CD Player/Transport Reviews

The current-to-voltage converter (I/V) is a discrete design first used in the Theorem, but with some modifications that reportedly result in better transient performance. This is followed by a fully discrete, class-A, direct-coupled analog output buffer that runs on ±35V rails (most output stages run on ±15V). The output low-pass filter is a newly designed third-order Bessel type. The I/V converter, low-pass filter, and output buffer are integrated into a single stage that uses very little feedback. Moreover, the output transistors are hefty TO-220 packages that can reportedly drive half an amp of current into low impedances.

Unusually, the Theorem II uses the Yamaha YM3623 input receiver, which is the device used in nearly every digital processor until the much better Crystal CS8412 input receiver took over the market. Where the Yamaha part had 3–5 nanoseconds (3000–5000 picoseconds) of jitter in its recovered clock (without any additional jitter-reduction circuitry), the Crystal has 200–400ps—a tenfold improvement in jitter performance over the Yamaha (footnote 2). Sumo reportedly kept the outmoded Yamaha part in the Theorem to maintain an upgrade path for Theorem owners (an original Theorem can be upgraded by Sumo to the Theorem II), and also to reduce the amount of support circuitry needed. Sumo claims that their "Crystal Crowbar" clock recovery circuit reduces clock jitter to less than 80ps, despite the Yamaha receiver's high intrinsic jitter.

This issue of recovered clock jitter is moot when the Theorem is driven by the Axiom transport with the separate clock line installed. Rather than try to recover the clock from the S/PDIF or AES/EBU datastream, the Theorem gets a pure, presumably low-jitter clock directly from the transport, bypassing the Yamaha's clock-recovery function. Most processor/transport combinations with a separate clock line send the clock from the processor to the transport, making the processor the master clock. Sumo takes the opposite approach, sending the clock from the transport to the processor—a theoretically less ideal method because the only place where the clock jitter really matters is at the DAC input pin.

A Burr-Brown DF1700 8x-oversampling digital filter—the same as the ubiquitous NPC SM5813—is used before the DAC. Sumo didn't think the Pacific Microsonics PMD100 HDCD decoder/filter offered a cost-effective improvement over the NPC filter (functionally equivalent to the Burr-Brown DF-1700), but notes that the Theorem's modular design allows future changes to the digital filter.

Although the Theorem II has XLR and RCA outputs as standard, the $899 version reviewed here simply connects pin 3 of the XLR connector to ground. This allows the Theorem II to be connected to preamplifiers with balanced XLR inputs, but the signal is single-ended, not balanced. For an additional $200, the Theorem II comes with four parallel analog stages for balanced output.

I experienced one minor problem with the Theorem II review sample: a bolt holding a circuit-board support came loose in transit and rattled around inside the chassis. Other than that, the Theorem II is a much nicer package then the original Theorem: the full-sized chassis, machined front panel, input switching, polarity-inversion switch, and AES/EBU input are all welcome additions.

Listening
It took a bit of getting used to the Theorem/Axiom combination; the last four processors I've reviewed were the Theta DS Pro Generation V, Sonic Frontiers SFD-2 Mk.II, Mark Levinson No.30.5, and Spectral SDR-2000 Pro, all driven by the Mark Levinson No.31 transport—tough acts to follow. Moreover, digital processors in the $1000 price range have gotten much better lately, thus presenting a greater challenge to the Theorem II than when I reviewed the original Theorem.

I began by listening to the Theorem II driven by the Mark Levinson No.31 transport, then in comparison with the $750 Adcom GDA-600. The Adcom turned out to be formidable competition for the Theorem II, and highlighted just how much affordable digital processors have improved in the past few years.

The Theorem II's bass was a sharp contrast with that of the GDA-600. The Sumo processor lacked the extension, power, weight, and authority heard from the GDA-600. The entire presentation sounded thinner, with bass-guitar lines having less rhythmic impact. Similarly, bass drum didn't have the depth or dynamic impact I've heard from other digital sources. The bass drum that comes in on the first track of Doug MacLeod's Come to Find (AudioQuest AQ-CD1027) can be startling. Through the Theorem, it didn't produce the same sense of surprise. On this same disc, the acoustic bass had less definition and detail through the Theorem II, making it harder to perceive pitch.

The Theorem II's treble was a little whitish, and tended to emphasize vocal sibilants. Robben Ford's vocals on his Robben Ford and the Blue Line (Stretch STD-1102) was a good example. The Sumo processor tended to add a bit of hash to sibilants, rendering "s" and "ch" sounds more intrusive. Cymbals were brighter-sounding through the Theorem II and emphasized more of the top-octave sheen rather than the lower-frequency "gong"-like component that gives the sound of cymbals its body. Consequently, high-frequency–rich instruments sounded a little bleached and threadbare. This characteristic was also apparent in the mids, with brass instruments sounding a little hard and brittle rather than rich and warm. The instrumental textures on Frank Zappa's orchestral The Yellow Shark (Barking Pumpkin R2 71600), for example, had an edginess that didn't encourage high playback levels.

