Making digital audio sound good appears to be a much more difficult job than its developers first realized. When digital audio was in its infancy, there was a tendency to think that digital either worked perfectly, or didn’t work at all. This belief led the engineering community to devise ill-considered and flawed standards that affect the musical quality of digitally reproduced music today.
The S/PDIF and AES/EBU interfaces used to connect a CD transport to a digital processor are a perfect example. When the digital interface was standardized, no one considered that its design could affect sound quality. If the interface could carry the binary ones and zeros without losing data, what else mattered?
It turns out that the timing of those ones and zeros is critical to a digital system’s sound quality. Unfortunately, the very design of the S/PDIF and AES/EBU interfaces creates such timing errors, called “jitter,” in digital audio. The problem with the interface is that it must carry the left and right audio data and the timing clock in the same signal. The digital processor must “lock” to this clock, which is buried within the digital audio data. Timing errors in the interface are thus passed to the digital processor, where the errors affect the accuracy of the D/A conversion process. This is why transports and digital cables can sound different, even though the binary ones and zeros representing the music are unchanged.
High-end audio designers have resorted to a number of techniques to overcome these inherent limitations in the digital interface. Examples of new devices and methods of reducing interface jitter include separate jitter-reduction boxes between a transport and processor, elaborate reclocking circuits in digital processors, low-jitter input receivers, CD players (with no interface, they don’t suffer from interface jitter), connecting a transport and processor via the Philips I2S bus (Audio Alchemy’s approach), and running a separate clock line between a transport and processor. All these devices and techniques exist for one purpose: to reduce the jitter introduced by the AES/EBU or S/PDIF interface and thus improve sound quality (footnote 1).
Which brings me to the new Sumo Axiom CD transport and Theorem II D/A processor reviewed here. The Axiom and Theorem can be connected conventionally to any other transport and/or processor, or will hook up together with a separate clock cable to reduce interface-induced jitter.
The Theorem II looks very different from the original Theorem I reviewed in October 1992 (Vol.15 No.10). The first Theorem’s slim chassis and thin front panel were scrapped in favor of a 2.5″-tall chassis with a full ¼” aluminum front panel. In addition, the removable rack mounts have been replaced by integral rack-mount holes in the front panel. The Theorem’s buttonless front panel has been revamped on the Theorem II to include a polarity-inversion button, power-standby switch, and a pair of input selector buttons.
The rear panel has also changed, with the addition of XLR analog outputssingle-ended as standard, balanced for an additional $200an extra pair of single-ended outputs on RCA jacks, two digital outputs on RCA jacks, and an AES/EBU digital input. These additional facilities are a significant departure from the original Theorem’s simplicity.
As with the Theorem, the Theorem II has a rear-panel clock input for locking the processor to a Sumo transport with a separate clock cable. When I first reviewed the Theorem, Sumo hadn’t yet brought their matching transport to market, which prevented me from reporting on this feature’s potential sonic advantages. Now that the Axiom transport, which has a compatible clock connection, is available, I can fully assess this transport/processor combination. Because no standards exist for these separate clock lines, products made by one manufacturer won’t lock through the separate clock link to products made by another manufacturer. (Any transport will, however, lock to any processor through the standard single-cable connection method.)
The Axiom offers all the standard CD-transport features in an attractive, cost-effective package. The remote control, Philips’s newest unit, has a curved, modern-looking profile.
The remote’s most interesting feature is the volume up and down buttons. Volume control on a CD transport? That’s right. The Axiom has a digital-domain attenuator at the output that lets you connect a digital processor directly to a power amplifier without the need for a preamp (provided you have no other source components). Alternately, the Axiom’s remote volume control allows you to fine-tune the volume when listening to CDs, even if your preamplifier lacks remote control.
The digital attenuation within the Axiom is performed by a new chip made by Philips, called the SAA7345. This chip’s main function is to encode the raw digital audio data into the S/PDIF or AES/EBU transmission format; digital attenuation is a sideline.
