This parameter is called heading drift. No matter what underlying technology, there will always be a drift when integrating the gyroscope date to get heading. The main factor for this drift is the bias stability. With BNO055, it is already close to the maximum stability allowed for a consumer electronic device. Sensors that are more precise than this are subject to export control.
Is there any reason why you would not want to use the NDOF mode ?
I used both bno055 usb stick and bno055 shuttle board (in NDOF mode) to measure the heading (in Eulers )in various points about 3 meters apart. I see lots of difference (upto 10 degrees) between the heading values. Calibrating the sensor again in that same point also gives the same value.
Thanks for replying,
You are right, I'm trying in indoor as it will be used on a robot running indoors.
Even tough I do dedicated calibration at a fixed point I do not get absolute value while facing towards the North, so can I please know what is calibration ? Why exactly do we calibrate?