headphones as well.
There will be no single 'correction' curve between individuals as that will also differ per headphone.
There will always be pinna interaction and there is also the ear canal which differs between individuals which are not included in this test so it is only a part. Then there is the actual hearing differences between individuals (hearing loss).
It's a fun experiment and it will give some extra insight but that will be all.
One would also need to characterize perception which is more difficult.
Perhaps playing around with Griesingers method could teach you more about differences in perception between individuals.
Maybe not using sound coming from the front but setting speakers up on the right and left of the head ?
There will be no single 'correction' curve between individuals as that will also differ per headphone.
There will always be pinna interaction and there is also the ear canal which differs between individuals which are not included in this test so it is only a part. Then there is the actual hearing differences between individuals (hearing loss).
It's a fun experiment and it will give some extra insight but that will be all.
One would also need to characterize perception which is more difficult.
Perhaps playing around with Griesingers method could teach you more about differences in perception between individuals.
Maybe not using sound coming from the front but setting speakers up on the right and left of the head ?
OP
- Thread Starter
- #11
Why not using loudspeakers from the front at LP? What info would speakers at 0 degree L-R position give other than the reduced head/body interaction of frontal arriving sound? After all there will always be a comparison headphone sound with sound coming from the front.
There is little quantifiable data out there in terms of how headphones behave and vary among individuals, or between individuals and ear simulators. We know they do, but by how much, particularly for anything that isn't a pair of old-school large, open dynamics over-ears, and for anything above 1kHz, is not superbly well known.
At low frequencies we have Harman's article on leakage (no paywall) - perhaps the reference right now : https://www.grasacoustics.com/files...mprovedMeasurementofLeakageEffects_Harman.pdf
And if you find acceptable that the five humans they test their HPs on rotate, Rtings has a more extensive dataset (albeit with no comparison with test fixtures).
At frequencies higher than that, among others :
https://www.aes.org/e-lib/browse.cfm?elib=16877 (some issues with that study though)
https://vbn.aau.dk/ws/portalfiles/portal/227875204/1995_M_ller_et_al_AES_Journal_b.pdf (no paywall)
Figure 9 here : https://www.aes.org/e-lib/browse.cfm?elib=16486
Probably a handful of additional articles, but that's all.
All have shortcomings one way or another.
We also have some data on how headphones "differ differently" across test fixtures (ex : Harman's latest 5128 research).
Just as importantly, we don't know very well whether or not this variation is of the desirable kind (ie express one's own HRTF in a "desirable" way, which could possibly mean "if you were in Harman's own reference listening room, what would the target look like for you ?") or the undesirable one (random, unwelcome variations). I would find it interesting, but assuredly outside the scope of what can realistically be done by forum enthusiasts, to compare individuals' HRTF sets and headphones' FR variation.
So I'm all in for more data, particularly given that there's already quite enough evidence IMO that, even when leakage isn't an issue, they'll differ in dB by more than minimum thresholds of audible difference.
I like @Thomas_A's selection of HPs a lot, this type of trio (large open dynamics over ears, passive closed back - presumed sensitive to leakage issues ? -, Bose ANC headphones) is exactly what I'd start with (the inclusion of the ANC headphones in particular, which haven't been extensively studied above 1kHz).
I'd love to, in addition to real ears measurements, see the relative differences between these same headphones (same sample) on real humans compared to how they differ on a typical "flat around the pinna" test fixture (by shoving the same mics in the ear simulator). None of the ear simulators, including the 5128, are anatomically correct outside the pinna and there is quite a bit of variation among individuals in that regard which may affect coupling even when leakage isn't an issue. And we already know that some of the wiggles one can see on ear simulators with closed headphones in particular, below 200Hz, aren't representative of how they behave on real humans (ex : the K550 in Harman's leakage article, which ear simulator measurements never properly matched real humans measurements, regardless of the pinna used).
Last edited:
solderdude
Grand Contributor
That would only be interesting if you want to know or imitate sounds coming in front of you. The pinna action differs a lot when sounds are coming from the sides. The latter is the case with headphones.