The SAA7345 is programmed in the Axiom so that it still provides 0.5dB of attenuation when the volume control is all the way up. Designer Michael Custer feels that full-scale digital signals (and even those near full-scale) can cause a processor’s digital filter to clip. By attenuating the signal by half a dB before it gets to the filter, the Axiom may have an audible performance advantage over transports without this feature. Note that the Pacific Microsonics PMD100 HDCD$r decoder/filter attenuates all signals by 1dB for the same reason.
The downside to the Axiom’s digital-domain attenuation is that digital attenuators throw away resolution and degrade the sound. Unless the overall signal path is of greater-than-16-bit capability, the greater the attenuation, the greater the reduction in resolution. For example, 6dB of attenuation is like reducing the resolution by one bit. Twelve dB of attenuation is equivalent to throwing away 2 bits of resolution. Not even very-high-quality digital attenuatorssuch as that used in the PS Audio Reference Linkare sonically transparent. You should thus set your preamplifier’s volume to the appropriate position, using the Axiom’s volume control only for small adjustments.
The Axiom is based on the Philips CDM-12 transport, a mechanism found in some transports costing $1500 (although with some modifications and addition of metal parts). The main decoder and servo board appear to be standard-issue, with only the output board designed by Sumo. This output board is, however, impressive. For example, each phase of the balanced AES/EBU signal is reclocked, then buffered by a discrete line amplifier. All this circuitry, along with the clock output driver, is located right next to the RCA and XLR output jacks. The Axiom has unusual digital output levels: 1V from the coaxial output, 2V from the AES/EBU output. The official S/PDIF and AES/EBU specifications call for an output voltage of 500mV (coaxial) and 5V (AES/EBU).
Build quality is good for $899: the chassis is made from 18-gauge steel, and the unit has a ¼” aluminum front panel. The cost-cutting measures include using stock decoder/servo and display boards, and mounting the digital outputs directly on the pcb rather than attaching them to the chassis. This technique avoids the labor of point-to-point wiring, but results in a less mechanically secure mounting. Many moderately priced products have jacks installed with this method.
The Theorem II has several important refinements over the original Theorem. In addition to the functional differences described earlier, the Theorem II now features AES/EBU input as standard. On the inside, the original Theorem’s Burr-Brown PCM67 DAC has been changed to the PCM69. Both these DACs are hybrid types, combining a multibit ladder converter with a 1-bit converter. Specifically, the upper 10 bits are converted to an analog output with a multibit resistor ladder converter, and the lower eight bits (the eight LSBs) are converted to analog by a 1-bit noise-shaping converter. The idea is to get the best of both worlds: the dynamics of multibit, and the intrinsic low-level linearity of 1-bit.
Looking at the data sheets for the PCM67 and PCM69 suggested that the only difference between the DACs is the PCM69’s ability to accept a wider range of clock frequencies. Other than that, the PCM69 appears to be identical in technical performance to the PCM67. Sumo reports that the PCM69 has a smoother sound. Only the highest “K” grade is used in the Theorem II. Note that very few manufacturers use this unusual, low-cost DAC in high-end processors.
The power supply has been refined, mostly with capacitor changes. My original review criticized the Theorem’s lack of bass extension, weight, and dynamics. These faults were reportedly corrected by these power-supply improvements. The power supply was changed after my first review but before the Theorem II was introduced, so later-production Theorems have the updated supply. Both processors feature 10 separately regulated supplies.
Footnote 1: For the technical details of why the interface induces jitter, I recommend Malcolm Omar Hawksford’s paper “Is the AES/EBU Interface Flawed?,” presented at the 1992 Audio Engineering Society Convention in San Francisco, and Rémy Fourré’s “Jitter and the Digital Interface,” published in the October 1993 Stereophile (Vol.16 No.10).Robert Harley