[basslova]

I have my subjective listening preferences (Don't we all?)

but I would like to know which headphones are the best at measurements such as CSD, Square Wave Response, Frequency Response (+ HRTF?), etc.

For example, many audiophiles claim Stax is the holy grail, but from some measurements such as Impulse response, I've seen them slower than a lot of dynamics. Again the point of this is not to bash the high end headphones, I just want to see which headphones perform the best at objective tests.

 

[NA Blur]

Try Tyll's data over at Innerfidelity.  He probably has the most complete list of available data so you can decide yourself.  Otherwise you are going to end up with a pile of opinion.
 
http://www.innerfidelity.com/headphone-data-sheet-downloads

 

[basslova]

Quote:

Try Tyll's data over at Innerfidelity.  He probably has the most complete list of available data so you can decide yourself.  Otherwise you are going to end up with a pile of opinion.
 
http://www.innerfidelity.com/headphone-data-sheet-downloads

I am downloading some of these. I dislike the format they are though and wish they were viewable in a web browser. Also, if it's objective, wouldn't opinions and subjective preferences have little meaning?

 

[mikeaj]

Plenty of web browsers have built-in (or via extensions maybe) pdf viewers, though likewise, I think the format at IF could be better.  Those are missing some information that could be useful; in addition, I'm kind of skeptical as to the value of the square wave plots.  Those are mostly just a snapshot of frequency response and phase response, sampled at just certain (arbitrary) frequencies, with some limited CSD-type information on top.  The combination of all these factors implies a sophisticated complexity, yet these cast a narrow net, only looking at a small set of particular frequencies.  In my opinion the complexity also obscures some of the information of interest.
 
 
Anyway, just because the tests result in numbers doesn't mean that there is only one way to analyze and interpret the results.  How do you weight the importance of FR vs. CSD, or anything else?  (actually, if you want to know an answer, maybe you need to supply an objective function...)  For some parameters, I'm not even sure if there is any real consensus or rock-solid indisputable research on the importance and effects, though there is probably more research along these lines than some naysayers assume.  I suspect that some few headphone designers out there know more about these things than any hobbyists or journalists commenting and making measurements of their own.
 
Furthermore, are you looking for the device with the best performance on the test dummy's ears (a reasonable goal, from an academic point of view) or on yours?
 
 
That said, if you're looking for a certain parameter only and ranking headphones by lowest THD or any other single category, I think that can be managed reasonably, though still there would be arguments over the importance of the distribution of harmonics and so on.
 
 
Taking a very casual approach and eyeballing which seem to do better with some kind of arbitrary weighing, you probably should look at some of the electrostatics (don't forget Senn Orpheus), modern high-end planars like HE-6 and LCD-3, and HD 800.  Actually, these happen to be the headphones you might look at based on peoples' subjective assessments.  But maybe somebody else has a better list.  In many ways, some older flagships or reference models like HD 600 and DT 880 seem to do a lot better than many more-expensive models these days.

 

[anetode]

Quixotically the benchmark for objective measurements in audiometry is the beyerdynamic DT48 A.00, which has horrid measurements if it's anything like the other DT48s.
 
Quote:

Anyway, just because the tests result in numbers doesn't mean that there is only one way to analyze and interpret the results.  How do you weight the importance of FR vs. CSD, or anything else?  (actually, if you want to know an answer, maybe you need to supply an objective function...)  For some parameters, I'm not even sure if there is any real consensus or rock-solid indisputable research on the importance and effects, though there is probably more research along these lines than some naysayers assume.  I suspect that some few headphone designers out there know more about these things than any hobbyists or journalists commenting and making measurements of their own.

 
This is why measurements can be deceptive. I'm skeptical of the value of impulse-response measurements in particular as people tend to ascribe any number of audible features to time-domain behavior. Consider the mythical speed of the electrostats.

 

[mikeaj]

Quote:

This is why measurements can be deceptive. I'm skeptical of the value of impulse-response measurements in particular as people tend to ascribe any number of audible features to time-domain behavior. Consider the mythical speed of the electrostats.

 
Agreed.  I mean, many of these lengthily-described differences in sound are being attributed to features in the low single-digits milliseconds range (or less, for some analysis of 300 Hz square waves).  I could be way off base, but that kind of rubs my intuition the wrong way.
 
