|Home | Audio Magazine | Stereo Review magazine | Good Sound | Troubleshooting|
A common question we get from Stereophile readers is why we don’t review products from Digital. The five-year absence from these pages of reviews of products by a major manufacturer invites all kinds of speculation). The explanation is simple: decided it was not in their best interests to have reviews in the US press, and declined our requests for review samples.
As we were contemplating buying products on the open market for review, the company had a change of heart, not coincidentally connected with a change of management. They provided review samples of their $8450 27 Digital Decoding Computer and $12,500 Wadia 7 transport.
27 DIGITAL DECODING COMPUTER
The 27 Digital Decoding Computer represents ’s latest thinking in digital processor design. Although the 27 costs less than the flagship $13,500 9, the newer 27 has some circuit refinements not found in the top-of-the- line processor.
The 27 is housed in a beautifully made chassis with no bolts visible from the top or sides. The chassis is built of separate panels of thick aluminum joined at the corners by foot rounded structures; these accept short spikes on the bottom for resonance control. If you don’t like the idea of spikes poking into your rack, the 27 comes with metal “coasters” that fit between the spike points and the rack shelf surface.
The 27’s front panel has no knobs, switches, buttons, or controls. Instead, an alphanumeric display shows the unit’s operating status, which is controlled exclusively by the remote control. The blue display indicates the output level (volume) in numbers between 0 (mute) and 100 (frill-scale). Input selection, phase inversion, and left/right balance adjustment are all made by the remote control and displayed on the 27’s front panel.
The supplied remote control also runs ’s transports. At the time of this writing was replacing the plastic remote with a cast aluminum remote more befitting the construction quality and appearance of their processors and transports. It’s a good thing, too: the plastic remote is an ergonomic nightmare, with lots of tiny, identically shaped, and poorly laid-out buttons. If your 27 was purchased with the plastic remote, will replace it with the aluminum one at no charge. I didn’t have the new controller in time for this review, but saw a prototype at ’s factory. It looks like a big improvement.
The 27’s rear panel provides six digital inputs: two ST-type glass-fiber optical, two S/PDIF coaxial on BNC jacks, one AES/EBU, and one TosLink. was a pioneer in applying glass-fiber transmission technology to consumer digital audio, and recommends using the ST type optical connection with their products. All the digital inputs are mounted on a separate panel attached to the machined chassis for easy upgrade if a new transmission technology comes along.
Balanced analog output appears on XLR jacks, single-ended on RCAs. The 27’s output stage was designed to drive both outputs simultaneously in systems (like mine) that use a separate woofer amplifier.
Using the 27 without a preamplifier: The 27’s adjustable output level lets you drive a power amplifier directly from the 27 without the need for a preamplifier. is a big proponent of this technique, which removes from your signal path a pair of interconnects, input and output jacks, a volume control, wires, switches, circuit-board traces, resistors, and the active gain stages of a preamplifier. No matter how good your preamplifier is, it can’t be better than no preamplifier.
Removing your preamplifier from the signal path has undeniable sonic ad vantages, but some drawbacks as well. First, the digital-domain volume control in the 27 reduces resolution at the rate of 1 bit for every 6dB of attenuation. The greater the attenuation (the lower the number on the 27’s front-panel display), the greater the degradation.
Second, removing the preamplifier poses the problem of source switching: how do you play your tuner or turn table? ’s solution is to digitize your other sources with their 17 A/D converter, whose digital output drives another input on your 27 processor. Or you could reconfigure your system each time you switched sources. Finally, some listeners may use only a CD source and not need input switching.
To minimize the amount of lost resolution resulting from digital-domain attenuation, the 27 has an analog- domain output level adjustment. A pair of tiny switches inside the chassis set the 27’s maximum output level from 0.41V to 9.02V, with 14 positions in between. If you have a power amplifier with a high input sensitivity (the input voltage required to drive the amplifier to clip ping) and high-sensitivity loudspeakers, you’d set the 27’s switches for a low out put level. This way, for the best sound, you’d run the 27 with very little digital- domain attenuation. Remember, the higher the number on the 27’s front- panel display during normal listening, the better.
