I repeated the test with other high-level combinations of high-frequency tones—11kHz+12kHz, 14+15kHz—and the noise floor remained low through the top octave. I then used a mix of 19kHz and 20kHz tones from a different test CD and again got the anomalous rise in the noise floor. However, to my surprise, when I repeated the test using the 48kHz or 96kHz sample-rate test signals on the Chesky DVD-V (fig.8), the spectrum revealed no top-octave problems. I can only assume that some aspect of the player's DVD engine, again most probably the Re-Master algorithm, is optimized for the 48kHz and 96kHz sample rates it will encounter on DVD-Audio discs, rather than the legacy 44.1kHz sample rate used on CDs.

Fig.8 Technics DVD-A10, HF intermodulation spectrum, DC-22kHz, 19+20kHz at 0dBFS into 100k ohms, 96kHz sample-rate data (linear frequency scale).

Another puzzle: When I checked the DVD-A10's S/PDIF digital output using the analytical capabilities of the dCS 972 digital/digital converter, the 96kHz tracks on the "official" DVD-A sampler disc were downsampled to 48kHz and truncated to 16 bits, as required by the DVD Forum. However, when I played 24/96 Chesky or Classic Records DVD-Vs, the player's digital jack output a 96kHz datastream with a full 24 bits of resolution. This suggests that the DVD Forum's restrictions on digital output bit depth and sample rate are implemented as instructions contained on the discs, not the players.—John Atkinson