Earthquake Detection On A Chip

If you’ve ever been in an earthquake you’d assume it would be pretty easy to detect one. If things are shaking, there’s an earthquake. In reality, though, a lot of things can shake a sensitive instrument that is detecting shaking, so — for example — mechanical sensors will produce a lot of false positives. Now, however, you can filter out errant vibrations and reliably detect earthquakes on a chip.

The Rohm BP3901 has two primary features. First, it supposedly eliminates false detections due to things like a heavy truck rumbling by. In addition, while most sensors must be mounted completely flat, the BP3901 has a compensation method for angle which lets you mount it as much as 15 degrees rotated in either direction and still get good results. That’s because the BP3901 is based on the combination of an accelerometer and a microcontroller in one package to detect movement, characterize it based on an algorithm and reacting through an I2C bus and an INT pin.

Rohm suggests you could power the BP3901 for about 5 years with two AA batteries with the example of averaging 10 three-minute wake up events a month. We aren’t sure why we want to detect an earthquake, but we think we do. Imagine a large sensor network sending back real-time data as an earthquake happens — something we saw last year using Raspberry Pi. That project used a Geophone as the detector, which could be replaced by this chip. Rohm plans to have “OEM quantities” for sale next month which we hope means we can get smaller quantities from distributors.

A lot of people spend a lot of time thinking about how to predict earthquakes, as we’ve seen before. Of interest, the ancient Romans may have had a way to deflect earthquakes, so they probably didn’t care as much about detecting them.

14 thoughts on “Earthquake Detection On A Chip

  1. Unless I miss something about this hackvertisement, all that’s special about this “chip” is the intellectual property locked away in it, which allows a bunch of normal sensors to produce a more useful result.

    I can imagine zero reason for the Hackaday community to care except for reverse-engineering the thing to produce an open-source pure-software implementation that’ll run with whatever sensors you happen to have sitting around.

    1. How much do you want to bet this “algorithm” simply measures the duration like the image implies? Turns out a lot of trains look like earthquakes! ;-)

      1. Trains probably have a periodic property based on wheel spacing over track imperfections… I wonder if they detect that?

        Based on SI units doesn’t tell me much. Most algorithms dealing with sensors probably have an SI unit or two at some point even if it’s just voltage from an adc.

        1. Given the obvious puffery with this marketing ploy, my bet is that they haven’t bothered to install and calibrate an FFT filter looking for those (or any other) vehicular periodicity. I’m sure the detection from a truck-filled bridge is quite amusing.

        2. They say their algorithm characterizes Spectral Intensity aka they apply FFT and match the spectrum to known earthquake patterns. And just like regular seismometers aren’t usually placed right beside train tracks, busy streets or heavy industry, you shouldn’t do it with this sensor if you plan to detect earthquakes. It should be relatively easy to replicate their algorithm based on, but I have no clue how you would simulate earthquakes to test the whole system.

    2. What would be more interesting is talking to the seismologists at Cal Tech and the USGS who build instrumentation to measure earthquakes and find out what works and doesn’t from them.

  2. Is not something new, the D7s from omron does the same.
    Very useful not to predict or send an alarm before an earthquake but to assess if the building was submitted to to much stress and if it’s safe to stay in it.
    Or to automatically cut gas pipe or other dangerous process.

  3. Well I would want a few of those. I live in Groningen, in the Netherlands and we have lots of (small) earth quakes.
    But as with all earth quakes, there is a relation between the number of small quakes and the number of stronger ones.
    For example every 1000 quakes with up-to 1 on the Richter scale will give 100 up-to 2 on the Richter scale and 10 up-to 3 and so on…
    The quakes here are induced by the decades of gas pumped out of the ground and their origin is at 3 km depth.
    So from about 1.5 on the Richter scale a human can feel them here.
    We had recently another one of about 3.4 on the Richter scale which resulted in about 4000 – 5000 damage reports of houses that got damaged. (on top of the 10’s of thousands of reports which have not yet been handles)

    So I wonder if these sensors can also detect the quite different quakes we have here, which are “surface quakes”.

    Our houses are going to be taken down and rebuilt (the whole street) and the new built houses will have base isolation to prevent new damages after light quakes. These base isolation pendulums should be able to slide up-to 125 mm in each direction, so I also plan to install some sensors to measure those. As soon as I have them, I will also write a blog post about them and link it here :)

  4. New earthquake detector? Really? From the graph, the serious shaking portion of the earthquake detected lasted about 40 seconds. I lived through the Loma Prieta earthquake in California. It lasted 15 seconds and was magnitude 6.9 with $6 billion in damage. For an earthquake lasting 40 seconds, a good detector is your house falling down.

  5. To be fair, it may be common sensors with a unique algorithm, but I think the 3.5uA standby current could be difficult to replicate from common parts. That alone is worth it in my book.

  6. I’d like to believe there are a lot of us visiting hackaday, are unlike the conservative to the point beyond being useful, commenters. Over the years publications that cater to the DYI communities have always feature new components, and tool they thought, customers and reader would find useful. I’d go as far to say, it was something many of the paying customers expected. In the event if hackaday is counting, I for one, appreciate posts as this.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.