So people don't get confused by basic misunderstandings of the research boldly and erroneously stated as fact in this thread...The below frequency response of the DT990 is Harman's measured frequency response of the actual headphone,
not the AKG K712 'replicator' headphone EQed to the DT990 up to 12 kHz - the same goes for the other FR measurements on slides 31-36 of
this presentation I linked to previously, and in Appendix 2 of
the AES paper it references, clearly entitled 'Headphone Measurements and Error Response Curve' (with no mention of EQed or 'virtualized' headphone measurements here).
How do we know this is a measurement of the real DT990 and not the 'virtualized' equivalent on the K712? Take a look at comparisons of Harman's real vs virtualized measurements of various headphones below (right, red, and left, blue channels offset by 10 dB for clarity, from
here):
Due to 'not attempting aggressive equalization above 12 kHz' (stated in the above linked paper), the very similar medium to high-Q variations in this upper treble region in the virtualized (solid) curves for each headphone are clearly obvious, indicating they are measurements of the same physical replicator headphone. Contrast this with the quite different responses for each headphone in this >12 kHz region for the real (dashed) curves, of course due to them being measurements of physically different headphones. Now take a look at the frequency response graphs in Appendix 2 of the paper (or
here in the presentation), and it's blatantly obvious that they are indeed real measurements of physically distinct headphones, due to their very different responses at frequencies >12 kHz. The claim that these graphs (including the above one of the DT990) are not measurements of the real physical headphones has zero basis.
That cleared up, back to the question of the apparent discrepancy between the high 91/100 rating given by listeners in Harman's blind tests and some users' impressions of the DT990, and it seems
my initial suspicion when learning of this surprisingly high rating was correct - the DT990 in Harman's study was a somewhat used pair (Oratory states he was told this), which due to the relatively fast and significant wear and compression of Beyerdynamic's pads over time changing the acoustic properties and so frequency response of the headphone, results in the treble peaks becoming notably subdued. (I still suspect the last couple of factors I mentioned in
my previous post - head size, which like wear also affects pad compression, and listening volume, come into play with people's impressions of these headphones too.) Now it could be argued that due to this fast pad wear, Harman's measurements will actually be more representative of the average DT990 owner's pair at any one time, seeing as most users won't be changing pads very frequently. Nevertheless, I don't think any headphone that undergoes such significant changes in sound (and likely comfort) in such a relatively short period of time is a good buy, even if there is an optimum level of 'ageing' at which they sound good. End of story.