CD Player/Transport Reviews

I used 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") to measure the AVM Evolution C9 for both analog and digital inputs. For some tests, I used my vintage Audio Precision System One Dual Domain. To test the C9 as a CD player, digital processor, and phono preamplifier and to avoid overloading the power amplifier stage, I took the measurements from its Line Output jacks, which have a fixed output, and with the volume control set to mute the speaker outputs. The C9's various adjustable parameters were tested as set by Art Dudley for his auditioning, with the exception of the balance control, which was reset from 0.5dB in favor of the left channel to identical levels in both channels.

The Apple USB Prober utility identified the C9 as a "USB Audio DAC" manufactured by "Burr-Brown from Texas Instruments Japan" and operating in isochronous adaptive mode with a polling interval of 1ms. The C9's USB input was restricted to 16-bit data and sample rates of 32, 44.1, and 48kHz. The TosLink S/PDIF input operated with data having sample rates from 32 to 96kHz; although the coaxial input is specified as operating up to 192kHz, it wouldn't lock to datastreams with a sample rate greater than 96kHz. Tested with the Pierre Verany Test CD, the CD transport successfully handled gaps in the data spiral up to 1mm in length, but muted for longer gaps—these days, that is only average error correction, though I note that AD didn't have any problems playing damaged discs. Digital data at 0dBFS gave a level at the Line Output jacks of 2.14V, with the correct absolute polarity.

Fig.1 shows the C9's frequency response with CD data (cyan and magenta traces) and 96kHz S/PDIF data (blue and red). With both sample rates, the response drops rapidly just below the Nyquist frequency, but there is a slight rise in output above the audioband at the higher sample rate. Channel separation at 20kHz for digital sources (not shown) was good from R to L, at 90dB, but less good from L to R, at 67dB. In both cases, the amount of crosstalk decreased as the frequency decreased, reaching –110dB L–R and –120dB R–L, the latter buried in the noise floor.

The top pair of traces in fig.2 shows a 1/3-octave spectral analysis of the C9's outputs while it decoded dithered 16-bit CD data representing a 1kHz tone at –90dBFS. Other than an odd peak of unknown origin at 9kHz and some power-supply–related spuriae at 180, 300, and 420Hz, the trace is dominated by the dither noise used to encode the data. The bottom pair of traces was taken with 24-bit S/PDIF data. The drop in the noise floor has unmasked the supply-related spuriae; more important, it has also unmasked peaks at 3 and 5kHz, especially in the left channel. Repeating the spectral analysis with an FFT technique (fig.3) shows that there is also some seventh harmonic present with 24-bit data, and that supply-related components are present up to the 10kHz limit of the graph, albeit at a very low level. The presence of the odd-order harmonics suggests that the C9's S/PDIF input truncates 24-bit data, more so in the left channel than the right. This is confirmed by looking at the waveform of an undithered 24-bit tone (fig.4). Although the shape of the wave is obscured by the analog noise, discrete bit transitions are visible rather than a fairly smooth representation of a sinewave, as with the Arcam FMJ D33 processor that I review elsewhere in this issue. With 16-bit data, the three DC voltage levels were obscured by noise (fig.5)

There was also more harmonic distortion present with digital playback than I like to see, a full-scale tone at 50Hz being accompanied by the second harmonic at –76dB (0.015%) and the third at –64dB (0.06%), as well as a picket fence of low-level, higher-order harmonics (fig.6). However, intermodulation distortion (fig.7) was relatively low in level.

The Burr-Brown PCM 2704 USB audio chip, operated in isochronous adaptive mode, has never been very good at rejecting word-clock jitter, which was confirmed by a high-resolution plot of the C9's analog output while it decoded 16-bit J-Test data via its USB port (fig.8). As well as a strong pair of sidebands at ±450Hz, and another pair at the polling-interval–related frequencies of ±1kHz, many other sideband pairs can be seen. Surprisingly, feeding the C9's S/PDIF input with 24-bit J-Test data gave an almost identical picture (fig.9), the only difference being the absence of the sidebands at ±1kHz and the odd-order harmonics of the LSB-level low-frequency squarewave that overlays the high-level tone at one-fourth the sample rate. This is poor performance, I feel.

