Whole lotta shakin’ goin’ on!

I live in Northern California, about halfway between the San Andreas and Hayward faults, in an earthquake-prone area.  I’m also fascinated by science and computers. So when I saw the http://raspberryshake.org site, where you could get your own personal seismometer station and network it together with others to form a net capable of locating and measuring earthquakes, I was very interested.  I have no training in seismology, but this was too interesting to pass up.  It wasn’t long before I placed my order.

I got a Raspberry Shake 4D, and I have installed it in my garage.  My garage has some crude plywood cabinets next to the walls.  I cut out a hole in the bottom of one of the cabinets, in order to bolt the seismometer directly to the concrete slab of my garage, for the lowest noise and best sensitivity.  The hole, and the unit itself, are at a bit of an angle to the wall of my garage, because this seismometer measures lateral motion, and is designed to be properly oriented toward true north.  The walls of my house are oriented several degrees east of true north.

The big black box partially shown in the lower left corner foreground of the photo is a 35 amp-hour 12V lead-acid battery, acting as a backup power supply in case of utility power failure.  It should be capable of running the unit for several days at least.  The small box in the right foreground is a Powerwerx USB Buddy, a DC-DC switching converter that will take 10-32VDC in, and produce 5VDC out at up to 3 amps, which goes directly to the Raspberry Shake’s input connector.  I supply power to the USB Buddy switching converter using either the battery or an old spare laptop adapter I had hanging around (not shown in this photo).  The laptop adapter takes AC mains power and produces 20VDC output.  The laptop adapter and the 12V battery are both simultaneously connected to the switching converter via diodes, so that whichever source has higher voltage will power the Raspberry Shake.  I can disconnect one or the other at any time, leaving the Raspberry Shake running uninterrupted.  I’ve had a similar system running some Raspberry Pi units I use for security cameras, and it has kept them running uninterrupted for years.

Silly me, I thought that, by anchoring the seismometer to the concrete slab of my garage foundation, it would be very steady, and not influenced much by people walking around in our wooden house.  I underestimated how sensitive the seismometer is!  It detects people walking on the sidewalk 25 feet in front of our house.  It detects cars driving by.  It detects people walking inside our house.  It detects the washing machine spin cycle.  I suppose I shouldn’t have been so surprised — it detects distant earthquakes that are very much smaller than I can feel, and it is simply accurately showing me that nearby footsteps cause disturbances that are comparable in magnitude to the earthquakes it measures.  I think my seismometer is placed about as well as it can be placed in my house, but nevertheless, it senses a lot of random noise due to everyday human activities.  It might be a little quieter if it were buried in the back yard, but my neighbors are close enough that it would pick up noise from their houses as well as my own.  It’s a lot quieter at night.

But to put this in perspective, here’s a chart recording made during the fifth day I had my seismometer.  It has one trace across the chart for every 15 minutes.  The chart covers 12 hours, starting a bit more than an hour before most of us get up at my house.

You’ll notice the top few lines are fairly quiet, when we and our neighbors were sleeping.  Then most of the rest of the chart has various blips on it, some possibly representing tiny earthquakes in the area of The Geysers, near Calistoga California, about 90 km away from my location, but most representing random noise from the activities of daily living in my home.

But there’s one big black squiggle, starting just after 16:00, and continuing for the better part of 15 minutes.  What was that?  A 7.8 magnitude earthquake across the Pacific near Fiji, about 8700 km away from my location!  It took about 11 minutes for the first waves to arrive across the ocean.

None of the squiggles on this chart represent earth motion that was perceptible to any humans at or near my house.  Not even the big squiggle representing the distant Fiji earthquake was perceptible here.  So it gives you an idea of how sensitive the detector is, that an imperceptible distant quake is so much larger than the background noise caused by the daily activities of the occupants of my house.

The unit I purchased is the model 4D.  Four dimensions?  Well, sort-of.  It has four channels of data.  Three are using MEMS sensors, tiny microelectronic accelerometers, not that much different than what are in fitness watches or smartphones.  The MEMS sensors are relatively cheap, and capable of measuring large accelerations well.  The fourth channel is a vertical geophone, which is an incredibly sensitive way of measuring the up/down motion of the ground.  The geophone is more expensive than the MEMS sensors, and it also has a more limited range.  It “saturates” with strong motions.  With this combination, I’ll get three full channels of motion (north/south, east/west, and up/down) for a nearby strong earthquake, but only one channel (up/down) for a small tremor.  I’m not sure exactly where the dividing line is.  So far, I haven’t seen motion sufficient to produce a noticeable signal on the MEMS sensors.  I also haven’t felt any earthquakes, but my geophone has picked up a few each day.

This seismometer is the motivation for setting up the stratum-1 GPS-based NTP server in my house.  This seismometer feeds data into a network of similar units.  Having the data accurately time stamped is important, because the varying times of arrival at varying locations are used to determine the locations of earthquakes.  That said, a local GPS is probably overkill.  The seismometer takes 100 samples per second, in other words, one sample every ten milliseconds.  My computer’s clock is now maintained with much better precision than that.  And there are uncertainties of the local ground conditions that cause timing errors much larger than the errors in my computer’s clock.  But still, having precision a couple of orders of magnitude better than strictly required never hurt anyone, did it?