NXP targets timing applications with MEMS based frequency generator

The approach replaces a quartz crystal with a bare silicon die that is more than 20 times smaller than the smallest crystal available today. The device does not require dedicated packaging and can be integrated with other parts into a low cost plastic package. Joost van Beeck, programme manager for NXP's MEMS timing devices, said: "Quartz crystals are very difficult to integrate for several reasons and are very different to silicon. The MEMS approach brings four benefits: smaller size; less expensive; scalability; and the ability to integrate them into other devices." NXP has developed a manufacturing process which uses standard cmos technology, except for one stage. "This requires a special MEMS tool," van Beeck noted, "which is readily available." The resonating element is contained in a cavity which is at less than atmospheric pressure; a move which reduces drag. This, in turn, brings a high Q factor. "Because we also use monocrystalline silicon," he continued, "the Q factor is more than 40,000 at frequencies in excess of 50MHz." However, monocrystalline silicon on its own has a high temperature drift. "It's around 30ppm/°C," van Beeck pointed out, "which is too much for many applications. By engineering the silicon, we have reduced this by a factor of 10 to produce a stable oscillator with respect to temperature." The company is targeting manufacture at 8in wafers and expects to produce 100,000 parts per wafer. "That means they will be cheaper to produce," van Beeck continued. "And standard back end processes can be used." While it is theoretically possible to integrate the device monolithically, NXP is following an integration path. "As a demonstrator," said van Beeck, "we have glued a resonator on top of an asic and connected it using standard wire bonding."