Friday, February 19, 2016

Linux - Mcube's Amazing Accelerometers

I am trying to build a cheap and sensitive accelerometer using the Raspberry Pi2, moveable over to the Zero, if it ever comes available.  You need extreme sensitivity if you want to calibrate these things with distant earthquakes.  You also need a lot if you want to have a picture of the earthquake.  I don't think 'early warning' will ever work if you put some weight on false alarms, which nobody seems to be doing right now.

Accelerometers used to be monstrous things, with big weights and coils.  I fooled with them enough, and they were horrible.  Then they came up with MEMS or micro-machined accelerometers.  This involved a complicated arrangement of putting down layers on silicon and etching away little bridges with acid.  People provided 'custom etchings' for a high price, but these had a large mass and were very good with low noise.

Then they came out with 'chip' versions which are in every phone.  The one I'm working with now is the 5 year old adxl345, with only 10 bit sensitivity at 2g.  I'm trying to sum two of them for greater sensitivity.  The physics is against these small things because the proof mass is so small and subject to thermal noise.  I can see the noise in the lower bits.  I don't think this will work that well.

It's about time for a new revolution, and maybe mcube is it.  They claim a different process and low noise at 14 bits for 2 g.  Whoopee!  This is seismometer grade.  Eventually you get to Earth Noise, and it's no good going any lower.  The type of installation I am considering will have problems with the washing machine, and heavy snoring.  :)

I have ordered a 3610 breakout.  It is 2x2mm, surface mount, so you need something that can go on a  breadboard.  The super-new 3575 is the same process (I think) at 1x1mm.  But I am wondering about the physics.  Computer chips get faster as they get smaller, I'm hoping that there is some physics that no longer requires a heavy proof mass to deal with thermal noise.  Perhaps a suspended quantum dot?

This is exciting, and I really hope that other people do it as well, since I am soooo slow.

ps.  I just summed 2 adxl345's (the maximum together) and it is really good.  Gives me another bit of sensitivity.  I can't wait to see what 4 bits will do.

No comments: