Digital Processor Reviews

I measured the second sample of the Brinkmann Nyquist with my Audio Precision SYS2722 system (see the January 2008 "As We See It"), using both the Audio Precision's optical and electrical digital outputs, and USB data sourced from my MacBook Pro running on battery power with Pure Music 3.0 playing WAV and AIFF test-tone files. Apple's USB Prober utility identified the DAC as "Nyquist" from "Brinkmann Audio," and confirmed that the USB port operated in the optimal isochronous asynchronous mode. Apple's AudioMIDI utility revealed that, via USB, the Nyquist accepted 16- and 24-bit integer data sampled at all rates from 44.1 to 384kHz. The optical input locked to datastreams with sample rates up to 96kHz, the AES/EBU and S/PDIF inputs locked to streams of up to 192kHz-sampled data.

The Brinkmann's maximum output level at 1kHz with the volume control set to "0,0dB" was 1.7V from the balanced output jacks, 848mV from the unbalanced jacks, and 7.47V from the headphone jack, with the volume set to "88" out of a maximum of "90." (When you select the headphone output with the front-panel button, the main outputs are muted and the volume display is set to "1," with the maximum being "90.") With the volume control set to its maximum of "10,0dB," the levels were 5.1V balanced and 2.54V unbalanced; ie, 9.5dB higher. Although the Nyquist's manual says that the volume control can be set to offer up to 20dB of attenuation, the control on our sample could not be set below "0,0dB." The balanced and unbalanced outputs preserved absolute polarity (ie, were non-inverting) with Phase set to "0°," the XLR jacks being wired with pin 2 hot. The headphone output inverted polarity with Phase set to "0°." The output impedance from the unbalanced jacks was a low 9 ohms at 1 and 20kHz, increasing slightly to 23 ohms at 20Hz. The balanced output impedance was higher but still low, at 15 ohms at 1 and 20kHz, 29 ohms at 20Hz. The output impedance from the headphone jack was a relatively high 30 ohms across the audioband.

Fig.1 shows the Brinkmann's impulse response with 44.1kHz data. Like every other MQA-capable processor I have measured, it is typical of a short, minimum-phase reconstruction filter. Tested with 44.1kHz-sampled white noise (footnote 1), this filter rolls off slowly above the audioband (fig.2, magenta and red traces) until well above the Nyquist frequency of 22.05kHz (vertical green line). Because of the filter's slow rolloff, the aliased image of a 19.1kHz tone at –1dBFS (fig.2, cyan and blue traces) is suppressed by just 18dB. I usually use a 19.1kHz tone at 0dBFS for this test, but the Nyquist overloaded at this level. All MQA-enabled processors seem to have a reconstruction filter that with 44.1kHz data trades off reduced rejection of aliased images at the top of the audioband for an optimized time-domain performance.

The Brinkmann's frequency response, tested with spot tones at –12dBFS with data sampled at 44.1, 96, 192, and 384kHz, is shown in fig.3. (The balanced and unbalanced, outputs behaved similarly, the headphone output rolled off a little faster, reaching –1dB at 20kHz with data sampled at 44.1kHz.) All the rates feature a very slight rise in the low bass and an output that rolls off above 20kHz, with the steepest rolloff occurring with 44.1kHz data (green and gray traces). At the higher sample rates, the rolloff is both slow and a little premature, reaching –3dB at 37kHz with 96kHz data (cyan and magenta traces), and at 54kHz with 192kHz data (blue, red). The response with 384kHz overlays the 192kHz response in this graph, but looking at it in more detail, it reaches –20dB at 95kHz compared with 88kHz with 192kHz data. Channel separation (not shown) was excellent, at >110dB below 20kHz, though the Nyquist's random noise floor (fig.4) was higher in in level than I usually find, with some spuriae present at 120Hz and its harmonics.

This noise was also higher in the left channel (blue trace) than the right (red) when I examined how the noise floor dropped in level as I increased the bit depth from 16 to 24 with a dithered 1kHz tone at –90dBFS (fig.5). Perhaps the tubes in the left channel were noisier than those in the right? Even so, the Brinkmann's resolution in the right channel in this graph appears to be around 18 bits' worth. With an undithered 16-bit tone at exactly –90.31dBFS (fig.6), the three DC voltage levels described by the data were well distinguished, though a little bit of waveform asymmetry is apparent in this graph. With undithered 24-bit data, the result was a noisy if somewhat asymmetrical sinewave (fig.7).

I usually test a processor's linearity with a full-scale, 24-bit 50Hz tone, but, as can be seen in fig.8, the Nyquist produced a surprisingly high level of odd-order distortion, with the third harmonic the highest in level, at –54dB (0.2%). The volume control was set to "0,0" for this graph; increasing it to "10,0" didn't change the picture. If I reduced the level of the 50Hz tone by 3 or 6dB, the level of the third harmonic dropped by the same 3 or 6dB and the spectrum looked the same. Repeating the test with a 1kHz tone gave a more respectable picture (fig.9), with the third harmonic now lying at –70dB (0.03%). With the punishing 600 ohm load (fig10), the right channel's distortion didn't change appreciably, but the left-channel distortion both increased and was now dominated by the second harmonic, at –60dB (0.1%).

With an equal mix of 19 and 20kHz tones, the combined waveform peaking at 0dBFS, the Nyquist's "leaky" reconstruction filter produced a slew of aliasing and intermodulation products (not shown). Reducing the signal level by 1dB produced a much cleaner spectrum (fig.11), with very low intermodulation distortion. However, the aliased images of the two tones are not suppressed by much. Fortunately for the Nyquist, music almost never has content approaching 0dBFS at the top of the audioband.

The Nyquist's relatively high level of random noise obscured the result when I tested for the DAC's rejection of word-clock jitter. With 16-bit J-Test data via the AES/EBU input (fig.12), it's difficult to see whether the odd-order harmonics of the low-frequency, LSB-level squarewave are at the correct level (sloping green line). However, of more concern in this graph are the sideband pairs at ±120, ±240, and ±1376Hz. The first two pairs are obviously related to the frequency of the full-wave–rectified power supply, the third pair of unknown origin—but all persisted, regardless of what input I used and whether I fed the Brinkmann 16- or 24-bit data. If the power-supply–related sidebands were due to inadequate jitter rejection, they would decrease in level if I reduced the signal frequency. However, when I analyzed the Nyquist's output while feeding it 20, 10, and 5kHz tones at –6dBFS, these sidebands remained at the same level. This means that they are more likely due to inadequate supply filtering on the DAC chip's voltage reference pin.

I admit that it's difficult to predict how these measured shortfalls will affect the sound quality, as the spuriae are mainly low in absolute terms. I do note that Michael Fremer wrote that he suspected that "John Atkinson's measurements will show that the noise floor of the Nyquist's tubed output stage, though inaudible as hiss, results in less than full resolution of hi-rez files." And when he described the Nyquist's balance as "soft and warm," I wasn't surprised.

Overall, I was disappointed by the Brinkmann Nyquist's measured performance. The higher-than-usual levels of random noise, the increase in distortion at low frequencies, and the supply-related sidebands all bothered me. You shouldn't have to make excuses for a DAC costing $18,000.—John Atkinson