I am thinking of two single-axis accelerometers, one sensitive to vertical motion and the second one inverted. Then connect the two via an instrumentation amplifier.

 

Here are some more research papers:

https://www.iris.edu/hq/instrumentation_meeting/files/pdfs/MEMS_Seismology.pdf

http://www.sercel.com/products/List...EMS-basedAccelerometer_article_Sercel pdf.pdf

A Low-Noise DC Seismic Accelerometer Based on a Combination of MET/MEMS Sensors

 

But, using multiple sensors would allow cleaning away a lot of the noise. Only a signal appearing on all sensors would pass.

Think about the means of realizing such a function at the microvolt level and doing it inexpensively. Possibly with some MEMS devices and custom circuitry, but that would be neither cheap nor that quiet. Low noise is seldom low priced.

 

But, using multiple sensors would allow cleaning away a lot of the noise. Only a signal appearing on all sensors would pass.

Hola @wayneh

Could you please elaborate a little? Any block or simple circuit to grasp the idea?

 

We were trying to build a seismometer using only an affordable +-1g accelerometer. We were trying to test if its possible to use an accelerometer as alternative sensor for a seismometer.

Depends on what you mean by "affordable": under $10,000? Under $100? Assuming the latter, even a premium MEMS accelerometer such as the ADXL354 (≈ $52 at Digi-Key) will not give you the performance you're after due to excessive output noise.

The data sheet gives the ADXL354's sensitivity as 400 mV/g typical, and its output noise density as 20 μg/√Hz. Over a 100 Hz bandwidth, per your top post, this translates to an output voltage noise level of

En = 0.4 V/g * 20 μg/√Hz * √100Hz = 80 μV rms,​

or about half a millivolt peak-to-peak. So it makes little sense to speak of nanovolt-level resolution in your amplifier/digitizer.

Looks to me like this is a classical "you can't get there from here" situation; "affordable" and "extreme low-noise performance" simply do not go hand-in-hand.

 

Hola @wayneh

Could you please elaborate a little? Any block or simple circuit to grasp the idea?

I read some years ago that geologists had figured out how to use networked computers (which contain accelerometers) as seismographs. No one laptop was useful alone but if 100 of them all saw a tiny blip, it was obviously a signal and not noise.

 

I read some years ago that geologists had figured out how to use networked computers (which contain accelerometers) as seismographs. No one laptop was useful alone but if 100 of them all saw a tiny blip, it was obviously a signal and not noise.

I read that article also.. But there were several flaws, starting with the application using up folks data quota reporting every little bump. But it might be able to work, at least in theory. In reality though, data latency and system overloads from reporting an actual event, in addition to that pesky data sending charge, prevented it from advancing. Besides, would you really want the program constantly sending your location to some unknown group??

 

I am still thinking that a mechanical resonant system to amplify the small earth motion is a big part of making it work. You need to examine the mechanical works of a recording seismograph to learn how it does the mechanical amplification.

 

I read that article also.. But there were several flaws, starting with the application using up folks data quota reporting every little bump. But it might be able to work, at least in theory. In reality though, data latency and system overloads from reporting an actual event, in addition to that pesky data sending charge, prevented it from advancing. Besides, would you really want the program constantly sending your location to some unknown group??

My point was just that using multiple sensors can help sort out signal from noise. I think 3 local sensors would be much better than one. But nanovolts are still the needles in the haystacks.

 

Noise analysis for starters of transducer so that you know
where the problem areas are.

CDS (Correlated Double Sampling) can get rid of offsets and 1/f.
Single digit nanovolt performance might/would be challenging.

http://www.cypress.com/documentatio...-5lp-correlated-double-sampling-reduce-offset

https://e-lab.github.io/data/papers/TBioCAS-lowcurrnoise-2012.pdf

https://mediatum.ub.tum.de/doc/601628/601628.pdf


Regards, Dana.