Mark Levinson No.39 CD player Measurements
I measured the No.39's performance from both its balanced and unbalanced outputs. Except where noted, my comments refer to the balanced outputs. The fixed output is set to 3.5V. Using the variable output, with the volume control set to its maximum position, the No.39 fed with a 1kHz, 0dBFS signal gave out an enormous 17.75V RMS (balanced), 8.87V RMS (unbalanced). Dropping the volume by an indicated 12.9dB on the display to "60.3" resulted in a -12.95dB drop to a more reasonable 4V (balanced) and 2V (unbalanced). Used without a preamp, the No.39 will have more than enough output to drive any power amplifier to clipping. The output impedance was low, at a measured 10 ohms (unbalanced) and 20 ohms (balanced). With pin 2 hot, the balanced output was in the correct polarity.
The No,39's balanced frequency response at 0dBFS is shown in fig.1 (the unbalanced response was identical so is not shown). Perfectly flat over almost the entire audio band, the response does feature an inconsequential droop of just under 0.25dB at 20kHz. Note the excellent channel balance in this graph, better than 0.03dB. This was maintained at almost all volume settings, a testament to the well-engineered volume control. The No.39 had zero deemphasis error across the audio band (not shown). The channel separation (also not shown) was a superb 119dB at 1kHz, decreasing to 107dB at 16kHz.
Fig.1 Mark Levinson No.39, balanced mode, frequency response (right channel dashed, 0.5dB/vertical div.).
Fig.2 shows a spectral analysis of the No.39's output while it decoded 16-bit data from a test CD representing a dithered 1kHz tone at -90.31dB. The plot is dominated by the dither noise, with no power supply or distortion components evident. This is excellent performance. Driving the '39's digital input from the Audio Precision System One's digital generator and setting the word length to 20 bits gave the spectrum shown in fig.3. Despite the dramatic reduction in the dither noise level, no distortion components are unmasked, and the only power supply artefacts making their presence known are hints of 60Hz and 180Hz in the right channel. Again, this reveals excellent engineering at work in the No.39's design.
Fig.2 Mark Levinson No.39, balanced mode, spectrum of dithered 1kHz tone at -90.31dBFS, with noise and spuriae (16-bit data, 1/3-octave analysis, right channel dashed).
Fig.3 Mark Levinson No.39, balanced mode, spectrum of dithered 1kHz tone at -90.31dBFS, with noise and spuriae (20-bit data, 1/3-octave analysis, right channel dashed).
The Levinson's linearity, measured using dithered 16-bit data from a test CD, is shown in fig.4. There is a just-perceptible negative error of around -0.2dB between -80dBFS and -100dBFS. But note the overall low error right down to -110dBFS and the very low noise, this already shown in the spectral analysis plots. The waveform of an undithered 1kHz tone at -90.31dBFS (fig.5) was reproduced with its three voltage levels easily identifiable, not obscured by the noise present. This is about as good digital performance as you can buy.