mbl 1611HR D/A converter & 1621 CD transport Measurements
Interestingly, while Kal thought the mbl to have a higher output level—4.038V balanced and 2.024V single-ended from the fixed output jacks—than the Burmester and Mark Levinson DACs that he also reviews in this issue, it was actually only 0.08dB higher. The analog output was sourced from 200 ohms (balanced) and 100 ohms (unbalanced). The big mbl doesn't invert signal polarity from either set of outputs (the XLR jacks are wired with pin 2 as hot).
This sample of the 1611HR wouldn't lock to data with a sample rate higher than 48kHz. Its frequency response on digital replay was identical from either set of outputs: flat within the audioband, but featuring a slight droop to –0.5dB at 20kHz (fig.1, top traces). Via the analog input, the response was ruler-flat across the entire band. With a pre-emphasized data input, the top two octaves shelved down by 0.25dB (fig.1, bottom traces), which will be just audible in direct comparisons. Any crosstalk was buried in the noise floor, even at 20kHz.
Fig.1 mbl 1611HR, balanced frequency response at 0dBFS (top) and response at –12dBFS with de-emphasis (bottom) at 44.1kHz sample rate. (Right channel dashed, 0.5dB/vertical div.)
Sweeping a 1/3-octave-wide analog bandpass filter across the mbl's output signal while it decoded data representing a dithered 1kHz sinewave at –90dBFS resulted in the spectra shown in fig.2. The top traces were made with 16-bit data, the lower with 24-bit data. While all traces are free from spuriae and AC supply components, there is only about a 10dB lowering of the noise floor with the greater bit depth, suggesting that the mbl's ultimate dynamic range is going to be more limited than with the Burmester and Levinson DACs. This is what I would have expected from its use of a noise-shaping delta-sigma DAC.
Fig.2 mbl 1611HR, spectrum of dithered 1kHz tone at –90dBFS, with noise and spuriae, 16-bit (top) and 24-bit (bottom) data. (Right channel dashed.)
Repeating the measurement, but this time using a 16-bit signal representing a 1LSB DC offset and extending the analysis bandwidth to 200kHz, gave the lower pair of traces in fig.3. (The small peaks in all of these graphs at 200Hz, 2kHz, and 20kHz are spuriae from the Audio Precision System One and should be ignored.) Note how much lower the noise floor is than in fig.2. In order to maximize their performance on a conventional test of S/N ratio, many manufacturers program their DACs to turn off their outputs when fed digital "black" data—hence my use of a signal comprising a 1LSB DC offset to fool such chips. Now it looks as if manufacturers are wise to this trick also. That the Crystal DAC turns off its output when fed a 1LSB signal is shown by the top pair of traces in fig.3, a spectral analysis of the mbl's output while it decodes 24-bit data representing a dithered 1kHz tone at –120dB. The bottom traces represent the analog output stage's noise floor, the top trace the true 24-bit noise floor. Note the large rise in ultrasonic noise, due to the noise-shaped, high-oversampling DAC topology. Note also the slight positive error in the level of the –120dBFS tone, which is worse in the right channel than in the left.
Fig.3 mbl 1611HR, spectrum of dithered 1kHz tone at –120dBFS, with noise and spuriae, 24-bit data (top); and of a 16-bit –1LSB (bottom). (Right channel dashed.)
This level error can be seen in the mbl's linearity plot (fig.4). The top traces were made with dithered 16-bit data, and the 1611HR has virtually no amplitude error down to –1115dBFS, which is excellent. But increasing the word length to 24 bits gave only a small increase in dynamic range in the left channel, and none at all in the right. Despite this performance, the mbl still did a good job of decoding 16-bit data. Fig.5, for example, shows the waveform of an undithered 1kHz sinewave at –90.31dBFS—the three discrete voltage levels can be easily seen. Feeding the processor undithered 24-bit data at the same level gave a good facsimile of a sinewave (not shown), though this was somewhat noisier than with the Levinson and Burmester DACs.
Fig.4 mbl 1611HR, departure from linearity, 16-bit data (top), 24-bit data (bottom). (Right channel dashed, 2dB/vertical div.)
Fig.5 mbl 1611HR, waveform of undithered 1kHz sinewave at –90.31dBFS, 16-bit data.
Whether processing digital or analog input signals, the mbl offered vanishingly low levels of harmonic and intermodulation distortion. Fig.6, for example, shows the balanced output spectrum while the processor decodes data representing a full-scale 50Hz tone. All harmonics are below –100dB (0.001%), and the unbalanced output (not shown) was as good. Fig.7 shows the processor's balanced spectrum while playing a mix of 19kHz and 20kHz tones at 0dBFS. (Each of this signal's individual tones lies at –6dBFS, as is apparent from this graph.) Only the 18kHz and 21kHz components poke their heads above the –100dB line, and then only by a couple of dB!
Fig.6 mbl 1611HR, balanced spectrum, DC–1kHz, 50Hz at 0dBFS, 100k ohm load (linear frequency scale, 20dB/vertical div.).
Fig.7 mbl 1611HR, balanced HF intermodulation spectrum, DC–22kHz, 19+20kHz at 0dBFS, 100k ohm load (linear frequency scale, 20dB/vertical div.).
I used the Miller Audio Research Jitter Analyzer to examine the effect of word-clock jitter on the mbl's analog outputs. The processor under test is driven with a signal consisting of a combination of a high-level tone at one quarter the sample rate and the 16-bit data's LSB toggling at approximately 229Hz. This signal was stored on a CD-R with low time-base error and played back on a PS Audio Lambda transport, this connected to the mbl with an ST-optical datalink. Any jitter will manifest itself as symmetrical pairs of sidebands to either side of the primary tone.
Fig.8 shows a narrow-band FFT-derived spectrum of the mbl's analog output signal, centered on the signal frequency of 11.025kHz. The weighted sum of the jitter components was 361.5 picoseconds peak–peak—a little higher than I'm used to seeing these days from top-quality DACs and CD players. Most of the jitter was in the form of sidebands at the data-related frequency of ±229Hz, these indicated with red "5" markers. A pair of AC supply-related sidebands was present at ±120Hz (purple "2" markers), but, more interestingly, note the rise in the noise floor to either side of the central peak. This seems typical of the Crystal delta-sigma DAC chip used by the mbl, and it has been hypothesized that this presence of very-low-frequency random-noise jitter leads to a slowing of musical pace in the bass and a very slight smearing of stereo imaging specificity.
Fig.8 mbl 1611HR, high-resolution jitter spectrum of analog output signal (11.025kHz at –6dBFS with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.
Finally, I checked the overload margin of the 1611HR's analog inputs. The input seems to be fed first to the volume control, meaning that the effective overload margin is infinite. But with the volume control set wide open, the mbl's circuit didn't clip until an output level of 23V! This can be seen in fig.9.
Fig.9 mbl 1611HR, balanced analog input, THD+noise (%) vs output voltage.
The mbl offers excellent measured performance overall. While its dynamic range is not quite as wide as the best competing D/A processors, this will be a moot point for CD playback.—John Atkinson