Linn Unidisk SC universal disc player Measurements

Sidebar 3: Measurements

Despite its small size and unprepossessing appearance, the Unidisk SC is one of the most versatile components I have encountered, hence one of the most complicated to assess on the test bench. Its excellent handbook was a big help, but you do need to hook the Linn up to a TV or video monitor in order to access its setup menus. Doing so added hum, so I disconnected the TV once I had the player functioning as I needed. As well as using S/PDIF data fed to one of its TosLink digital inputs, I burned a CD-R of 16-bit test tones and a DVD-Audio–formatted DVD-R of 24-bit test tones. I also used the "provisional" test SACD from Sony.

The volume control operated in steps that varied from 0.5dB to 1dB, with the average over the range "40" to "90" equal to 0.8dB. One complicating factor was that with digital sources, the maximum setting of "100" resulted in significant overload; the maximum usable setting (which I defined as 1% THD+noise) was "96" for CD but "90" for SACD, DVD-A, and external data, which will interfere with A/B comparisons between CD and hi-rez media. The output level at these settings was 2.85V, or 3.1dB higher than both the CD standard and the specification. The analog inputs overloaded at the same output level, and offered a maximum gain of 13.15dB. These inputs offered an impedance of 5.8k ohms, which is higher than the 300 ohms specified but will give a bass-shy sound with typical capacitor-coupled, tubed source components. The Unidisk's analog inputs offered a wide measured bandwidth, –3dB at 129kHz, meaning that these are not digitized. The "TV Out" jacks gave a fixed output level, peaking at 1.97V.

Data fed to the digital inputs was not passed through to the digital outputs, these offering CD data for CD playback, nothing for SACD and DVD-A playback, and 48kHz or 96kHz data with DVD-Vs, depending on the disc and the Unidisk's settings. The digital input successfully locked to 32kHz, 44.1kHz, 48kHz, and 96kHz datastreams, but would not lock to 88.2kHz data. The analog output was noninverting, with a lowish source impedance of 297 ohms. Error correction on CD playback appeared to be excellent, the Unidisk playing through gaps in the data spiral of up to 1.25mm without audible glitches, though the longest gaps in the data spiral activated the error flag in the S/PDIF output. However, despite this fine performance I did note some persistent dropouts with some of my older test CDs, which have many fine scratches on their surfaces but play without problem on other players.

Playing CDs, the Unidisk's response rolled off slightly and inconsequentially at the frequency extremes (fig.1, top pair of traces), but suffered from a 0.5dB error in the mid-treble with pre-emphasized discs (fig.1, bottom traces). This rolloff continued smoothly above the audioband with playback of both DVD-As (fig.2, top traces) and SACDs (fig.2, bottom traces). The player's output dropped like a stone above 45kHz with DVD-A; with SACD, the ultrasonic output slowly continued to drop above that frequency, reaching –26dB at 120kHz. The shape of these responses indicates that the Linn has a reconstruction filter optimized for time-domain performance, this generally felt to correlate with good sound quality. Channel separation measured through the front-channel outputs (not shown) was superb, at better than 100dB below 2kHz, though this degraded to 80dB (right–left) and 85dB (L–R), due to the usual capacitive coupling.

Fig.1 Linn Unidisk SC, frequency response at –12dBFS into 100k ohms, with de-emphasis (bottom) and without (top). (Right channel dashed, 0.5dB/vertical div.)

Fig.2 Linn Unidisk SC, DVD-A frequency response at –12dBFS into 100k ohms (top), SACD response at –3dBFS (bottom). (Right channel dashed, 0.5dB/vertical div.)

Spectral analysis of the Unidisk's analog output while it played back CD data representing a dithered 1kHz tone at –90dBFS (fig.3, top traces) revealed that the noise floor was a little higher than usual. Playing a DVD of 24-bit data representing the same signal gave the lower pair of traces in fig.3: while the increase in word length resulted in an improvement of just 6dB in dynamic range in the treble, there was more like 15dB improvement at lower frequencies, which is odd. Repeating the analysis with SACD data (fig.4, top traces) showed similarly improved dynamic range in DVD-A playback, though the Unidisk only just resolved a dithered 1kHz tone at –120dBFS (fig.4, bottom traces).

Fig.3 Linn Unidisk SC, 1/3-octave spectrum of dithered 1kHz tone at –90dBFS, with noise and spuriae, 16-bit CD data (top), 24-bit DVD-A data (bottom). (Right channel dashed.)

Fig.4 Linn Unidisk SC, SACD data, 1/3-octave spectrum of dithered 1kHz tone at –90dBFS (top), –120dBFS (bottom), with noise and spuriae (right channel dashed).

