Audio Research DAC1 D/A converter Measurements

Sidebar 2: Measurements

It's always reassuring to find excellent bench performance in a product that has first proven itself in the listening room. Conversely, products that sound good and measure poorly are cause for consternation: measured performance aspects that allegedly affect certain musical qualities are questioned. I had no such concerns, however, after measuring the DAC1: it exhibited excellent performance on the bench.

The DAC1's output level was the lowest I have measured, producing 1.74V when decoding a 1kHz, 0dB full-scale signal. This resulted in not quite enough level when driving the VTLs through the passive EVS Stepped Attenuator. Owners of passive attenuators are therefore advised to audition the DAC1 in their systems before buying to determine whether the sound will be loud enough. There were only a few instances where I wanted more level, however.

The CD standard is 2V output level, but many processors put out much higher voltages. The VTL, for example, puts out 3.26V, while the Theta DSPro Basic produces a whopping 7.2V. When making comparisons between processors, these differences must be accounted for. Level matching to less than 0.5dB is essential. Beware the trap of matching levels with a 1kHz tone that has not been pre-emphasized, then playing CDs recorded with pre-emphasis. Although their respective levels may be matched with the de-emphasis circuit out, they may not be matched with the de-emphasis circuit switched in.

Frequency response, shown in fig.1, was ruler-flat, with an insignificant 0.15dB rolloff at 20kHz. Note, however, how the left and right channels have exactly the same output level, the two traces nearly appearing as one. This is the best interchannel amplitude tracking I've measured.

Fig.1 Audio Research DAC1, frequency response at 150;12dBFS (right channel dashed, 0.5dB/vertical div.).

Not surprising in light of the DAC1's digital domain–performed de-emphasis, the unit exhibited no de-emphasis error (fig.2). Interchannel crosstalk was low, measuring –115dB at 1kHz, dropping to –96dB at 16kHz (L–R). The R–L was slightly better, achieving a remarkable 120dB of isolation at 1kHz, dropping to 111dB at 16kHz. These curves are shown in fig.3.

Fig.2 Audio Research DAC1, deemphasis error at 150;12dBFS (right channel dashed, 0.5dB/vertical div.).

Fig.3 Audio Research DAC1, channel separation (10dB/vertical div.).

Looking at the DAC1's spectral content when decoding a dithered –90.31dB, 1kHz sinewave (fig.4), we can see an extraordinarily low levels of noise, harmonics of the 1kHz signal, and power-supply–related junk at 60 or 120Hz. In addition, a hint of the DAC1's low-level linearity is provided by the fact that the traces just reach the –90dB horizontal division. What is also remarkable about this plot is the nearly exact tracking between channels, indicating the two channels within the UltraAnalog DAC are performing identically.

Fig.4 Audio Research DAC1, 1/3-octave spectrum of dithered 1kHz tone at –90dBFS with noise and spuriae (10dB/vertical div.).

Plotting the departure from linearity (fig.5) reveals in more detail the excellent performance indicated in fig.4. Linearity error was virtually nonexistent, the trace remaining straight to –100dB. The apparent error below this level is due to noise swamping the extremely low signal level. This is extraordinary performance, and, along with the Stax DAC-X1t and VTL, the best performance I've measured. D/A converters don't get any better than this. It should be noted that this performance is intrinsic to the DAC and not dependent on external adjustment. Consequently, this linearity will not drift over time or become misaligned.

Fig.5 Audio Research DAC1, left-channel departure from linearity (2dB/vertical div.).

The DAC1 produced a 1kHz, full-scale squarewave with a shape typical of the linear-phase NPC digital filter chip with a slight leading edge overshoot and ringing (fig.6). A –90.31dB, 1kHz sinewave as decoded by the DAC1 is shown in fig.7. Although the three discrete levels (0, +1, –1) are less distinct than seen with the Stax, the waveform reveals the signal transitions to be of equal-amplitude steps, overlaid by a small amount of audio-band noise.

Fig.6 Audio Research DAC1, 1kHz squarewave at 0dBFS.

Fig.7 Audio Research DAC1, undithered 1kHz tone at –90.31dBFS.

Fig.8 shows the spectrum produced when the DAC1 is decoding an equal-amplitude mix of 19kHz and 20kHz at full scale. The 1kHz intermodulation product, sometimes seen in other processors, is notably absent. Looking at a positive-going impulse revealed the DAC1 to be non-inverting when the front-panel switch is in the "normal" position, inverting when in the "invert" position. Output impedance was quite low, measuring 29 ohms at 17Hz and 1kHz, dropping slightly to 24 ohms at 20kHz, in line with the specified output impedance of 30 ohms. No measurable amount of DC offset was present at the output jacks.

Fig.8 Audio Research DAC1, HF intermodulation spectrum, 300Hz–30kHz, 19+20kHz at 0dBFS (linear frequency scale).

Overall, the DAC1's bench performance was exemplary. Incidentally, there was no measurable difference in the DAC1's performance whether driven by a coaxial or optical signal.—Robert Harley

Audio Research Corporation
3900 Annapolis Lane N.
Plymouth, MN 55447-5447
(763) 577-9700
Share | |
volvic's picture

I bet you that if you were to drop that in any system against any of the newer competitors it would still hold its own.  Okay it doesn't have HDMI, USB inputs for todays computer audio systems but knowing how good their gear is I am sure it still sounds great.

Pro-Audio-Tech's picture

Listening to these new digital audio servers and expensive DAC's is  like going to a high price steak house and ordering a big 1-1/2" steak made out of Oscar Meyer bologna.

Get thee analog my son!

ARC makes a superb phono preamp use it!

Enter your username.
Enter the password that accompanies your username.