[headradio]

I'm in an auditory perception lab at MIT, and we're looking for the most "objectively flat" headphones for doing experiments: that is, the headphones that by objective measurements have the flattest frequency response as measured at the eardrum. We also want to minimize any colorations like harmonic distortion.

Today we measured the frequency response of my Senn HD650 using a silicone ear/ear canal and found a peak at 2-4khz, otherwise flat. It seems like the Denon AH-D2000 seem to be the way to go given the frequency responses as measured on headphones.com but I was wondering if anyone would have any advice on this.
 
http://www.headphone.com/headphones/denon-ah-d2000.php#tabs
 
Thanks!

 

[MalVeauX]

Quote:

I'm in an auditory perception lab at MIT, and we're looking for the most "objectively flat" headphones for doing experiments: that is, the headphones that by objective measurements have the flattest frequency response as measured at the eardrum. We also want to minimize any colorations like harmonic distortion.

Today we measured the frequency response of my Senn HD650 using a silicone ear/ear canal and found a peak at 2-4khz, otherwise flat. It seems like the Denon AH-D2000 seem to be the way to go given the frequency responses as measured on headphones.com but I was wondering if anyone would have any advice on this.
 
http://www.headphone.com/headphones/denon-ah-d2000.php#tabs
 
Thanks!

 
Pffft. I would expect MIT to make their own headphone to suit the need of the experiment.
 
Half serious. Joking a little. Kinda serious. Kinda joking. Can't decide.
 
You're not going to find a flat headphone friend. The D2000 is less flat than the HD650 probably. Graphs don't tell you hardly a thing. Your measuring setup and the setups used by HeadRoom & InnerFidelity are all different and thus show different graphs with values that exist and don't exist on each one. As an experiment, that should tell you the first problem with the bias control...
 
If your experiments are not crucial, then maybe look at things like the KRK KNS 8400, Mr Speaker Mad Dog, Ultrasone Signature Pro, Fostex TH900 (order of cost increasing...).
 
Honestly one of the new Mad Dogs with Alpha Pads would probably be the best "flat" headphone for $300 that you're going to find.
 
But that's my ear talking. Not a microphone in a dumby. And ears suck, so not really a good judge compared to a graph that tells... oh wait... 
 
Very best,

 

[bluepumpkin]

Why do you want a flat response at the eardrum? Sounds in a normal environment are filtered by the different parts of the ear (canal, concha, and pinna), as well as the head and chest, resulting in a frequency response that deviates very significantly from flat at the eardrum. The exact frequency response and depends on the location of the sound source as well as reflections from the environment (see diffuse-field and free-field curves). I don't think that there are any headphones that produce a flat response at the eardrum because they wouldn't sound very good, since they would be severely lacking in high midrange and low treble.
 
Typically, headphones are designed to produce sound similar to properly set-up loudspeakers. The diffuse-field curve is a decent approximation for this.
 
If you *do* actually want a flat response at the eardrum, your best bet would be to find headphones that have a very smooth, reproduceable frequency response and then equalize them (IEMs would be better for this).

 

[AzraelDarkangel]

Beyerdynamic DT48A is probably the flattest frequency response headphone I've heard of.

Joshua

 

[headradio]

@Malveaux
 
Measuring setups can be quite different from place to place, mostly because they might be measuring different things. We were measuring the response of a microphone placed inside of a tube simulating the ear canal with a silicon ear attached, whereas as far as I can tell, they were measuring the response of a microphone right next to the headphone.
However, I don't see any significant differences between the graphs shown at inner fidelity and headphone.com for the headphones mentioned here. Thanks for pointing out inner fidelity to me.
In addition, if you assume linear time invariance of the setups it's pretty easy to make informed comparisons. HD 650 was almost perfectly flat except for a bump at 2-4khz, so if I find a website where the response of the HD650 is identical to the response of another headphone except for less between 2-4 khz, then that's probably the headphone I'm looking for.


@Bluepumpkin
The idea is to be able to drive the auditory system in a predictable manner using synthetic sounds without having to compensate for the frequency response of the headphones. We played a sine sweep and measured the response of the Senns at a simulated eardrum. It was actually quite close to flat, except for the aforementioned bump at 2-4khz. Because of the fragility and difficulty of insertion of IEMs we've kinda ruled those out.

 

[KamijoIsMyHero]

try a planar magnetic, maybe the HE500

 

[bluepumpkin]

Quote:

as far as I can tell, they were measuring the response of a microphone right next to the headphone.

