Kavehrad and Fadlullah built the experimental system using a low-power infrared laser to prevent possible damage to the eyes or skin. They focused the light through a lens, creating an elliptical spot on the ceiling; they then used a high-sensitivity light detector called an avalanche photodiode to pick up the light reflected off the ceiling. They used a plastic holographic lens to collect enough reflected light from the ceiling spot and focus it on to the photodiode's active area. By using the lens, Fadlullah and Kavehrad could transmit a one-gigabit-per-second optical signal across a room about eight meters long by four meters wide.

Free-space optical networks have previously been used to transmit broadband data over long distances, but the high power of the lasers and need for a clear line of sight and extremely precise alignment between the transmitter and receiver have limited their usefulness. The low-power, diffuse light approach that Kavehrad and Fadlullah chose doesn't require such precise alignment and is much more practical for indoor communications. Kavehrad says that their system should work for visible and ultraviolet light as well as infrared.

Companies such as Intel, InterDigital, Siemens, Sony, Samsung, Mitsubishi, and Sanyo are all pursuing research on optical wireless networks, say Kavehrad and Xu. Several of these companies are members of the Infrared Data Association (IrDA), an industry organization that is developing technical standards for infrared wireless communications. IrDA recently announced the GigaIR standard for very short range, line-of-sight infrared communication links operating at one gigabit per second. And the IEEE 802.15 working group, which sets standards for wireless personal area networks, is working to create standards for wireless networks that use visible light, says Fadlullah.

Kavehrad says that "a lot of engineering has to happen" before optical wireless networks are a reality. He and Fadlullah used lasers, transmitters, and receivers not designed for communications in their experimental system; all of that equipment must be optimized for data networking. However, Kavehrad says, if development of white LEDs for indoor lighting continues at its present pace, it should be possible to have practical wireless optical networks within three years. "The main limiting factors will be industries and their politics, as well as consumer demand," he says.