well I have the same problem but I was not able to solve it by using the ARef. As I understand the postings, the ARef thing is only relvant for boards prior Rev.B. Hope I'm right.
I was trying to read MEM gyro sensor values using the analog port. I've got a nearly stable reading for the neutral value of the sensor. The majority of measurements have correct values while some of them are wrong. Only for one step and always the same value.
Is there a way to avoid such a behaviour? If not, does anybody know a filtering to remove those artefacts?
Here are a some of my readings
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:385 1.246V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:385 1.246V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:384 1.242V
U:385 1.246V
U:384 1.242V