Since this is
Sound Science, I'm not going to say any more in here about what I think (subjectively) about how the HM-801 sounds than this: I like it. A lot. (I do own other digital audio players, including a variety of iPods, and iPhone, a Sansa Clip, and others.)
The main point of my post in this thread is to discuss frequency response, as represented in the graphs, and which (for many) seems to be the primary determinant (within the context of this thread) of performance. Here's one that's been used in here (below):
Here's are some frequency response plots of the Ayre QB-9 (a USB DAC that I've heard and used, and that has been pretty universally praised, at least from what I've read), taken from an Ayre white paper:
Using the QB-9's "Measure" setting, intended (per Ayre) for better accuracy in the frequency domain (below):
Using the QB-9's "Listen" setting (which is the setting it was on when Ayre handed it to me), and which is intended by Ayre to produce more accuracy in the time domain (below):
To put numbers to it, it's -6 dB at 22.025 kHz (exact numbers provided by Ayre). Eyeballing it at 20 kHz suggests to me that it's about -4 dB at 20 kHz.
This is the Ayre QB-9's frequency response, as measured by John Atkinson in
Stereophile (below):
A quick glance at the Ayre-provided graphs (when compared to
Stereophile's) might suggest that Ayre's FR representation is flatter. Independent of Y-axis scale, it certainly does look flatter; but take into account the actual Y-axis scale, and you see that they're actually pretty close.
My point? Take into account the Y-axis scale. In my opinion, -4 dB at 20 kHz does not constitute a major roll-off, but it can be made to look like a giant drop. And the roll-off does
look steep on that graph, and does start off at a lower frequency than the Ayre QB-9, but (and this is an opinion) it does not constitute a major roll-off to the ears at all.
If someone would rather listen to the Sansa Clip than the Ayre QB-9 in "Listen" mode (based on it being flatter to 20 kHz), I'm not going to argue the point. If someone's going to argue it's better for that reason, I'm not going to argue that either.
Some time ago, a headphone manufacturer asked me to listen to a headphone, and give my opinion on whether or not they should drop the response centered at around 8 kHz, about 2.5 dB down (the width of the notch approximately 0.75 octave). What I thought isn't relevant without revealing much more information (which I can not do, for obvious reasons). To approximate the effect, I used the parametric equalizer in Amarra. What might be relevant to this discussion is that it did have an effect (a lot of us are rather sensitive to things going on at 8 kHz), but this whole notion of
major roll-off, especially when we're talking primarily at 20 kHz is interesting.
While I can not say whether or not the reasons for the HM-801 being that way are analogous to Ayre's reasons for rolling off, or Wadia's reasons for rolling off, etc., I can say that, in my opinion, again, this is not a major roll-off. Yes, yes, argue that point all you want (and some of you will), but that's my opinion.
As I said in another post, I'm not much of a measurements guy, but admit it can make for interesting discussions, as it has here. As I state in that other post, I haven't heard any digital I've preferred to a good turntable setup--and, yes, that Sansa Clip just might slaughter even the finest of turntable rigs in most measurements, most certainly including frequency response, crosstalk, dynamic range, wow-and-flutter (which only the turntable will be measurable for), noise floor, etc., etc. And if you've heard a good vinyl rig, and feel the Sansa Clip sounds better, that's something else I'm not going to argue about.
As someone who bought the HM-801 and listens to it a lot, I'm not at all perturbed by these measurements.