This might be of interest (wild speculation on my part but just happened to be reading this yesterday sounded familiar)
From Sound on Sound review of Echo Audiofire 12 interface
"It happened at every sample rate, and I eventually managed to track down the problem to the Audiofire's most unusual noise performance. Rather than the roughly flat noise spectrum one would expect, the Audiofire's output noise level starts to rise from about 25kHz, and when it got to the highest frequency I could measure (96kHz — half of the 192kHz sample rate) it had risen by a huge 70dB! When you choose the 176kHz or 192kHz sample rates, the Console utility input meters also display a continuous low-level input signal, and once I'd recorded some of this it turned out to be similar ultrasonic background noise.
A swift email exchange with Echo's boffins confirmed that this noise isn't a fault at all, but a by-product of the rather unusual Cirrus Logic converters used in the Audiofire range, as well as in the 3G and Indigo IO. These particular converters use IIR (Infinite Impulse Response) filters instead of the more usual FIR (Finite Impulse Response) designs adopted by most competitors like AKM. Much research has been carried out to determine why many people hear an improvement in digital audio quality at 96kHz and higher sample rates compared with 44.1kHz, and many experts now seem to agree that it's not the extended frequency response that provides the audible improvements, but the gentler high-end filtering, resulting in better impulse response with less 'time smear' and cleaner transients.
IIR filtering is more efficient, providing lower converter latency, and it also has a better impulse response, which according to the argument above should transfer some of the benefits of higher sample rates (better stereo imaging and improvements in playback of instruments with lots of transients, such as percussion) even to 44.1kHz recording/playback. The only disadvantage is this ultrasonic noise, if this is a disadvantage — you can't hear it, and the A-weighting filters used in dynamic range measurements automatically exclude it as well. I wouldn't have even known it was there except that my analogue mixer has a wide bandwidth with a -3dB point above 125kHz, while the Echo Console has no filtering at all, which is the only reason that it showed up on the meters. On other gear you might not even be aware of its existence."
Hope it helps
..