Lexicon MC-1 preamplifier/surround processor Measurements

Sidebar 3: Measurements

With such a versatile component as the Lexicon MC-1, it's hard to know what to measure first. But as almost everything it does in terms of ambience regeneration and surround-sound processing can be assessed only by listening to the result, I concentrated my examination of the MC-1 in terms of its basic signal-path functionality. All the measurements were performed with the Lexicon set to 2-Channel mode, the tone controls set to flat, and with the left and right front analog output jacks putting out full-range signals.

As the Lexicon performs all its processing in the digital domain, it has to digitize analog signals. I therefore looked first at the analog input's behavior. The input impedance was a usefully high 70k ohms, which will not unduly load down source components. The Lexicon can be set to automatically adjust its analog input sensitivity, in order to avoid clipping the ADC. Fig.1 shows the THD percentage plotted against input voltage with this sensitivity set to "00." The ADC can be seen to start to overload at an input level of 1.4V; CD players with their standard 2V output should be manually set to apply some gain reduction, or be used with the Lexicon set to Automatic. The A/D converter's linearity error (not shown) was very low, within 1dB down to below –110dBFS, which is excellent.

Fig.1 Lexicon MC-1, ADC THD+N (%) vs input level in volts (gain set manually to "00").

Channel separation via the analog input was also superb, and buried beneath the noise floor below 1kHz. As expected, the analog inputs are processed by an anti-aliasing filter, which sharply cuts off all content above 20kHz (fig.2). What surprised me about this graph was the 0.5dB channel imbalance. I therefore checked that the balance control was centered (it was), meaning that this imbalance was either coming from the ADC or somewhere further down the signal path.

Fig.2 Lexicon MC-1, ADC frequency response at 1V input (gain set manually to "–3"; 0.5dB/vertical div., right channel dashed).

Turning to the frequency response via one of the digital S/PDIF inputs (fig.3), the channel imbalance is still apparent, suggesting that it arises somewhere in the MC-1's D/A or analog output circuitry. The response shown in fig.3 was taken with a normal signal; feeding the Lexicon with pre-emphasized digital data gave a response that was no different, just as it should be.

Fig.3 Lexicon MC-1, DAC frequency response at –12dBFS (right channel dashed, 0.5dB/vertical div.).

Channel separation was as good through the digital inputs as it was through the analog. Fig.4 shows spectral analyses of the MC-1's analog output while it decoded digital data representing a dithered 1kHz tone at –90dBFS with both 16- and 24-bit word lengths. The increase in word length results in dropping the noise floor by 15dB, which is excellent. Note also the absence of spurious tones, harmonic distortion, and power-supply artifacts, indicative of good analog and digital design practice. Extending the measurement bandwidth to 200kHz and repeating the spectral analysis with data representing "digital black" (not shown) again revealed a low noise floor, and only a moderate rise in the ultrasonic noise that is due to the processor's delta-sigma DAC topology.

Fig.4 Lexicon MC-1, 1/3-octave spectrum of dithered 1kHz tone at –90dBFS, with noise and spuriae, 16-bit data (top) and 24-bit data (bottom). (Right channel dashed.)

The level of the 1kHz peak in fig.4 just touches the –90dBFS line, implying low DAC linearity error. This is confirmed by fig.5, which indicates that any amplitude error lies below 2dB to –115dBFS, and is actually dominated by noise to the bottom limit of the graph. As a result, the Lexicon's reproduction of an undithered 24-bit/1kHz tone at –90dBFS is nigh on perfect (fig.6).

Fig.5 Lexicon MC-1, DAC left-channel departure from linearity, 16-bit data (2dB/vertical div.).

Fig.6 Lexicon MC-1, waveform of undithered 1kHz sinewave at –90.31dBFS, 24-bit data.

Looking at the analog output stage, this was absolute-polarity–correct and had a low source impedance of 101 ohms. The maximum analog output level was 1.62V at 1kHz for a 0dBFS digital signal with the volume control set to "0.0." The control operated in accurate 1dB steps (though, again, with the 0.5dB channel mismatch), and an indicated gain of "+12" gave a maximum output of 6.45V—exactly 12dB above the reference level, and more than enough to drive the partnering amplifiers to their maximum power.

Harmonic distortion was vanishingly low, even at maximum level into a punishing 600 ohm load (fig.7). However, when I looked at how the Lexicon handled the demanding high-level mix of 19kHz and 20kHz tones, I was puzzled to see that, while the 1kHz difference component was very low in level (fig.8), a number of sidebands appeared around the fundamental tones, along with some noise-floor modulation. (For this test, the test signal peaked at 0dBFS, the volume control was set to "0.0," and the load was 100k ohms.) Given Lexicon's pedigree of digital signal processing, I would be very surprised to learn that this was due to DSP limitations, as had apparently been the case with the Technics DVD-A10 DVD-Audio player that Stereophile reviewed last November. It's possible, therefore, that wordclock jitter was to blame.

