0

I am using the following code to generate an allen variance plot from some accelerometer data

https://github.com/nmayorov/allan-variance/blob/master/allan_variance.py

It seems a sound implementation looking at Allan variance definitions I can find online

How do I interpret this graph?

![enter image description here

00__00__00
  • 181
  • 3

1 Answers1

1

This is the result of drift (non-stationary data). The horizontal axis is the averaging time $\tau$, and the vertical axis is the standard deviation (ADEV) of differences between successive blocks of data each $\tau$ long, and separated by $\tau$ seconds. If the data had a stationary white noise statistic, the ADEV would instead be going down at the square root of tau. For more on ADEV please see these posts:

Bias instability in Gyroscopes : AVER / ADEV

How to interpret Allan Deviation plot for gyroscope?

Is Allan variance still relevant?

Dan Boschen
  • 50,942
  • 2
  • 57
  • 135
  • indeed, my data was not stationary. I guess I should look again at this plot after taking first temporal difference of the data to make them stationary? when doing that, indeed the slope changes to negative – 00__00__00 Dec 08 '22 at 07:26
  • That is what the ADEV computation does.. .the point is that gyroscopes drift and are not stationary. The ADEV is very useful to quantify and compare non-stationary processes since it will do that difference you describe, so you don't need to do that first. – Dan Boschen Dec 08 '22 at 14:51