The floating current limiter circuit would make things better electrically but from a manufacturing and test cost standpoint it would not really not be an improvment. A pot would have to be adjusted and the heatsinking of Q2 would eliminate a 100% surface mount board. Right now the board is all surface mount and assembled with a pick and place machine that does the job fast. They make about 3000 boards a year and having to go with some hand assembly would most likely be a problem. I'll suggest it and see what the responce is.

They are saying they only have the common mode problem with about 2% of their units but it is something that is know and needs correction. The A/D they are using in U1 does not need 22-bits of resolution. That part is used elsewhere and they decided to use here also.

If U1 with a 2.5V reference is the only chip sacred in the circuit what would be a good approach to accomplishing inputs for 4-20 mA, or 0 - 10 V or 0-5V with protection against connecting 24 V input?

How did you learn about the real floating current limiter?