Linn Majik DS-I D/A integrated amplifier Measurements

Sidebar 3: Measurements

To perform the measurements on the Linn Majik DS-I, I mostly used Stereophile's loan sample of the top-of-the-line Audio Precision SYS2722 system (see the January 2008 "As We See It" and; for some tests, I also used my vintage Audio Precision System One Dual Domain and the Miller Audio Research Jitter Analyzer. Before I did any testing of the Majik DS-I, I ran it at one-third power into 8 ohms for an hour, which imposes the maximum heat stress on an amplifier with a class-AB output stage. At the end of that time, the chassis was only warm.

The Majik DS-I's MM-compatible phono input offered a gain of 39.3dB at 1kHz, measured at the fixed-level Line Out jacks. The maximum gain from the variable Preamp Out jacks was 59.2dB, confirming that the Majik DS-I's preamp section offers up to 20dB of gain, as specified. The phono input preserved absolute polarity (ie, was non-inverting), and the input impedance varied from 40k ohms at 20kHz to 76k ohms at 20Hz.

The RIAA-equalized response was well matched between the channels and basically very accurate, though with a slight (0.25dB) downward tilt (fig.1). This response incorporates the IEC low-frequency rolloff, reaching –3dB at 17Hz, and is also sensibly rolled off above the audioband. Overload margins, ref. 1kHz at 5mV, were excellent, at >21dB at high and midrange frequencies, increasing to 25dB at 20Hz. Noise levels were low, with the unweighted, wideband signal/noise ratio measuring 69dB (ref. 1kHz input at 5mV), improving to 77.8dB when A-weighted.

Fig.1 Linn Majik DS-I, MM input, RIAA response at 5mV (left channel blue, right red), measured at Line Out jacks (0.25dB/vertical div.).

Turning to the Majik DS-I's performance as a D/A converter, I assessed this using the TosLink S/PDIF input. (I will publish a Follow-Up next month comparing its S/PDIF behavior with that via a wired Ethernet connection.) The digital input preserved absolute polarity and successfully locked to data with sample rates of up to 192kHz. Data representing a full-scale 1kHz tone gave a result at the Line Out jacks of 1.92V. The maximum distortion-free volume-control setting with this signal was "71," which gave a level of 40.5W into 8 ohms at the speaker jacks. A setting of "72" gave 49W, but with a THD figure of 2.33%.

Playing a dithered 1kHz tone at –90dBFS from CD while sweeping the center frequency of a 1/3-octave bandpass filter from 20kHz to 20Hz and looking at the signal at the Line Out jacks gave the spectra shown in fig.2. The traces peak at exactly –90dB, suggesting minimal linearity error, and with 24-bit data (bottom traces), the noise floor was approximately 10dB lower than it was with 16-bit data (top pair of traces). This implies a resolution of around 18 bits, which is sufficient to allow the Majik DS-I to correctly decode a dithered tone at –120dBFS (bottom traces). Repeating the analysis with an FFT technique gave the same result (fig.3), with almost no harmonic spuriae unmasked by the increase in bit depth.

Fig.2 Linn Majik DS-I, 1/3-octave spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS with 16-bit data (top) and 24-bit data (middle), plus dithered 1kHz tone at –120dBFS with 24-bit data (bottom). (Right channel dashed.)

Fig.3 Linn Majik DS-I, FFT-derived spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS with: 16-bit data (left channel cyan, right magenta), 24-bit data (left blue, right red).

The low noise floor and excellent linearity allowed the three DC levels described by an undithered sinewave at exactly 90.31dBFS to be readily resolved, with good waveform symmetry (fig.4), and 24-bit data gave a good representation of a sinewave (fig.5). Distortion via the digital input was very low, a full-scale tone being accompanied by just the second and third harmonics, at –98dB (0.0012%) and –106dB (0.0005%), respectively (fig.6). Similarly, intermodulation distortion was vanishingly low, the 1kHz difference product resulting from a full-scale mix of 19 and 20kHz tones lying at –103dB (0.0007%) in the right channel and –109dB (0.00035%) in the left (not shown).

Fig.4 Linn Majik DS-I, waveform of undithered 1kHz sinewave at –90.31dBFS, 16-bit data (left blue, right red).

Fig.5 Linn Majik DS-I, waveform of undithered 1kHz sinewave at –90.31dBFS, 24-bit data (left blue, right red).

Fig.6 Linn Majik DS-I

The Linn's rejection of jitter via its S/PDIF input was dependent on the source component. Via the RME soundcard in one of my test-lab PCs, a TosLink connection gave 454 picoseconds peak–peak of wordclock jitter, as assessed by the Miller Analyzer housed in the same PC. This was mostly due to sidebands at the data-related frequencies of ±229 and ±689Hz. However, these sidebands disappeared when the same 16-bit data were fed via TosLink from the Audio Precision SYS2722 (fig.7).

Fig.7 Linn Majik DS-I, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz, 16-bit S/PDIF data via 15' TosLink. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right red).

