The tighter bit density of 40-Gb/s signals makes them more vulnerable to chromatic and polarization mode dispersion as well as nonlinear optical effects. But other than laying more fiber – a solution with problems of its own – the only alternative to increasing bit rates is increasing the number of channels a fiber can carry. Researchers at Essex Corp. are reporting a prototypical hyperfine wavelength division multiplexing (WDM) device that can deliver 1600 channels, each separated by 2.5 GHz. At that channel density, Essex's technology leapfrogs past the industry's current aspirations to fit 160 wavelengths on a fiber. But Terry Turpin, chief technology officer for the company, claims Essex's prototype only scratches the surface. "The technology came out of an attempt to build a radar receiver capable of delivering 16 GHz of bandwidth divided into parallel channels separated by 50-MHz resolution." Turpin said. "Right now, we're aiming for 25-GHz channel spacing because everyone seems to be going there, but we also make a 5-GHz device for OC-48, OC-12 and Gigabit Ethernet applications." In any case, if proved in trials, Essex's multiplexer could circumvent single-channel OC-768 transmission by delivering more channels at less technologically challenged OC-192 and OC-48 protocols in the same bandwidth. "Conventional dense WDM requires pulse shaping to do 40 Gb/s, which makes sending OC-768 a real pain," Turpin said. "But if you take our 5-GHz widget, one device delivering OC-48 on each channel will give you the same bandwidth as OC-768 on a fiber." Until the patent is awarded, Turpin can provide little detail as to how the passive module works. It is analogous to a Bragg cell, he said, which diffracts light using acousto-optic modulation. Instead of an acoustic grating, however, an electro-optic modulator applies signals to the light beam, and a passive optical device separates the modulation frequencies spatially. There are trade-offs in the prototype, most visibly a 2-dB insertion loss that, by itself, isn't so bad but that could eat up a network's loss budget quickly if several modules were used. Besides reducing loss, the task is to produce a prototype by this summer to allow users to characterize it for their application and to optimize the technology. Essex can ramp up production quickly, Turpin said, and the module may be a commercial product by early next year. The company is marketing it for applications in access networks typified by surges in demand and diverse protocols, but it also has potential as an add/drop multiplexer in metropolitan systems or as a capacity booster on long-haul networks. "What it allows is small business or home data rates, and it allows those data rates without the overhead of multiple levels of SONET hierarchy," Turpin said.