The Relationship between Perception and Measurement of Headphone Sound Quality

[tonmeister86]

We recently conducted some controlled double-blind listening tests and measurements on six popular headphones to study the relationship between perception and measurement of headphone sound quality. The question we posed was can listeners agree on what headphones they prefer based on sound quality? And can their headphone preferences be predicted based on a set of objective measurements?

 

I've post a summary of the paper with a link to the slides we presented at the recent AES Convention:

 

Audio Musings by Sean Olive: The Relationship between Perception and Measurement of Headphone Sound Quality

[wgb113]

Which was the AKG K701?

 

Bill

[souptin]

As always a very interesting read and clearly very well implemented tests.

[sandyk]

[You will be hard pressed to find a manufacturer who will supply these measurements of their products/QUOTE]

 

Sean

You will be hard pressed to even find a manufacturer that tells you if their headphones comply with the old IEC61938 directive, which specifies that all headphones, irrespective of their impedance, should be driven from a 120 ohm source impedance,

as the AKG K701, Audio Technica W1000 and many other older types do.

 

Regards

Alex

[wkhanna]

Headphones are notoriously difficult to measure accurately.

 

Interesting to see the correlation between the measured & perceived results.

[dalethorn]

Much more difficult to do blind tests because of the feel and fit of the headsets, which give clues to the test subjects.

[tonmeister86]

[You will be hard pressed to find a manufacturer who will supply these measurements of their products/QUOTE]

 

Sean

You will be hard pressed to even find a manufacturer that tells you if their headphones comply with the old IEC61938 directive, which specifies that all headphones, irrespective of their impedance, should be driven from a 120 ohm source impedance,

as the AKG K701, Audio Technica W1000 and many other older types do.

 

Regards

Alex

 

Alex,

I'm wondering how important this is? How many input sources have such high output impedances? The iPod/iphone are much less than 120 ohms, as are most receivers.

[tonmeister86]

Much more difficult to do blind tests because of the feel and fit of the headsets, which give clues to the test subjects.

 

Yes, this is true. We considered anesthetizing subjects from the neck up but the lawyers wouldn't allow us to for some reason

 

There are virtual headphone listening test methods where you can simulate the different headphones over a reference headphone that is equalized to the measured linear magnitude response of the target phones. That makes the comparisons blind w/o tactile/comfort cues; of course, it ignores the nonlinear behavior of the headphone but binaural recordings can capture that.

 

We tried this with some success, and are presenting a paper on this at the upcoming AES 51st Loudspeaker/Headphone conference in Helsinki in August.

[tonmeister86]

As always a very interesting read and clearly very well implemented tests.

 

Thanks. We have learned a lot from doing these listening tests. They are a pain in the butt to do, which is probably why so few people try.

[sandyk]

 

Alex,

I'm wondering how important this is? How many input sources have such high output impedances? The iPod/iphone are much less than 120 ohms, as are most receivers.

Hi Sean

It's very important with headphones such as the K701 which have steeply rising impedance at higher frequencies.

See http://www.innerfidelity.com/images/archivesart/806AKGFIG1.jpg

Headphones like these have exaggerated HF response which can be very fatiguing to listen to for extended periods unless driven from a source impedance of 120 ohms. When doing that, you also need to use an amplifier with high voltage swing capabilities such as one using + and -20V supply rails for best results. I have previously posted a project in a U.K. based forum for a Class A headphone amplifier, where quite a few members report vastly improved results with their AKG 701s when using this amplifier .

In fact, another group buy is presently being organised by a couple of members to make more of these amplifiers. I also get markedly improved results from this HA with my Audio Technica W1000 with 120 ohm series output resistors.

I won't post a link to this project, but I can PM you a link if you wish to see the reports by those using this amplifier with their

K701s. The ebay Panda kit or ready made Panda HA, when used with 120 output resistors also gives excellent results with these difficult to drive headphones, as it also uses +-20V supply rails. Varying the output impedance of headphone amplifiers to improve tonal balance with certain types of headphones is quite an old technique.

Regards

Alex

 

P.S.

It's an 80 page thread !!!

[dalethorn]

Alex, I'm wondering how important this is? How many input sources have such high output impedances? The iPod/iphone are much less than 120 ohms, as are most receivers.

 

I remember for a long time the Beyerdynamic DT48 series recommended a 120 ohm or so output impedance. Sennheiser says their HD800's 'flat' response curve was measured freefield, which I think might also have something to do with the 120 ohm output. Then on the flipside, the Meridian DAC made in Britain that sells for $300 USD (pricy for a mini-DAC) started out with a high output impedance on the headphone out, and they were so heavily lobbied for a low output impedance for 'modern' headphones that they capitulated and reissued it with the new spec.

[sandyk]

Sennheiser says their HD800's 'flat' response curve was measured freefield, which I think might also have something to do with the 120 ohm output.

Dale

I have seen several reports in that other forum about the HD800 sounding best when fed from a 120 ohms source impedance.

Alex

[dalethorn]

Dale - I have seen several reports in that other forum about the HD800 sounding best when fed from a 120 ohms source impedance.

Alex

 

I'm glad to see these posts because this information tends to get pushed aside with the ongoing avalanche of i-devices and smartphones and fashion headphones.

[sandyk]

Dale

That garbage from Nwavguy about all headphones should be fed from a very low impedance, hasn't helped either. Even the very conservative Solderdude disagrees with him on this issue.

Alex

[Gregor Samsa]

I'm glad you learned a lot. By not revealing which headphone was which, Mr. Olive, this is useless to the rest of us.

 

I certainly haven't heard everything, but I've heard a few. So far, Audez'e blows everything else into the weeds.

[dalethorn]

Dale - That garbage from Nwavguy about all headphones should be fed from a very low impedance, hasn't helped either. Even the very conservative Solderdude disagrees with him on this issue.

Alex

 

NwAvGuy doesn't get much love beyond his amp design for many of the newer headphones with lower impedances. But since the classic designs and measurements done by Sennheiser, Beyer and others are being ignored by the fashion set today, most people can't connect the dots. It will be most interesting to see how this topic runs, and whether the AES people will address these points.

[mav52]

Headphones, some work for some and some don't. Only the ears know not to mention as with anything audio a good match up with a headphone amp.. Nice write up.

[sandyk]

One thing that many people seem to agree on , is that headphones with the flattest frequency response up to 10kHz, and no huge dips or peaks below 200Hz usually sound best.

The Stax SR507 is a good example. I have enlarged and tidied up the original graph a little.

Regards

Alex

 

[dalethorn]

Seeing that graph immediately made me think of the Stax SR009 curve, and a couple of the planar headphone curves. Most of the dynamic headphone curves don't look similar to that, as measured by the popular sites anyway.

[Kiwi2]

One thing that many people seem to agree on , is that headphones with the flattest frequency response up to 10kHz, and no huge dips or peaks below 200Hz usually sound best.

 

And you would assume that this applies to loudspeakers in a room as well.