BNO055 Accelerometer Expected Repeatability in NDOF Mode?

[dustin_]

I'm attempting to keep track of the velocity of a slow-moving object (about 6" / sec) over the span of 30 seconds to a minute. I'm trying to keep accelerometer noise under control by averaging 32 samples before sending the averaged data point over to Matlab for processing, which seems to be working fairly well. The issue I'm seeing is that after picking up the sensor and setting it back down, the zero-acceleration offset seems to shift, this is illustrated in the attached screenshot. Ignore any data after 210 on the X axis, nothing has been collected from there on. Data is from the LIA_DATA_* registers and divided by 1000. I'm using NDOF mode.

So I have two questions:

1. Would a more precise calibration procedure eliminate this issue? I'm simply standing an Adafruit breakout board on its ends at the moment, which isn't perfectly stable.
2. Is it reasonable to try to keep track of an object moving this slowly with this sensor or have I been doomed to fail from the start?

Thanks!

[BSTRobin]

Hello dustin_,

Generally, the zero drift is evaluated when the sensor is stationary, not when the data changes during the movement.