[Jon Sonne]

Lately I have been looking at a lot of headphones measurements, and suddenly I noticed that no measurement of THD+N exceeds 10 kHz. Of course it makes perfect sense that you cannot measure harmonic distortion above 10 kHz, as the first harmonic of those frequencies would lie outside the traditional sensing system limited to 20kHz audio bandwidth. But what about noise? Why should we not be able to measure noise above 10 kHz? 
 
Another thing I was thinking about is why don't we have an established standard for intermodulation distortion measurements? Given that intermodulation is not overtones, I think it would tend to be more objectionable to the listener than harmonic distortion (if it is at an audible level) and thus important to consider.
 
Headphones are notorious for being bad at reproducing treble. One reason might be that the treble interacts with the ear and this alters the frequency response perceived by the user. And while bass and midrange fidelity can, to some extend, be predicted by measurements, treble performance is in my experience much more difficult to predict given the currently available measurements.
 
What kind of advancements in headphones measurements do you think could be valuable for analysing treble fidelity?
 
/J

 

[pinnahertz]

  Lately I have been looking at a lot of headphones measurements, and suddenly I noticed that no measurement of THD+N exceeds 10 kHz. Of course it makes perfect sense that you cannot measure harmonic distortion above 10 kHz, as the first harmonic of those frequencies would lie outside the traditional sensing system limited to 20kHz audio bandwidth. But what about noise? Why should we not be able to measure noise above 10 kHz? 

Distortion and noise can and is measured well above 20kHz. There is no measurement limit, only that of the equipment in hand. In the analog days a distortion analyzer had bandwidth to 100kHz, with a 20kHz low pass filter available for meaningful noise measurements, along with a set of weighting filters.  Digital analysis systems today are similarly unlimited in bandwidth, and do all the analysis using FFT and post processing.  But the results are pretty much the same.  

Another thing I was thinking about is why don't we have an established standard for intermodulation distortion measurements? Given that intermodulation is not overtones, I think it would tend to be more objectionable to the listener than harmonic distortion (if it is at an audible level) and thus important to consider.

There are several IMD measurement standards that have been in place for many decades. SMPTE IMD (SMPTE RP120-1983) uses a test signal of 60hz and 7kHz mixed 4:1. The CCIF IMD standard uses a test signal of a pair of high frequency tones mixed 1:1, separated by 80Hz. DFD (Difference Frequency Distortion) is also tested with two tones with known frequency separation. As with CCIF, the result is intermod products at the sum and difference frequencies of the two tones. Each measurement standard has its specific application and sensitivity to different IMD generation mechanisms.  
Probably one of the more advanced techniques is described in the 1988 AES paper, "SPECTRAL CONTAMINATION MEASUREMENT" by Deane Jensen and Gary Sokolich in which an digital arbitrary signal generator produced a signal comprised of many (like 100) closely spaced high frequency tones which were then carefully filtered, presented to the device under test, and the products of which could be seen as spectrum outside that of the bank of tones.
 
And yes, IMD can be more objectionable than THD, though it's quite rare to have one apart from the other.  Nonlinearity gets you into both THD and IMD at about the same time. There are very few exceptions.

Headphones are notorious for being bad at reproducing treble. One reason might be that the treble interacts with the ear and this alters the frequency response perceived by the user. And while bass and midrange fidelity can, to some extend, be predicted by measurements, treble performance is in my experience much more difficult to predict given the currently available measurements.

Not too sure about any of that. Those are subjective observations, and there remains some disagreement among manufacturers as to what the ideal headphone target curve should be. And it's different for open, closed, and IEMs of course.

What kind of advancements in headphones measurements do you think could be valuable for analysing treble fidelity?

I've been impressed with the results of the testing Audyssey has done, with "tunings" available in their AMP app on IOS. Sadly, the app is klunky, but the tunings are pretty amazing. They use an artificial head/ear, and their "fuzzy clustering" measurement methods to high their own idea of the ideal target curve. There are others doing similar work.

 

[Joe Bloggs]

Headphones are notorious for being bad at reproducing treble. One reason might be that the treble interacts with the ear and this alters the frequency response perceived by the user. And while bass and midrange fidelity can, to some extend, be predicted by measurements, treble performance is in my experience much more difficult to predict given the currently available measurements.

What kind of advancements in headphones measurements do you think could be valuable for analysing treble fidelity?

/J

Yes they are, but the badness does not lie in terms of distortion, it's mostly in terms of how greatly the response magnitude fluctuates with frequency.

See a relevant recent discussion here:
http://www.head-fi.org/t/831012/why-is-treble-all-over-the-place

 

[PETEBULL]

  Headphones are notorious for being bad at reproducing treble.

I'd call it a rather strong claim. Heard plenty of headphones and i'd say that mids distortion is more common in over-ear headphones than treble. Bad treble may be encountered in poor headphones. Moreover, speakers on the contrary need good positioning in order not to screw up treble and a silky dome tweeter at least.