Anablog

Had a meeting with Bob Frostholm, VP at Microbridge Technologies a few weeks ago. They make an adjustable resistor they have dubbed rejustors. They suspend a polysilicon resistor on an under-etched tiny bridge (hence the Microbridge name). They run another resistor along side the first resistor that is used to heat the first resistor. This is why the whole shebang is suspended above the silicon die—to provide a huge thermal resistance so the heater resistor can get the polysilicon resistor hot enough to anneal it. This permanently changes the resistance value. Yes, you can heat it differently to change the resistance the other way as well.

There are a couple very neat things about this. The resistance change is continuous. You can get 0.1% accuracy as opposed to setting bits or laser or zener zapping that just gets you close but is not a continuous change like laser trimming. The adjustment can come after you have soldered the chip onto the board so any stresses in the mounting will not be a factor. They work at 125 C and will function well up to 200, unlike digital pots. Better yet, they can function well at high frequencies; a GHz is not a problem. It has all the benefits of a plain dumb resistor only it can be trimmed to tight tolerances.

OK, now the bad side (it is analog after all). Because the rejustor is mounted on the thermally high resistance microbridge, the resistors cannot take a lot of power—0.5 or 1 mW. Above that and you might change the resistance. Since power is V²/R, that V² term means you will need a lot of R to keep under a mW. A 5-volt circuit would need a 25k resistor. This is not too bad, but it will add some noise to a delicate input. The other headache is the time to adjust them— about a second. If you are doing high-volume manufacturing with a very expensive test setup, well, waiting around a second to adjust a resistor may not be the best use of your capital expenditures.

Unlike a lot of MEMS-based start-ups, Microbridge can actually send you parts that are in production today. For instance, the MBD-333-AL (pdf) has two equal-value rejustors in a 3x3mm QFN and the resistance can be set from 23k to 33 kohms. They have parts with 3:1 and 5:1 ratios as well. In addition they welcome any semiconductor manufacturers that would like to discuss using the technology. Bob tells me that doing the under-etch is pretty basic in CMOS process flows and should not add very much cost. It is a well-understood process and many fabs can do it now. Jingle up Bob and he can tell you all about rejustors. They may be a lab curiosity or they may be the next big thing. Here is a rejustor whitepaper  and here is a rejustor primer.