Quality Lies in the Details Page 8
As described in the January 1993 Stereophile (Vol.16 No.1), we are now able to assess the level of word-clock jitter in digital processors and CD players, measured where it matters: at the pins of the DAC. The technique uses the LIM Detector designed by Ed Meitner of Museatex. I won't go into the details here—they're in the aforementioned Stereophile article—but the procedure produces an overall measurement of the jitter (expressed in picoseconds), and also shows the spectral content of that jitter (figs.15 and 16). There's a good correlation between jitter and sound quality in that processors with high measured levels of jitter tend to sound musically uninvolving, sometimes even annoying. However, the spectral content of what jitter there is appears to be even more important for sound quality, as two high-quality transports with almost equally low levels of word-clock jitter—the Mark Levinson No.31 and C.E.C. TL 1, for example—still sound radically different from one another.
Fig.15 Average CD player, word-clock jitter spectrum, DC-20kHz, when processing 1kHz sinewave at -90dBFS (linear frequency scale, 10dB/vertical div., 0dB=1ns).
Fig.16 Excellent CD player, word-clock jitter spectrum, DC-20kHz, when processing 1kHz sinewave at -90dBFS (linear frequency scale, 10dB/vertical div., 0dB=1ns).
Tracking & error correction
A test we perform on CD players and transports checks the tracking and error-correction ability of the unit. The Pierre Verany Test Disc (Pierre Verany PV-784031) has a section with data dropouts of increasing length (you can easily see these by looking at the disc). The higher the track number in this series, the longer the dropout. By listening for when the test tone is interrupted and noting the track number, the dropout length the player can correct without error can be determined. The higher the track number, the better; track 35 is the typical limit.
When a processor under test is on the bench, we make sure it will lock to 32kHz and 48kHz sampling frequencies by driving it with the System One's digital signal generator at those sampling frequencies.
I must reiterate that good bench performance doesn't guarantee high sound quality. Moreover, there's absolutely no way to predict a component's sonic character from looking at test results. Nevertheless, a product that sounds good and measures well inspires more confidence than one that performs poorly on the test bench.