Due a low 3V battery and power off situation all the saved data on the Tac20 CPU board was lost, so I downloaded the job files to the board via hyperterminal, including the job configuation. The is only lower and top floor at the job. IMS screen shows position=2, Floor Ref=999999999, Pos Count= 731564543. A new hoist learn was necessary, but when starting the learn, the hoist moves to the lower floor, IMS Position=0, hoist goes down 3 inches below lower floor, stops and then moves up, but instead to move to the top floor to finish the learning, it stops right there at lower floor level, and fails the learning. For a brief moment (half a second) Pos Count becomes 10024 and return to 731564543. Fails the learning and gives me SELECTOR error (even on IMS hoist learning screen), errors 1031 and 1056. I already checked and measured positions of magnets and selector boards, connectors, cables. Interesting that IMS shows everywhere the hoist is at the floor #2, not #1. IMS inputs shows correctly BPP at lower level and BP1 at top floor so it is not this. Still it thinks it is at top floor (2) and this (I think) is the reason why it stops at lower level and not complete the learning. Any help will be appreciated.

Sounds like a bad cpu card. I'm sure Dirt will correct me if I'm wrong here, but a tac20 doesn't have volatile memory. If your 3v battery was dead it just wouldn't remember it was in fire service after cycling power, not loose settings. Sounds like you had a bad 188e card from a power issue

Thank you EElevator for you answer.
At the IMS Car IO Screen, I have FSH1 gray and FSH2 green. As no fire service is active, they both should be the same, isn't it? I never paid attention to that before. I am waiting for another CPU card for testing.

Another CPU board installed, was programmed with the 5 files from the job, it behaves exactly the same. Hoist learning moves down to the bottom limit, starts to move up and stops 2 to 3 inches from there with setup failure, never complete the hoist learning. Floor Ref=999999999, Pos Count = 73156543, exact numbers as the other cpu board. We just move the down limit magnet up and down and test everytime, the only thing that changes is where the car stops bottom on hoist learning, moves up 2 to 3 inches and stops. Every time it stops it briefly (less than 1/2 second) shows a number 10025 or around this on the Pos Count. I noticed that moving the down limit magnet further down, that number increases, what would be the distance from the down limit to the floor level (by the addressing magnets on other side of the tape), so it is reading and calculating correctly. We tested other two boards on the hoist, same issue. So, apparently this is not the hoist boards, not CPU board. What is curious is that IMS shows the car to be at floor 2, not 1, all the time, but BPP at lower level, and BP1 at top tells me the sensors are reading correctly, but everything else is telling second floor. "Being at second floor", it would stop, not further goes up for the hoist learning, but hoist learning will not pay attention to the floor level, but the top limit magnets, doesn't it ? I really don't know what else to do...

Eelevator is right about there not being volatile memory on a TAC20. If your parameters are lost it wasn’t because of the battery. FSH2 is the off position monitor for the phase 1 switch so should be on and FSH1 should be off. It sounds to me like your bottom directional limit is set too far down. The learn will fail if that’s the case. It should be 1” to 1-1/2” below the floor. With the car level at the bottom floor, mark the bottom of the selector on the tape the. Set the top of that magnet 5/8” below the mark. With IMS on 8 bit ports and positive logic and the car level at the bottom floor you should have every IO high in port 3 except NTSB.

We found it... the MCD relay (moving up - up to speed) was not latching, so VC2A and VC2B relays not latching either, car stops after moving 1 inch up, coincidently to the lower floor level or sometimes a bit above that. Fixed the wire to MCD and then "hoist learning" worked, and everything else after that.