Wavelength Audio Proton USB D/A converter Measurements
Sidebar 3: Measurements
I used Stereophile's loan sample of the top-of-the-line Audio Precision SYS2722 system to perform the measurements on the Wavelength Proton (see www.ap.com and the January 2008 "As We See It"); for some tests, I also used my vintage Audio Precision System One Dual Domain and the Miller Audio Research Jitter Analyzer. Source for the test files was my MacBook, fitted with 4GB of RAM and running Mac OS10.6.8 and Pure Music V1.8.
Wavelength warns that the Proton's line-level output will clip with full-scale signals, and advises owners to set the Proton's internal volume control to 90% in AudioMidi Set-Up. (The remaining 10% of the volume control's range of adjustment is to allow a full-scale signal to be output from the Proton's headphone jack.) I made sure the Proton's internal battery was fully charged before I performed any tests.
The Mac's USB Prober utility reported that the product was the "Proton USBDAC" from "Wavelength Audio, ltd.," and that it operated with 24-bit resolution in "Isochronous asynchronous" mode. Sample rates handled were 44.1, 48, 88.2, and 96kHz. With the volume control at 90%, the maximum output at 1kHz from both sets of outputs was a lowish 899mV, and both outputs preserved absolute polarity (ie, were non-inverting). The headphone output still clipped with the Proton's volume control set to 100%, with an output level of 1.4V. Backing off the volume control to 94% eliminated the clipping, at which point the output level was 1.13V.
The line-level jacks featured a low output impedance of 33 ohms at high and middle frequencies, but this rose to 1504 ohms at 20Hz, presumably due to the presence of an output coupling capacitor. The Proton should be used with a preamp having an input impedance of at least 10k ohms if the bass is not to sound a little lean. The impedance from the headphone jack was an appropriately minuscule 1.5 ohms at high and middle frequencies, rising slightly to 4.6 ohms at 20Hz.
The Proton operated correctly with data sampled at rates from 44.1 to 96kHz, the appropriate LED on the rear panel illuminating for each rate. The frequency response into 100k ohms with 44.1kHz data (fig.1, blue and red traces) was flat almost to the top of the audioband but then began to roll off, reaching 0.9dB at 20kHz. With 96kHz data (fig.1, cyan and magenta traces), the output was 0.5dB at 30kHz and 3dB just above 40kHz. Channel separation (not shown) was superb, at 110dB in both directions at 1kHz, and still 89dB (LR) and 99dB (RL) at 20kHz. The Proton's impulse response (fig.2) indicated that it uses a conventional linear-phase FIR filter.
Linearity error with 16-bit data was negligible down to below 105dB (fig.3), the traces in fig.4 representing a dithered 1kHz tone at 90dBFS peaking at that level and being commendably free from either power-supplyrelated or harmonic spuriae. However, both this graph and fig.5, which was derived using an FFT technique, also show that the increase in bit depth gives only a slight increase in dynamic range. The use of a battery for the Proton's power supply does reduce the maximum output level at the expense of dynamic range. Consequently, the waveform of an undithered waveform at 90.31dBFS (fig.6) is overlaid with analog noise, obscuring the three DC voltage levels described by the data.
With the Proton's volume control set to 90%, a full-scale 1kHz signal gave a regular series of distortion harmonics from the Proton's line output (fig.7), though these are all at or below 93dB (0.0028%). Commendably, the level of these harmonics didn't increase significantly when the load impedance was dropped to just 600 ohms (not shown). With a 1kHz signal at 10dBFS (fig.8), all the higher-order harmonics disappeared, leaving the second and third harmonics below 114dB (0.0002%). Intermodulation distortion was also low, with all products at or below 90dB (fig.9). But this graph also shows that the Proton uses a fairly slow reconstruction filter, with the aliasing products at 24.1 and 25.1kHz easily visible.
The Miller-Dunn J-Test signal is not really diagnostic for a USB datalink, where the clock is not embedded in the data. However, fig.10 shows the spectrum of the Proton's output while it decoded 16- and 24-bit versions of the J-Test signal. The odd-order harmonics of the low-frequency squarewave almost all disappear when the 16-bit data (cyan and magenta traces) are replaced by 24-bit data (blue and red traces). However, two pairs of data-related sidebands are emphasized and don't disappear: those at ±229 and ±689Hz. The measured level of these sidebands, according to the Miller Analyzer, was equivalent to just 130 picoseconds peakpeak of jitter, which is low enough to have no audible consequences. But these sidebands shouldn't have been present at all, given the Proton's asynchronous USB connection. Perhaps they stem from some kind of interaction within the circuit akin to Meitner and Gendron's "Logic-Induced Modulation" . . . ?
Overall, the Wavelength Proton performed well on the test bench. Its only obvious fault was its limited ultimate dynamic range, due to the use of a battery supply with limited voltage capability.John Atkinson