- Written by Stephen Hornbrook
- Published on 30 June 2014
The Sony UDA-1 USB DAC Amplifier On the Bench
Measurements Section by Dr. David A. Rich
No schematic was supplied so it was not possible to correlate the measurements to the parts used inside the unit. Measurements were made with Audio Precision APx 585 and APx 582 test instruments by Chris Heinonen. Use of these was generously provided by Audio Precision.
Digital measurements are made with an SPDIF input on a coax cable. Except as noted below the test signal was dithered with a 24bit depth and a sampling rate of 192k samples /sec. The analog output was taken from the fixed line output jacks. The volume control and associated preamp gain stage are not part of this measurement and this should yield better SNR and distortion.
The output level (1kHz) at digital full scale was 2.3VRMS which is slightly higher than the standard value of 2.0VRMS. Channel balance was 0.6dB. Signal to Noise ratio was disappointing at 100 dB (16.5 bits equivalent) relative to 2.0VRMS. For the SNR test was band-limited from 20Hz to 20kHz.
The frequency response is extended to 50kHz. The strange peak above 10kHz will not be audible but it is hard to understand at what point in the analog chain it is introduced. Crosstalk (graph not shown) was an excellent -100dB at 1kHz rising to -70dB at 20kHz
Full scale THD is low at 0.002% up to 500Hz but then starts to increase significantly to 0.05% at 20kHz.
The distortion components of the 1kHz full scale signal (8 dBV) are seen above. The 2nd harmonic is at -91dB (0.003%) relative to full scale. The 3rd harmonic is at -95dB relative to full scale (0.002%).
Note the strange spectral lines spaced 1kHz apart starting at 500Hz. At -125dB (0.00006%) relative to full scale they would not appear significant. Still the unexplained origin of the signals leads to the possibility that they would become significant using tests designed to look for non-harmonic behavior in an oversampling DAC (IC). More information can be found on this in my introductory article on DAC ICs.
The noise floor of the graph is -130dB. As explained in other reviews I have written this is not the SNR but is related to it. The noise floor of the spectral plot is dependent on the frequency span (bin width) used to produce each point on the X axis. So many points appear on the X axis it looks continuous.
Since the bin width is not standardized you cannot compare this noise floor between measurements made on different equipment even if the test instrument comes from the same company. For the Audio Precision the bin width is adjustable. Other options on the AP can also change the appearance of the noise floor in a spectral plot.
You can compare the plot with other measurement sets made by Chris Heinonen. The SNR value is independent of the test instrument used but you need to scale it to the referenced full scale level which for our reviews is 2VRMS. I am bring this up now since DAC boxes have become the most coming component for which detailed measurements are presented in reviews.
Power supply noise was low with the maximum level at -110dB at 60Hz (graph not shown).
The spectra of a mix of 19kHz and 20kHz -6dB down shows low levels of IM products but unexpected spectral lines, not predicted from the IM equation, are seen across the band at -115dB. These may be related to the spectral lines spaced 1kHz apart starting at 500Hz seen in the 1kHz spectral plot above.
A 1kHz signal reduced to -60dB full scale is relatively free of discrete spectral lines above the higher than average noise floor reflective of the 101dB SNR of this product. The strange spectral lines seen in the full scale plots above are not seen.
Moving to an investigation of the DACs linearity performance for low signal levels we see the performance of the DAC is below average in the level linearity plot above. The deviation is up 1dB for a signal level of -98dB and is up 3dB at -102dB. These numbers indicate the linearity is starting to degrade for signal slightly below 16 bits equivalent. A mid-grade DAC (IC) would have the curve shifted to the left by 3dB, and a top grade DAC would shift the curve by more than 10dB.
In the time domain, a 16 bit un-dithered signal at -90dB distinctly shows the 3 discrete levels of this signal. Some of the ringing is the impulse response of the digital reconstruction filter and some is it is noise.
The effect of the relatively low SNR of 101dB is more clearly seen in the time domain plot above which extends the bit depth of the signal shown in the last plot to 24bits.
Only one graph is needed here. The amp produces 0.02% distortion until the clipping point of only 12 Watts average into 8 ohms. The graph above is for a 1kHz signal. Increasing the signal to 10kHz produced a similar graph with little increase in THD before clipping (not shown).