Calibration drift based on temperature

[PorschefanRoel]

I'm doing an experiment on calibration drift. Every 5 minutes I'm doing a on-chip calibration to see what the on-chip calibration values are.
During this a saw a fluctuation during day/night. So I've placed the camera in the dark, to exclude (day)light being the cause of this.
And added monitoring of the ASIC temperature and Sensor Temperature.

The spike to 0.1 of the calibration data is when the realsense camera was off, during the time I was editing the python script to start monitoring ASIC & Sensor data. You can clearly see that it's related to the temperature of the sensor/ASIC.

Is this fluctuation on the On-Chip Calibration normal?

[MartyG-RealSense]

Hi @PorschefanRoel Temperature can have an effect on sensor calibration, though it is not usually necessary to perform a re-calibration of the camera as a result of temperature unless there is a severe temperature event.

Temperature changes at different times of day can cause drift in depth measurement accuracy. The external environment's heating or chilling of the camera's outer casing can have an effect on internal temperature. For example, Intel lab tests once found that the internal temperature should not exceed 60 degrees C as long as the casing's temperature is below 44 degrees C.

The D45x class of RealSense cameras, such as D455, have a mechanism called thermal compensation that can automatically adjust depth values to take account of temperature variations. The subject is discussed at #9561

[PorschefanRoel]

Hi Marty,

Thank you for the quick response. So the drift I'm seeing here is not out of the ordinary and should be OK?
I know that the calibration currenlty not needed, but I was interrestest in how much it did oscilate, als is was a range of 0.25, I was a bit worried, as starting from 0.25 the realsense viewer will tell us that calibration "Can be improved"

The Thermal Compensation, sounds interresting, but unfortunately, the D45x, isn't a solution for us, as it's minimum distance is 0.6m instead of 0.2 (We need around 0.3m, at minimum).

[MartyG-RealSense]

It is normal to have variation in on-chip calibration 'health check' score when repeating calibration. How it is intended to be used is that you repeat the calibration process until you obtain a low score that you are satisfied with (the closer to 0, the better the calibration) and then save that calibration to the camera hardware. You should not then need to check the calibration score again until a later date, unless you want to keep trying for a lower score and save the improved calibration.