We recently made 70 devices containing the BME280 and as part of our production test procedure we cool them to about -25C and then bring them back up to room temperature. We have found that 5 of the sensors give bad reading during this cooling/warming cycle, typically when the rate of temperature change is greatest. They otherwise read correctly at room temperature.
The unit powers up the BME280 every minute to take a reading in forced mode, with oversampling = 1 for temperature, humidity and pressure.
On further investigation we find sometimes the calibration data read is corrupt, but the raw readings may also be corrupt on occasion.
We’ve added a soft reset when we detect bad calibration data and then read again, but that doesn’t ensure the subsequent readings are good on these particular units.
Is this a known problem when temperature changes quickly? The surprising thing is that the calibration data is corrupted on reading, which suggests its an electrical problem within the device. We are confident that our I2C communication is good, as is our firmware.
