Plinius Audio SA-103 power amplifier Measurements
To perform the measurements on the Plinius SA-103, I 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). As Erick Lichte listened to the SA-103 only in balanced stereo mode, I didn't test the amplifier in bridged mono mode.
Before doing the testing, I ran the Plinius SA-103 at one-third its rated power for 60 minutes, which thermally is the worst case for an amplifier with a class-A/B output stage. At the end of that period the chassis was only slightly warm, with a temperature of 97°F (36°C), while the heatsinks were actually cooler, at 91.4°F (33°C). Repeating the test with the SA-103's output stage to class-A bias drastically changed the picture, with the heatsinks ending up at 115°F (46°C) and the chassis at 111°F (43.6°C). Running this amplifier in class-A will require good ventilation.
The voltage gain into 8 ohms was the same in both modes, at 31.5dB for the balanced input, which is 6.5dB lower than specified, but the same 31.5dB for the unbalanced input. The XLR jacks are wired with pin 2 hot, meaning that the amplifier preserves absolute polarity through both the balanced and unbalanced inputs. The input impedance is specified as 47k ohms; I measured 45k ohms for the unbalanced input at low and middle frequencies, this dropping slightly but inconsequentially to 38k ohms at 20kHz. The balanced input measured 60k ohms at all frequencies.
The output impedance in class-A/B mode at low and midrange frequencies was 0.095 ohm (including 6' of speaker cable); switching to class-A lowered this very slightly, to 0.086 ohm. (Both modes rose very slightly, to 0.1 ohm, at 20kHz.) As a result of this low value, the modification of the amplifier's frequency response due to the Ohm's Law interaction between this source impedance and the impedance of our standard simulated loudspeaker was less than ±0.1dB (fig.1, gray trace). The response rolled off above the audioband, reaching 3dB at 61kHz into 8 ohms (fig.1, blue and red traces), which slowed the edges of a 10kHz squarewave (fig.2). There is also the faintest hint of overshoot in this graph, but no ringing. The 1kHz waveform is superbly square (fig.3).
Fig.1 Plinius SA-103, class-A bias, 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.2 Plinius SA-103, class-A/B bias, small-signal 10kHz squarewave into 8 ohms.
Fig.3 Plinius SA-103, class-A/B bias, small-signal 1kHz squarewave into 8 ohms.
Channel separation via the balanced inputs was only fair, at 77dB across the audioband. The unweighted, wideband signal/noise ratio (ref. 1W into 8 ohms), measured with the input shorted, measured 65.9dB, improving to 79.5dB when the measurement bandwidth was restricted to the audioband, and to 85dB when A-weighted.
These measurements were taken with the amplifier operating in class-A; switching to class-AB didn't change its frequency-domain behavior. However, when it came to distortion, the SA-103 behaved quite differently in its two modes. Fig.4 shows how the THD+noise percentage in the amplifier's output varied with output power into 8, 4, and 2 ohms in class-A/B; fig.5 shows its behavior in class-A. The clipping powers (at 1% THD+N) with both channels driven are identical in both modes into 8 and 4 ohms, at 127W (21dBW) and 235W (20.7dBW), respectively. With one channel driven, a little more power is available into 2 ohms in class-A/B, at 368W (19.6dBW), than in class-A, 345W (19.35dBW). However, while the shapes of the traces in these two graphs are similar, there is less distortion in class-A at powers below a few tens of watts.
Fig.4 Plinius SA-103, class-A/B bias, distortion (%) vs 1kHz continuous output power into (from bottom to top at 100W): 8, 4, 2 ohms.
Fig.5 Plinius SA-103, class-A bias, distortion (%) vs 1kHz continuous output power into (from bottom to top at 100W): 8, 4, 2 ohms.
This can also be seen in the plots of THD+N percentage against frequency (fig.6, class-A/B; fig.7, class-A), taken at a level where the measurement will not be dominated by noise. However, not only does the THD rise into lower impedances in both modes, there is a rise in THD in the top three octaves that is actually greater in class-A. In this respect, the SA-103 behaves in a very similar manner to the Plinius SA-Reference that Paul Bolin reviewed in May 2006 (see figs.4 & 5 in that review).
Fig.6 Plinius SA-103, class-A/B bias, THD+N (%) vs frequency at 18.75V into: 8 ohms (left blue, right red), 4 ohms (left cyan, right magenta), 2 ohms (green).
Fig.7 Plinius SA-103, class-A bias, THD+N (%) vs frequency at 18.75V into: 8 ohms (left blue, right red), 4 ohms (left cyan, right magenta), 2 ohms (green).
The harmonic signature of the distortion is also very different in the two modes. Fig.8 shows the waveform of the spuriae in class-A/B; even though the THD+N percentage is low, the third-harmonic nature of the distortion is overlaid by sharp spikes at the waveform's zero-crossing points. This is classic crossover distortion, suggesting a lack of output-stage bias. By contrast, fig.9 shows the waveform of the spuriae at the same level in class-A mode: There are now no zero-crossing spikes, the THD+N percentage has halved, and the signature is predominantly the subjectively innocuous second and third harmonics.
Fig.8 Plinius SA-103, class-A/B bias, 1kHz waveform at 20W into 4 ohms (top), 0.015% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).
Fig.9 Plinius SA-103, class-A bias, 1kHz waveform at 20W into 4 ohms (top), 0.0075% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).
FFT analysis shows even more clearly the improvement wrought by class-A operation (fig.10). In class-A mode (blue and red traces), the only harmonics visible are the second at 94dB (0.002%) and the third at 100dB (0.001%). Switching to class-A/B mode (cyan and magenta traces) not only increased the levels of the second and third harmonics to 90dB (0.003%) and 86dB (0.005%), respectively, but now also visible is a picket fence of higher-order harmonics, these associated with the spikes in the distortion waveform in fig.8. Note that some AC supplyrelated spuriae can be seen, especially in the right channel (red and magenta traces). I couldn't eliminate these by experimenting with the grounding between the Plinius and the test system, though it is fair to note that, above 240Hz, these are all at a very low level.
Fig.10 Plinius SA-103, spectrum of 1kHz sinewave, DC10kHz, at 10W into 8 ohms with class-A/B bias (left cyan, right magenta) and class-A bias (left blue, right red). (Linear frequency scale.)
Finally, the decrease in linearity at high frequencies noted earlier results in a somewhat disappointing performance on the high-power, high-frequency intermodulation test (fig.11). While the 1kHz difference product into 4 ohms lies at a low 80dB (0.01%) in the left channel and 77dB (0.014%) in the right, the higher-order products at 18 and 21kHz are almost 10x higher in level. This graph was taken in class-A mode; class-AB actually gave slightly lower levels of the higher-order products (fig.12).
Fig.11 Plinius SA-103, class-A bias, HF intermodulation spectrum, DC24kHz, 19+20kHz at 127W peak into 4 ohms (linear frequency scale).
Fig.12 Plinius SA-103, class-A/B bias, HF intermodulation spectrum, DC24kHz, 19+20kHz at 127W peak into 4 ohms (linear frequency scale).
The Plinius SA-103's measured behavior clearly shows that class-A operation is to be preferred and I am not surprised EL liked the sound of the amplifier in this mode. Though class-A/B operation is definitely more environmentally friendly, there appears to be insufficient output-stage bias current for optimal operation.John Atkinson