Hello. It's almost 2 years now that I keep taking some strange behavior from ADISes mentioned in caption: in a time range of ~1200...3600 seconds from power on they start giving out some abnormal gyro data while actually laying still at the table. Pictures attached show typical case, that particular data set was taken right now.
We're using our own board to communicate to IMU, it's based on STM32F4 family microcontroller. All timing requirements concerning SPI communication are met, actual working SPI freq is 11.4 MHz (but experiments show that there's no difference if it's 5, or 2, or 0.5 MHz), EKF is disabled (for 16480), data ready set to positive polarity at DIO2, FIR filters are bank 2 for all sensors, decimation rate is 12 (now applied to 2400 external clock via DIO3, but most of that 2 years it was applied to internal clocking), all other settings are default, no signs of error are in SYS_E_FLAG, starting self-test is OK. Data reading was experimented to be series of singe-register reading (like "low, hi, 0", or "hi, lo, 0"), and kind of burst read (like "lo acclx, hi acclx, ..., lo gyrox, hi gyrox, ..., magnx, ..., 0") with all proper chipselect pauses between words. Could you give me any ideas what could be the reason for that?
Unfortunately I don't have ADIS eval board accessible right now, but AFAIR i didn't manage to repeat that behavior on it when it was avaliable.
ADIS1.jpg is 11000-seconds-long record of ADIS16480 communication, ADIS2.jpg is zoomed version of the moment where disturbance happens. ACCLs are in g, GYROs are in deg/sec, MAGN are mgauss, TEMP is centigrade.