ADXL103: Noise in accelerometer

Q 

We are currently using some of your accelerometers to measure subsea pipe
motions for offshore applications (extremely small motion amplitude ~ 0.01 m
and reasonably low frequency 0.03 Hz). I have a few questions regarding your
accelerometer's specifications, and would be most grateful if you could respond
to my queries as soon as possible.

Take ADXL105 for example,

Resolution 2 mg - what does this really mean? Is it possible to measure
accelerations smaller than this?

Noise floor is 225 ug (Hz)^0.5? How is this defined and obtained? We measured
the noise of the accelerometer in time domain when it is stationary and worked
out the amplitude spectrum using DFFT. Is the noise floor defined as the
magnitude of this white-noise spectrum.

In your technical note (Rev.A), this is an expression for calculating rms noise
from the noise floor level.

What is then the relationship between resolution and rms noise?

 

A 

A resolution of 2mg means that, given the noise level of 225ug/rtHz, you can
realistically reduce the noise by filtering to such an extent that you can
determine the acceleration to 2mg. The p-p acceleration noise is about 6.6x the
rms value (this gives a statistical probability that the signal is within this
range for 99.9% of the time. Therefore if you filter to 1Hz with a single pole
filter you have 225ug x 1.6 = 360ug rms noise which is 2.376mg peak-to-peak
noise. 2mg is an approximation of the resolution you can reasonably achieve
using the device.

By digital filtering after the accelerometer you can get better resolution at
the expense of longer response time.

The noise spectral density is 225ug/rtHz - integrate this over the noise
bandwidth of interest to get the rms noise.

The ADxL103 has a noise spectral density of 110ug/rtHz - about half of the
ADxL105. But it also has a lower full scale so the dynamic range is a bit less.