Columns Retired Columns & Blogs |
hum? in a 21st century amplifier? front end which overloads due to too much gain? this amp is garbage.
Before performing tests on an amplifier, I thermally stress it by running both channels for 60 minutes into 8 ohms at one-third the specified power—in the case of the Naim Supernait, 27Wpc. The THD+noise percentage was 0.025% at the beginning of that period; it hadn't significantly changed at the end, by which time the amplifier's case was quite hot, though not so that I couldn't keep my hand on it.
Looking first at the Naim's digital input, it locked on to data with sample rates ranging from 32 to 96kHz, and the input was non-inverting. A signal at 0dBFS gave a level that clipped the Supernait's main output with the volume control at 10 o'clock. With the volume control full, a digital signal at –26.5dBFS clipped the speaker outputs, meaning that the amplifier as a whole has way too much gain for its digital input. Surprisingly, the frequency response above the audioband with 96kHz-sampled data, assessed at the tape output (fig.1, blue and red traces) merely continued the slight top-octave rolloff evident with CD data (fig.1, yellow and green traces), with a –3dB point of 31kHz. The output with CD data was down 0.25dB at 20kHz, and the channels were very well matched in level.
Fig.1 Naim Supernait, frequency response into 100k ohms with external data sampled at 96kHz (left, blue; right, red) and at 44.1kHz (left, yellow; right, green). (0.25dB/vertical div.)
Spectral analysis of the Supernait's tape output while it decoded dithered 16-bit data representing a 1kHz tone at –90dBFS have the top pair of traces in fig.2. The tone peaks at the correct level, and the high-frequency noise floor is that of the recorded dither. However, at lower frequencies, there is a bump to –117dBFS at the supply-related frequency of 60Hz. Extending the word length to 24 bits and driving the Supernait with dithered data representing 1kHz tones at –90dBFS and –120dBFS gave the middle and bottom pairs of traces in fig.2. The noise floor drops by 15dB at high frequencies, suggesting a true D/A resolution of around 18 bits, but a series of odd-order AC-supply harmonics are unmasked in the bass and midrange.
Fig.2 Naim Supernait, 1/3-octave spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS, 16-bit data (top) and 24-bit data (middle); and of 24-bit dithered tone at –120dBFS (bottom). (Right channel dashed.)
I experimented with the grounding between the Supernait and both of my measurement setups—Stereophile's vintage Audio Precision System One Dual Domain, and a new Audio Precision SYS2722 we have on loan (see www.ap.com)—and couldn't eliminate this low-level hum. As the hum occurs at the supply frequency of 60Hz and its odd harmonics rather than at the full-wave–rectified supply frequency of 120Hz, I suspect it is due to magnetic coupling from the large toroidal transformer rather than some kind of grounding problem. Yes, these components are at a very low level, but I would rather not have seen them at all.
DAC linearity error, assessed with 24-bit data, was negligible (fig.3), and the Naim's reproduction of an undithered tone at –90.31dBFS was superb, with good waveform symmetry and a clean depiction of the three DC voltage levels (fig.4). Word-clock jitter, assessed with the Miller Audio Research Analyzer, was quite low, at 367 picoseconds peak–peak. The spectrum of the Naim's output while it reproduced the jitter test signal (fig.5) was dominated by data-related sidebands, these close to the residual level.
Fig.3 Naim Supernait, left-channel departure from linearity, 24-bit data (2dB/vertical div.).
Fig.4 Naim Supernait, waveform of undithered 1kHz sinewave at –90.31dBFS, 16-bit data (left channel blue, right channel red).
Fig.5 Naim Supernait, high-resolution jitter spectrum of analog output signal (11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz), 16-bit external data sourced from PC via 15' TosLink. Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz (left channel blue, right channel red).
Turning to the analog domain (there was no digital connection during these tests, to make sure the D/A section was turned off) and looking at the speaker-level outputs, the Supernait's line-level inputs gave a maximum gain into 8 ohms of 49.5dB, which is significantly higher than usual for an integrated amplifier. The input impedance was a usefully high 64k ohms at 1kHz, dropping slightly but inconsequentially at the frequency extremes, and the amplifier preserved absolute polarity; ie, was non-inverting.
