M2Tech Young D/A processor & Palmer Power Station battery power supply Measurements
Sidebar 3: Measurements
For the testing, I powered the M2Tech Young from the Palmer Power Station, which had been fully charged. I repeated some of the measurements powering the Young with its wall-wart supply. The testing was primarily performed with Stereophile's loan sample of the top-of-the-line Audio Precision SYS2722 system (see www.ap.com and the January 2008 "As We See It"), feeding data to the Young via an AES/EBU link. To test its behavior as a USB DAC, I installed the driver program on my MacBook Pro; the computer identified the Young as "M2TECH 384/32 async USB," with a vendor ID string of "Cypress Semiconductor" and the interface defined as "vendor specific." The USB input operated correctly with data sampled at all rates from 44.1 to 384kHz, including 88.2kHz and 176.4kHz, and with a bit depth identified as "2ch-32 bit integer." The AES/EBU and S/PDIF inputs locked to data sampled at all rates from 44.1 to 192kHz.
As Jon Iverson found, the Young's maximum output at 1kHz was higher than is usual, at 2.62V, which is 2.35dB higher than the CD Standard's 2V. The output impedance was very low, at <1 ohm, and the output preserved absolute polarity (ie, was non-inverting).
With 44.1kHz data, the Young's impulse response (fig.1) suggested that the M2Tech's digital reconstruction filter is a conventional linear-phase type rather than the minimum-phase type claimed in the specification, but with fewer than usual of the coefficients that are "mapped" by the ringing before and after the pulse. That this filter has an unusually slow rolloff is shown by the blue trace in fig.2, which was taken with white noise sampled at 44.1kHz. This indicates that the filter is "leaky"; that is, it does not completely reject ultrasonic image energy, as would a conventional reconstruction filter. This is demonstrated by the red trace in fig.2, which is the spectrum of the Young's output as it decoded data representing a full-scale tone at 19.1kHz. The image of the tone at 25kHz (44.1 minus 19.1) is suppressed by just 18dB, and while the distortion harmonics of the tone are low in level, a slew of aliasing and image products can be seen. This looks terrible, but, as I explained in my 2011 Richard Heyser Memorial Lecture, listeners seem to prefer this kind of behavior to a conventional fast-rolloff reconstruction filter.
Fig.3 examines the Young's frequency response with data sampled at 44.1kHz (green and gray traces), 96kHz (cyan, magenta), and 192kHz (blue, red). At the two higher sample rates, a smooth rolloff above the audioband is interrupted by a fast rolloff just below the Nyquist frequency (half the sample rate), though in both cases the rolloff is a little premature; the rolloff with 96kHz data, for example, lies at 3dB at 30kHz. By contrast, the response at 44.1kHz suffers from ripple in the passband. Although never greater here than ±0.1dB, such ripple is generally believed to negatively affect sound quality. The responses in fig.3 were taken into a high (100k ohms) load, and didn't change into a very low (600 ohms) load. Channel separation (not shown) was superb in the midrange, at 120dB or greater, but decreased at the frequency extremes to a still-superb 110dB.
With the Palmer battery supply, the Young's noise floor was clean and very low in level. Changing to the wall wart introduced a series of spectral components at 60Hz and its odd harmonics (fig.4), but at 110dB and below (<0.0003%), these won't be audible as hum. I'm not sure, therefore, why JI found the sound with the Palmer so much better. For consistency with my tests of digital products going back more than three decades, my first test of a DAC's resolution is to feed it dithered data representing a 1kHz tone at 90dBFS with 16- and 24-bit word lengths and sweeping a 1/3-octave bandpass filter from 20kHz down to 20Hz. The result is shown in fig.5, with the 16-bit spectrum the top pair of traces and the 24-bit spectrum the middle pair. The increase in bit depth drops the noise floor by 21dB in the treble, suggesting that the Young has a resolution of between 19 and 20 bitswhich is excellent, especially considering the M2Tech's affordable price. This resolution is easily enough to allow the Young to resolve a 24-bit, 1kHz tone at 120dBFS (fig.5, bottom pair of traces).
Repeating the analysis with a modern FFT technique (fig.6) suggested that the resolution in the treble is closer to 20 bits; no distortion harmonics are evident in this graph. These spectra were taken with the Palmer battery supply; changing to the wall wart gave fundamentally the same picture, but with the 60Hz components visible at and below 110dB (not shown). Repeating the 24-bit spectrum with the USB input confirmed that the Young correctly handles 24-bit data via USB. With its very low level of noise and high resolution, the Young correctly handled undithered 16-bit data representing a tone at exactly 90.31dBFS (fig.7), with a symmetrical waveform and the three DC voltage levels described by the data clearly resolved. With undithered 24-bit data, the result, even at this very low signal level, was a clean sinewave (fig.8).
Whether driven by the Palmer Power Station or the wall wart, the M2Tech Young offered very low levels of static distortion, even into 600 ohms (fig.9). The second harmonic is the highest in the left channel (blue trace), the third in the right channel (red), but both are vanishingly low in absolute terms. Although with 44.1kHz data the Young apparently performed poorly when handling an equal high-level mix of 19 and 20kHz tones, the messy spectrum in fig.10 is not actually the result of intermodulation distortion but of aliasing and image productsmore consequences of the Young's leaky reconstruction filter.
Looking at the Young's rejection of word-clock jitter via its AES/EBU input, the 16-bit version of the Miller-Dunn J-Test data was handled correctly, though with some slight accentuation of the sidebands at 11.025kHz, ±229.6875Hz, and the appearance of a pair of low-level sidebands of unknown origin at ±358Hz (fig.11). With 24-bit J-Test data, there was still a pair of sidebands present at ±229.6875Hz, which is very unusual, as well as sidebands at ±358 and ±690Hz (not shown). The J-Test signal is not diagnostic for transmission of audio data via USB, as the behavior of the interface will not be affected by the values of the data being transmitted. Nevertheless, with 24-bit J-Test data transmitted to the Young via USB (fig.12), the sidebands at 11.025kHz, ±229.6875Hz, are still present, as are pairs at ±690 and ±1380Hz. Their presence in this graph suggests that the data-related sideband pair might arise from the Logic-Induced Modulation first described two decades ago by Meitner and Gendron. However, it is fair to point out that all of these spuriae are very low in level, and that the spike that represents the high-level tone at 1/4 the sample rate is well defined, with no spectral spreading at its base.
Overall, the M2Tech measured very well, even when driven by its wall-wart supply rather than the Palmer Power Station. The main exception is that leaky reconstruction filter, which, with 44.1kHz data, allows aliasing products to be reflected into the audioband. But as much as it horrifies my inner engineer, that is something that, for some reason, listeners tend to prefer.John Atkinson