A general method (with Python scripts) for calibrating accelerometer sensors.
Developed By: Michael Wrona, B.S. Aerospace Engineering
GitHub: @michaelwro
YouTube: @MicWro Engr
Blog: mwrona.com
$ pip3 install numpy matplotlib pandas # pip for Windows(myenv) $ conda install -c conda-forge numpy matplotlib pandasIMPORTANT: Before following these steps, I highly recommend watching the video I created about this process. You can watch it at this link.
record-data.py expects to read comma-separated accelerometer data from a serial connection. Each sensor is different, so you will need to write your own microcontroller code to output comma-separated accelerometer measurements to a serial port, similar to this format:
0.0642208,-0.05490976,1.02357024Open record-data.py in a text editor. Change the variables at the top as required.
Once you can output comma-separated raw accelerometer measurements over a serial connection, you can begin logging data. Run record-data.py to begin logging data. Have the accelerometer flat and stationary and press ENTER as prompted. Then, type 'm' as prompted to take a measurement. Move the accelerometer to a different orientation, then take another measurement. Repeat for many accelerometer orientations (sideways, upside down, left, right, etc.).
Once you are satisfied with the number of measurements, type 'q' to save the measurements to a tab-delimited file.
Magneto is an ellipsoid-fitting software used to calibrate accelerometer and magnetometer sensors. Magneto expects raw measurements to be input as a tab-delimited text file. The norm of the gravitational field will be the ideal magnitude of your accelerometer measurements. For example, my accelerometer output data in G's, so my norm/magnitude would be 1. Load your text file generated by record-data.py, then click 'calibrate.' BAM! It's that easy!
Open plot-calibration-data.py in a text editor. Copy the A^-1 matrix and bias vector values to the Python code and specify the tab-delimited text file of uncalibrated measurements. Run the code and compare uncalibrated and calibrated data!
