Digital Processor Reviews

Sidebar 3: Measurements

I measured the Auralic Altair with my Audio Precision SYS2722 system (see the January 2008 "As We See It"). As well as the Audio Precision's optical and electrical digital outputs, I used Pure Music 3.0 to play WAV and AIFF test-tone files sourced via USB from my MacBook Pro running on battery power. To test the Altair's performance via its streaming connection, I loaded Auralic's Lightning DS app onto an iPad Mini and sourced the same WAV and AIFF files from an i7 Mac mini running Twonky server and with my NetGear router connected to the Altair by 30' of CAT-6 Ethernet cable. (I usually have bad network karma but I didn't encounter the problems setting the Altair up that Jon Iverson had experienced.) Apple's USB Prober utility identified the Auralic DAC as "ALTAIR Outputs" from "AURALiC" and confirmed that its USB port operated in the optimal isochronous asynchronous mode. Apple's AudioMIDI utility revealed that, via USB, the Altair accepted 16- and 24-bit integer data sampled at all rates from 44.1 to 384kHz. The optical and coaxial S/PDIF inputs locked to datastreams with sample rates of up to 192kHz; the AES/EBU input appeared to be limited to 96kHz.

Looking first at the performance of the Altair's analog outputs, the maximum output levels at 1kHz were 4.44V from the balanced jacks and 2.22V from the unbalanced and headphone jacks. All three sets of inputs preserved absolute polarity with Phase set to Normal, meaning that the balanced XLRs are wired with pin 2 hot. The output impedances didn't vary with frequency and were low, at 10 ohms balanced, 51 ohms unbalanced, and 3 ohms headphone.

The impulse responses of the Altair's digital reconstruction filters with 44.1kHz data indicated that the Precise and Dynamic filters are conventional linear-phase FIR types: fig.1 shows the Precise impulse response; Dynamic was almost identical. The Balance filter is a linear-phase type with fewer coefficients (fig.2), while the Smooth filter, which JI preferred, is a short minimum-phase type (fig.3). With 44.1kHz-sampled white noise, the Precise filter offered a rapid rolloff above 21kHz (fig.4, red and magenta traces) with almost complete elimination of the aliased image at 25kHz of a full-scale 19.1kHz tone (blue, cyan). The Dynamic filter offered an even faster rolloff, while Balance and Smooth gave slower ultrasonic rolloffs (fig.5).

These graphs were taken with 44.1kHz USB data, as I needed to set the Audio Precision to its maximum sample rate of 200kHz in order to analyze the spectrum up to 100kHz; sourcing the data from the network gave identical results. However, when I tried to measure the Altair using the Audio Precision's digital S/PDIF outputs (both coaxial and TosLink), I got anomalous results. While the Altair would lock to the datastream, its display would not show the sample rate (though it did with USB and Ethernet data). If I changed the rate of the Audio Precision's output data, the Auralic lost lock and muted until I either rebooted the DAC or reselected one of the filters with the front-panel menu. And even then, regardless of which filter I selected, all were identical to the Precise filter in both the time and frequency domains.

When I repeated the tests using the TosLink output of my Astell&Kern AK100 portable player I got the same problematic behavior. Then I tried the TosLink output of my MacBook Pro. Mirabile dictu—now the Altair's display showed the sample rate, I could change the sample rate without the Altair losing lock, and the four filters behaved as they had with USB data. All I can think is that instead of monitoring the incoming S/PDIF datastream's sample rate, which is the sensible way to do it, the Altair simply looks to see if the sample rate channel-status bit has been set. If the bit is undefined, the Altair will still lock to the datastream when a filter is selected, and appears to work correctly, but the DSP engine is unaware of the sample rate and so can't apply the appropriate filter (footnote 1). JI reported that the Altair didn't lose lock with his Meridian Sooloos system sending data to the Auralic via S/PDIF, so I assume the Sooloos was setting the channel-status bit correctly. On the other hand, he did find the audible differences between the filters to be, at best, subtle.

The Precise filter's frequency responses with sample rates from 44.1 to 384kHz are shown in fig.6. The response at each of the four rates shown conforms to the same basic shape, but with a sharp rolloff just below half of each of the three lower sample rates. Channel separation was simply superb, at >120dB in both directions below 3kHz and still 116dB at the top of the audioband. The Altair's noise floor was free from any power-supply–related spuriae—this is excellent analog engineering.

The noise floor was extremely low in level, meaning that when I changed the bit depth of the incoming data from 16 to 24 with a dithered tone at –90dBFS, the floor dropped by 23dB (fig.7), implying resolution close to 20 bits, which is state-of-the-art DAC performance. This graph was taken with Ethernet data; you can see a slight spreading of the spectrum at the base of the 1kHz tone and some low-level odd-order harmonics, both which were absent with USB or S/PDIF data (fig.8). With its very low level of analog noise and superb DAC linearity, the Altair's reproduction of an undithered tone at exactly –90.31dBFS was essentially perfect, with a symmetrical waveform, the three DC voltage levels well defined, and the ringing of the reconstruction filter clearly visible (fig.9). With undithered 24-bit data, the Altair output a well-formed sinewave (fig.10).

The spectral analyses in figs. 4 and 5 suggest that the Altair produced very low levels of harmonic distortion, and this was confirmed when I looked at the result with a full-scale 50Hz tone. Even into 600 ohms (fig.11), the highest-level harmonic, the third, lay at –110dB (0.0003%). Intermodulation distortion was similarly minimal, though the suppression of the aliased images of the 19 and 20kHz tones I use to test for intermodulation depended on the filter chosen: Precise and Dynamic gave almost total suppression (fig.12), while Balance and Smooth gave very little suppression (fig.13).

The Altair was effectively immune to the effects of jitter in the incoming data. Fig.14 shows a narrowband spectral analysis while the Altair decoded 16-bit/44.1kHz J-Test data: all the odd-order harmonics of the Fs/192, LSB-level squarewave are close to the correct levels (sloping green line), and no other sidebands are visible around the sharply defined spectral spike that represents the high-level Fs/4 tone. This test was taken with Ethernet data; the spectra with USB and S/PDIF data were identical to the Ethernet data. With 24-bit J-Test data, the resultant spectrum (fig.15) was free from spuriae of any kind.

Other than its anomalous behavior with S/PDIF and AES/EBU data, the measured performance of Auralic's Altair reveals superb audio engineering.—John Atkinson

---

Footnote 1: In his "Manufacturer's Comment," Wang Xuanqian, President & CEO of Auralic Ltd., wrote that the Altair "has a complex internal processing algorithm, as well as a unique way of handling clock synchronization: it does not compare or lock to the input system, but instead uses the DAC as a master clock, to achieve real zero-jitter performance. For this reason, the circuit has to know the input data's correct sample rate. With USB and streaming input, the processor can get the correct sampling rate from the datastream; for S/PDIF and AES/EBU inputs, however, we have to rely on the sampling-rate information in the data channel-status bits—any digital stream lacking this information will not be accepted."