Digital Processor Reviews

The $799 Theorem was originally shown at the 1992 WCES in a very small chassis that prohibited adding features or upgrades. Sumo has since become more ambitious, putting the Theorem in a full-sized chassis and offering several upgrade options that would have been impossible in the truncated version.

There's not much to say about the Theorem's features; the front panel has a power "On" LED and nothing else. Selection between the unit's single coaxial input and Toslink optical input is automatic. If a digital signal is present at both inputs, the system defaults to the optical input. The rear panel has a single RCA digital input jack, a Toslink optical input, a "clock" input, and a pair of RCA jacks on which the analog output signals appear. A line fuse and captive AC line cord finish the rear panel. Sumo plans to announce an upgraded version of the Theorem, with balanced outputs and AT&T ST-type optical inputs. No release schedule or prices were available at press time.

The Theorem is the only outboard D/A converter I'm aware of (except the Sony DAS R1 and Linn Numerik) to provide a method of independently synchronizing the converter to the transport. This is the "clock" input just mentioned. When connected to a CD transport with a clock output, the Theorem will lock to the separate clock signal and not need to recover it from the S/PDIF signal. In all other digital processor/transport combinations, the clock signal is carried in the audio data down the interconnect cable. This method, however, is a significant source of jitter in the recovered clock. By locking the processor to a separate clock, the entire problem of clock jitter introduced by the digital interface is avoided (footnote 1).

While this sounds good in theory, there are no transports with a 256x clock output on the market, and as far as I'm aware, none in development. Sumo included the clock input just in case they make their own transport, or another manufacturer decides to include a 256x clock output.

Technology
Under the skin, the Theorem is an unusual design. A small printed circuit board containing all the digital and analog circuitry is mounted on a much larger board that holds some of the power supply. This arrangement provides more board space for options and makes wholesale upgrades (replacing the entire circuit) easier. Moreover, the DAC board will plug into Sumo's Athena II and Diana preamplifiers, turning those preamps into a DAC/preamp combination.

The power supply consists of a toroidal transformer and single bridge rectifier supplying ten regulation stages. Six of these stages are active (three-pin ICs), four passive. The analog supply rails are an unusually high ±35V.

Although Sumo's brochure suggests the Theorem uses the new Crystal CS8412 input receiver, my sample had the usual Yamaha YM3623 16-bit chip. Sumo originally designed the Theorem with the Yamaha chip, switched to the Crystal (which had problems in early versions), then switched back to the Yamaha. Digital filtering is provided by the 8x-oversampling Burr-Brown DF1700. This filter is identical to the more common NPC 5803.

The digital-to-analog converter (DAC) is a dual-channel Burr-Brown PCM67P, the so-called "hybrid" DAC. It gets this name from the fact that the upper 10 bits of the 18-bit digital word are converted to analog by a multi-bit resistor ladder converter, and the lower 8 are processed by a 1-bit converter with noise shaping. This hybrid approach theoretically combines the best of both conversion technologies: good low-level linearity from the 1-bit portion, with superior dynamics and other qualities associated with multi-bit converters. Sumo chose the PCM67 after listening to a variety of DACs. The Theorem is the first product I've reviewed to use this DAC; it doesn't appear to have caught on among high-end designers.

The Theorem is unusual in that there are no ICs in the analog circuit after the DAC. In fact, the whole output stage is both tweaky and impressive. The current-to-voltage converter, gain stage, and output buffer are integrated into a single stage. The circuit is an all-discrete, fully complementary class-A design, with very little feedback. The output drivers are a complementary pair of hefty (TO-220 type package) bipolar transistors, which can reportedly drive high current (500mA) into low impedances. The circuit is direct-coupled, with a 741 op-amp acting as a DC servo. The low-pass Bessel filter and de-emphasis circuit are built into the analog stage's feedback loop, with de-emphasis switched in by a relay. A second relay mutes the output when the unit isn't locked to a digital input (footnote 2). Sumo believes that the level of performance provided by their output section can't be matched by op-amps.

I had mixed feelings about the Theorem's build quality. The chassis metal is quite thin, and the unit lacks the ¼"-thick front panel found, cor example, on the Forté on the DAC 50 and PS Audio Digital Link II. I was very impressed, however, with the output stage's design and execution; this sophisticated a circuit is unusual in a digital processor, particularly one of such a modest price. Given the choice between a thicker chassis and a better-sounding analog stage (if it is indeed better-sounding), I'll take better sound any day.

Listening
As I began to listen to this processor, I was reminded of a common criticism of subjective reviewing: that the reviewer knows the products' identities and is therefore biased. The word "bias" implies either favoritism to some brand name(s) or preconceived ideas about how the products will sound. When listening to music, the last thing on my mind is the product's brand name: it has no effect on my musical experience. And as for favoritism, I don't care which of the processors I review—if any—fares well, and which doesn't. Let the chips fall where they may.

Because the Theorem was the last converter to arrive, I listened to it only after spending the previous week auditioning the other three processors that I review this month. I was in for a pleasant surprise.

---

Footnote 1: See my article in Vol.15 No.1, p.166 (January 1992) for a more complete discussion of this technique.

Footnote 2: I had one functional problem with the Theorem: when the Theta Data transport automatically turned off its digital output after five minutes of nonuse, the Theorem put out a very loud glitch through the loudspeakers. Apparently the muting circuit isn't fast enough to ground the analog output when it loses lock with the incoming signal.