I changed the original posters code to stop the estimated angle from drifting. This example only has one degree of freedom - the tilt which I believe is roll. it is damned confusing to get proper information on this device even from the data sheet. However, this example uses a complementary filter to combine the gyro and accelerometer readings. This is because one works well at high frequencies and one at low!! So an integrator is no longer required, instead it is a low pass filter and high-pass combined into a simple recursion. It is a clever yet simple approach used by many Segway builders.
y(k)=0.8(y(k-1)+gyro_rate*dt)+0.2(acceleration_angle)
(dt is the sample interval which we are told is 1/1000. I am confused if this is the sample rate for the sensor or for the control loop - because I have 10kHz sample loop and it still works).
of course 0.8+0.2 adds to 1 and different combinations can be used. the alternative is a Kalman filter which isn't that difficult but I am led to believe there is little difference in the result. (don't take my word for it). The acceleration angle is found using arctan2 as acceleration is a vector. (that was the confusing part). You can do this for all three axis of course but I am only interested on one since my application at present is in Segways and balancing carts. Connections are as previously mentioned. I also added a trim for the gyro velocity.
3.3V --> VCC
GND --> GND
VIO --> GND
SDA --> 34
SCL --> 32
This is for the blue break out board (not the red one that is often shown), Read previous postings about it as it needs an extra 3.3v I believe. This will sit for a long time without drifting.
