McIntosh MS750 music server Measurements

Sidebar 3: Measurements

To test the McIntosh MS750, I ripped the CD-Rs carrying my assortment of test signals as FLAC lossless-encoded "CD Quality" files to the MS750's hard drive, then played them back as needed. (The MS750's CD drive rips CDs at a rate of about 15x.) Measurements were performed with Audio Precision's System One and System SYS2722 setups (see "As We See It" and Audio Precision's website.

The MS750 offers only unbalanced outputs. These turned out to invert signal polarity, but offered a very low source impedance of 1 ohm. The maximum output level was 2.07V RMS. The McIntosh's frequency response was basically flat, with negligible rolloffs at the top and bottom of the audioband (fig.1). Channel separation (not shown) was only moderate, but the measurement was somewhat compromised by low-frequency analog noise. (The MS750 was very sensitive to grounding issues between it and my test gear; floating its power-cord ground gave the lowest level of AC supply noise.)

Fig.1 McIntosh MS750, frequency response at –12dBFS into 100k ohms (right channel dashed, 0.5dB/vertical div.).

This noise reduces the player's dynamic range, as can be seen in fig.2, which analyzes the MS750's output with a swept 1/3-octave bandpass filter as it decodes data representing a dithered 1kHz tone at –90dBFS. The spectrum peaks at –90dB as expected, and above that frequency the measured noise floor is that of the recorded dither. But below 500Hz or so, the noise floor is that of the MS750. Fig.3 shows an FFT spectral analysis of the same signal: again, the signal peaks correctly at –90dB, but the greater resolution of this technique uncovers some spurious tones in the noise.

Fig.2 McIntosh MS750, 1/3-octave spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS, 16-bit data (right channel dashed).

Fig.3 McIntosh MS750, FFT-derived spectrum of 1kHz sinewave at –90dBFS, 16-bit data (linear frequency scale; left channel blue, right channel red).

Is this typical of the MS750's performance? Had I inadvertently set the MS750 to rip the test-CD data to lossy MP3 files rather than to lossless FLAC versions, which would also produce this kind of behavior? I fed the McIntosh's digital output to the digital input of the RME soundcard that resides in my test-lab PC and recorded the data to the PC's hard disk using Adobe Audition. A bit-for-bit file comparison with the original test-CD data indicated that I had correctly ripped the CDs at "CD Quality," and that both the ripping and the FLAC encoding/decoding were transparent. The lack of lower-frequency resolution was indeed characteristic of the MS750's analog outputs.

Fig.4 McIntosh MS750, left-channel departure from linearity, 16-bit data (2dB/vertical div.).

The higher-than-usual noise can be seen affecting the McIntosh's plot of linearity error against absolute signal level (fig.4). Although the error is vanishingly small down to –100dBFS, it becomes increasingly positive below that level due to the presence of noise. It also obscures the waveform of an undithered tone at exactly –90.31dBFS (fig.5), which ideally should comprise just three clearly defined DC voltage levels. (See the review of the Onkyo DX-7555 CD player elsewhere in this issue for an example of good performance on this test.)

Fig.5 McIntosh MS750, waveform of undithered 1kHz sinewave at –90.31dBFS, 16-bit data (left channel blue, right channel red).

The MS750 produced relatively high levels of THD. Fig.6, for example, shows a spectral analysis of its output while it reproduced a 50Hz tone at 0dBFS into 100k ohms. The third harmonic is the highest in level, at –66dB (0.06%). Although this is low enough to be inaudible by itself, it is accompanied by a regular series of higher-order harmonics, both odd- and even-order. This behavior was unaffected by a drastic reduction in load impedance, however.

Fig.6 McIntosh MS750, spectrum of 50Hz sinewave at 0dBFS into 100k ohms (linear frequency scale; left channel blue, right channel red).

The MS750 also offered disappointing performance when I looked at how it handled an equal mix of 19 and 20kHz tones, the combination's waveform peaking at 0dBFS. Fig.7 shows the spectrum of the player's output when loaded by a high 100k ohms. Again, this behavior didn't change significantly into low impedances; although the 1kHz difference product lay at a moderately high –75dB (0.02%), and the higher-order products at 18 and 21kHz at –69dB (0.04%), there are many other products visible. More unusually, a considerable amount of spectral spreading is evident with respect to the fundamental tones, something that tends to indicate high levels of random word-clock jitter.

Fig.7 McIntosh MS750, unbalanced HF intermodulation spectrum, 19+20kHz at 0dBFS peak into 100k ohms (linear frequency scale; left channel blue, right channel red).

To be sure that I hadn't inadvertently done something wrong, I redid the measurement, using the MS750 to play back the original test CD. There was no change in the result. I then abandoned the regular test systems and recorded the McIntosh's output on my laptop using a Echo Indigo PC card running at 24 bits and 96kHz. FFT analysis of the data file gave the same spectral spreading as before. (Repeating this test with a player with known low-IMD gave an excellent result, validating the Indigo's A/D performance.)

Looking at the MS750's word-clock jitter using the Miller Audio Research Analyzer and the diagnostic tone that I'd confirmed I had ripped at "CD Quality" gave a very high measured level of jitter: nearly 14 nanoseconds, or two orders of magnitude higher than the best players I have measured. Fig.8 shows the spectrum of the McIntosh's output, derived using the Audio Precision SYS2722 set to the same FFT analysis parameters as the Miller Analyzer. Again, the significant spectral spreading can be seen, as well as pairs of sidebands at high levels and low frequencies, and at higher frequencies but lower levels.

Fig.8 McIntosh MS750, high-resolution jitter spectrum of analog output signal (11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz), 16-bit CD data. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right channel red).

Overall, this is very disappointing measured performance, and not what I had expected from a product bearing the McIntosh label. Audiophiles might argue about sound quality, but until now, my experience of the veteran Binghamton brand's technical performance had been universally impressive. To take advantage of its superb features, the MS750's owner will need to use its digital output to feed an A/V receiver or a separate D/A processor (such as McIntosh's own MDA 1000, which Sam Tellig raved about in July 2005).—John Atkinson

McIntosh Laboratory, Inc.
2 Chambers Street
Binghamton, NY 13903-2699
(800) 538-6576