The output impedance was a little higher than usual for a solid-state design, at 0.3 ohm across most of the audioband. As a result, the modifications of the amplifier's response by the usual Ohm's Law interaction between its output impedance and the impedance of the speaker was a mildly audible ±0.3dB (fig.6, magenta trace). Into resistive loads, the output rolls off quite quickly above 20kHz, resulting in relatively leisurely risetimes on the amplifier's reproduction of a 10kHz squarewave (fig.7). Channel separation (not shown) was only moderately good, at 70dB or higher below 2kHz, while the Supernait's unweighted signal/noise ratio, assessed with the input shorted but the volume control at its maximum, was compromised by the high gain and by the residual 60Hz hum, at 56dB ref. 2.83V into 8 ohms. The A-weighted figure was better, at 64.5dB left and 67.5dB right.
Fig.6 Naim Supernait, left-channel frequency response at 2.83V into: simulated loudspeaker load (magenta), 8 (red), 4 (blue), 2 (green) ohms (1dB/vertical div.).
Fig.7 Naim Supernait, small-signal 10kHz squarewave into 4 ohms.
Fig.8 shows how the THD+N percentage in the Naim's output varies with output power into 8, 4, and 2 ohms. The downward slope of the traces below 20W indicates that the reading is dominated by noise rather than distortion. Defining clipping as 1% THD, the Naim more than meets its specified power into 8 ohms, at 89Wpc (19.5dBW) with both channels driven. It clipped at 140Wpc into 4 ohms (18.5dBW), but turned off at 141W into 2 ohms to protect itself, which is why the trace in fig.8 ends at that point. Reducing the drive signal allowed the amplifier to turn itself back on.
Fig.8 Naim Supernait, distortion (%) vs 1kHz continuous output power into (from bottom to top at 10W): 8, 4 ohms.
Fig.9 plots how the THD+N percentage changed with frequency at 8V, equivalent to 8W into 8 ohms (blue), 15W into 4 ohms (red), and 32W into 2 ohms (green). It stays commendably low at all frequencies, partly because the background noise still contributes to the reading. At a slightly higher power into 8 ohms (fig.10), the distortion is primarily the benign second harmonic, but still overlaid with noise despite my having averaged 32 separate readings to produce this graph.
Fig.9 Naim Supernait, left-channel THD+N (%) vs frequency at 8V into: 8 (blue), 4 (red), 2 (green) ohms.
Fig.10 Naim Supernait, 1kHz waveform at 18W into 8 ohms (top), 0.02% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).
Even at high powers, the distortion remained predominantly second-harmonic in nature (fig.11). The second harmonic in this graph lies at –72dB (0.025%) and the third at –84dB (0.006%). But note that the power-supply component at 180Hz is almost as high as the second harmonic, especially in the right channel (red trace), and that a regular series of off-order power-supply harmonics extends across the spectrum. As I conjectured earlier, I suspect that this behavior is due to magnetic interference from the toroidal transformer that dominates the Supernait's interior. These spuriae can also be seen, albeit at a low level, in the spectrum of the amplifier's output while it reproduced an equal mix of 19 and 20kHz tones into 4 ohms at a level just below visible clipping on the screen of the oscilloscope with which I monitor my measurements (fig.12). Actual intermodulation is commendably low, however, with the 1kHz difference component lying at –83dB and the higher-order components at 18kHz and 21kHz at –79dB.
Fig.11 Naim Supernait, spectrum of 50Hz sinewave, DC–1kHz, at 110W into 4 ohms (linear frequency scale; left channel blue, right channel red).
Fig.12 Naim Supernait, HF intermodulation spectrum, DC–24kHz, 19+20kHz at 100W peak into 4 ohms (linear frequency scale; left channel blue, right channel red).
In many ways, the Naim Supernait offers excellent measured performance. Its D/A section is also good, and not just an afterthought. But the presence of the 60Hz AC-supply components in its output, even though they lie at a low level, did concern me, and the amplifier's ultimate noise performance is compromised to an extent by the higher-than-necessary gain.—John Atkinson
hum? in a 21st century amplifier? front end which overloads due to too much gain? this amp is garbage.