Burmester 001 CD player Measurements
As it has an external data input, I was able to test the Burmester 001 with both 16-bit CD data and with external 24-bit data, sampled at 96kHz. The volume control operates in 60 steps, but with digital data at 0dBFS (the maximum level), the player's output clipped at volume-control settings above "54." The output level at this setting was 6.495V from the balanced jacks, 3.234V from the unbalanced, which is 4.2dB higher than the standard 2V. The output impedance was commendably low, at 66.2 ohms from the unbalanced RCAs, this doubling to 132 ohms from the balanced, as expected. Neither figure varied across the audioband. Error correction was good, the Burmester coping with data dropouts up to 1.5mm in length without audible glitches.
The top pair of traces in fig.1 show the CD response with the reconstruction filter set to Linear. It is perfectly flat. Switching the filter to Soft drops the output at 20kHz by 0.55dB (fig.1, middle traces). Feeding the Burmester with external 96kHz-sampled data, the response trends were continued, the Soft filter's -0.6dB at 20kHz reaching -1.53dB at 45kHz. With the Linear filter, the flat response is maintained to 32kHz, above which it starts to roll off, reaching -0.42dB at 45kHz. The audioband response with de-emphasized data (fig.1, lower traces) is basically the same as with regular data.
Fig.1 Burmester 001, balanced outputs, frequency response at -12dBFS into 100k ohms, with de-emphasis (bottom) and without (top). (Right channel dashed, 0.5dB/vertical div.)
Channel separation (not shown) was better than 115dB in both directions below 10kHz, with just a slight decrease evident above that frequency. Fig.2 shows spectral analyses of the player's balanced outputs while it decoded a dithered 1kHz tone at -90dBFS with 16-bit, CD data (top) and 24-bit external data (bottom). The increase in bit depth drops the noise floor by 15dB or so in the treble, implying DAC performance close to 19 bits, which is excellent. There might be a faint hint of some second-harmonic content, but the traces in this graph are otherwise free from spuriae and power-supply spikes. Good engineering.
Fig.2 Burmester 001, balanced, 1/3-octave spectrum of dithered 1kHz tone at -90dBFS, with noise and spuriae, 16-bit (top) and 24-bit data (bottom). (Right channel dashed.)
Linearity error, measured with CD data (fig.3), remained below 2dB down to -110dBFS, which is excellent, implying low noise. This, along with the excellent DAC bit matching, is confirmed by fig.4, which clearly shows the correct three-level shape of an undithered 16-bit, 1kHz sinewave at -90.31dBFS. Increasing the word length to 24 bits gives an excellent sinewave shape, even at this very low level (fig.5).
Fig.3 Burmester 001, balanced, left-channel departure from linearity, 16-bit data (2dB/vertical div.).
Fig.4 Burmester 001, waveform of undithered 1kHz sinewave at -90.31dBFS, 16-bit data.
Fig.5 Burmester 001, waveform of undithered 1kHz sinewave at -90.31dBFS, 24-bit data.
Not only was the 001's noise floor very low, so was distortion. A 1kHz tone at 0dBFS reproduced with just 0.0017% THD from the left channel, 0.001% from the right. (This was with the volume control set to "54.") The spectrum of this signal is shown in fig.6; the second and third harmonics hover at or just above the -100dB level, with the fourth at -110dB. "Not bad," as we say in my native UK when presented with outstanding performance. Intermodulation distortion was also very low, though it differed slightly according to which filter was used. Linear (fig.7) was slightly better than Soft (fig.8) with respect to the level of the higher-order components at 18kHz and 21kHz.
Fig.6 Burmester 001, volume control at "54," unbalanced spectrum of 1kHz sinewave, DC-1kHz, at 0dBFS into 8k ohms (linear frequency scale).
Fig.7 Burmester 001, unbalanced, Soft filter, HF intermodulation spectrum, DC-24kHz, 19+20kHz at 0dBFS into 600 ohms (linear frequency scale).
Fig.8 Burmester 001, unbalanced, Linear filter, HF intermodulation spectrum, DC-24kHz, 19+20kHz at 0dBFS into 600 ohms (linear frequency scale).
Possibly of more concern than conventional intermodulation is the appearance of what appear to be symmetrical sidebands at non-signal-related frequencies to either side of the fundamental tones in these two graphs. Yes, these are at a very low level, but they imply insufficient rejection of word-clock jitter. It was with some surprise, therefore, that when I tested the Burmester 001 for jitter with the Miller Audio Research Analyzer, I got an extraordinarily good result: just 104.1 picoseconds peak-peak of jitter.
This is the lowest I have ever measured. Even driving the Burmester 001 with external data via a TosLink connection only increased the level of jitter-related sidebands to 118ps, which is the second-lowest I have measured! Fig.9 shows a narrowband spectrum of the 001's analog jitter while it decoded the Miller test signal. Other than a smattering of low-frequency sidebands around the central 11.025kHz tone, at 15.6Hz and its first four harmonics, the only other sidebands visible are at ±2933Hz and ±3210Hz (purple "6" and "7" markers).
Fig.9 Burmester 001, unbalanced, high-resolution jitter spectrum of analog output signal (11.025kHz at -6dBFS sampled at 44.1kHz with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.
But I was still worried about those sidebands in the intermodulation graphs. I therefore repeated the Miller Jitter Test using 48kHz- and 96kHz-sampled external data. I haven't shown the graphs here, because the spectra were taken with different hardware and aren't directly comparable with fig.9. However, the important point to note is that the very low measured jitter level of 118ps with 44.1kHz-sampled data rose to a high 900ps with the other sample rates. It appears that the Burmester 001's rejection of word-clock jitter has been optimized for CD-sampled data (footnote 1). This makes sense—it is a CD player—but my eyebrows did rise a little.
Nevertheless, for CD playback, the Burmester 001 offers superb measured performance. I was not surprised to learn that Brian Damkroger liked how it sounded in his system. And I didn't find any significant differences between its performance in unbalanced and balanced modes that would explain either why Brian preferred the latter or found the former to sound too warm.—John Atkinson
Footnote 1: There is another implication of these results, which is that the Burmester 001 has been engineered to perform well on the specific test signal used by the Miller Jitter Test. Such gamesmanship is not unknown—witness how many CD players mute their outputs when they sense "digital black" data, in order to give an unrealistically good measured signal/noise ratio. But it's hard to see how something similar could be done with the Miller signal without affecting performance elsewhere.—John Atkinson