# Spikes in acceleration data

I am trying to quantify the noise of the accelerometer inside the Parrot AR Drone 2.0, but the signal contains odd spikes. I simply placed the drone on a table and recorded the acceleration for about 30 seconds, without turning the rotors on. The measurements then produce the following image:

Without mean subtraction:

There seems to be a fairly horizontal line, that contains very large spikes (up to +-0.1 m/s^2). This holds for all axes of the accelerometer. I expect some kind of normal distribution, but something odd is going on. Does anyone have any ideas? Simply removing the noise using filters is not what i'm looking for...

• What is the digital resolution of the measurement the accelerometer makes? – Andy aka Sep 7 at 10:34
• @Andyaka not sure, but the peaks from the "flat line" don't seem to be equal in height, you can see that better here. – Thomas W Sep 7 at 10:42
• There is no acc_z on the last graph. Which sensor you use, and how the data are sampled (synchronously or channel by channel)? Which program is the UI? Can you export data to any math program like MATLAB, OCTAVE, SCILAB? – Marko Buršič Sep 7 at 11:11
• @MarkoBuršič oops, I forgot to add that the second graph shows the data without any mean subtraction. The program is MATLAB. Accelerometer is the BMA150 I believe. – Thomas W Sep 7 at 11:16
• OK, if you have a MATLAB, then you could do a cross-correlation of each axis VS other one: x VS. y, x VS. Z and y VS. z. You would get a result if the spikes are related each to other and also a time shift. A strong relation would indicate a power supply noise. – Marko Buršič Sep 7 at 11:32