An eye toward the future

Close your eyes and imagine building electronic products without PCBs. Instead of connecting a device's micron-sized transistors with gargantuan bond wires and copper traces, the edge of one chip would touch its neighbor, transmitting signals directly by capacitive coupling. Eliminating the wires would increase circuit speeds dramatically (perhaps two orders of magnitude). Systems would shrink. Power consumption would drop.

Sound farfetched? At the Custom Integrated Circuits Conference in San Jose in late September, Sun Microsystems announced just such a project. Dubbed "proximity communication," Sun's patented technique will permit packing hundreds of chips in a face-to-face checkerboard arrangement much denser than current designs. Developed as part of a supercomputer project for DARPA, this new chip design promises to dramatically change the way the industry makes everything from handheld devices to communication satellites.

Two chips positioned for proximity communication by capacitive coupling rather than large off-chip wires. Courtesy of Sun Microsystems.

Of course, no good deed ever goes unpunished. Such a breakthrough will bring its own set of challenges. Chief among these will be the need to line up the tiny contacts in adjacent devices almost perfectly. Misalignment would compromise signal strength by reducing coupling area and increasing distance. Crosstalk would increase as well because voltage swings on adjacent transmitter and receiver pads would couple onto the receiver pad. Although offering few specifics, the Sun researchers propose solving this potential problem with an on-chip measurement system.

Heat dissipation becomes another issue, because of increased circuit density. Also, without bond wires and PCB traces, the geometry of adjacent devices in a circuit would have to match, so initial applications might primarily involve ASICs and other custom designs.

Naturally, someone will have to test these products to be sure they work. Conventional bed-of-nails access would be virtually nonexistent. Inspection—even x-ray inspection—would have difficulty identifying misalignments and other relatively subtle failures. Circuit complexity would present new challenges to functional test. Solutions would likely take the form of built-in self-test (BIST) and comprehensive defect prevention. At the same time, because circuit construction would not require soldering devices together, a faulty device could simply be removed from the failing circuit and a good device dropped into its place.

Proximity communication is currently in the experimental stage, with real products probably more than 5 years away. Still, DARPA rarely backs pipe dreams, and any innovation that could so fundamentally change the nature of electronic manufacturing deserves careful watching.

Steve Scheiber, Contributing Technical Editor, sscheiber@aol.com