The Lynx Hilo's published technical specifications make impressive reading, and I was largely able to confirm their veracity with an Audio Precision test system. Measuring the performance of the A-D from the analogue line inputs to the AES3 outputs, the AES 17 dynamic range figure is 117.4dB with the recommended CCIR-2k weighting, rising to 121.3dB with A-weighting. This figure slightly exceeds the published specifications, as did the THD+N results of -114dB measured with the +22dBu trim level selected for an output of -1dBFS. That equates to just 0.00025% distortion!
The published crosstalk figure is given as -140dB but at the unusual test frequency of 1kHz! Crosstalk is usually due to capacitive coupling between channels which rises with frequency, which is why the norm is to measure it at 10kHz. However, I obtained figures of -138dB at 1kHz and -137dB at 10kHz which, although not quite up to the claimed figures, are still extraordinarily good and remarkably consistent across the spectrum, indicating excellent circuit board design and build quality.
Looking at the D-A converter to the analogue line outputs I measured an AES17 dynamic range of 116.5dB (CCIR-2k) and 120.5dB (A-weighted) — again within a whisker of the published specs. THD+N was slightly inferior at -104dB instead of the claimed -109dB, but given that this equates to 0.0006% distortion I don't plan to lose any sleep over such a minor discrepancy. Crosstalk measured 2dB better than the claimed -135dB at 1kHz and 1dB better at 10kHz.
The monitor outputs (with the default +10dBu maximum setting) gave AES 17 dynamic range figures of 112.7dB (CCIR-2k) and 116.4dB (A-weighted). The published specs claim the same 121dB figure as the line outs, but given that the converter topology used should give a 3dB higher noise floor, I think the actual figure is quite acceptable and comparing the CCIR figures it is 3.8dB worse which seems exactly right to me. The THD+N figure came out at -105.6dB or 0.0005% and crosstalk was -130.5dB at 1kHz but a noticeably worse -118dB at 10kHz. This probably indicates slightly less stringent circuit board layout for the monitor outputs, but the figures are still excellent and would easily put most other products to shame.
All of the digital I/O (with the exception of the USB interface) is transformer or optically coupled, and the wordclock input is permanently terminated with 75Ω. Both 32 and 64 bit USB drivers are available for the current Windows OS, and native support is provided in the Mac OS X.