Open ayang923 opened 2 years ago
thinking of using floating point decimals stored in nvram
In prior work with the same temperature IC, we found that a linear fit (with slope and intercept coefficients) reduced RMSE from the IC's factory calibration from ~ 0.1 to close to 0.01 C. I can dig up exact numbers if helpful but recall marked improvements. Our calibration used five or six points from 5–25 C and a well-calibrated reference thermometer.
@SUPScientist this issue is more referencing the process of saving these calibration constants after they are obtained. In regards to the temp calibration, we performed a similar calibration and I believe we achieved similar results. If it isn’t too difficult, it might be helpful if we could compare the previous constants to the constants we got this summer.
@ayang923, understood, thanks for the context. I realized it was about the process on the FW side rather than the physical procedure, just thought I might make a note on the procedure here in case it could help. Glad you achieved similar results. Sure, wouldn't be too hard to make such a comparison.
Configure particle board to store and send calibration constants along with ADC counts. These calibration constants are used to convert uncalibrated SI unit readings to calibrated SI unit readings and these constants are specific to different particle boards.