All the ST inertial sensors seem to have crappy data sheets from my point of view. They're very detailed about the stuff of concern to an embedded software guy, but their electrical specifications seem pedestrian, and the specifications on their inertial sensors are downright vague, and missing some critical bits if you want to do serious measurements.
I assume that ST does this to satisfy the people who are going to complain the loudest, or because they aren't paying their datasheet writer very much.
Here's their L3DG20 data sheet:
In Figure 5 they call out a 10uF cap and a 100nF cap, both to be placed as closely as possible to the chip "common design practice" they say. In other inertial sensor data sheets they call out a 10uF aluminum cap, with the other wording the same.
Can anyone see why this should be necessary? My faith in the data sheet having been reduced by the vague specifications for the inertial sensor itself, I'm wondering if the writer was told about bulk board decoupling and individual chip decoupling and misunderstood.
I'm about ready to send files off to the board house. I've pretty much decided that since I've got a 100nF decoupling cap on the part in an 0603 package I can just get the damned boards, and if I have problems I can stack an 0603 10uF ceramic on top of the 100nF.
But if such extreme decoupling measures have, indeed, become common practice while I was napping, I'd love to hear about it.