Head-Fi.org › Forums › Equipment Forums › Sound Science › AES 2012 paper: "Relationship between Perception and Measurement of Headphone Sound Quality"
New Posts  All Forums:Forum Nav:

AES 2012 paper: "Relationship between Perception and Measurement of Headphone Sound Quality" - Page 4

post #46 of 130
Quote:

Originally Posted by MeatusMaximus View Post

 

[snip, emphasis added]

 

As for flat being preferred, yes it turned out that in this limited study that was more or less the case, although the Audeze is not really flat (you are in fact saying it is DF).  That fact and the limited scope of our first set of tests, would not allow us to say that "flat is preferred".  We were only speaking of the six headphones we studied.  We are currently working on a new test, and so far, the "flat" (that is, equalized absolutely flat at DRP) headphone is NOT likely to be preferred over the other eq's we are investigating.  As Sean said, "Stay Tuned"!!

 

I guess the thing I was wondering about "why not use same headphone + EQ to remove other confounding variables?" has been answered.  I'll stay tuned, thanks!

 

 

Correct frequency response curve aside, do you have any thoughts regarding crossfeed or more sophisticated compensation for playing back stereo recordings on headphones, to simulate L speaker -> R ear (and R speaker -> L ear) transfer function? 

post #47 of 130

We are also planning to do some "virtual" tests, using a single headphone with multiple eq's.

 

As for crossfeed etc, that is another subject.  We are focusing on simple target curves for now.

post #48 of 130

I've been a follower of Dr.Toole and Dr.Olive for a long time. It is truly good to know Harman Kardon is investing their resources in technological advancement of headphones. Thank you.

 

Your plot is a full -10 dB from 2 kHZ up, with respect to the ISO DF you reference.  For me that would be a large enough difference to say that it is not "DF" equalized, with respect to the ISO DF, or any other DF curve (there are a number of other ones and not all the same).  Your "tolerance mask" may be larger than mine smile.gif

 

The tendency of high-frequency radiation, which is distinctive for a random-incidental sound field, is there, and the ear canal resonance is there too. I'd say LCD-2 v2 has more of a negative sloped DF response, well within the "tolerance mask" in accordance with ITU-R rec BS.708. There is absolutely nothing subjective with my tolerance mask.

 


As for flat being preferred, yes it turned out that in this limited study that was more or less the case, although the Audeze is not really flat (you are in fact saying it is DF).  That fact and the limited scope of our first set of tests, would not allow us to say that "flat is preferred".  We were only speaking of the six headphones we studied.  We are currently working on a new test, and so far, the "flat" (that is, equalized absolutely flat at DRP) headphone is NOT likely to be preferred over the other eq's we are investigating.  As Sean said, "Stay Tuned"!!

Due to much decisive manner of representing some of his conclusions, many of my acquaintances were confused in regards to what to follow when it comes down to predicting subjective preference. As you may know, the sub clause 8.6 of IEC 60268-7 clearly states simulator measurements to be only considered beneficial for industrial usage; a subjective tonal quality shall not be of a concern due to much deviation between simulator measurements & subjectively assessed results. Yes, we were having quite a difficulty with this year's AES papers. biggrin.gif 

 

 


On another note, I am going to start a thread on measurement of headphones.  The subject will be "to use pinna or not to use pinna".  Not trying to cross-post, but I though I'd mention it so you can look out for it.  

For that, I'd say "nay"! Not to mention a free-air measurement is not applicable with many headphones, even a simplified artificial ear measurement does not reveal anything related to subjective preference. I think Type 3.3 pinna simulation is vital.


Edited by udauda - 11/18/12 at 4:37pm
post #49 of 130
Quote:

I'd say LCD-2 v2 has more of a negative sloped DF response, well within the "tolerance mask" in accordance with ITU-R rec BS.708. There is absolutely nothing subjective with my tolerance mask.

 

I think perhaps you are referring to the DF-normalized plot when you say that about the slope (?)  I am thinking abut the raw measured response (no assumptions about how it is normalized, corrected, etc.).  The DRP/coupler response definitely slopes up slightly as you go up in frequency, with only a very shallow 3 kHz resonance.  When I get to work tomorrow I could put up a plot.

 

The tolerance mask of the ITU-R BS.708 would not allow the deviation you showed, would it?  I'm seeing  +/- 2 dB at mid frequencies, and a max of +/- 4 dB at 16 kHz in the spec.  How would that accommodate the broadband -10 dB difference your plot shows?  Did ITU release a new version with a larger tolerance?  wouldn't surprise me!

 

As for our conclusions being decisive, I don't think they were intended to make conclusions outside our own results.  For example:

 

 

Quote:
..The most preferred headphones generally had the smoothest and flattest measured amplitude response when measured on GRAS 43AG ear simulator.

 

The operative word is "had", not "have".  We are not making larger inferences yet, just reviewing our own results.   I would be the first to say that much more tests need to be done to make broader conclusions.  As for simulators, I don't think there has been enough VALID subjective data gathered to say how well objective and subjective measurements match.  That is what we are trying to contribute.  Anyone can make coupler measurements, but getting good subjective data in a controlled double blind test paradigm is not easy.  

 

I hope you will respond about the pinna or no pinna in the other thread, which I started.

