Accelerometer wiki page erroneous axes ?

[rixed at happyleptic.org]

-[ Wed, Apr 08, 2009 at 04:26:02PM +0200, giacomo giotti mariani ]----
> If you accelerate upwards your phone (with the screen pointing up) with
> an acceleration a  you see a greater acceleration a_tot (along the z
> direction) that is the sum of gravity acceleration g and your imposed
> acceleration a (a_tot=a+g).

Well, some confusion here.

OK, let's imagine a _perfect_ accelerometer.
This device would report only acceleration. So, when the phone lies on the
table, it would report 0 (neglecting the rotation of earth here :-)).

Then, to take your phone up to your ear, you first accelerate it upward,
then downward (to stop the movement). If this perfect accelerometer
were reporting accelerations along a Z axis pointing up, it would thus
report first a positive value, then 0, then a negative value (the other
way around if the other convention is used).

Now, lets suppose that this perfect accelerometer is not calibrated correctly,
and offset every values by 900. It will then report first a value >900,
then 900, then a value <900.

This is what you get with the actual accelerometer (*).
So, Z is actually pointing Up.

(*): our accelerometer is not perfect, but not because of a random
offset. It is not perfect because it has no mean to distinguish between
an actual acceleration from the effect of gravity. This is not a
defect and there is no need to send the phone back to OpenMoko :-) because
it is just how nature work (http://en.wikipedia.org/wiki/File:Elevator_gravity.svg)
The effect of gravity (downwards) is the same as an acceleration (upwards!!)

The freerunner is really a great thing. First it remembers me the good old days
when I was hacking open all the guts of my 8bits computer ; then it remembers me
the good old physic classes ! :-)