MSB Link DAC III D/A converter Measurements
For most of these measurements I upsampled the 44.1kHz data to 96kHz, which was how KR preferred to audition the MSB Link III. When I stuck with 44.1kHz data, I say so in the text.
Although KR had problems getting the Link III to lock on to AES/EBU datastreams, I had no problems with the Audio Precision's XLR digital output. The Link III also had no problems locking on to the 96kHz-sampled S/PDIF data output of a Panasonic A-120 DVD player playing DADs or the data output of a RadioShack 3400 portable CD player.
As seems always to be the case these days—presumably because no one wants to be the quietest player in the store—the Link III's maximum output level of 2.275V was higher than the CD standard's 2V RMS. This was sourced from a low 51 ohms across the audioband, and the unit didn't invert signal polarity.
The Link III's frequency response with 44.1kHz sampling, with and without pre-/de-emphasis, is shown in fig.1. With a pre-emphasized signal, there is a tiny presence-region boost apparent and slightly more rolloff at 20kHz, but this is inconsequential. What is more significant is the rise in the very low bass. Will this be audible? Hard to say, as human hearing is insensitive in this region. However, the "contour lines" in the human sensitivity curve are very close together in this region; this slight boost might be heard on appropriate program material.
Fig.1 MSB Link III, CD frequency response with (top) and without (bottom) de-emphasis at -12dBFS, 44.1kHz sampling (right channel dashed, 0.5dB/vertical div.).
Upsampling to 96kHz didn't change the response—although, of course, neither did it add any information above the CD's Nyquist frequency of 22.05kHz. Channel separation (not shown) was excellent at better than 115dB below 2kHz, and with a slight rise above that frequency.
Fig.2 shows a 1/3-octave spectral analysis of the MSB's analog output while decoding data representing a dithered 1kHz tone at -90dBFS. The peak at 1kHz just kisses the -90dB mark, implying low linearity error. Increasing the data word length from 16 bits (top traces) to 24 bits (bottom traces) gave about an 8dB lowering of the noise floor above 1kHz, less below 1kHz, and no change below 70Hz or so. The Link III accepts 24-bit data, but its dynamic range is ultimately limited by the output stage's analog noise compared with state-of-the-art (and much more expensive) digital processors.
Fig.2 MSB Link III, 1/3-octave spectrum of dithered 1kHz tone at -90dBFS, with noise and spuriae, 16-bit (top) and 24-bit (bottom) data, 44.1kHz sampling (right channel dashed).
Extending the measurement bandwidth to 200kHz and changing the data to -1LSB DC gave the graph shown in fig.3. The MSB's D/A chips use noiseshaping to achieve their resolution, as revealed by the rapid rise in ultrasonic noise. The left-hand trace was taken at 44.1kHz sampling, and the noise floor reaches -50dB at 180kHz. But with the unit set to upsample to 96kHz, the ultrasonic rise in the noise floor is pushed an octave higher, which might have positive subjective consequences with some amplifiers.
Fig.3 MSB Link III, 1/3-octave spectrum of -1LSB, with noise and spuriae, 16-bit data and 44.1kHz sampling (top), 96kHz sampling (bottom) (right channel dashed).