 

Cheers

post #50 of 130

I think perhaps you are referring to the DF-normalized plot when you say that about the slope (?)  The tolerance mask of the ITU-R BS.708 would not allow the deviation you showed, would it?  I'm seeing  +/- 2 dB at mid frequencies, and a max of +/- 4 dB at 16 kHz in the spec.  How would that accommodate the broadband -10 dB difference your plot shows?  Did ITU release a new version with a larger tolerance? 

 

Just checked the reference, and yes- the ITU tolerance has to be doubled in order to accomodate the slope. I was thinking about the good old DIN tolerance curve, sorry about causing the confusion. Regardless, This plot shall provide the general idea on what I've been saying:

 

700

Even if it comes down to a matter of perspective, I would still confidently say it is a sloped diffuse-field response with -5dB / oct tilt, of which Etymotic Research & Sennheiser implement on their ER-4P & HD650. This type of equalization has been around for a long period of time in the market, so there is nothing new about it.

 

As for our conclusions being decisive, I don't think they were intended to make conclusions outside our own results.  

And that's what we've been wondering. Thanks for making that clear! 


Edited by udauda - 11/18/12 at 10:01pm
post #51 of 130

OK, it could be called a DF with -5dB tilt I suppose.  I think if you listened to the ISO DF and the Audeze and switched back and forth you would hear a huge difference though.

 

FWIW, here is the curve we get for the Audeze on Gras 43AG coupler.

 

Nevermind, I don't have permission yet!

post #52 of 130
Thread Starter 

Another question for Tonmeister and Meatus along the lines of "virtual" subjective testing:

 

Has there been thought given to measuring with a dummy head the response from loudspeaker playback, as a reference for desired frequency response? I know Harman uses a related technique to "virtualize" a vehicle's interior listening environment as a way to subjectively assess the vehicle's playback system, but what about the other way around? That is, to use Harman's own world-class speakers and listening rooms to assess what the ideal headphone response should be?

 

I'm basically thinking of what the Smyth Research Realiser A8 does, but restricted to frequency response only, without their other processing (for a start). Many users have reported great results with this system, and I can imagine that it can make for a good portrayal of what headphones would ideally output.

post #53 of 130

We are doing looking into that now, JMS!

post #54 of 130

I had a thought today and I was wondering what everyone thought of this:

 

I think that a reason that the preferred target curve for most people seems to be rolled off treble, slightly more rolled off than "standard diffuse field eq curves," is a phenomenon which I've read about and have confirmed for myself while listening to a sine wave generator with several different pairs of on/over ear headphones: There's a peak I can hear with every pair of headphones at around 7.5 kHz "due to the acoustic impedance mismatch between transducer, ear canal and ear drum causing a half wavelength resonance in the canal." Source on quote: http://www.linkwitzlab.com/reference_earphones.htm

 

The peak is usually a 7-10 decibel peak to my ears.

 

The funny thing is... I've never seen this peak show up on any headphone target curves.

 

...should headphone manufacturers be targeting this resonance in their headphones' frequency response? Should HRTF compensation curves be accounting for this resonance?

 

InnerFidelity's dummy head doesn't seem to be picking up this resonance in its raw measurements. What is different about the dummy head compared to a real human head which is causing that resonance to not show up? How can we create a measurement system which reliably reproduces said resonance?

 

Thoughts, all?

post #55 of 130

The only spike I've heard is a 4-5 dB jump around 12-14 kHz with low bitrate lossy files. A friend of mine calls it digital whistle.

post #56 of 130

I've posted a summary of the paper in my blog yesterday. There is a link to the slides from our talk as well.

 

http://seanolive.blogspot.com/2013/04/the-relationship-between-perception-and.html

post #57 of 130
Quote:
Originally Posted by Tonmeister2008 View Post

I've posted a summary of the paper in my blog yesterday. There is a link to the slides from our talk as well.

 

http://seanolive.blogspot.com/2013/04/the-relationship-between-perception-and.html

 

Thanks a lot.  Reading through right now.

 

All I'll say at this point is that I approve of the classical music selection, ensemble, and conductor.

post #58 of 130

 

 

guessing game cheat sheet? - clearly the head/ear simulators don't match perfectly

 

HeadRoom graphs only plot 4 at a time - so QC15 added below

 

there appear to be 3 models called "Crossfade", M-80, LP2 are the ~$200 models on HeadRoom's database


Edited by jcx - 4/23/13 at 3:55pm
post #59 of 130

Guessing game?

 

1 - Audeze LCD-2  (flat subbass extension, treble extension)

2 - AKG K701   (bass rolloff at 100 Hz, upward slope)

3 - AKG K550   (elevated bass but otherwise somewhat diffuse-field, 9 kHz spike)

4 - Bose QC15  (like 3, but not 9 kHz spike)

5 - Beats Studio  (hole in midrange, sharp rolloff under 40 Hz)

6 - V-Moda Crossfade  (hole in midrange, subbass issues, treble extension issues)

 

Wait, that's just the order they're listed, except I swapped 4 and 5.  Beats Studio should be the one that rolls off sharply under 40 Hz.

post #60 of 130

I'm wondering how you came up with the "perceived spectral balance" in slide 25.

 

For example HP4 clearly shows boosted bass and treble, but the perceived spectral balance looks like HP2.

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › AES 2012 paper: "Relationship between Perception and Measurement of Headphone Sound Quality"