Digital Processor Reviews

Spectral is a bit of an enigma in the high-end audio world. Although nearly 20 years old and one of the founders of the American high-end audio industry, Spectral isn't a name that comes quickly to mind when considering the best of the best in high-end.

Spectral's low profile is of their own choosing. They advertise very little, their products are demonstrated in a small number of stores, they almost never send products to magazines for review, and they are very quiet about their accomplishments. Rather than push their products on customers, they seem to let a small segment of the market discover on their own what a Spectral system can do. Consequently, Spectral relies heavily on the demonstration expertise of their dealers.

One reason for Spectral's reticence to have their products reviewed is their belief that one of their components cannot be evaluated properly when put into a system of other components. Theirs is a holistic approach to audio system design, and they hold that an entire Spectral system—including cables—is needed to hear the full musical capabilities of just one component. This requirement has meant that the only full reviews of Spectral products to have appeared in Stereophile have been Larry Greenhill's of the DMC-10 preamplifier in the February 1984 issue (Vol.7 No.2), and Arnis Balgalvis's of the MCR-1 phono cartridge in July 1989 (Vol.12 No.7).

Technical description
The SDR-2000 Pro is a two-chassis unit composed of the processor itself and an outboard power supply, designated the DPS-2000. The supply is a smallish (6" by 3.4" by 13.3") box that feeds DC to the processor through a ribbon cable and D-connectors.

The SDR-2000 Pro's front-panel layout, controls, and display are simple and intuitive. Six large buttons in a row each select one of six digital inputs. A second row of three identical-looking buttons inverts absolute polarity, mutes the output, and dims the display. A larger button on the front panel's right-hand side switches the unit between the operate and standby modes. This switch turns off the digital circuits and front-panel display, leaving the analog stages continuously powered. Small squares beneath each of these controls illuminate when the corresponding control is activated. The area to the left of the control buttons displays the input sampling rate, and the HDCD$r logo illuminates when the SDR-2000 Pro is decoding an HDCD-encoded disc. The sampling-rate display also serves as a lock indicator, illuminating only when the SDR-2000 Pro is locked to a digital source.

The rear panel holds three RCA digital input jacks, two AES/EBU inputs, one TosLink input, and the "Spectralink" data bus. Spectralink is a proprietary method of connecting a Spectralink-equipped digital source to the SDR-2000 Pro. In this format, the audio data are transmitted on separate lines, and the SDR-2000 Pro serves as the master clock, forcing the transport to slave to the processor, reducing jitter. A Spectral transport with Spectralink output will be available later this year. Finally, balanced and unbalanced analog outputs are provided on XLR and RCA jacks.

Looking inside the power supply revealed two transformers—one for the SDR-2000 Pro's digital circuits, one for its analog electronics. In fact, only the AC-line cord, AC-line switch, and fuse are shared by the two supplies. A pair of small isolation transformers and line filters prevents noise from getting back on the AC line, further isolating the digital- and analog-stage supplies. Rectification is provided by discrete diodes, and all the regulators are discrete. Keith Johnson doesn't use three-pin IC regulators because he feels they have too much of a propensity to generate noise.

This attention to reducing noise and isolating the two supplies extends to the flat cable connecting the DPS-2000 supply to the SDR-2000 Pro processor. The DC voltages for the analog and digital circuits in the SDR-2000 appear on opposite sides of the flat cable, with a ¾" gap between them to increase isolation. Once inside the SDR-2000 Pro, the DC voltages enter a decoupling circuit (a kind of two-way trap) to prevent noise from getting back on the flat cable. Twenty-one 1500µF filter capacitors are distributed throughout the SDR-2000 Pro. Although regulated in the DPS-2000 supply, the analog and DAC supplies are re-regulated with discrete shunt regulators next to the DACs and analog output circuits. This level of care and attention to the power supply is worthy of comment.

The SDR-2000 Pro itself is divided into two sections separated by a shield running down the length of the chassis: the left-hand side houses the digital electronics (input switching, input receiver, digital filter), the right-hand side the DACs, reclocking circuit, and analog output stages.

Digital input signals are selected by optically isolated, low-capacitance relays. The selected digital signal is buffered by a transformer and video amplifier before being sent to the input receiver. The transformer isolates the source's ground from the SDR-2000's ground, preventing noise from entering the SDR-2000 Pro. Moreover, the isolation transformer prevents ground loops from occurring between the SDR-2000 Pro and the digital source. An equalization circuit in the input section processes the digital input for optimum waveshape. A switch on the input board selects between two equalizations, with the correct position determined by listening. Specifically, the switch boosts low frequencies to compensate for the low-frequency rolloff introduced by transformers at the outputs of many transports (footnote 1).

The SDR-2000 Pro uses the ubiquitous Crystal CS8412 input receiver, implemented with a "smart loop filter" to reduce jitter. (The input receiver largely determines the jitter performance of a digital processor; the receiver has its own intrinsic jitter, and it either passes or rejects incoming jitter on the S/PDIF interface.)

