|
Removing the top cover (which is very heavy and dead-sounding, thanks to a lining of vibration-damping
material) reveals four circuit boards that occupy virtually all of the Processor 3's interior. One,
behind the front panel, holds the panel controls and display. Two boards are at the rear of the
chassis, one for the digital inputs and input receiver and the other for the tube output stage. The
fourth and largest board holds the DAC modules and their power-supply regulators, the new I-V
converter modules, and the passive LC low-pass output filters. The power supply's interior is
occupied by one circuit board that carries all of the rectifiers, filter capacitors, primary voltage
regulators, and ancillary parts. Parts and build quality are of a high order in both chassis.
Measurements:
The frequency response curves in Fig. 1 were taken at the Processor 3's balanced outputs. Switching
from instrument to IHF loading with unbalanced output dropped the level only about half as much,
because the unbalanced outputs' impedance is half that of the balanced outputs. De-emphasis error was
essentially zero. On 1kHz square waves, ringing was symmetrical and the peaks were unclipped. The
symmetry is characteristic of finite-impulse-response (FIR) digital filters; the absence of peak
clipping is characteristic of the Pacific Microsonics PMD-100 digital filter/HDCD decoding chip used
in the Processor 3. As can be seen in Fig. 2, total harmonic distortion plus noise (THD + N) at I kHz,
distortion is higher for unbalanced output than it is for balanced output and with IHF rather than
instrument loading. Even at its worst, however, THD + N won't be very noticeable: Its maximum level,
-70 DBFS, amounts to a mere 0.03%. And that maximum is attained only near digital full scale
(0 DBFS), well above typical music levels. (The distortion rise occurred in the tube output stage,
as I determined by tracing it back through the circuit.) Furthermore, as Fig. 3 reveals, distortion
does not rise much with frequency and doesn't even begin to climb until about 2 kHz.
Deviation from linearity for a 500-Hz input signal (Fig. 4) is very good; linearity is nearly perfect
down to 105 DBFS. This was borne out by a noise-modulation test, devised by Richard Cabot of Audio
Precision, ill which noise spectra are analyzed for a 41-hz signal at five levels from -60 to -100
DBFS. The Processor 3's noise spectra (not shown) were almost identical at every level.
Interchannel crosstalk was outstandingly low: -120 dB or better from 20 Hz to 15 kHz, increasing to
about - 1 05 dB at 20 kHz.
S/N ratios are listed in Table 1. The excellent wideband results testi@, to the effectiveness of the
LC filters that follow the Processor 3's DACS. Because noise common to both phases of the balanced
outputs cancels out, the noise at those outputs is noticeably lower. This can be seen in Fig. 5, an
analysis of the noise for a 1 -kHz signal at -80 DBFS. Whether I used balanced or unbalanced output,
quantization noise was no more than -93.4 DBFS, and dynamic range was 96 dB or better.
For all the tests that I've described so far, I fed signals to the Sonic Frontiers Processor 3 via an
AES/EBU balanced interconnect. 1 found that there were no significant differences when I repeated
several tests via the 12S-e bus connection, although some noteworthy differences emerged when I
compared jitter performance.
To determine the I2S-e link's effect on jitter, I first obtained a Sonic Frontiers SFT- I CD transport
that had been modified to include an I2S-e output. I put the SFT-1 in pause mode and connected it to
the Processor 3's AES/EBU and I2S-e inputs. Next, I measured jitter at each DAC module for each type
of connection. My test points were the modules' deglitch inputs, which work with the digital filters'
deglitch outputs to prevent switching noise during transitions from one bit to the next.
As Fig. 6 reveals, switching from the Processor 3's AES/EBU to its I2S-e connection considerably
reduces random jitter over most of the audio range, even though the 12S-e curve has slightly more
prominent peaks at 120 and 240 Hz (related to the AC line frequency) and slightly higher jitter above
10 kHz. (With AES/EBU connections, the digital input receiver's phase-locked loop, whose cutoff
frequency occurs at about I kHz, causes a high-frequency rolloff in the jitter spectrum.) Note that
the jitter levels in Fig. 6 are peak to peak; the equivalent sine-wave jitter would be 3.54
nanoseconds rms for 10 nanoseconds peak to peak, 1. 1 2 nanoseconds rms for 3 (actually, 3.16)
nanoseconds peak to peak, and so on. When I played a jitter-prone -90 DBFS signal on the CBS CD- I
test disc, the Sonic Frontiers' jitter was again lower in IIS-E mode. The reduction I recorded was
about 26 dB for the 1 -kHz test signal and 14 dB for its second harmonic; that's impressive.
Use and Listening Tests
When I first received the Sonic Frontiers Processor 3, 1 hooked it up in my normal CD setup. Right
from the beginning, I was very impressed with the sound and thought it was more lifelike than that
of the Class6 Audio DAC-1, one of my longtime favorite D/A converters. The Processor 3's definition,
clarity, and spaciousness all struck me as superior.
Next, I replaced my PS Audio CD transport with the modified Sonic Frontiers SFT- I so that I could
hear how the Processor 3's AES/EBU and I2S-e connections compared. The I2S-e connection did improve
transparency, detail, and spaciousness, but the differences weren't large. However, this precluded my
using the DGX Audio digital preamp (which provides EQ for my B&W speakers) or any other ancillary
digital components that have only the common types of digital connections.
The Sonic Frontiers Processor 3 performed flawlessly in my lab and in my listening room. Not only was
its measured performance superb, it also gave me the best digital sound I've heard in my home system.
Needless to say, I very much enjoyed its stay in that system, and I enthusiastically recommend it. A
|
