About 2 years ago I added some automation to the Southern Pacific staging yard using some Azatrax IR detection modules. The idea was to have a train enter the yard and stop on one a particular track and another train automatically leave on another track. The post I made on this in November 2020 is HERE.
This has worked well but occasionally when after coasting to a stop the space between 2 locomotives lands right over the receiver that is under the track. These particular Azatrax units close a relay contact momentarily but will continue to show detection as long as the beam is blocked. When the beam is no longer blocked, the circuit is reset. When it is time for the train to go out again, as soon as it moves the beam is blocked by the second locomotive and the circuit thinks the train has just arrived and switches to the next train in the sequence.
After giving this problem some thought over a period of time I came up with some ideas on how to overcome this and recently I put those ideas to the test. First I removed the IR transmitter stand and carefully bent both the 1/8 inch rod and the 1/8 inch tube near where they go through the deck. This would hopefully prevent the beam from passing through between the locomotives. After re-installing them I re-tested by slowly pushing a pair of flat cars over the IR receiver. This helps as but did not totally correct the problem. Next I pushed the IR receiver down deeper into it's hole between the ties. This again was an improvement but did not totally prevent the beam from passing through to receiver. Last, I made a lens from .250 wide by .010 Evergreen strip with a hole of about 1/16 inch. The hole was aligned over the spot where the receiver is. This seemed to do the trick.
To test the arrangement I was using a Kato SD40-2 and an SD40 set back to back and slowly pushed them across the path of the sensor. These have the most open back porch and when running back to back provide the largest path for the IR signal to get through.
While these adjustments seem to have corrected this issue, the real test will come when the upper helix and return loop are finished. Then I can setup several trains to run in continuous operation.