Audio Note CD-4.1x CD player Measurements
Sidebar 3: Measurements
I used Stereophile's loan sample of the top-of-the-line Audio Precision SYS2722 system to measure the Audio Note CD-4.1x (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 Audio Note player's error correction was excellent, there being no audible glitches in its output with the Pierre Verany test CD until the gaps in the data spiral reached 2mm in length. The player muted its output when the gap reached 2.4mm. The CD-4.1x's maximum output level at 1kHz was 2.45V, an audible 1.8dB higher than the CD standard's 2V RMS. The output inverted polarity and was sourced from a moderate output impedance of 650 ohms at 1kHz and 622 ohms at 20kHz. The output impedance rose at low frequencies, however, reaching a ridiculously high 12k ohms at 20Hz. This is due to the anode-follower topology and the necessarily finite size of the output coupling capacitor, and will result in lean-sounding low frequencies with preamps having low input impedances. Even with the 100k ohm input impedance of the Audio Precision SYS2722, the response was down by 2.5dB at 20Hz (fig.1, blue and red traces). When the CD-4.1x played back preemphasized data (fig.1, cyan and magenta traces) there was virtually no difference in response. The channels matched to within 0.1dB, which is good; channel separation (not shown) was >90dB above 800Hz, which is excellent, but decreased to just 50dB at 20Hz.
From here on, the CD-4.1x's measured performance was drastically affected by Peter Qvortrup's decision to use a non-oversampling (NOS) D/A converter topology without the usual reconstruction low-pass filter. The lack of a filter, either analog or digital, means that ultrasonic image energythe "negative frequencies" to which I refer in the title of my Richard M. Heyser Memorial Lecture to the Audio Engineering Society in October 2011, "Where Did the Negative Frequencies Go?"is present to its full extent in the DAC's output, which in turn means that its time-domain reproduction is optimal. Fig.2, for example, shows the impulse response of the CD-4.1x; it is indeed a perfect, if inverted, impulse.
But as I also explained in my AES lecture, the reconstruction filter plays a fundamentally important role in converting digital data to analog. Fed a series of DC voltage steps by the D/A converter, the filter literally reconstructs what the original analog signal would have been. Do away with that filter and, as you can see in fig.3, you get a raw, staircase waveform rather than what should be a perfect sinewave. Ironically, this is the meaningless criticism expressed by some analog diehards of conventional PCM systems: that "digital" results in a staircase analog waveform. With conventional players, it doesn't. However, that criticism does become meaningful in the context of something like the Audio Note CD-4.1x's non-oversampling technology.
My primary test for assessing a digital product's resolution is to sweep a 1/3-octave bandpass filter from 20kHz to 20Hz while the product decodes dithered data representing a 1kHz tone at 90dBFS. The resultant spectrum with the Audio Note player is shown in fig.4. The level of the noise floor is a little higher than the dither noise used to encode the signal. In addition, the traces fail to reach the 90dBFS line, suggesting a degree of negative linearity error, which was confirmed by the usual linearity test (fig.5), and there is a noticeable amount of second-harmonic distortion. (The negative linearity error is due to DAC bit errors transferring energy an octave higher.) Also visible are some power-supplyrelated spuriae, particularly in the left channel.
Higher-resolution FFT analysis gave an identical picture (fig.6), with the right channel (red trace) a little noisier than the left (blue). Repeating the noise-floor analysis for frequencies below 1kHz (fig.7) revealed AC spuriae in the left channel at 120, 180, and 240Hz. What fig.6 doesn't show is some very low-frequency noise in the CD-4.1x's outputI measured approximately 8mV of LF noise in the left channel, 36.5mV in the right. This won't give rise to any problems, but it did move the Audio Note's reproduction of an undithered tone at 90.31dBFS away from the 0V line (fig.8). Note that this waveform should be reproduced as symmetrical DC steps on either side of the time axis. (See fig.5 in the review of NAD's M51 DAC, elsewhere in this issue, for a perfect example of what it should look like.) The CD-4.1x's reproduction of this waveform is poor.
Distortion for full-scale signals was fairly high but dominated by the subjectively innocuous second and third harmonics (fig.9). Commendably, these harmonics increased only a little into lower load impedances. However, when it came to the high-frequency intermodulation test, the Audio Note's lack of a reconstruction filter allowed the negative-frequency image energy to play havoc with the spectrum (fig.10). Not only are the images of the 19 and 20kHz tones at 24.1 and 25.1kHz (44.120 and 44.119kHz, respectively) reproduced at almost full level, but very high-level, higher-order products occur at low frequencies.
I expected the poor intermodulation performance from Audio Note's NOS topology; what I did not expect was the very high level of jitter shown in fig.11, with high-level sidebands around the Fs/4 tone at the frequencies of the odd-order harmonics of the J-Test signal's LSB-level low-frequency squarewave. This is unusual with a conventional CD player, because there is no serial pathway in which the word clock is embedded in the data, as there is with an S/PDIF link. However, Audio Note states that the CD-4.1x comprises its CDT-Two/II transport and a slightly upgraded version of its DAC 2.1x; could the company have kept a jitter-prone S/PDIF link between the two rather than use a discrete clock connection?
It is fair to note that Art Dudley liked the sound of the Audio Note CD-4.1x. As I pointed out in my AES lecture, people do tend to like the sound of an NOS DAC, negative frequencies be damned. But I don't comprehend why that should be accompanied by higher levels of noise and distortion than good engineering would require, poor rejection of word-clock jitter, and a premium price. The CD-4.1x is a paradox: does it sound good because of how it measures or despite it?John Atkinson