My system worked best with the 27 set for a maximum output level of 2.63V. The next lower setting, 2.03V, didn’t have quite enough output to pro duce satisfying levels with CDs re corded at low levels, or those with a high peak-to-average ratio. I calculated the resolution loss between these set tings at about a third of a bit—probably not audible. Most of my listening was with the display showing between 80 and 100. With some discs I ran the 27’s volume control wide open, with no attenuation and thus no degradation.
Some users may not take hill advantage of the 27’s variable analog output level feature. doesn’t provide instructions on setting the switches; you must rely on your dealer for this information. If you find yourself listening most of the time with the display showing below 70, you’re needlessly throwing away resolution. It’s well worth the trouble to open the chassis and set the switches to get the best possible performance from the 27. recommends having your dealer set the output level, but you need to adjust it while it is in your system to find the right level. If your dealer won’t come to your home to correctly set the analog gain switches, get the information on the switch set tings and do it yourself. Taking the time to play with this adjustment to get it just right is well worth the effort.
While we’re on the subject of digital-domain volume controls, I should mention the exemplary technique used in the Enlightened Audio Designs DSP-9000. That unit automatically switches-in analog-domain attenuation so that the digital volume control never attenuates the signal by more than 6dB. For example, as you reduce the volume from full-scale output the DSP-9000 attenuates the signal in the digital domain until 5.9dB of attenuation is reached, then switches-in 6dB of analog attenuation and puts the digital volume control back to no attenuation. Consequently, if you had 10dB of attenuation, 6dB of it would be in the analog do main, 4dB in the digital.
Inside the 27: During a visit to ’s factory I spent some time with the 27’s principal designer, Jim Kinne. The unit uses some innovative design techniques, the first involving the power supply. Two, large toroidal transformers are housed in a sub-enclosure within the 27’s chassis. The transformers are sandwiched .between layers of Neoprene damping material on the sub-enclosure bottom and the machined aluminum plate that covers the transformer sub-enclosure. This technique isolates the transformers mechanically, acoustically, and electrically from the rest of the chassis.
The transformers’ outputs are rectified, filtered, and regulated on a separate pcb. Unusually, the regulation stages for the I/V converter and analog output stages are current regulators, not the typical voltage regulators. These constant-current supplies are then regulates with voltage regulators on the audio board next to the audio circuit.
In a current regulator the output current is held constant. When followed by a shunt voltage regulator any current not use by the supplied circuit is dumped to ground. believes that this power-supply topology produces superior sound quality, primarily because a current source is less affected by inductance and capacitance of the power-supply traces.
In all, the 27 uses a whopping 36 regulators, most of which form cascaded regulation stages (the output of a pre-regulator feeds the input of the final regulator) for a cleaner and more stable power supply. The eight current pre regulators are in TO-220 packages with 2”-tall heatsinks. Many of the cascaded voltage regulators next to the audio circuits are surface-mount devices.
Aiding these 36 regulators are loads of filter capacitors, some of them distributed near the audio circuits. The 27 has more than 30,000uF of filter capacitance — as much as some small power amplifiers. This huge power supply makes the 27 run very hot.
The pre-regulated power-supply rails aren’t connected to the second regulators on the audio board through conventional circuit-board traces. Instead, the power supply is routed to the audio board by a separate dedicated board underneath the audio board; the supply voltages are “popped up” to the audio board where they’re needed. This separate power-sup ply board can have huge traces and direct routing, in contrast with the smaller traces and indirect routing of conventional power-supply distribution on a single board shared with the audio circuits. The 27’s power-supply design and execution are impressive, to say the least.
