Wadia Digital 1000 Decoding Computer Page 2

Oversampling, or, more correctly, the resampling of the original 16-bit x1 digital audio code, is conventionally performed in the frequency domain. It is claimed by Wadia that time-domain curve-matching provides superior accuracy with asymmetric music waveforms, resulting in improved transient definition, better intertransient silence, and a more solid low-frequency range. The "Frenchcurve" code is resampled at a 16x rate and processed at 36MHz on a 36-bit–wide number path, the computations running at 36MIPS (million instructions per second) per channel. CD replay does demand quite powerful processors; a 1989 IBM AT PC runs at just 3MIPS.

The 36-bit filtered code employs the preferred techniques of dither and rounding to present the best 16-bit data to the DACs. Parallel-to-serial conversion is performed to interface the processors to the DACs.

The DSP section is built on a module basis, and the Frenchcurve software is contained on programmed chips which may be changed or upgraded at any time. The filter operates at a high level of 192 taps/channel, and virtually any conceivable filter characteristic may be executed with the appropriate software, including C Spline, digital de-emphasis, convolution, deconvolution, and correlation. Certain FIR and JIR filters may also be used.

The Frenchcurve is worthy of some detailed attention: the abandonment of a steep rolloff above 20kHz is bound to result in poorer ultrasonic filtering and a poorer measured performance. A brickwall or "taut" filter has excellent frequency-response specifications but generates some ringing on test- and music-signal impulses due to its attempt to "template-interpolate resample" in the frequency domain. Frenchcurve seeks to join the resample "dots" or information points with a best-fit interpolation in the time domain. Instead of linear dot-to-dot interpolation, the Frenchcurve algorithm generates, in real computing time, unique best-fit interpolation curves, which for each new dot acquired spans a total of 13 dots.

The algorithm is quoted as a 12th-order polynomial of the form: Ax12 + Bx11 + Cx10...Lx2 + Mx + N

Thus the curve fit is a smooth envelope, considered to be a closer construct of the original music waveform prior to sampling. The consequences of poor filtering above 20kHz will not be too serious with modern high-linearity, wide-band analog electronics, but some older equipment may suffer on such a diet. More experience needs to be gained in interfacing such units as the Wadia.

With 16x resampling, very little analog post-filtering is required, and the signal path is short and direct, following the integrated circuit buffer. This should help clarity and dynamics over and above any contribution made by the Frenchcurve filtering.

Wadia sound
Notwithstanding some inevitable variability of results with different arrangements and systems, from the start it was clear that the WD1000 was indeed a state-of-the-art digital decoder.

I have not yet heard either the new Krell or the Theta digital units, and thus cannot compare their performances with that of the Wadia. However, on the basis of the Wadia's showing, this comparison will surely be worth making. My own researches over a 30-month period indicate that, even with existing recordings, much better digital replay is within reach. It has been frustrating to see so many commercial designs fall so far short of the absolute. If my own open-ended scoring system is used for comparison with a notional present state of perfection scored at 25 (this attained by a well-installed Goldmund Reference turntable system), then the CD-player industry average score, for all prices, is a disappointingly low 9.

"Audiophile" players average a score of 13, the odd exception reaching 14. Two years ago, the Accuphase 80/81 two-box player reached a then-amazing 16. Now we have the Wadia 1000, and when it is installed as part of a compatible system, it can score 20. Make no mistake, this represents a major improvement, on a scale comparable to an upgrade from Krell KSA50 to KSA80 or Audio Research D70 to Classic 60, D115 to a D125.

A unit scoring at the 20 level is immediately identifiable, so startling is the improvement over present norms. First impressions are worth recording: for me, foremost among these was a striking level of clarity and transparency, akin to hearing a Wilson WATT for the first time. Such clarity provided an abundance of information about the individual musicians and the total performance, about the acoustic of the hall in which they were playing, and also a good idea of where the performers were located.

Such high clarity was not confined to the midrange. Compared with established standards, the treble was exemplary, imbued with pinpoint focus, detail, and definition. The bass was also exceptional by digital replay standards, sounding tight, powerful, and very tuneful.

