I believe that my 6 accelerometers now have the sensitivity of seismometers that I have worked with. All you really want is accuracy at some fraction of the background noise, perhaps a factor of 5. I am pleasantly surprised at the power of a Raspberry 3. It has handled everything I've thrown at it.
My next step will be to plot an accurate velocity graph. For that I need to maintain a moving average before integration. Otherwise the velocity tends to drift. I'm going to be using the mathematics library 'numpy' for that. Then I'm going to stream out the results to the main computer. I hope the raspi3 can handle all that.
In the future, I'll have a bunch of raspi zeroes, each with 2 or 4 accelerometers. There should be enough power to pre-process the accelerations to velocities. The raw accelerations are very spiky and mostly meaningless. Integrating to velocity brings us within the realm of physics. Those poor engineers who stick with accelerations are sad. :)
It is my wish that everybody sticks one of these on their buildings. But it won't happen, because no trumpy will ever admit that nobody wiretapped the building. :(
ps. signal to noise is very important. For the seismometer network, we had very expensive seismometers and had a hard time finding places where the noise was low enough. But these seismometers had the best dynamic range for strong ground motion. I think my new seismometers would work just as well, and if not, just add more accelerometers. :) This is like a factor of 1000 cheaper.