The input section is based on a Crystal CS8412 receiver chip implemented on the DSP board. Signals present at unselected inputs (if you have ST-type optical and AES/EBU connected simultaneously, for example) are disconnected from the circuit to prevent contamination. Although the C58412’s jitter rejection is only mediocre by today’s standards, the clock that drives the DACs isn’t generated directly by the 8412. Instead, the 8412 is used to demodulate the incoming S/PDIF signal, and the 8412’s recovered clock synchronously drives a Xilinx programmable gate array. The gate array generates a clock at its output that is de-jittered by a Voltage Controlled Crystal Oscillator (VCXO), which is de-jittered again by another VCXO located near the DACs. In all, three stages of VCXO-based re-clocking reduce clock jitter. This scheme is a more sophisticated implementation of ’s “Rock-Lok” de-jittering circuit, first used in the 2000 Digital De coding Computer. By the way, a single Xilinx gate array—a square chip about half the size of the bar code on the front, of a magazine — replaces literally dozens of separate ICs used in earlier products.
All the clock lines are differential (two traces carry the clock and its inverted version) to minimize the effect of noise on the crucial timing signals. Moreover, the clock lines are not just traces on the PCB, but are miniature transmission lines, with a precise output impedance from the driver and termination resistors on the receiver to create a matched input impedance. A carefully executed transmission line prevents reflections in the conductor that would add noise to the clock signal, making the edges less precisely timed and introducing jitter.
was at the forefront of using a custom digital filtering algorithm, a tradition continued in the 27. Two Motorola 56004 DSP chips, under the control of ’s software contained in Read-Only Memory (ROM) chips, form the digital filter. The filter runs at 16x.oversampling, meaning it interpolates 15 new samples for every incoming sample read from the CD. The DSPs execute ’s patented “Digimaster” filtering algorithm, an update of the algorithm used in ’s original 2000 Decoding Computer. The two Motorola DSPs replace the four AT&T DSPs used in the 2000 while nearly doubling the amount of computing horsepower.
The Digimaster “spline” filter algorithm is based on mathematics first developed in the 18th century for shipbuilding. Shipbuilders needed to fit curved horizontal planks of wood to the fixed points represented by the vertical hull members. Similarly, digital-interpolation filters attempt to “fit the curve” by creating new points along an arc (a musical waveform) between existing points (the original sample values) (see Note 2 below).
The Digimaster filter is optimized for time-domain performance. Consequently, the 27’s filter will repro duce an impulse with no pre- or post- echo, and its square-wave response has none of the ripple of conventional digital filters that follow a sin(x)/x function. However, there is a penalty in the frequency domain for this improved time- domain performance: a rolloff of nearly 3dB at 20kHz. (See Note 3 below).
Volume control and muting are also performed by the DSP chips. To attenuate a digital signal, each sample is multiplied by a number less than 1. For example, if you set the digital volume control for 6dB of attenuation (half the voltage), the DSP multiplies each sample by 0.5. Similarly, pressing the remote control’s mute button simply ramps down the volume to zero by software control.
The 16x-oversampled data are put through a Xilinx gate array that retimes the over-sampled data for output to the DACs. Specifically, the gate array creates four data-streams and delays three of them by slightly different amounts. To create a balanced signal, the gate array also inverts the second and fourth datastreams. Each datastream is converted to analog with its own DAC: The result is an increase in the effective oversampling from 16x to 64x (16x in the digital filter, multiplied by four by the time-staggered DAC technique). To look at it another way, 16x-oversampling is performed by the DSPs under control by the Digimaster software, and 4x-oversampling is performed by hardware (the DACs).
The DACs are Burr-Brown PCM 1702s in a fully balanced configuration, with the positive and negative phases of the balanced signal handled by two DACs each. The 1702’s intrinsically good linearity and lack of an MSB trimmer must have made life easier for the 27’s designers; setting the trimmers on previous DACs, so they all perform identically, must have been a challenge.
The DAC outputs of each phase are summed with their own current-to-volt age (I/V) converters, the super-fast (400MHz) Burr-Brown OPA-642 op amp chip. This is followed by another Bun-Brown device, the INA-103 instrumentation amplifier. The variable analog output level described earlier is implemented by varying this stage’s gain. The IMA- 103 processes each phase of the balanced signal separately, and also acts as a differential amplifier to convert the balanced signal to single-ended to drive the RCA jacks.
The output buffer is a Burr-Brown BUF-634 op-amp, which can reportedly swing more than 9V RMS and source a quarter of an ampere of current. Each phase of the balanced signal, along with the single-ended signal, is buffered by its own BUF-634. The output stage, with its high output level and specified out put impedance of less than one ohm, is designed to drive a power amplifier directly through any length of cable. Although calls this output stage the “Sledgehammer,” it is completely unlike the Sledgehammer stage used in earlier products.