Soundstaging was good, but not at the level of the GDA-600. The Adcom processor had greater openness, transparency, space, and differentiation between individual images. The space on John McLaughlin's Qué Alegria (Verve 837 280-2) was smaller through the Sumo processor, and I heard less air around the images. This is one area in which the original Theorem excelled in relation to its competition of the day, but has been overtaken by more recent designs.

I repeated these comparisons with the Theta Basic and Axiom (but no clock link) as the sources, and my listening impressions were similar.

Next I compared the Theta Data Basic transport to the Axiom driving the Theorem II and GDA-600. The Axiom had a similar sound to that of the Theorem II: slightly emphasized treble, lean bass, and less-than-optimum pitch definition in the bass. Without the separate clock link, the Axiom tended to exacerbate the Theorem II's shortcomings. Compared to the $1750 Theta transport (which is one of the best buys in high-end audio), it was no contest: the Theta won hands down. Note, however, that the Theta costs nearly twice as much as the Axiom.

I auditioned the Axiom and Theorem II together with the separate clock link installed. It's a bit of a pain to switch over when making comparisons: you have to unplug the unit (just turning off the front-panel power switch leaves some of the electronics powered), remove the top cover, and reconfigure a tiny jumper wire near the edge of the board to switch between normal clock recovery (for use with any transport) and locked mode for separate clock connection to the Axiom. A rear-panel switch would have been better.

What a difference the separate clock line made! The hashy character I heard on vocal sibilants was greatly ameliorated, with the treble sounding smoother and cleaner. The soundstage opened up, allowing me to hear the music as individual images, not one big continuum. Low-level detail was better portrayed, with the presentation taking on a newfound sense of resolution. The bass also sounded tighter and better defined, although the sound was still thin.

Because the only difference in performance between the Axiom/Theorem II combination with and without the clock link is in the amount of jitter, you can easily hear for yourself by disconnecting the separate clock line what jitter does to digitally reproduced music.

Although the clock line improved the Axiom/Theorem II's musical performance, I still preferred the GDA-600/Data Basic combination, which lists for $2500—$700 more than the price of the Sumo pair.

Finally, I used the Axiom's remote volume control to attenuate the signal by a few dB, then increased the preamplifier gain to match levels. The digital-domain attenuator was surprisingly good, with almost no degradation at small attenuation settings. Although I did all my listening with the volume control all the way up for the best sound, the remote level control is a great feature—particularly if your preamp doesn't have remote control.

Conclusion
Although the Sumo Theorem II appears to be a better processor than its predecessor, the new unit is less competitive than it was three years ago. I found the Theorem II to sound good, but not superlative. The $150-less-expensive Adcom GDA-600 beats the Theorem II in many areas—particularly bass performance, treble purity, and soundstaging. Similarly, the Sumo Axiom transport is eclipsed by the Theta Data Basic and the PS Audio Lambda (though each of those transports is nearly double the Axiom's price). Consequently, it's difficult to recommend the Axiom or Theorem II for use with other non-Sumo transports or processors.

But when used together as intended with the separate clock link, the $1798 Axiom/Theorem package provided much better musical performance and became worthy of a recommendation. Many of the problems I heard with the transport and processor under individual auditioning were ameliorated by the clock link. Although the treble was smoother, the soundstaging became more transparent, and the music had greater ease, the Sumo package's fundamental sound remained: a lean bass and slightly uptilted treble. System matching will be crucial with these products: a loudspeaker with an overly ripe midbass will benefit from the Sumo's sound, while one with an overdamped bass will suffer.

If you've got $1800 to spend on a digital front-end, give the Sumo Theorem II/Axiom processor/transport combo an audition. With the right combination of associated components and loudspeakers, the Sumo package may sing in your system.

---

Footnote 2: There is also the point to consider that the Yamaha YM3623 receiver chip only passes 16-bit data to the filter and DAC. Using the transport's volume control will, therefore, reduce signal resolution in the manner described earlier. If the Theorem II used a more recent Yamaha chip or the Crystal chip, a reduction in volume by, say, 6dB would move every bit in each data word one place to the right, resulting in 17-bit words. But as these longer words would be passed through to a DAC with 18–20-bit resolution intact, the signal would be preserved, albeit with an analog noise floor 6dB higher in level.—John Atkinson