Wavelength Cosecant v3 USB digital/analog converter By Art Dudley   •   June, 2009 While my enthusiasm for the long-discontinued Sony PlayStation 1 remains high (see the July 2008 Stereophile), I freely acknowledge that not every high-end audio enthusiast wants a CD player with an injection-molded chassis, a Robot Commando handset, and a remarkable lack of long-term reliability: Yes, the Sony sounds wonderful, but sound isn't everything.

Nor is an expensive high-end CD player the answer to everyone's needs. With world-class LP players available for a few thousand dollars and up, some hobbyists are, if anything, increasingly reluctant to spend that much or more on a medium they consider to be inherently inferior.

That's how I feel about it, too—in addition to which I don't like clutter, have no use for remote controls, and loathe conspicuous consumption.

And I already have a very good desktop computer...

That adds up to one thing: In 2009, I am ready once again—once and for all, I think—to use my desktop computer as a digital music source. As fate would have it, Gordon Rankin, of Wavelength Audio, recently revised his seminal computer-audio product, the Cosecant USB digital/analog converter. After John Atkinson used the new Cosecant v3 for a seminar on hi-rez music at last year's Rocky Mountain Audio Fest, he sent the review sample my way, with Rankin's blessing.

Description
The Cosecant v3 is the latest entry in a category all but invented by Wavelength Audio: a freestanding device for converting the digital-audio datastream from a computer's USB bus to a line-level analog signal appropriate for a perfectionist music system. Its first incarnation, launched in January 2004, was the original and slightly different-looking Cosecant, a zero-oversampling D/A converter—it lacked a digital filter, intentionally—with passive current-to-voltage conversion, a tube output stage, and an outboard power supply.

Although the original Cosecant was among the most-talked-about high-end products at the 2004 and subsequent Consumer Electronics Shows, it seems that a no-less-significant advancement was in the works. By early 2007, Gordon Rankin had become increasingly frustrated that outboard computer-audio devices were forced to adapt to the digital transfer rate established by the computer itself: conditions that were fertile ground for high levels of word-clock jitter. Given the availability of a new and unprecedentedly powerful USB controller chip, the Texas Instruments TAS1020, Rankin devoted the next year to writing his own controller software, so that his USB devices could establish their own transfer rates. The resulting asynchronous-mode control program allowed for a single master clock in the D/A converter, without the need to synchronize with the computer's own clocks (footnote 1).

Audio hobbyists—myself included—who consider single-box CD players superior to CD separates in terms of jitter performance (footnote 2) may recognize in Wavelength's asynchronous-mode controller the potential for a whole new wave of field-leveling.

The first commercial product to use Rankin's asynchronous-mode controller software, now trademarked Streamlength, was the upmarket Wavelength Crimson D/A, introduced at the 2008 CES. The Crimson was also Rankin's first modular digital design: Its architecture permits easy updating of software and hardware. In fact, it was the success of the asynchronous-mode, modular-design Crimson that prompted Gordon Rankin to redo the Cosecant in similar fashion—especially in light of customer requests for a Cosecant that could handle 24-bit/96kHz input data, which is beyond the abilities of Wavelength's original zero-oversampling D/A.

Thus the Cosecant v3 is available with a choice of two different converters: the Transcendental D/A board, which is the modular version of that zero-oversampling converter; and Wavelength's new Numerator board, which uses the Scottish-made Wolfson WM8716 chip to support word lengths of up to 24 bits. My review sample was supplied with the Numerator.

The Wolfson D/A chip incorporates selectable digital filtering; the slow rolloff characteristic that Rankin prefers is selected during initialization, by a Microchip 8-bit microcontroller, which then goes into sleep mode during actual music playback. The Numerator board also plays host to a Microchip 64-kB EEPROM, which is where Wavelength's proprietary USB controller program resides.

The TAS1020 controller chip is installed on the Cosecant v3's motherboard, alongside certain other notable bits: hefty low-noise regulators, a series of hand-selected tantalum resistors for current-to-voltage conversion, and—of course!—a vacuum tube. That last is an Amperex 6GM8/ECC86 dual-triode voltage amplifier, operated in parallel-feed mode with the aid of custom-wound transformers dedicated to that purpose—which also keep the output impedance to a low 500 ohms.

As I've come to expect from Wavelength Audio, the Cosecant's build quality is superb. The physical arrangement of motherboard and user-replaceable subboard made me think, fondly, of Naim Audio preamplifiers of years past, as did the clean, logical parts layout and the neat wiring. The chassis itself is a sturdy, smoothly finished box with a burled wood faceplate and a brushed-aluminum top plate. An additional, half-size plate of Plexiglas sits proud of the aluminum, with a hole just large enough for that 6GM8. (Removing the tube is tricky but not impossible.) Threaded into the bottom plate are three nice-looking carbon-fiber isolation cones, two of which need to be removed to access the screws that hold the bottom in place, in order to change DAC modules.

---

Footnote 1: Actually, the TAS1020 did not, in and of itself, signal the switch from adaptive to asynchronous modes. When Rankin first started using that controller chip, his DACs were still adaptive, though he did use the TAS1020's expanded memory buffer in an effort to reduce jitter.—Art Dudley

Footnote 2: One could argue that the multibox digital front end from dCS that I wrote about in Listener in November 2002, which offers superb measured jitter performance, is proof of the wrongness of that point of view. But I would argue that the dCS system, with its decidedly nonstandard clock connection, is in fact a two-piece one-box player, because its performance cannot be similarly maximized when mixed up and matched up with separates from other manufacturers.—Art Dudley