Digital Processor Reviews

Digital processing is based on the presence or absence of digital pulses. If we see a pulse, it's a "1"; if we don't, it's a "0." In all our discussions we assume perfect pulses all over the place. Well, the pulses happen not to be that perfect, and with good reason.

Some of the blame has to be laid at the execution of the digital output at the CD transport side. The spec calls for an output impedance of 75 ohms and a pulse amplitude of 0.5V p–p. Why anyone would choose an amplitude that low is a sweet mystery of standards committees; a higher amplitude would be far more capable of successfully driving longer cables. Wadia tells me that they have surveyed many CD transports and found that this spec is not met. The output impedance varies all over the place, and the signal quality is also poor.

The major concern is the output impedance of the transport. In order to transmit 2.8 million bits per second of digital code without disturbing the transmission quality, it requires 75 ohms throughout the system—at the source, in the cable, and at the load. Miss anywhere, and the shape of the pulses is disturbed. Wadia assures me that their input impedance is on the money, and I have several 75 ohm coaxial cables available. Unfortunately, all that is for naught if the output impedance, a parameter totally dependent on the manufacturer of the transport, is off. Standing waves in the cable are sure to be created, and the pulse characteristics will suffer.

Using the RCA phono jack for the output and the input connectors is also a no-no. The RCA jack has not been designed to handle frequencies in the megahertz region. How could we possibly expect the industry to put the proper connector there, when it's so much easier to cut costs with an RCA? Even though the BNC type, familiar from oscilloscope use, is specifically designed for the high transmission rates encountered here?

Incidentally, guess what connector Wadia uses at the inputs of their DigiLink 30 unit. BNC.

The point is that we can have problems in a seemingly insignificant area. Everyone appears to forget that any interface where a mix of manufacturers is likely to be encountered is a critical area. Murphy's Law says a problem will deteriorate in direct proportion to the number of components involved. Let's see now—we have the output stage of the CD transport, the output connector, the cable, the connector and detection circuitry on the processor...

This may well be part of why we have such wide variability in the performance characteristics of CD transports. Not to mention the corresponding sensitivity in cable selection.

That's the bad news. The good news is that, because we are dealing with discrete pulses, it's possible to analyze them and zero in on the problem. Pulses have to have the proper rise and fall times: they can't droop, they can't jitter, nor can they lose timing. There's no hocus-pocus here. It's a totally analytical situation.

Unfortunately, manufacturers do not realize how critical these parameters are; they cut corners. If they would only adhere to basic engineering principles, we'd be well on the way to better sound reproduction.

I can see that the front-runners in the processor field, namely Wadia, Krell, and Theta, are making sure they have their technical destinies under control—they provide their own CD transports, either under their own name or by distributing a suitable model.

So there you have it. For want of a nail, a kingdom was lost. Does anything ever really change?

2.8 million conversions/second
Anyone who didn't see add-on digital processors coming as soon as the CD's digital code was made accessible at the back of a player was very shortsighted. It was the green light entrepreneurs needed. That's why today we have Wadia, Krell Digital, and Theta. I'm sure many others will follow.

Though the introduction of processors didn't surprise me, I must confess I hadn't expected to see the complexity exhibited by the Wadia 2000 Decoding Computer quite so soon. It wasn't that long ago that we saw a single DAC multiplexed between the two channels to save a few bucks. Now, at 64x-resampling (footnote 2) the DACs have to click more than 2.8 million times every second.

If we assume that 64x or more resampling is the way to go, then the need for an efficient number cruncher is obvious. Hence the computer—though Wadia had to come up with their own design, using four dedicated digital signal-processing chips from AT&T running in parallel at 36MHz!

It was not a trivial design task—Wadia needed three years of R&D before unveiling their Decoding Computer at the 1988 CES. Add a few months to that, and that's when they were ready to ship product—roughly a year ago.

While other components like Krell Digital's SBP-64X are now available, the Wadia 2000 should be recognized for its groundbreaking efforts in the processor field—it was the first true 64x-resampling decoder (footnote 3). Unfortunately, it is difficult for those not familiar with digital concepts to appreciate what a quantum leap it really is to produce a commercially available 64x-resampling unit. Wadia lifted processing power and DAC technology to a new level at a time when others appeared to be content with 8x-resampling.

Readers of Stereophile already have a good idea of how some similar products improve CD sound reproduction. Lewis Lipnick reviewed the Theta DSPre D/A preamplifier last March (Vol.12 No.3), and Martin Colloms gave a close listen to the Wadia 1000 Decoding Computer in September '89 (Vol.12 No.9).

A whole lotta shakin'
It never ceases to astound me to what lengths designers will go to advance the state of the art of home sound reproduction. It had to happen sooner or later, but the Wadia 2000 heralds the arrival of the computer designed specifically to assist us in our playback endeavors.

And why not? Just look at the information on a CD. Did you know that to store the approximately 5 billion bits on a single CD would require a total of 15 Personal Computers, each with a 40-megabyte hard disc? That it takes 20 of the tracks storing the pits on a CD to match the width of a human hair? That the information is retrieved from the CD while it is spinning at speeds up to 500rpm? That the laser pickup retrieves data at a rate of 4.3218 million bits each second? That, folks, is a whole lotta shakin' goin' on.

---

Footnote 2: Although everyone uses the term "oversampling," "resampling" is actually the correct terminology.

Footnote 3: The Theta devotes its processing power to implementing a digital filter with a large number of coefficients, hence accuracy. It therefore runs at an 8x rate.