Note that the signal remains balanced from before the DACs to the XLR jacks (parallel signal paths for each phase of the balanced signal). In addition, the single-ended signal is created by combining the balanced signal with a differential amplifier, then buffering it independently. This technique provides single-ended users with the benefit of the 27’s balanced topology. The alternative is to simply connect the positive phase of the balanced output to the RCA jack.
This inferior method effectively throws away half the DACs and output stage when using the single-ended outputs.
The four-layer analog output board that contains the DACs, I/V converter, and analog output stages sits above the power-supply board. The four-layer digital board with the Xilinx gate array and DSP chips sits behind the audio board. The left and right channels on the audio board have been designed as mirror images of each other so that their electromagnetic fields cancel.
Most of the 27’s components are surface-mounted devices. The DSP chips, Xilinx gate array, DACs, I/V converter, and output buffer are all surface-mount devices. Surface Mount Technology (SMT) eliminates the long leads of “through-hole” construction, instead mounting the electronic component on the printed circuit board’s surface. touts the many advantages of SMT, including the removal of parasitic resistance and inductance of component leads, along with the ability to put components closer together on the circuit board. SMT resistors and capacitors are so tiny you almost don’t see them on a circuit board.
Previous products have, in my view, put the emphasis on the digital side of the design at the expense of the analog stages. For example, the original 2000 broke new ground with its sophisticated digital filtering and time- staggered DAC technique, but subjected the analog audio signal to decidedly audiophile push-on connectors. (The 2000 was designed by telecommunications engineers, not tweaky audiophiles.) Fortunately, ’s thinking seems to have shifted in the last eight years: the 27’s analog section appears to be as care fully realized as the digital sections.
Some might be perturbed by the 27’s use of an op-amp output stage rather than a discrete circuit. The conventional wisdom is that any contender for crown of State of the Art in digital playback — to which the 27 appears to aspire — must use a discrete output stage. Al though few would argue that an op- amp can outperform the best discrete stages, it should be noted that virtually all digital processors use an op-amp in the current-to-voltage converter. Every UltraAnalog DAC module, for example, has an integral op-amp current-to-voltage converter (except for the Spectral SDR-2000 Pro, which performs current-to-voltage conversion with a discrete circuit outside the DAC). Consequently, many superb-sounding processors —including the Mark Levin son No.30.5—have an op-amp in the signal path (the No.30.5 has a discrete output stage, however).
But what ultimately matters is how the product sounds. Any preconceptions engendered by the product’s de sign vanish in the listening room.
I found the 27 gorgeous in cosmetics and build quality—it has a substantial solidity and beautiful metalwork. My only complaint about its operation is the narrow window over which the remote control works. I had to point the remote almost straight-on to get it to accept commands — a real nuisance when listening. Moreover, if you push and hold the volume control while trying to find that window, the 27 doesn’t accept the command until you release the button, then activate it again while it is pointed directly at the front panel. This drawback won’t be a factor if you have a direct line of sight between the listening position and the 27, but I found it frustrating in my listening setup.
7 CD TRANSPORT
At $12,500, the 7 is the most expensive CD transport I’m aware of its look and feel, and its beautiful construction and finish quality certainly reflect its astronomical price.
The 7 comes in two chassis: the transport itself, and a power supply. The chassis construction is the same as that of other products, with machined aluminum panels connected at the corners with rounded structures into which spikes can be threaded. The drawer is a thin slit in the center of the main unit’s front panel, just above the display.
The front-panel display shows all the usual information (track number, playing time, etc.), and has a total of two controls: Play and drawer Open/Close. Track skip, stop, pause, search, and other functions are provided only on the remote control. The remote itself is a massive metal unit that requires two hands to operate. Rows of burtons of identical color, size, and shape nearly cover the remote’s front, though many of these are for controlling other products and don’t apply to the 7. I didn’t have this remote control for the review; advised me to use the 27’s remote, which also controls the 7. Nonetheless, the 7’s remote looked intimidating.