Looking at the Majik DS-I's behavior as a conventional integrated amplifier, it offered a maximum gain of 47.9dB into 8 ohms, which is higher than average. The speaker output, Line Out, and Preamp Out were all non-inverting. The line input impedance was a constant 10.5k ohms at all frequencies, which agrees with the specification. The Line Out's source impedance was 444 ohms and the Preamp Out 297 ohms, both figures constant at all frequencies.

The output impedance at the speaker terminals was 0.16 ohm at low and midrange frequencies, rising slightly to 0.18 ohm at the top of the audioband, which results in a +0.2dB/–0.1dB modification of the amplifier's frequency response with our simulated loudspeaker (fig.8, gray trace). This graph also indicates that the DS-I's response is curtailed at both ends of the spectrum, reaching –1.5dB at 10Hz and 29kHz, and –0.5dB at 20Hz and 16kHz. This was not affected by the volume-control setting, but as a result, the Linn's reproduction of a 10kHz squarewave has longer risetimes than normal (fig.9), though there is no sign of overshoot or ringing.

Fig.8 Linn Majik DS-I, frequency response at 2.83V into: simulated loudspeaker load (gray), 8 ohms (left channel blue, right red), 4 ohms (left cyan, right magenta), 2 ohms (green). (0.25dB/vertical div.)

Fig.9 Linn Majik DS-I, small-signal 10kHz squarewave into 4 ohms.

The volume control operated in accurate 1dB steps, with "80" the unity-gain setting. Channel separation is disappointing, at 75–80dB below 3kHz (not shown), while the unweighted, wideband S/N ratio, taken with the input shorted but the volume control set to its maximum, was a modest 56.8dB ref. 2.83V into 8 ohms. This improved to 63.5dB when A-weighted.

Figs. 10 and 11 show how the percentage of THD+noise present in the DS-I's output varies with power into 8 and 4 ohms, respectively. The downward slope of the traces below 20W in both graphs implies that the measured percentage is dominated by noise at low powers. This graph also indicates that the Majik DS-I clips (defined as 1% THD+N) at 45Wpc into 8 ohms (16.5dBW) and at 85Wpc into 4 ohms (16.3dBW), the latter an inconsequential 0.2dB below the rated power of 90Wpc.

Fig.10 Linn Majik DS-I, distortion (%) vs 1kHz continuous output power into 8 ohms.

Fig.11 Linn Majik DS-I, distortion (%) vs 1kHz continuous output power into 4 ohms.

I plotted the Majik DS-I's THD+N percentage at 10V, a level at which the distortion will dominate the measurement. Fig.12 shows that the Linn offers low distortion even into low impedances at this level, though there is the usual rise in THD in the top two octaves. There is also an odd rise in THD between 200 and 600Hz, particularly in the right channel and into lower impedances, though this is still low in absolute terms.

Fig.12 Linn Majik DS-I, distortion (%) vs frequency at 10V into 8 ohms (left channel blue, right red), 4 ohms (left cyan, right magenta), 2 ohms (green).

The distortion at low powers consists primarily of the subjectively benign third harmonic (fig.13), but regular series of higher-order harmonics can be seen at higher powers (fig.14). Though none of these components rises significantly above –100dB (0.001%), I'd rather they weren't there at all.

Fig.13 Linn Majik DS-I, 1kHz waveform at 8W into 8 ohms (top), 0.01% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).

Fig.14 Linn Majik DS-I, spectrum of 50Hz sinewave, DC–1kHz, at 53W into 4 ohms (left channel blue, right red; linear frequency scale).

Finally (fig.15), the second-order difference component resulting from an equal mix of 19 and 20kHz tones at a level just below visible waveform clipping on the oscilloscope lay at –68dB (0.04%), though higher-order products were all lower in level. However, I was puzzled by the noise present at low frequencies in this graph, which could not be eliminated by experimenting with the grounding between the test set and the amplifier.

Fig.15 Linn Majik DS-I, HF intermodulation spectrum, DC–24kHz, 19+20kHz at 30W peak into 8 ohms (linear frequency scale).

The Linn Majik DS-I offers a fairly respectable measured performance, let down only by its limited channel separation and the relatively high level of background noise.—John Atkinson

Linn Products Limited
US distributor: TC Group Americas
335 gage Avenue #1
Kitchener, Ontario, Canada N2M 5E1
(519) 745-1158

deckeda's picture

1) Linn says Ethernet sounds better but doesn't explain why.
2) AD concurs.
3) JA apparently theorizes about jitter and goes looking but doesn't find meaningful differences.
4) Linn says yes, it's lower jitter that makes RJ-45 et al better but JA's test equipment can't reveal it.
5) JA says actually, yes it can.


I didn't get a sense at all that the "stupidness" software issue has been overcome. And Linn---"open" software more often than not means everyone is free to design by committee, with predicable results. Don't tout that too strongly.

There are other facets not covered here, like why iTunes isn't a great ripper (it has a selectable error correction, giving the software more time for example) or why something that incorporates CD Paranoia while ripping wouldn't suffice.

This review is a good example of the perils of subjecting single components to scrutiny that demands investigation beyond the norm---so much more interaction occurs with computer-based music replay.