> An astounding frequency response of 20Hz to 100kHz makes the DMP3 ideal for today's 96k recording work.
On the contrary--a 96 kHz recording system absolutely has to filter out anything at or above 48 kHz, so about half of this "astounding" bandwidth is not only wasted, but a liability.
The original (and to my mind, the only sonically or scientifically honest) motivation for sampling rates greater than the minimum necessary (e.g. 44.1 kHz) was so that the required low-pass filtering could be gentler, thus improving impulse response within the audible band. See
http://www.nanophon.com/audio/antialia.pdf for example. With specially generated test clicks and electrostatic headphones, it is sometimes possible for some people to hear a difference between 44.1 kHz and higher sampling rates on the same material (limited to a 20 kHz bandwidth).
I tend to doubt whether that difference is audible on real-world speech or music, but significant signal energy beyond the audible range is always garbage, practically speaking, and the question is not
whether to filter it out, but how to do so in the least harmful way.
Just as a personal aside, it particularly amazes me when people who revere "vintage" tube microphones and other analog recording equipment are simultaneously convinced that ultra-wide bandwidth is a good idea; those two realms are mutually exclusive.