Micromega AS-400 D/A integrated amplifier Measurements
To perform the measurements on the Micromega AS-400, 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 www.ap.com); 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 Micromega AS-400 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. As the AS-400 is a class-D design, one of the benefits of which is very high efficiency, I wasn't expecting the amplifier to be hot at the end of that time. However, the chassis was warm, particularly on the left-hand side, where the top cover measured 108.5°F.
The Micromega's MM-compatible phono input offered a maximum gain of 50.1dB at 1kHz, measured at the variable Preamp Out jacks, which is a little on the high side. However, it is fair to note that this includes the gain of the line-preamplifier section, estimated at 10.9dB. The unweighted, wideband signal/noise ratio ref. 5mV input at 1kHz with the input shorted was an excellent 71.2dB in the left channel and 72.7dB in the right. A-weighting improved these figures to 80 and 89dB, respectively. The phono input preserved absolute polarity (ie, was non-inverting), and the input impedance varied from 45k ohms at low and middle frequencies to 11k ohms at the top of the audioband. The RIAA-equalized response is shown in fig.1; there is a 0.2dB mismatch between the channels at some frequencies, and the RIAA correction suffers from a slight lack of midrange energy and a little too much treble energy. Though the errors are small in absolute terms, the frequency regions affected are wide enough for the change in response to be just audible. When Art Dudley described the sound of the phono section's top end as "light and detailed but not bright," this is what I would have expected from the measurement.
Channel separation via the phono input was excellent, as was distortion, which lay below 0.1% at typical recorded levels. However, the overload margin was not as large as I would have wished, being just 12dB at 20Hz, 9.2dB at 1kHz, and 7.5dB at 20kHz. High-output moving-magnet cartridges are best avoided.
I tested the AirStream feature using Apple Lossless files played with iTunes on my Intel-based MacBook. Setting iTunes to stream audio to the AS-400 was as straightforward as AD described, and, as with the phono input, I assessed performance at the variable Preamp Outs. AirStream data were restricted to a 16-bit word length and sample rates of 48kHz and below. A full-scale digital signal at 1kHz clipped the AS-400's preamplifier at volume-control settings greater than "7," at which setting the THD+noise was 0.1% and the output level was 3.26V; the output preserved absolute polarity. (The output level was the same using my iPhone 3G and AirPlay as the source.) The AirStream frequency response was perfectly flat to 20kHz, so I haven't shown it. Channel separation at 1kHz was a good 82dB, RL, and 88dB, LR. The separation was almost 20dB better in the bass, but 15dB worse above 10kHz.
The DAC's linearity error (not shown) was less than 1dB down to 107dBFS, and in the spectrum of a dithered 1kHz tone at 90dBFS (fig.2) the peak representing the tone just touched the 90dBFS line. The noise floor in this graph was a little higher than usual with 16-bit data, and there were some power-supplyrelated spuriae evident, though at low levels. Both observations were confirmed by FFT analysis (fig.3). However, the waveform of an undithered tone at exactly 90.31dBFS was perfectly symmetrical, with the three DC voltage levels described by the data clearly resolved (fig.4).
Distortion via AirStream was higher at high signal levels than at low. This can be seen in fig.5, which compares the spectra of a full-scale 1kHz tone (red trace) with that of a 1kHz tone at 10dBFS (blue trace). (Both were measured at the preamp outputs with the volume control set to "20," which is equivalent to a level of 730mV at the preamp outputs, in order to be sure that the distortion products are due to the digital decoder rather than the preamplifier circuit.) A regular series of harmonics can be seen with the 0dBFS tone, with the highest in level, the second harmonic, reaching 54dB (0.6%). At the lower recorded level, the distortion decreased dramatically, the second harmonic now lying at 80dB (0.0.01%). The picture was similar with intermodulation distortion, the 1kHz difference product from a full-scale mixture of 19 and 20kHz tones reaching 66dB (0.05%, not shown).
Jitter via the AirStream feed was moderately high, and estimated by the Miller Analyzer software to be 887 picoseconds pp, left, and 911ps pp, right. As shown by the spectrum (fig.6), the main sidebands lay at the power-supplyrelated frequencies of ±120, ±180, and ±240Hz, but what can also be seen in this graph is a significant widening of the base of the peak that represents the high-level 11.025kHz tone. This suggests the presence of fairly high levels of random low-frequency jitter that are not included in the Miller Analyzer's estimated figure.
Turning to the Micromega AS-400's line-level analog inputs, these preserved absolute polarity at both the preamp and speaker outputs, and offered an input impedance of 24k ohms across the band. This is lower than the specified 100k ohms, but should have no practical consequences. The maximum gain from the preamp section was 10.9dB. Driving the amplifier section directly gave a voltage gain of 32.3dB into 8 ohms, meaning that the maximum gain for the AS-400 assessed as an integrated amplifier was 43.2dB. The subwoofer output, which appears to be available from just the lower of the two RCA jacks labeled "Sub," rolled off by 0.8dB at 200Hz, 5dB at 300Hz, which is a lower crossover point than the 3dB at 400Hz point that is specified.
The source impedance for the preamp outputs was a low 100 ohms at all frequencies. The input impedance for the AS-400's power amplifier input was 45k ohms. The source impedance for the speaker outputs was a low 0.07 ohm at 20Hz and 1kHz, rising inconsequentially to 0.1 ohm at 20kHz. As a result, the variation in response with our standard simulated loudspeaker was minimal (fig.7, gray trace). While the Micromega's preamplifier-section response was flat to 200kHz, the response at the speaker outputs rolled off above 10kHz, reaching 0.5dB at 20kHz and 3dB at 50kHz, which slightly reduced the risetimes of a 10kHz squarewave (fig.8). This graph reveals that some ultrasonic noise is being produced by the class-D output stage; with no audio signal present, this noise measured 230mV with a center frequency of 480kHz. I used Audio Precision's AUX-0025 low-pass passive filter for all subsequent measurements to avoid the possibility of this noise driving the analyzer's input stage into slew-rate limiting.
Channel separation for the amplifier as a whole was good, at better than 90dB below 20kHz and still 70dB at 20kHz. Fig.9 shows how the THD+N percentage in the AS-400's output changed with output power. The specified maximum power is 400W into 4 ohms with both channels driven (23dBW); the amplifier actually delivered 190Wpc into 8 ohms (22.8dBW) at clipping (defined as 1% THD+N), 325Wpc into 4 ohms (22.1dBW), and 460W into 2 ohms with one channel driven (20.6dBW). With some class-D amplifiers, the distortion rises with frequency. However, the AS-400's THD remained commendably constant with frequency (fig.10).
The spectral content of the distortion varied with signal level. At low powers, it was heavily second-harmonic in nature (fig.11). At high powers, the third harmonic rose above the second, and higher-order harmonics appeared (fig.12). The Micromega AS-400 did well on the high-frequency intermodulation test, both the 1kHz difference tone and the higher-order products at 18 and 21kHz resulting from an equal mix of 19 and 20kHz tones, appearing at 80dB (0.01%), or just below visible waveform clipping on an oscilloscope (not shown).
Micromega's AS-400 measured well in most respects, especially considering that it has a class-D output stage. The AirStream feature was easy to use, though I was a little bothered by a higher level of distortion at high signal levels than I'd expected to see.John Atkinson