 
Nope. Both InnerFidelity and headphone.com use a dummy head and measure the frequency response at the eardrum. The frequency response you see is also compensated. Headphone.com uses the diffuse field curve (I think), and InnerFidelity uses the ID curve (read more here: http://www.innerfidelity.com/content/headphone-measurement-proceedures-frequency-response).
 
If you look at the raw, uncompensated frequency response from InnerFidelity (the grey lines at the bottom of the frequency response graph), you'll notice that all good headphones have a peak centered at ~3k. As I said before, they would not sound correct if they measured flat at the eardrum.

 

[obobskivich]

IF puts up compensated data, so does Headphone.com. So the question is - do you want perceived flat, or raw flat? (Not that either exists, but it sounds like there's not a lot of refinement in your question - hence why I'm asking). Most headphones will have a few peaks on the upper-end to try and sound flat to human listeners. Raw flat will not sound flat to human listeners (due to equal-loudness and outer-ear interaction); there's ITU standards (specifics of which escape me right now) that define both phenomenon. Basically what you want, as you want it, cannot exist. My advice would be to look for headphones that are built and sold for audiometry, like the Beyerdynamic DT48A.00 (which, as far as I know, are no longer made), or Sennheiser HDA280, which will include individual measurements, and can be calibrated to a relatively "flat" response (and it sounds like you have the equipment to check this). Here's the page for the Sennheiser 280:
http://en-de.sennheiser.com/audiometers-closed-dynamic-ear-protector-headphones-high-frequency-testing-hda-280

I'd also suggest either consulting with an audiologist, or at the bare minimum doing a literature review related to audition and perception studies.

 

[MalVeauX]

Quote:

@Malveaux
 
Measuring setups can be quite different from place to place, mostly because they might be measuring different things. We were measuring the response of a microphone placed inside of a tube simulating the ear canal with a silicon ear attached, whereas as far as I can tell, they were measuring the response of a microphone right next to the headphone.
However, I don't see any significant differences between the graphs shown at inner fidelity and headphone.com for the headphones mentioned here. Thanks for pointing out inner fidelity to me.
In addition, if you assume linear time invariance of the setups it's pretty easy to make informed comparisons. HD 650 was almost perfectly flat except for a bump at 2-4khz, so if I find a website where the response of the HD650 is identical to the response of another headphone except for less between 2-4 khz, then that's probably the headphone I'm looking for.

 

 
Kinda missed my point, but I get what you're saying.
 
The beauty of all this measuring is that regardless of how it measures, it doesn't mean it sounds that way. The HD650 is hardly flat, to my ears.
 
Very best,

 

[DS21]

Quote:

I'm in an auditory perception lab at MIT, and we're looking for the most "objectively flat" headphones for doing experiments: that is, the headphones that by objective measurements have the flattest frequency response as measured at the eardrum. We also want to minimize any colorations like harmonic distortion.

Today we measured the frequency response of my Senn HD650 using a silicone ear/ear canal and found a peak at 2-4khz, otherwise flat. It seems like the Denon AH-D2000 seem to be the way to go given the frequency responses as measured on headphones.com but I was wondering if anyone would have any advice on this.
 
http://www.headphone.com/headphones/denon-ah-d2000.php#tabs
 
Thanks!

 
Have you checked out Dr. Olive's recent AES paper, "The Relationship between Perception and Measurement of Headphone Sound Quality"? http://www.aes.org/e-lib/browse.cfm?elib=16486
 
While expensive, it seems the Audeze LCD-2 would be a good choice. I don't have any personal experience with the device, so I can't speak to its reliability, etc.

 

[vid]

The Sennheiser HD 25 (whatever version, can't remember) seems often used in studies of speech perception. They measure flat-ish in the mids, but other than that, I'm not sure of the selection criteria there.

 

[chewy4]

The Paradox by LFF is the most objectively flat headphone that I've seen.

 

[vid]

Quote:

The Paradox by LFF is the most objectively flat headphone that I've seen.

 
Tyll's plots have them pretty ragged. The modded Aiwa 500, for instance, is much cleaner, but also the usual LCD and Hifiman and whatnot are straighter.

 

[chewy4]

Quote:

 
Tyll's plots have them pretty ragged. The modded Aiwa 500, for instance, is much cleaner, but also the usual LCD and Hifiman and whatnot are straighter.

I didn't see his measurements before, that really isn't that great. Plus a lot of distortion.

 

[vid]

It's just one pair, though. Maybe Tyll got a crud deal there.
 
Then again, modded headphones in general are probably not ideal for science - repeatability and all.