Comparing figs.4 and 3, SACD's rising noise floor is evident in the top two audio octaves. Extending the analysis out to 200kHz and using data representing a 1LSB DC offset (CD) or a dithered tone at –150dBFS (DVD-A, SACD) to keep the converters active, you can see SACD's ultrasonic noise peaking at –50dB in the 100kHz region (fig.5). However, you can also see that there is a somewhat lower peak in this frequency region with linear PCM playback, this identical for both 16- and 24-bit data. I assume this is due to the noise-shaping used by the Unidisk's D/A converters.

Fig.5 Linn Unidisk SC, 1/3-octave spectrum of –1LSB, with noise and spuriae, 16-bit CD data (bottom); of a dithered tone at –150dBFS with 24-bit DVD-A data (bottom) and SACD data (top). (Right channel dashed.)

Linearity error with PCM data was very low (fig.6), and the Linn's reproduction of a 24-bit, 1kHz sinewave at –90dBFS was good, if a bit noisy (fig.7). I haven't shown the undithered waveform with CD data because I got anomalous results: The waveform was slightly different each time I did the test, as was the level of the noise floor.

Fig.6 Linn Unidisk SC, departure from linearity, 16-bit CD data (top), 24-bit DVD-A data (bottom). (Right channel dashed, 2dB/vertical div.)

Fig.7 Linn Unidisk SC, waveform of dithered 1kHz sinewave at –90.31dBFS, 24-bit DVD-A data.

The Unidisk offered very low levels of distortion, even into low impedances. Decoding 24-bit data on DVD-A representing a full-scale 1kHz tone driving a load of 8k ohms, I got the spectrum shown in fig.8. The THD (true sum of the harmonics) was 0.00125%, and the third harmonic was the highest in level, at just –100dB (0.001%). The random noise floor in this graph is actually that due to the analyzer; however, some discrete tones that are not harmonically related to the test signal can be seen, these due to the Linn. I wonder if these are spuriae from the switch-mode power supply. They are very low in level, and I doubt that they affect the sound quality.

Fig.8 Linn Unidisk SC, spectrum of 1kHz sinewave, DC–1kHz, at 0dBFS into 8k ohms, 24-bit DVD-A data (linear frequency scale).

Repeating this analysis with CD data representing an equal mix of 19kHz and 20kHz tones gave what was, in general, a very clean-looking spectrum (fig.9), with very low levels of intermodulation products. However, I got a different result when I fed one of the Unidisk's S/PDIF inputs with the same 16-bit data (fig.10). There is now a noticeable modulation of the noise floor around the tones, with some high-order products evident. In this respect, the Linn's behavior is reminiscent of the Esoteric X-01 SACD player, which I reviewed in the May 2005 Stereophile (p.125). I doubt that this behavior will in itself have audible consequences, but it does suggest some shortfall either in the player's digital signal-processing resources or in its data receiver circuit.

Fig.9 Linn Unidisk SC, HF intermodulation spectrum, DC–25kHz, 19+20kHz at 0dBFS into 8k ohms, CD data (linear frequency scale).

Fig.10 Linn Unidisk SC, HF intermodulation spectrum, DC–25kHz, 19+20kHz at 0dBFS into 8k ohms, external 16-bit data (linear frequency scale).

This difference between the Unidisk SC's behavior when playing discs and when fed external data can also be seen in fig.11, which shows a narrowband, FFT-derived analysis of the player's analog noise floor produced by the Miller Jitter Analyzer while it decoded an analytical tone at exactly one-quarter the sample frequency. The black trace in this graph shows the result with CD data. The calculated jitter level was a low 257 picoseconds peak–peak, this mainly coming from sideband pairs at ±15Hz (purple "1" numeric markers), at the data-related frequencies of ±229Hz (red "3"), ±794Hz (purple "5", and ±914Hz (purple "6").

Fig.11 Linn Unidisk SC, CD data, high-resolution jitter spectrum of analog output signal (11.025kHz at –6dBFS sampled at 44.1kHz with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz. Grayed-out trace was taken with external 24-bit data under identical conditions.

By contrast, the grayed-out trace in fig.11 shows the spectrum taken with the Unidisk's data input fed 24-bit data representing the same analytical signal. (The data source was a PC with an RME soundcard connected to the Linn with a TosLink S/PDIF cable.) Usually, these test conditions result in a drop in measured jitter. However, in the case of the Unidisk the jitter has risen by a factor of five, to a high 1.29 nanoseconds, with data-related sidebands at ±229Hz contributing significantly to this total, despite the increase in bit depth. An enigma, even if you disregard the noise-floor modulation for external data, similar to that seen in fig.10.

The Linn Unidisk SC's measured behavior when fed external digital data raised my eyebrows, and I admit to some puzzlement at the occasional inconsistency of the results I got using CD as a source. However, its performance with hi-rez SACD and DVD-Audio discs is beyond reproach.—John Atkinson

US distributor: Linn Inc.
8787 Perimeter Park Boulevard
Jacksonville, FL 32216
(904) 645-5242