As I adjusted to these qualities, I could stand back and take a more critical attitude. Compared with my model of perfection, some of whose facets have been illuminated by elements of existing technology, the Wadia 1000 showed a hint of mid-"glare" and treble brightness, plus a touch of "chromium plating"—the sound being just a shade brighter and shinier than necessary. In this respect it sounded more "digital" than analog, and required some acclimatization on my part.

A sympathetically balanced listening system, free from glare or hardness, partnered the WD1000 very well. Skillfully set up, such a system showed that a breakthrough has been made in digital replay, one which at last shows digital to be a worthy alternative to high-class vinyl. The results for the two media were not the same, but definitely had their own merits. Moreover, this comparability was obtained despite the present inferior state of most digital recordings, very few mastered on anything but a Sony 1610 or 1630, with their brickwall-filtered, linear, 16-bit encoding. Unquestionably the WD1000 sounded better on the better-encoded recordings, and considerable scope for improvement remains on the recording side. Nevertheless, the Wadia 1000 and its rivals sounded very good even on established material, which counts for a lot in my book.

Older recordings came up as new: for example, Rickie Lee Jones's Rickie Lee Jones, or the better Chesky reissues. Rock-jazz was startling in its impact and dynamic power: Miles Davis's Tutu or Flim and the BB's. Wadia replay gave the feeling that a more direct link had been established with a recording master, a considerable veiling of detail and a typical softening of transient definition appearing to be stripped away. What had sounded like an analog tape copy now sounded more like the original.

Stereo soundstages were wide and deep, with clarity well maintained to the back of the soundstage, while stereo focus was exceptionally good, maintained right into the perspective layering of good recordings. In some respects the effect was almost holographic.

On rock material the decoder established a convincing bass line, with driving rhythms and fast, clean articulation. The Apogee Duetta Signatures excelled at resolving such nuances. This bass performance was the finest I have yet experienced from a commercial digital replay.

Working from absolutes, the Wadia character tended to the clinical and very slightly "mechanical." Another critic might, however, reach an opposing conclusion, describing this as neutrality and a lack of false bloom or richness.

Sound-quality differences between the transports were evident, with the Philips-Marantz providing the best results for both clarity and rhythm. Interestingly, the Sony transport showed good treble clarity but was more subdued in terms of dynamics and rhythm. Different digital coupling cables affected the sound in more subtle ways, while a symmetrical buffer, built for me by Peter Wittersand, significantly improved the sound using the Philips and Marantz models. In my north London location, the Wadia suffered from occasional RF breakthrough from portable telephones, radio cabs, and CB, resulting in momentary audio muting. Muting and de-emphasis time constants were rather slow, and the odd note was sometimes missed at the start of a selected track. These matters need attention in view of the Wadia's high price.

Despite some interfacing problems involving spurious responses and, in my location, spurious muting due to RF breakthrough, the WD1000 proved its point. Digital audio replay is at long last beginning to approach the standard set long ago by the best vinyl replay. In this I include listener involvement, rhythm, pace, transparency, depth, and, not least, a natural, dynamic quality. At last the listener is to be offered a chance to enjoy the known benefits of the digital cake—wide dynamic range, zero wow, no noise, etc.—and enjoy eating it, too!

The WD1000's design contains many contradictions. Built to an impressive level of extravagant structural solidity, its mains transformers are nevertheless popped in right under the sensitive DAC sectors. The emphasis is on short signal paths and good sound, yet the power-supply lines are routed in a circuitous fashion from board to board. Ordinary cabling is used inside, the boards (including those for audio) linked by industrial-grade IDCs (insulation displacement solderless connectors). There is also evidence of an odd mix of military, industrial, and domestic thinking.

I would like to see lowered levels of spurious output, better internal screening, higher DAC stability at the 15th and 16th bits, and better immunity to RF interference.

The de-emphasis and mute relay time constants also deserve attention. The idea is sound, however, and the results would appear to prove the case for Robert Moses. This is a great-sounding digital decoder, and Wadia is to be commended for bringing it to the market. With more units like this, the audiophile digital arena is beginning to look more and more interesting; who said that all digital components must sound the same!?!

Wadia Digital Corp.
1556 Woodland Drive
Saline, WI 48176
(734) 786-9611