You are correct about the change in the force rebalance design at Analog Devices after the original XL50. All subsequent designs have been open loop. The major design issue on the XL50 was the uncertainty of the polysilicon spring constant over time. There just was no good data. So the servo system gain was 10X the spring constant's force thus reducing the effect of any changes in the polysilicon spring over life by a factor of 10. After many years and millions of devices it became clear that polysilicon was stable and therefore the scale factor of the accelerometers would not change. It is easier to build an open loop device since first of all we did not need to put a high value resistor on the chip to feedback to the input. The performance of the accelerometer did not need to be that high therefore we were not concerned about excessive displacement of the proof mass or nonlinearity so open loop was a path to a smaller die and therefore less cost. It resulted in reducing the die size by 50%. If you make stiffer springs you get less signal to work with and the errors show up as a larger percent of the signal. We also have a self-test on all of our parts and if you stiffen the spring it you have a hard time getting enough force to move the mass electrostatically. It's not to hard for a low-g device of a fee g's full scale but we make parts up to 125g's and like to have significant movement to insure good diagnostics. Hope this helps. Bob Sulouff Analog Devices -----Original Message----- From: Yong Xu [mailto:yong@mems.caltech.edu] Sent: Friday, November 15, 2002 4:03 PM To: mems-talk@memsnet.org Subject: [mems-talk] a question about force-balanced sensing Force-balanced sensing is used in capacitive accelerometers to improve the linearity. the whole point is to used negative feedback (e.g., electrostatic force) to decrease the displacement of the proof mass. My question is: why not simply increase the spring constant? e.g., by making stiffer supporting beams of the proof mass? My guess is that the spring constant is not easy to control so there might be nonuniformity problem. But this argument does not convince me 100%. the force-balanced sensing was employed in ADXL50. but I didn't find it in ADXL150's data sheet. did AD abandon force-balanced sensing? any input will be highly appreciated. Yong _______________________________________________ MEMS-talk@memsnet.org mailing list: to unsubscribe or change your list options, visit http://www.memsnet.org/mailman/listinfo/mems-talk Hosted by the MEMS Exchange, providers of MEMS processing services. Visit us at http://www.memsnet.org/