Using an accelerometer to measure speed

[lucvh]

Hi,

I am interested in using an accelerometer to calculate the speed of a car. The accelerometer is a triaxial one and gives 3 values of acceleration; x y and z. These can be assumed to be accurate as i have already worked on the calibration of these values so that they come out in gs. Is there a way to calculate the acceleration of the device independent of its alignment to the motion of acceleration?
For example, the accelerometer would be mounted to the dashboard of the car, but as the car accelerates the rear axle of the car dips and therefore the angle of the vector has changed. This is fine to calculate, if it weren't for gravity. I can't understand how to remove the acceleration due to gravity when the exact alignment of the sensor is un known. Does anyone have any ideas?

 

[sennyk]

You need that angle. Try adding an angular rate sensor and integrating to find the angle. Then you can remove the acceleration due to gravity and integrate to find the velocity. http://www.analog.com/UploadedFiles/Data_Sheets/778386516ADXRS150_B.pdf They are about $50 on digikey. However, I know you can find them cheaper.

 

[sennyk]

Or you could find the angle by having a frictionless POT and using an ADC. I'm not sure a close enough to frictionless POT exists.

 

[lucvh]

The idea was to only use the available hardware. Is there no way to do it knowing that the vector i need to resolve it into is at 90 degrees to the acceleration of gravity?