Besides with most stereo systems (which one wants to mimic) the sound is not coming from directly in front of you but rather at an angle (speaker distance and listening distance) where the pinna interaction differs yet again.
Nearfield also has a different angle than home music enjoyment.
So the question would be what the purpose of your experiments is and what you plan to do with the obtained info (out of curiosity on my part).
Why exclude the ear canal where it obviously is part of differences between individuals.
Last edited:
The idea would be that while there may not be a constant transfer function across headphones for each individual overall, for anything that happens past the ear canal entrance, there would be, whether the canal is blocked or open.
To my knowledge there are quite a few articles on the subject of measurements at various points between the ear canal entrance and the eardrum, often related to HRTF measurements. I recall reading an article by Hammershøi and Møller on that subject for example.
Here's is a little bit of data on occluded ear canal entrance measurements vs. DRP measurements (figure 2) : https://www.researchgate.net/public...mmission_From_Sound_Sources_Close_to_the_Ears
For headphones I'm not so certain that it totally holds, but Mad_Economist replicated, on average, that figure's findings with headphones, on his own HATS : https://audiosciencereview.com/foru...-ears-as-from-a-41-000-hats.15359/post-488798
While I personally think that blocked ear canal measurements may introduce some inaccuracies vs DRP (in terms of the relative difference between headphones), possibly above minimum thresholds of detection, I'm tempted to think that, in light of how impractical it seems to be to measure at the eardrum on real humans, with commonly available, medically certified probe tube microphones, it's better than nothing, and possibly even better if compared with measurements done on an ear simulator with the same occluded ear canal entrance microphone.
solderdude
Grand Contributor
I agree. A lot of this has been researched already. I can understand to try to find personal HRTF's. Maybe for 3 persons when the all want to.
Might be interesting to, when they are at it, to compare results from a mic in ear compared to results obtained with Griesinger method (at 70dB SPL as this method is level dependent ?) and between that method with speakers from the side and in front.
Cool. If you are just looking for some kind of a neutral reference point for comparison to your own measurements, then I'd suggest doing some measurements of neutral speakers with a normal stereo in-room configuration for that (rather than directly to the front or sides). Or replicating some of Harman's in-room experiments along a similar lines.
If you're doing measurements from a different reference point than the DRP, then as Maya has already said, you won't be able to compare them directly with the raw measurements made on other rigs that use that reference point. It would still be somewhat interesting to see how the measurements compare to your own neutral targets though. And to compare EQ curves based on that to EQ curves based on the Harman target on other rigs.
Last edited:
If you're uncertain whether the acoustics of your room or speakers are up to this, then you could also try what Harman did in their reference room. And EQ the speakers to a flat in-room response at the listening position using a standard omni-directional mic. Then do some measurements at the same listening position with your in-ear or in-pinna mics. That will give you just the tonal response for each individual's pinnae, minus most of the room-related effects.
Combine the above pinnae responses with the in-room response curve of your choice using something like the stack feature in EAPO's Config Editor, and you should have a reasonably good reference curve for the in-pinnae response of your headphones.
If you want to save a couple steps, then you could simply try to EQ your speakers to match your preferred in-room response curve (rather than a flat response) at the listening position using the omni mic. And then just re-measure their response at the same position with your in-pinna mic, to get your in-pinna target for the headphones. Separating the pinnae's response from the room effects, as in the first example, might make it a bit easier to experiment with some different kinds of in-room response curves though.
There is a wealth of in-room data from which to derive a reference in-room response curve in the spinorama data here btw...
Or you could just use a simple slope that's somewhere in the -1 dB per octave range for this. And then adjust the tilt a bit until it sounds right in your EQ tests.
Last edited:
If you want your target to be extended well enough into the lower frequencies for closed-backs and maybe some planar magnetic headphones, then you might also consider using a sub-woofer with your in-room/in-pinna speaker measurements.
Last edited:
Similar threads
