I've watched that video before and it seems you do a bit of autodetecting as the equipment sets up the experiment? It seems that would add a fair bit of variability between tests as the run time and torque is different between tests. Obviously a rigidly standardized test can favour one design over another too. I can imagine some designs being more susceptible to oblique loading or pressure put on the switch. You do mention taking enough time to dissipate heat which should at least eliminate most of that factor.
Yes, as shown, the torque and end limits are determined by the software. The stepper's output shaft needs to be perpendicular to the meter. There is no downward force and the side loading is keep very low. I am trying to see how the switch by itself behaves. In the real world, the person has a firm grasp, pushes down, pulls up, turns them past the dead stops..... Lots of variables I wanted to try and remove. My tests also run non-stop as where in the real world, apparently people will spit shine their contacts every now and then.
You can see that the fingers that turn the knob are very loose (on purpose). The knob is never placed in a bind nor does it ever reach a hard stops. I could have a 100Nm or torque available and it wouldn't make a difference. I limit the torque for the case where something goes wrong (the test runs over several days). I don't want the stepper motor to damage the meter. This never turned out to be a problem. The clamp holding the stepper motor and box securing the meters has never moved enough to cause any sort of bind or misalignment problem. The Labview code that runs it is predictable as well. It's not going into the weeds and trying to take over the world.
Actually, my CEM was damaged when I loaned it out and the idiot tried to turn the knob past the dead stop (and succeeded). I assume these highly educated people are rare and did not consider running a test to determine how much torque was required to defeat the dead stops on the meters I tested.
My test also doesn't require yellow sticky notes with obscure equations to try and decode the data.
There was mention of putting the Toggle Bot back together to run some tests but I never heard anymore about it. I had assumed this was for the new meters. No time to run it? No product to sell, so no reason to run it? It did very poorly, so no reason to make a video about it? One thing you get with my tests is pretty much in your face data. I could care less how a particular meter holds up.