Digital Processor Reviews

I examined the Arcam FMJ D33's electrical performance with Stereophile's loan sample of the top-of-the-line Audio Precision SYS2722 system (see www.ap.com and the January 2008 "As We See It"); for some tests, I also used my vintage Audio Precision System One Dual Domain. The maximum output level at 1kHz was 4.732V from the balanced XLR jacks, 2.184V from the single-ended RCAs, the latter 0.76dB higher than the CD standard's 2V. Both pairs of outputs preserved absolute polarity (ie, were non-inverting). The XLRs are wired with pin 2 hot. The output impedance was a low 47 ohms from the RCA jacks, 142 ohms from the XLRs.

The D33's TosLink input locked to data with sample rates up to 96kHz, the coaxial S/PDIF input up to 192kHz, though the AES/EBU input wouldn't lock to the SYS2722's AES/EBU output. The Apple USB Prober utility revealed the USB1.1 input port as "ARCAM USB Audio 1.0," operating in isochronous asynchronous mode and handling 24-bit data with sample rates up to 96kHz, including 88.2kHz. The USB2.0 input port was identified as "ARCAM USB Audio 2.0" using "ARCAM DFU" firmware, and again operating in isochronous asynchronous mode. This input successfully operated with 24-bit data and sample rates up to 192kHz, including 176.4kHz.

Fig.1 shows the D33's impulse response with 44.1kHz data (one sample at 0dBFS) and with the Burr-Brown DAC chip's default filter engaged. The symmetrical ringing either side of the pulse reveals this filter to be a conventional linear-phase type. Switching to Arcam's Filter 1 gave the impulse response shown in fig.2. The ringing now occurs after the pulse, meaning that the filter is a minimum-phase type. By contrast, Filter 2 has a single cycle of damped ringing (fig.3). The frequency response with the Burr-Brown filter (fig.4) is conventional, with a sharp rolloff just below the sampling frequency. Not only is Filter 1 a minimum-phase type, its frequency response (fig.5) confirms that it is an "apodizing" filter; ie, it has a null at half the baseband sample rate, something that correlates with improved sound quality. With 192kHz data (green and gray traces), the response follows the smooth rolloff seen with lower sample rates, reaching –3dB at 60kHz. As suspected from its impulse response, Filter 2 is a slow-rolloff type that optimizes the time-domain performance at the expense of allowing some aliasing energy to leak into the audioband (fig.6).

The 0.2dB channel imbalance in these response graphs is larger than I would have expected. Channel separation, however, was excellent, at >125dB below 1kHz, and still 112dB at the top of the audioband. Other than some supply-related spuriae at 60Hz and its third and fifth harmonics, these all lying at or below –130dBFS (0.00003%), the D33 had a superbly low noise floor (fig.7). Increasing the bit depth from 16 to 24 with a dithered 1kHz tone at –90dBFS therefore dropped the noise by >14dB (figs.8, and 9), implying resolution of at least 18 bits. These graphs were taken via the coaxial S/PDIF input. The USB ports gave identical performance, revealing that they do indeed operate without truncating the 24-bit data. (See the "Measurements" sidebar of the review of AVM's Evolution C9 elsewhere in this issue for what happens with word-length truncation.) The peaks in these graphs just kiss the –90dBFS line, implying good low-level linearity, which, in conjunction with the low noise, allowed the D33 to reproduce a good facsimile of a sinewave with undithered 24-bit data at –90.31dBFS (fig.10). With undithered 16-bit data (fig.11), the Arcam DAC clearly distinguished the three DC voltage levels described by these data.

The D33 gave low levels of harmonic distortion, even with a punishing 600 ohm load (fig.12). The second and third harmonics are the only harmonics visible in this graph, which would be subjectively innocuous even if they lay higher than –80dB (0.01%). Tested for high-frequency intermodulation distortion with an equal mix of 19 and 20kHz tones and the D33 set to Filter 1, the second-order difference product lay at a very low –106dB (0.0005%) (fig.13), with the images of the primary tones at 25.1 and 24.1kHz well suppressed at below –90dB (0.003%). With the slow-rolloff Filter 2, however (fig.14), though intermodulation remains very low, the ultrasonic images are only 10dB lower in level than the primary tones, and some low-level aliasing products are now visible in the audioband. The steep rolloff of the Burr-Brown filter gave the lowest intermodulation and image suppression (fig.15).

The Arcam demonstrated excellent jitter rejection on its S/PDIF inputs. Operating in the preferred asynchronous mode, the D33's USB inputs were equally good, with a well-defined central peak representing the tone at one-quarter the sample rate and a clean noise floor free from jitter-related sidebands (fig.16).

Its measured performance indicates that the Arcam FMJ D33 is as well-engineered a product as I have come to expect from Arcam.—John Atkinson