Although the SDR-2000 Pro's implementation of the Crystal input receiver appears excellent, Spectral has taken jitter reduction even further in the SDR-2000 Pro with a circuit they call "Time Star." The Time Star circuit consumes the entire real estate of a 2.5" by 10" circuit board that runs vertically between the analog output stages. The 11.3MHz clock generated by the Crystal CS8412 is carried to the Time Star circuit by balanced, Teflon-insulated, shielded cables. The Time Star circuit reclocks the clock output from the CS8412. Half the Time Star circuit works for the CD sampling rate of 44.1kHz, the other half for 48kHz. (The SDR-2000 Pro will not decode 32kHz sources.) The four-layer circuit board buries the critical clock signals between ground planes to increase isolation between the clock and the rest of the circuit.

The only point where clock jitter matters in a digital processor is at the DAC. A low-jitter clock can be degraded by circuit-board traces, point-to-point wiring, and even solder joints. Putting this elaborate reclocking circuit literally an inch away from the DACs ensures that the newly created low-jitter clock won't be degraded before it can do its job. To prevent noise from being radiated into the nearby analog circuits, the Time Star board is surrounded by shielding.

The Time Star circuit has three R/C time constants in the servo loop for a 100Hz jitter-attenuation corner frequency. This means that jitter with a frequency above 100Hz will be maximally attenuated (specifically, jitter is attenuated by 3dB at 100Hz). For comparison, the Crystal CS8412 on its own has a corner frequency of 25kHz, and the UltraAnalog AES20 has a corner frequency of 1kHz. Spectral claims that the Time Star circuit achieves a whopping 40dB-greater jitter rejection than the CS8412 by itself. Further, Spectral claims that the clock jitter at the DAC is less than 8 picoseconds—an astonishingly low value.

While we're on the subject of jitter measurements, I should point out that many jitter claims are rife with misinformation. Most manufacturers have no instruments for measuring jitter, and sometimes pull a number out of thin air. Measuring such tiny time variations is challenging, to say the least. For example, 100ps (0.0000000001 second) is the time it takes light to travel an inch.

Moreover, the exact point where the jitter is measured is critical; a low-jitter clock can be degraded on its way to the DAC. Another problem is that the RMS jitter measurements you see in ads don't specify the measurement bandwidth; a narrower measurement bandwidth makes the jitter number lower.

Finally, an RMS jitter number doesn't tell you about the nature of the jitter: is the jitter sonically benign in that it is random in nature, or is the jitter energy concentrated at specific frequencies—a condition that is much more sonically detrimental? Until these issues are resolved and some standards set, take all published jitter numbers with a grain of salt.

Having said that, it is probable that Spectral's extraordinarily low jitter claim for the SDR-2000 Pro is accurate. While at Pacific Microsonics to interview the inventors of HDCD (including Spectral designer Keith Johnson; see interview elsewhere in this issue), I got a hint of their extremely sophisticated jitter-measurement techniques and instrumentation. Reducing jitter is a big concern at Pacific Microsonics; they've invested a lot of engineering time and money in quantifying it. Although Pacific Microsonics and Spectral are separate companies, one would expect that Keith's insights into jitter measurement and reduction would not be restricted to one company's products.

In short, the SDR-2000 Pro has one of the most sophisticated jitter-reduction techniques I've yet seen in a digital processor. Note, however, that, despite this extraordinary attention to isolating the clock from the incoming signal, the digital interconnect between a transport and the SDR-2000 Pro still makes an audible difference, as I found in this review.

Spectral's SDR-2000 Pro differs from the earlier SDR-2000 in having the PMD100 HDCD decoder/filter in place of an NPC filter. The PMD100 is located on a small daughterboard that fits in a socket on the main digital board. The Pro version is shipped with the HDCD decoder/filter as standard. Owners of the SDR-2000 can upgrade to the PMD100 for $695.

The DACs are four UltraAnalog 20-bit types, made of discrete and monolithic components encapsulated in a potted module, in a balanced configuration (left +, left –, right +, right –). The SDR-2000 Pro is unique among UltraAnalog-based processors in that the current-to-voltage (I/V) converters and sample-and-hold circuits are not incorporated into the potted DAC module. Instead, UltraAnalog makes a custom version exclusively for Spectral that allows Spectral to build discrete current-to-voltage and sample-and-hold circuits. The capacitors in the sample-and-hold circuit are the result of Keith's years of searching for the ideal capacitor for this ultra-critical application. The I/V is an ultra-fast-slewing, direct-coupled, FET-based circuit that reportedly achieves fast settling time and low energy storage. According to Keith, this section of a digital processor is critical to sound quality.

A DAC's conversion precision is affected by the ambient temperature in which it operates. Any temperature change could disturb the hand-calibrated and -trimmed precision inherent in the UltraAnalog DACs. Madrigal addresses this situation in their Mark Levinson No.30.5 by mounting the DACs in sealed, heatsinked chambers whose thermal characteristics strike a perfect balance between the heat generated inside the chamber and the heat removed by the heatsinks, thus keeping the DACs at their optimum temperature. In the SDR-2000 Pro, a servo-controlled fan cycles on and off to maintain the correct operating temperature for the DACs. The chassis is sealed, with no vents.

---

Footnote 1: See Rémy Fourré's article in the October 1993 Stereophile (Vol.16 No.10) for a full explanation of why low-frequency rolloff of the S/PDIF signal creates jitter.