Loudspeaker designs have long wandered between the technically exotic to the visually exciting and back again, from the sinuous speakers perched atop early Atwater Kents, to the remarkable B &W Nautilus units that were on our February '98 cover. However they appear, all have shared a common element at their heart -- cone-shaped drivers. Originally introduced at Comdex '98, Benwin speakers replace conventional speaker cones with 0.25 mm flat panels powered by a 12V exciter to generate 250 to 20,000 Hz reproduction. Benwin., City of Industry, CA, says that their NXT flat panel technology enables the units to project sound in a full 360-degree radius. There are no 'hotspots,' so the sound permeates the room. While most of the components are proprietary, Benwin indicates that panel material in the model recommended for computer/multimedia applications is similar to PC board material found in consumer-grade computers. The exciter is a four-layer voice coil that uses neodymium magnets, and is directly epoxied to the center of the flat panel. Overall, the finished thickness of the panel speaker is about 7mm (1/8 in.). A separate, compact subwoofer unit fills in the 20 to 250 Hz range. Benwin is producing NXT-based speakers for computers, small audio systems and multimedia setups.

As sensors become easier and less expensive to manufacture, they have grown to be near-ubiquitous in products in consumer, medical and industrial arenas. Unfortunately for these devices and the processes they support, life is not so inexpensive when a sensor fails. SmartSignal, a start-up software company in Lisle, IL, is offering a microprocessor-based detection engine that will provide the earliest possible warning of problems in equipment and industrial processes, and simultaneously increase instrument sensitivity. Originally developed at Argonne National Laboratory for nuclear power applications, SmartSignal technology gathers information from the sensor signals in a piece of equipment, then applies signal analysis and fault detection algorithms to detect very subtle changes, subtle enough to be within the "normal" bandwidth of the monitored signal. A modeling tool creates an empirical model in seconds from a set of "good" historical data, and that is overlaid with live data to create virtual signals of a normal system. These virtual signals are then compared to real-time signals, and abnormalities are picked out with very fine resolution. When sensors fail or lose their calibration, the estimated sensor signal can be used as a virtual signal to permit continued operation of systems. The basic software and models occupy as little as 100K of RAM.