Given the way (that I think I understand) people hear, I'd think that most of the differences are really just in the frequency response magnitude, with nonlinear distortion mostly evidenced by THD levels as a secondary factor, other things being even less important.

 

[liamstrain]

One of the problems is in mapping a measurement to perceived sound quality. 
 
We can easily crunch numbers and say, X headphone has a flatter FR than Y headphone - but it is harder to say for certain that a perfectly flat FR is what sounds best.*
 
 
*not the best example, we do have some data to support mapping FR to quality - but you get my point. 
 
With that said - an argument can be made for objective measurements as - as close to reproducing the original signal input as possible. 

 

[gnarlsagan]

Including iems the obvious choice is the ER4S.

 

[ultrabike]

It probably depends on what the baseline for neutral (flat) frequency response is. However, AFAIK the best (un-modded) open headphones in terms of measurements are: STAX SR-009, Orpheus HE90, and the HE-6:
 
STAX SR-009:

 
 
HE90:

 
HE-6:

 
 
Some IF datasheets for convenience: 
http://www.innerfidelity.com/images/StaxSR009.pdf
http://www.innerfidelity.com/images/HiFiMANHE6.pdf
 
AFAIK, the FR measurements from IF and ER differ mainly on the absolute flat frequency response baseline. For this reason, it is usually advised to use a more relativistic approach when making an assessment, by comparing measurements from the same source (or obtained using similar methodology and baseline), and further correlate to our own personal experience. That said, it seems both sources agree in that the above headphones measure "better" than their peers. These "reference" headphones are not perfect, again, just well measuring relative to their peers.
 
Regarding CSDs, clean frequency response magnitude usually corresponds to clean CSD behavior. However, the HE6 does seem to have some ringing at 5kHz (note there is a slight discontinuity in the FR at that frequency.) Regarding non-linear distortion, the above have very little across the audio band (relative to other headphones.) Square waves are a function of frequency response / impulse response. If FR is flawed, one can have certain square waves do great if the frequencies of these particular square waves are "lucky" enough to avoid the FR problem regions, other square waves will go very wrong.
 
To complicate matters, there are personal preferences, and media is not always recorded with neutral transducers in mind (or just poorly recorded and/or intentionally "colored".)

 

[OJNeg]

Quote:

AFAIK, the FR measurements from IF and ER differ mainly on the absolute flat frequency response baseline. For this reason, it is usually advised to use a more relativistic approach when making an assessment, by comparing measurements from the same source (or obtained using similar methodology and baseline), and further correlate to our own personal experience. That said, it seems both sources agree in that the above headphones measure "better" than their peers. These "reference" headphones are not perfect, again, just well measuring relative to their peers.
 
Regarding CSDs, clean frequency response magnitude usually corresponds to clean CSD behavior. However, the HE6 does seem to have some ringing at 5kHz (note there is a slight discontinuity in the FR at that frequency.) Regarding non-linear distortion, the above have very little across the audio band (relative to other headphones.) Square waves are a function of frequency response / impulse response. If FR is flawed, one can have certain square waves do great if the frequencies of these particular square waves are "lucky" enough to avoid the FR problem regions, other square waves will go very wrong.
 
To complicate matters, there are personal preferences, and media is not always recorded with neutral transducers in mind (or just poorly recorded and/or intentionally "colored".)

 
I notice the IF measurements show square wave response at 30Hz and 300Hz. So it stands to say that if a headphone has better response at these two frequencies, it will at least appear to have good square wave response, correct?
 
More specifically, let's say we look at the HD598. http://www.innerfidelity.com/images/SennheiserHD598.pdf
 
I notice that below 1kHz, there are two "major" notches. One at ~80Hz, and another at ~600Hz. As I understand, this is due to some sort of deadening caused by the size of the headphone chamber or earcups. What would the square wave response look like at those two frequencies? In other words, outside of the simple damping conditions, how much do other factors (such as earcups and positioning on the head) affect square wave response?

 

[ultrabike]

Quote:

 
I notice the IF measurements show square wave response at 30Hz and 300Hz. So it stands to say that if a headphone has better response at these two frequencies, it will at least appear to have good square wave response, correct?
 