Assessed at the Line Output jacks, the line inputs preserved absolute polarity (ie, were non-inverting) and offered slightly less than unity gain. The input impedance was to specification but low, at 6800 ohms at all audio frequencies, which might result in lean-sounding low frequencies with tubed source components. The phono input also preserved absolute polarity and offered a gain of 47.8dB, the latter a little on the high side for moving-magnet cartridges but appropriate for high-output moving coils. The input impedance was an appropriate 44k ohms at low and middle frequencies, dropping slightly to 39.5k ohms at the top of the audioband. RIAA error was low, with excellent channel matching (fig.10), but with a slight upward tilt at high frequencies and a low-frequency rolloff reaching –3dB at 20Hz. Channel separation via the phono input was okay, at 75dB in both directions at 1kHz, though the unweighted, wideband signal/noise ratio, ref. 1kHz at 5mV input, was moderate, at around 48dB in both channels. A-weighting improved the S/N to 67.5dB.

With its slightly higher-than-usual gain for MM cartridges, the C9's phono-input overload margins were only just adequate, at 7dB at 20kHz, 7.75dB at 1kHz, and 11.7dB at 20Hz, all figures ref. 1kHz at 5mV. Distortion via the phono input, however, was low, and dominated by the third harmonic at –80dB (0.01%, not shown).

As the AVM C9 uses a class-D output stage, the chassis got only slightly warm during the testing. The volume control operated in accurate 0.5dB steps, up to a maximum of "99.5," with excellent matching between the channels. The maximum gain for line-level signals was on the high side for an integrated amplifier, at 47.4dB, and all inputs were non-inverting when measured at the speaker terminals.

The amplifier's output impedance was very low at all audio frequencies, at 0.07 ohm including 6' of speaker cable. As a result, there was very little modification of its frequency response with our standard simulated loudspeaker (fig.11, gray trace), and very little reduction of its output level as the load impedance dropped. This graph was taken with the volume control set to "63.0"; there was no change in the response at other settings of the volume control. The C9's output rolled off sharply above the audioband, presumably due to the presence of a low-pass filter to reduce the level of the output stage's ultrasonic switching noise. However, because this noise still lay at 418mV in the left channel and 368mV in the right, for all subsequent measurements I placed an Audio Precision AUX-0025 passive low-pass filter between the dummy load and the Audio Precision analyzer's input. This filter has a maximum input level of ±200V peak, and rejects noise above 200kHz without affecting the measured performance. Fig.12 shows the C9's reproduction of a 10kHz squarewave into 8 ohms with the AP filter in circuit. The C9's own filter slows the waveform's risetime, and there is a well-damped half-cycle of overshoot evident, which can also be seen on a 1kHz squarewave (fig.13).

Channel separation was 80dB in both directions across the audioband (not shown). Without the AP filter, the ultrasonic noise restricted the C9's unweighted, wideband S/N ratio to just 18dB (ref. 1W into 8 ohms, with the line input shorted), but the audioband ratio was much better, at 83dB. A-weighting gave an even better result, at 87dB, though some supply-related components were visible, mostly below –100dB (fig.14).

Testing the AVM C9's maximum power proved a little tricky because it became apparent that the line input stage overloaded with around 4V input. I therefore had to make sure that the input signal didn't exceed this level during the testing by keeping the volume control at or above "75.0." The C9's maximum power is specified at 180Wpc into 8 ohms (22.55dBW) or 300Wpc into 4 ohms (21.75dBW). Using our standard definition of clipping as the power when the THD+noise in the amplifier's output reaches 1%, the C9 exceeded its specified power with both channels driven, clipping at 205Wpc into 8 ohms (23.1dBW, fig.15) and 365Wpc into 4 ohms (22.7dBW, fig.16).

I tested how the THD+N percentage changed with frequency at a level, 15.5V, equivalent to 30W into 8 ohms, where I could be sure that I was measuring actual distortion rather than noise. The result is shown in fig.17. While the THD rises into lower impedances, this isn't by much; more notably, the THD doesn't rise to any significant extent at the top of the audioband, which is usually a class-D amplifier's Achilles' heel.

Fig.18 indicates that for the left channel, the distortion is predominantly the subjectively benign third harmonic, although in the right channel (fig.19, red trace) there is almost as much second harmonic as third. With its good high-frequency linearity, the C9 performed well on the demanding high-frequency intermodulation test (fig.20), with both the 1kHz difference tone and the higher-order tones at 18 and 21kHz lying at –83dB (0.007%). However, some enharmonic products can be seen in this graph.

Although it is a versatile and well-made product, I was disappointed by the AVM Evolution C9's measured performance, and particularly by that of its digital inputs. The C9 is not inexpensive at $5750, and at the price, I feel it should have a better-performing USB receiver chip than the PCM2704. And while CD playback was okay, the S/PDIF inputs don't appear to be optimized for playing back files with bit depths greater than the Compact Disc's 16.—John Atkinson