Reducing accelerometer noise using sensor array?

[mbertoni]

Posted on July 27, 2017 at 18:53

Hi to all, I was taking a look to LIS344ALH accelerometer for a project where I need to place MEMS accelerometer into Raspberry board.

I've heard there is a way to reduce acceleration noise placing sensors in array (e.g. using 4 LIS344ALH instead single one).

Is this correct? How does it work?

Thanks.

[Legacy member]

Posted on July 28, 2017 at 07:33

Hello,

What is your target application? What is your desired sensitivity and maximum acceptable noise?

If you search for 'acceleration array' on Google, this IEEE article comes up: 

http://ieeexplore.ieee.org/document/7456267/

 , but it describes quite a specific application. However, the article is only accessible for members with a subscription.

I don't think this you get a significantly better results if you place the sensors on the same board.

David

[Miroslav BATEK]

Posted on July 28, 2017 at 09:43

As David said, it would be good to know requirements of your target application in terms of noise, bandwidth, current consumption, full scale range, ...

In case you would prefer acceleromter with digital output please take a look at

http://www.st.com/content/st_com/en/products/mems-and-sensors/accelerometers/lis3dhh.html

 which is modern low noise accelerometer.

[mbertoni]

Posted on July 28, 2017 at 10:41

My target application is accelerometer embedded on Raspberry Pi (or similar) SBC board, fullscale +/- 1g, 16 or 18 bit of resolution,  banwidth 0-100Hz.

I used digital MEMS accelerometer in a test, compared with professional seismometers (noise about hundred of ng/√Hz).

Of course I don't mean to pretend same noise for MEMS and professional seismometer.

Using a professional accelerometer I got acceleration values follow normal distribution law (let's suppose sigma=0,25), while MEMS give me sigma=0,8 (I don't remember exact values) so I can see MEMS is much more noisy.

I've seen using 4 MEMS accelerometers (on same board) to decrease noise; ok, LIS344ALH has low noise nominal value of 50 ug/√Hz, but if I can get lower noise placing 4 accelerometers it would be better.

I'm making simulations on Matlab: adding 4 noise signals (sigma=1,0) I got as result output noise with sigma value=0,5.

I don't know if this is correct way or if theris something better.....

I'm taking analog accelerometer so I can decide in a second time which ADC to use.

Thanks.