You have a 3-axis accelerometer, which means it is sensitive to the linear acceleration vector that it experiences. Assuming it is sitting on a bench top and not visibly moving, it will experience a constant acceleration of 1g in the direction pointing down to the center of the earth. Changing its orientation will change the measured acceleration (measured in units of "g") from (0, 0, -1) (flat on benchtop, assuming it is level) to, say, (1, 0, 0) when the X axis points straight down. As long as the accelerometer is "stationary" (i.e. its center of mass isn't moving), the magnitude of the measured acceleration (sqrt (x^2 + y^2 + z^2), where x, y, z are the acceleration components in the X, Y, and Z axes) will always be 1. Moving the accelerometer will alter the acceleration by adding the "motion-in-space" acceleration vector to the "pointing-to-the-center-of-the-earth" Gravity acceleration vector, giving you a net acceleration that differs from (0, 0, -1).
Of course, to get velocity, you integrate acceleration (adding "noise" in the form of rounding and calibration errors), and to get displacement, you integrate again (adding another layer of errors).
Note that 6-axis accelerometer chips that include angular accelerations (to distinguish "rotation" from "translation") are also readily available.
Bob Schor
