Sidebar 3: Measurements
For logistical reasons I performed a full set of measurements on a different sample of the Devialet Astra than that reviewed by Ola Björling. After I hard-wired the Astra to my network and installed the Devialet app on my iPad mini, the app informed me that the Astra had the serial number U07A00001PH09 and was running firmware version 3.2.9. The Astra is Roon Ready; the Roon app recognized the Astra, and I selected it as a destination for network audio data.
I used my Audio Precision SYS2722 system for the measurements, repeating some of the testing with the magazine's higher-resolution APx555 analyzer. As the Devialet uses a class-D output stage that emits ultrasonic noise, which would drive my analyzer's input into slew-rate limiting, I inserted an Audio Precision AUX-0025 passive low-pass filter between the test load and the analyzer. This filter mitigates noise above 80kHz and eliminates noise above 200kHz. Without the filter, there was just 13mV of ultrasonic noise with a center frequency of 320kHz present in the amplifier's speaker outputs. I used this filter for all the power amplifier tests other than the frequency response. I preconditioned the amplifier by running it at a moderate power into 8 ohms for 30 minutes before starting the testing. At the end of that time, the temperature of the top of the chassis was 94.6°F/34.8°C.
Looking first at the Devialet Astra's single-ended line-level inputs, the amplifier preserved absolute polarity, ie, was noninverting, at the loudspeaker outputs. The unbalanced line input impedance was a usefully high 44.8k ohms at 20Hz and 1kHz, dropping to 39k ohms at the top of the audioband. The volume control operated in accurate 1dB steps. With the volume control set to the maximum of "+12dB," the voltage gain at 1kHz was 36.1dB into 8 ohms.
I used my Audio Precision SYS2722 system for the measurements, repeating some of the testing with the magazine's higher-resolution APx555 analyzer. As the Devialet uses a class-D output stage that emits ultrasonic noise, which would drive my analyzer's input into slew-rate limiting, I inserted an Audio Precision AUX-0025 passive low-pass filter between the test load and the analyzer. This filter mitigates noise above 80kHz and eliminates noise above 200kHz. Without the filter, there was just 13mV of ultrasonic noise with a center frequency of 320kHz present in the amplifier's speaker outputs. I used this filter for all the power amplifier tests other than the frequency response. I preconditioned the amplifier by running it at a moderate power into 8 ohms for 30 minutes before starting the testing. At the end of that time, the temperature of the top of the chassis was 94.6°F/34.8°C.
Looking first at the Devialet Astra's single-ended line-level inputs, the amplifier preserved absolute polarity, ie, was noninverting, at the loudspeaker outputs. The unbalanced line input impedance was a usefully high 44.8k ohms at 20Hz and 1kHz, dropping to 39k ohms at the top of the audioband. The volume control operated in accurate 1dB steps. With the volume control set to the maximum of "+12dB," the voltage gain at 1kHz was 36.1dB into 8 ohms.
Fig.1 Devialet Astra, line input, 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) (1dB/vertical div.).
Fig.2 Devialet Astra, line input, small-signal, 10kHz squarewave into 8 ohms.
The loudspeaker output impedance was extremely low, at 0.003 ohm from 20Hz to 20kHz. The modulation of the Devialet Astra's frequency response due to the Ohm's law interaction between this impedance and the impedance of our standard simulated loudspeaker was therefore negligible (fig.1, gray trace). The amplifier's response into resistive loads was flat in the audioband and –4dB at 40kHz into 8, 4, and 2 ohms (fig.1, blue, red, cyan, magenta, and green traces). Both the very close channel balance and the overall response were preserved at lower settings of the volume control. There was a single cycle of overshoot and ringing on the Devialet Astra's reproduction of a 10kHz squarewave into 8 ohms (fig.2).
Fig.3 Devialet Astra, line input, frequency response into 8 ohms with the treble and bass controls set to the maximum and minimum (left channel blue, right red) and with the "Bass Reducer" control on (left green, right gray) (2.5dB/vertical div.).