The rear panel holds one each of the four main output categories: AES/EBU, ST-type glass-fiber optical, TosLink, and coaxial on a BNC connector. Three small screwdriver adjustments adjust the front- panel display’s brightness and drawer open and close speeds, respectively.
The 7’s high cost is due largely to the expensive Esoteric P-2S transport mechanism. The P-2S is the top-of-the-line mechanism made by Esoteric (a division of TEAC). (The P-2 transport I reviewed in Vol.13 No.12 used the transport one model down from the P-2S.) This transport is probably the best engineered and most solid yet devised. It is also the most costly—by a wide margin. It was reported to me that the raw P-25 mechanism alone costs $2500 —and that’s the wholesale OEM quantity price.
The device uses the Vibration-Free Rigid Disc Clamping System (VRDS) first seen in the P-2 transport sold in North America under the Esoteric name. The VRDS clamping system, just larger than the CD itself, holds the entire disc in a viselike grip. Compare this method with the usual technique of holding the CD in the center by a small spindle. Al though all the Esoteric VRDS mechanisms use this clamping technique, the various models are built very differently. The P-2S used in the 7 features a machined aluminum and bronze turn table and clamp, the clamp mounted on a massive structure spanning the mechanism. Lower models in the Esoteric VRDS line are less massive, use stamped rather than cast parts, and plastic elements where the P-2S uses metal ones. sent me a raw mechanism so I could get a better look at its construction. This is one serious piece of mechanical engineering—it makes the ubiquitous Philips transport mechanisms look like toys.
modifies the standard P-2S mechanism by replacing the oscillator, changing the digital output board, and adapting the unit for their external custom power supply. They also mount the VRDS assembly on an integral suspension system to isolate it from vibration. has established a tighter jitter tolerance for their mechanisms than the standard P-2S transports.
Note that damper discs, rings, and other mechanical CD tweaks interfere with the VRDS mechanism. The VRDS clamping system would seem to obviate the need for such tweaks anyway.
I listened to the 27 and 7 in a variety of configurations. Before the 7 arrived, the 27 saw lots of action driven by the Mark Levinson No.31 transport. Conversely, the 7 fed the Mark Levinson No.30.5 processor and a Classe DAC-1 so it could be compared with the Levinson No.31 transport. Of course, the 27 and 7 were also used as a pair, connected through the -recommended glass- fiber optical (ST-type) cable. Other digital links included the Illuminati coaxial and AES/EBU cables — my references in digital interconnects.
Loudspeakers were Genesis 11.5 dynamic/ribbon hybrids. Most of my time listening to the 27 was with the processor driving the Audio Research VT15O tubed mono-block power amplifiers directly from the 27’s balanced outputs. The Genesis woofer amplifier was fed from the 27’s single-ended outputs.
To judge whether the sound quality I was getting was due to the 27’s intrinsic quality or to the removal of the Sonic Frontiers SFL-2 preamplifier, I experimented with the SFL-2 in and out of the signal path. For the matched-level comparisons between the 27 and Mark Levinson No.30.5 processor, both units fed the SFL-2.
The long balanced interconnect between the 27 and VT15Os was Wire-World Gold Eclipse, with Magnan Type V feeding the Genesis woofer amplifier’s unbalanced inputs. Other interconnects called into play when using the SFL-2 included shorter runs of Gold Eclipse or AudioQuest Diamond X3. The loudspeaker cables were also Gold Eclipse.
The products under review sat on a Billy Bags 5500 series rack or a Merrill Stable Table. My Well Tempered Turn table was temporarily ousted from its place atop the Stable Table because I’d eliminated the SFL-2 from the system and because I needed the rack space to house all these digital products. I removed the Genesis Digital Lens from the system so I could evaluate the products on their own merits. Power to the system was conditioned by the entire MIT Z-System, including the Z-Stabilizer II, Z-Iso-Duo, Z-Center, and Z-Cord II power cords.
Reviewed by Robert Harley (courtesy Stereophile, Oct 1996); Related articles at Stereophile: 2002 Follow-up