More specifically, let's say we look at the HD598. http://www.innerfidelity.com/images/SennheiserHD598.pdf
 
I notice that below 1kHz, there are two "major" notches. One at ~80Hz, and another at ~600Hz. As I understand, this is due to some sort of deadening caused by the size of the headphone chamber or earcups. What would the square wave response look like at those two frequencies? In other words, outside of the simple damping conditions, how much do other factors (such as earcups and positioning on the head) affect square wave response?

 
So it stands to say that if a headphone has better response at these two frequencies, it will at least appear to have good square wave response, correct?
Not necessarily. A square wave is made up of its fundamental frequency (30Hz or 300Hz in this case) and it's odd harmonics. If the headphone does not have a flat magnitude response and linear phase at the odd harmonics relative to the fundamental of the square wave, the square wave will not appear good. Here is a 300Hz square wave along with it's FR.



I notice that below 1kHz, there are two "major" notches. One at ~80Hz, and another at ~600Hz...
The HD598 80Hz notch is fairly close to the 90Hz component of a 30Hz square wave (third harmonic - first and most influential odd harmonic) and will influence the shape of the HD598 30Hz square wave response. Also note that, in the HD598 uncompensated FR at IF, the 30Hz component is down (on average) about 5dB from 90Hz. This difference will contribute to the shape of the 30Hz square wave response of the headphone. The 600Hz notch will probably have a lower impact on both the 30Hz and 300Hz square wave responses because these square waves do not have frequency components at that frequency (no even harmonics on a square wave).
 
Outside of the simple damping conditions, how much do other factors (such as earcups and positioning on the head) affect square wave response?
The response and limitations of the driver will more than likely contribute to the shape of the square wave as it contributes to the shape of the FR.
 
EDIT: Given that earcups and positioning on the head affect FR, they will affect square wave response... Based on the HD598 IF results, intuitively the 30Hz square wave would probably be more affected by positioning than the 300Hz square wave. My reasoning for this is that I believe the uncompensated FR responses are taken at different positions, and they show relatively large variation close to 30Hz (a little over 5dB) where as 90Hz, 300Hz and 900Hz show relatively small variations.

 

[Eisenhower]

Wouldn't it make sense to use a well known headphone as a benchmark reference? Like the HD600?
The point of a benchmark is that it is something people are very familiar with, and hence is valuable to compare headphones to it.
 
If you are going to use a super accurate headphone that most people have never even seen, you might as well make up a fictional headphone with a flat freq. response and zero distortion for a benchmark.

 

[ultrabike]

Quote:

Wouldn't it make sense to use a well known headphone as a benchmark reference? Like the HD600?
The point of a benchmark is that it is something people are very familiar with, and hence is valuable to compare headphones to it.
 
If you are going to use a super accurate headphone that most people have never even seen, you might as well make up a fictional headphone with a flat freq. response and zero distortion for a benchmark.

 
I think that even if one has not heard the most super accurate headphone there is, one would benefit from the objective and subjective characterizations of it, by relating this information to the objective and subjective characterizations of the headphones one has heard or owns. 

 

[Eisenhower]

Quote:

 
I think that even if one has not heard the most super accurate headphone there is, one would benefit from the objective and subjective characterizations of it, by relating this information to the objective and subjective characterizations of what one has heard or have. 

Subjective characterizations are, by nature, very personal. So reading someone's opinion about an uncommon headphone, like the Stax-009, is not all that useful since their subjective experience may not be same as yours. This is especially true of headphones that have stringent amp requirements, and if the company's manufacturing tolerances are not well known.
 
The whole point of a benchmark is that people have lots of experience with it, and it is well characterized. The HD600 meets this requirement with flying colors.
It is such a popular headphone, that even if individual people might differ on how it sounds, there is enough experience with it that on a macroscopic level there is an average opinion that has been converged on.

 

[gnarlsagan]

Subjective characterizations are, by nature, very personal. So reading someone's opinion about an uncommon headphone, like the Stax-009, is not all that useful since their subjective experience may not be same as yours. This is especially true of headphones that have stringent amp requirements, and if the company's manufacturing tolerances are not well known.

The whole point of a benchmark is that people have lots of experience with it, and it is well characterized. The HD600 meets this requirement with flying colors.
It is such a popular headphone, that even if individual people might differ on how it sounds, there is enough experience with it that on a macroscopic level there is an average opinion that has been converged on.

Semantics really. Define benchmark either way. Just be specific when doing so.