With the Tone controls activated, the treble and bass controls offered a maximum boost and cut of 12dB above 7kHz and below 100Hz (fig.3). The traces in this graph rapidly roll off above 40kHz, which suggests that the analog input is digitized with a sample rate of 96kHz. With the "Bass Reducer" control activated (fig.3, green and gray traces), the output was down by 7dB at 10Hz.
Fig.4 Devialet Astra, line input, spectrum of 1kHz sinewave, DC–1kHz, at 1Wpc into 8 ohms with volume control set to the maximum (left channel green, right gray) and to –12dB (left blue, right red) (linear frequency scale).
Channel separation was a good >80dB below 2kHz, reducing to 57dB at the top of the audioband. With the volume control set to its maximum, the single-ended line inputs shorted to ground, and the Audio Precision ultrasonic filter, the wideband, unweighted signal/noise ratio at the loudspeaker outputs was a very good 78.6dB in both channels, ref. 2.83V, which is equivalent to 1W into 8 ohms. This ratio improved to 81.1dB when the measurement bandwidth was restricted to the audioband, and to 83.5dB when A-weighted. The random noise level in the Devialet Astra's low-frequency noisefloor at 1Wpc into 8 ohms with the volume control set to "+12dB" (fig.4, green and gray traces) dropped by around 10dB when I set the volume control to "+0.0dB" and increased the level of the input signal so that the output was the same 1Wpc into 8 ohms (blue, red traces).
Fig.5 Devialet Astra, line input, THD+N (%) vs 1kHz continuous output power into 8 ohms.
Fig.6 Devialet Astra, line input, THD+N (%) vs 1kHz continuous output power into 4 ohms.
Devialet specifies the Astra's maximum power as 150Wpc into 8 ohms and 300Wpc into 4 ohms, both equivalent to 21.76dBW. Figs.5 and 6 respectively plot how the THD+noise percentage in the Devialet Astra's output varied with power into 8 ohms and 4 ohms with both channels driven. Stereophile defines clipping as when the THD+N reaches 1%. The Devialet Astra slightly exceeded its specified powers, clipping at 165Wpc into 8 ohms and at 325Wpc into 4 ohms (both equivalent to 22.1dBW). The FTC's updated "Amplifier Rule" states that maximum power should also be assessed at frequencies other than 1kHz. I therefore repeated the power test with a 20kHz signal. The amplifier clipped at 160Wpc into 8 ohms at this frequency (22.04dBW).
Fig.7 Devialet Astra, line input, THD+N (%) vs frequency at 25V into 8 ohms (left channel blue, right red) and 4 ohms (left green, right gray).
The downward slope of the traces in figs.5 and 6 indicate that actual distortion lies beneath the noisefloor at powers below the onset of clipping. Fig.7 shows how the Devialet Astra's THD+N percentage changed with frequency at 25V, which is equivalent to 78W into 8 ohms (blue, red traces) and 156W into 4 ohms (green, gray traces). (The amplifier muted its output after a minute or so at higher voltages into 4 ohms with this test.) The distortion+noise at low and medium frequencies into both loads is extremely low, with a slight rise in the top audio octaves.
Fig.8 Devialet Astra, line input, 1kHz waveform at 100W into 8 ohms, 0.00097% THD+N (blue); distortion and noise waveform with fundamental notched out (red, not to scale).
Fig.9 Devialet Astra, line input, spectrum of 1kHz sinewave, DC–10kHz, at 100Wpc into 8 ohms (left channel blue, right red, linear frequency scale).
Fig.10 Devialet Astra, line input, HF intermodulation spectrum, DC–30kHz, 19+20kHz at 200Wpc peak into 4 ohms (left channel blue, right red, linear frequency scale).
Because the distortion is so low, all that can be seen in the distortion+noise waveform at 100W into 8 ohms is random noise, though there is a hint of low-order harmonics (fig.8, red trace). Spectral analysis at 100W into 8 ohms (fig.9) reveals that while the second and third harmonics are the highest in level, these lay close to –110dB (0.0003%). Even at the same peak voltage into 4 ohms, intermodulation distortion with an equal mix of 19kHz and 20kHz tones was extremely low in level (fig.10).—John Atkinson