Fig.7 Lexicon MC-1, spectrum of 50Hz sinewave, DC–1kHz, at 0dBFS into 600 ohms (linear frequency scale).

Fig.8 Lexicon MC-1, HF intermodulation spectrum, DC–24kHz, 19+20kHz at 0dBFS into 100k ohms (linear frequency scale).

Accordingly, I then assessed the effect of such jitter by hooking up the Lexicon's analog output to the Miller Audio Research Jitter Analyzer and feeding the MC-1 digital data from a CD-R (played in a PS Lambda transport) representing a high-level 11.025kHz tone, over which had been superimposed the LSB toggling on and off at 229Hz. (The volume control was set to "+04" for these measurements, to maximize the resultant graph's dynamic range.)

The Analyzer software averages sixty-four 32k-point FFTs to produce a high-resolution spectrum of the analog signal; it then searches the FFT bins for symmetrical sideband pairs to the sides of the 11.025kHz fundamental. The result with the Lexicon connected to the PS transport via 6' of Apature S/PDIF cable is shown in fig.9. A large number of sidebands can be seen, resulting in a weighted jitter reading of 7044 picoseconds (7 nanoseconds)—almost 50 times worse than with the best digital gear I have tested!

Fig.9 Lexicon MC-1, high-resolution jitter spectrum of analog output signal (11.025kHz at –6dBFS with LSB toggled at 229Hz, CD data). PS Audio Lambda transport connected via 6' Apature S/PDIF cable. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.

The higher-level sidebands include a pair at ±113Hz (purple "7" markers) that contribute almost 1ns on their own to the total, with the harmonic series of these sidebands fully represented. Very strong sidebands can be seen at ±1250Hz (purple "72") and twice that frequency (purple "134"). While data-related jitter sidebands can be seen (red numeric markers), these are below the other forms of jitter present. And note the symmetrical rise in the noise floor around 1.4kHz away from the fundamental: this looks very similar to what is shown in the intermodulation plot (fig.8).

There is no way of knowing what, specifically, is happening inside the MC-1 to produce this dreadful-looking plot, but there was one thing I could examine—the Lexicon allows you to turn off its front-panel display. Doing so produced the plot shown in fig.10. While the broad sweep of lower-level sidebands remains unchanged, the high-level sidebands are either drastically reduced in level or eliminated entirely, which results in the weighted jitter level being more than halved, though this is to a still-high 3.16ns p–p. As a result of the decimation of the higher-level sidebands, a strong data-related pair of sidebands can be seen at ±229Hz (red "6" markers), which contributes 289ps of jitter.

Fig.10 Lexicon MC-1, display turned off, high-resolution jitter spectrum of analog output signal (11.025kHz at –6dBFS with LSB toggled at 229Hz). PS Audio Lambda transport connected via 6' Apature S/PDIF cable. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.

Whenever I get suspiciously good or bad test results, I repeat the measurement under identical conditions with a component I have measured before. This way, I can see if something has gone wrong with my measurement technique or equipment. The grayed-out trace in fig.11 is the spectrum of the MC-1's jitter taken with the PS Lambda connected to the processor via 15' of plastic TosLink cable. (The jitter level has dropped to 5.8ns.) For comparison, the black trace in this graph is the spectrum of the Musical Fidelity X-24K D/A processor driven from the PS transport via the same S/PDIF cable used to generate figs.9 and 10. The jitter level is a very low 200ps p–p, the main sidebands are all data-related, and the analog noise floor is a couple of dB lower—all of which confirm that the behavior seen in figs.9 and 10 is indeed representative of the Lexicon.

Fig.11 Musical Fidelity X-24K high-resolution jitter spectrum of analog output signal (11.025kHz at –6dBFS with LSB toggled at 229Hz, CD data). PS Audio Lambda transport connected via 6' Apature S/PDIF cable. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz. (Grayed-out trace was taken under identical conditions for the Lexicon MC-1 connected to the PS transport via 15' of TosLink cable.)

What does this mean? It should be remembered that this jitter behavior is only typical of what happens when the MC-1 handles a streamed digital medium, such as CD or a DVD-V with a PCM soundtrack. With compressed audio data (Dolby Digital or DTS), the data are transmitted in packets that need to be decoded, which will decorrelate any jitter from the audio. It has also been argued (by Dolby and others) that the sonic effects of word-clock jitter have been exaggerated. (LG was impressed by the unit's sound quality, even with CD sources.) But I have to say that, given the improvement in the measurement that resulted from turning off the MC-1's front-panel display, there is either far too much stray RF energy floating around inside the processor's chassis, or despite the extensive use of separate power supplies for the digital and analog circuitry, this still doesn't minimize the effect of RF noise on the audio circuitry.

I have mixed feelings about the Lexicon MC-1. Its fundamental analog and digital circuits seem very well implemented, but its susceptibility to jitter worries me. If in doubt, Toslink connections are to be preferred over electrical for CD playback—and definitely turn off that front-panel display!—John Atkinson

Lexicon Inc.
3 Oak Park
Bedford, MA 01730-1441
(781) 280-0300