Some of our customers have reported occasional sudden jumps in their data, that are cleared up by rebooting our device. These are fairly rare events. There are generally months between an event in a population of hundreds of channels.
We came up with a debug command to dump the ADS1212 register values and have correlated this with a change in the low byte of the CMR register to FF. It always seems to effect byte zero, and it always seems to change to FF. Here is a sample of our output....
Read A/D C 0 O 184 S 6309351 CMRA 4208426f CMRX 4208426f
Read A/D C 1 O 125 S 6317784 CMRA 4208426f CMRX 4208426f
Read A/D C 2 O 170 S 6316688 CMRA 420842ff CMRX 4208426f *
Read A/D C 3 O 42 S 6312138 CMRA 4208426f CMRX 4208426f
CMRA is the Actual reading from the A/D.
CMRX is the eXpected value that was written in. The * indicates that they don't match.
Byte zero changes the rate at which the A/D operates so the zero and full scale correction values are no longer valid, hence the jump in the data.
This generally happens after many readings of the A/D, so I first suspected that a timing glitch was causing a write to the register instead of a data request. I noticed that the command to write an FF to CMR byte zero could be found in my bit pattern to read the data output registers if it is shifted by three bits, as shown in the attachment image. We changed the last 24 bits to zeros, to avoid this, but it had no effect.
We are contemplating rounding off the A/D data rate settings, so byte zero is always intentionally set to FF. That way, the event that changes the register to FF wouldn't have an effect. That may work, but it doesn't seem like an ideal solution.
Any suggestions?
Thanks
Chris