Digital Processor Reviews

Sidebar 3: Measurements

After I connected the Eversolo DMP-A6 Master Edition Gen 2's Ethernet port to my router and installed the Eversolo Control app on my iPad mini, the app informed me that the processor was running software version v1.5.75. Roon recognized the Eversolo as an MQA decoder and renderer that would handle 32-bit network PCM data with sample rates up to 384kHz. I connected my MacBook Pro to the Eversolo's USB port; Apple's USB Prober app identified the processor as "Eversolo DMP-A6" from "MQA" and confirmed that the USB port operated in the optimal isochronous asynchronous mode. Apple's AudioMIDI utility revealed that this port accepted integer data with a bit depth up to 32, sampled at all rates from 44.1kHz to 768kHz.

I tested the Eversolo DMP-A6 Gen2 with my Audio Precision SYS2722 system, repeating some tests with the magazine's higher-resolution APx555 system. I performed the digital input testing using the TosLink S/PDIF input, which accepted data sampled at all rates up to 192kHz, as well as via the USB and network ports.

The Eversolo's analog outputs preserved absolute polarity (ie, were noninverting) from all of the DMP-A6's digital inputs. The balanced output impedance was 189 ohms and the single-ended output impedance was 51 ohms, both values consistent from 20Hz to 20kHz. With the volume control set to its maximum, the level with full-scale 1kHz data was 5.25V from the balanced outputs and 2.63V from the single-ended outputs. The volume control operated in accurate 1dB steps.

The DMP-A6 offers a choice of seven digital reconstruction filters, labeled "Fast Roll-Off, Linear Phase," "Slow Roll-Off, Linear Phase," "Fast Roll-Off, Minimum Phase," "Slow Roll-Off, Minimum Phase," "Apodizing, Fast Roll-Off, Linear Phase," "Hybrid, Fast Roll-Off, Minimum Phase," and "Brickwall."

I identify a processor's reconstruction filter by examining its impulse response, which I test by sending it 44.1kHz PCM data representing a single full-scale sample. Fig.1 shows the DMP-A6's impulse response with the Fast Linear filter; it indicates that the reconstruction filter is a long linear-phase type with equal amounts of ringing before and after the single sample at 0dBFS. The impulse responses of the Apodizing and Brickwall filters were identical to the Fast Linear's. The Slow Linear filter's impulse response (fig.2) was a significantly shorter linear-phase type.

The Fast Minimum filter's impulse response was typical of a long minimum-phase type (fig.3), with all the ringing following the single sample at 0dBFS. There was less ringing with the Slow Minimum filter (fig.4). The odd man out was the Hybrid filter, which had a small amount of ringing before the full-scale sample and a greater amount of ringing after it (fig.5).

With 44.1kHz-sampled white noise, the DMP-A6's ultrasonic response with the Fast Linear and Fast Minimum filters reached full stop-band suppression at 24kHz (fig.6, red and magenta traces). An aliased image at 25kHz of a 19.1kHz tone at 0dBFS (blue and cyan traces) is suppressed by 85dB, and the distortion harmonics of the 19.1kHz tone are all very low in level. As expected, the ultrasonic rolloff was slower with the Slow Linear and Slow Minimum filters (fig.7), with only mild suppression of the aliased tone at 25kHz. The Apodizing, Brickwall, and Hybrid filters all reached full stop-band suppression at exactly half the sample rate, 22.05kHz (fig.8, red and magenta traces).

Fig.9 shows the Fast Linear and Fast Minimum filters' frequency responses with TosLink data sampled at 44.1kHz, 96kHz, and 192kHz. The responses at all three sample rates are flat in the audioband with a fast rolloff just below half of each rate. The Apodizing and Brickwall filters' responses were very similar, though there was a small amount of passband ripple visible (not shown). The ultrasonic responses with the two Slow Roll-Off filters rolled off earlier (fig.10), with the output down by 3dB at 20kHz with data sampled at 44.1kHz (green, gray traces). The rolloffs with the Hybrid filter were fast (fig.11), though the 44.1kHz response was –9dB at 20kHz (green, gray traces).

Channel separation (not shown) was superb, at>120dB in both directions below 3kHz. It decreased to a still-excellent 113dB at the top of the audioband. The low-frequency noisefloor with a 1kHz tone at –12dBFS and the volume control set to the maximum (fig.12, green and gray traces) was very low in level and free from any power supply–related spuriae. Reducing the volume by 12dB and increasing the signal level by 12dB dropped the noisefloor by 3dB (blue, red traces).

Linearity error (fig.13) was 0.1dB down to –120dBFS, which implies superb measured resolution. This was confirmed by fig.14, which shows that an increase in bit depth from 16 to 24 with dithered data representing a 1kHz tone at –90dBFS dropped the DMP-A6's noisefloor by 30dB, which suggests that the measured resolution was a state-of-the-art 21 bits. When I played undithered 16-bit data representing a tone at exactly –90.31dBFS, the waveform was symmetrical, and the three DC voltage levels were well resolved (fig.15). Repeating the measurement with undithered 24-bit data gave a clean sinewave (fig.16).

The DMP-A6 produced very low levels of harmonic distortion. With a full-scale 1kHz tone into the high 200k ohm load, the third harmonic was the highest in level in the balanced left channel output at just –117dB (0.00014%, fig.17). When I reduced the load impedance to 600 ohms, the levels of the harmonics were even lower. Intermodulation distortion with an equal mix of 19 and 20kHz tones, each lying at –6dBFS, with the Fast Linear, Fast Minimum, Apodizing, Hybrid, and Brickwall filters was vanishingly low in level, even into 600 ohms (fig.18), with the aliased products at 24.1kHz and 25.1kHz lying at –90dB. The levels of the two aliased products were much higher with the Slow Roll-Off filters (fig.19), but intermodulation distortion was still superbly low in level.

Fig.20 shows the spectrum of the DMP-A6's output when it was fed high-level, optical 16-bit J-Test data sampled at 44.1kHz. The odd-order harmonics of the undithered low-frequency, LSB-level squarewave all lie at the correct levels, indicated by the sloping green line, and the central spike that represents the high-level tone at one-quarter the sample rate is very narrow. Repeating the analysis with 24-bit J-Test data resulted in a clean spectrum (fig.21).

The Eversolo DMP-A6 Master Edition Gen 2 did superbly well on the test bench.—John Atkinson