Mark Levinson No.30.6 Reference D/A processor Measurements
The No.30.6 didn't invert polarity from either set of outputs (the XLRs are wired with pin 2 hot). Its unbalanced source impedance was a very low 5.5 ohms in the midband; balanced was 14 ohms. Maximum output level was 2V unbalanced, 4V balanced, and the channel matching was within an excellent 0.05dB across the band. Crosstalk was unmeasurable.
The frequency response was essentially flat across the audioband from both outputs, while the de-emphasis correction added an imperceptible (-0.05dB) swayback in the mid-treble. The processor locked on to all sample rates up to 96kHz, and the higher sample rate effectively doubled the reproduced bandwidth, as expected (fig.1).
Fig.1 Mark Levinson No.30.6, frequency response at 0dBFS at 44.1kHz and 96kHz sample rates (5dB/vertical div.).
Fed with 16-bit data, the Levinson's reproduction of a dithered 1kHz tone at -90dBFS was textbook-perfect (fig.2, top traces), the noise floor in this graph basically being that of the test signal. Increasing the word length to 24 bits resulted in the lower pair of traces in fig.2, there being a 20dB drop in the processor's analog noise floor and not a trace of power-supply noise. This is the best reading I have seen on this measurement, suggesting that the No.30.6's resolution is close to 20-bit performance! Though many DACs are specified these days as being "24-bit," their analog noise floors limit their dynamic ranges to only somewhat better than the CD's 16 bits. True 24-bit dynamic range is not achievable in real circuits having real impedances—see fig.1 in the DSD sidebar (p.97) accompanying this issue's review of the Sony SCD-1 Super Audio CD player—but the Levinson Reference processor is about as good as it can currently get.
Fig.2 Mark Levinson No.30.6, spectrum of dithered 1kHz tone at -90dBFS, with noise and spuriae, 16-bit (top) and 24-bit (bottom) data (right channel dashed).