Good points you made, here. I'm learning a lot from this thread. Up until the big valley in the highs that shows up on Tyll's graph, they sound very flat to my ears....and even more important....realistic....true to the way acoustic instruments sound in real life. However what I do notice when comparing them to the 240DF and the hd580, is that there's a lot more energy in that lower treble area with those other phones. With high pitched horns and strings I'm hearing significantly more brightness with the DF and 580 compared to the dt48. So perhaps it's simply be the difference between free field equalization and DF equalization that accounts for that. I think the 580 is DF equalized, as well as the 'DF', obviously. The dt48 corresponds closer to how I hear horns in real life, having played in large jazz bands and orchestras throughout my early years, and also attending a lot of live performances. So, as you say, the valley in the lower treble on Tyll's graph may very well be intentional. Perhaps I simply prefer the sound of free field equalization over DF equalization. The dt48 sounds even more 'real'...true to life... than those other two phones, which are quite good in their own rights. It's definitely a very unique listening experience compared to anything else I've heard. If I could only keep one phone for the rest of my life it would be a very tough choice between the 580 and the dt48..
This is consistant with my impressions as well... The 2.5khz peak can be a bit nasty (ringing ?), depending on your HRTF. I read that the 12/13khz frequency range is used by the ear for spatial location in a FF soundfield which would explain the peak as well.
What you have to understand is that those peaks and valleys ideally DO follow an intention... Do NOT try to eradicate them by equalizing (or very slightly to fine tune to your own HRTF). They are not necesarly a bad thing. In fact, what I see is that the DT48 FR follows pretty closely the theorical Free Field frequency response... For some it will work better than others, because the FF frequency response is much more sensitive to HRTF variations than the DF for example.