Steve,

I think you're real close to having the problem solved when you speak of a master and a slave. My opinion is that instead of calling one servo a master, let both servos be slaves and create a virtual master which consists of a software simulation of where the door should be. Let this virtual master be the desired trajectory and let the two slaves follow it. (Known in the trade as a "model follower"). Let the virtual master slew at a rate less than what the actual doors are capable of slewing (gives the slaves the ability to catch up to the master). Set the P, I and D gains of the slave servos to follow the master such that neither slave falls too far behind the other (25 mm). If you wish, cross couple the two slave speeds so they can anticipate an expanding differential speed (integrating into crab). To cross couple, bring the speed of one servo into the position setpoint of the second servo via a scaling term (gain).

Using vertical cable displacement to detect overload conditions might work out pretty well. I think the cable would need to be preloaded in the up or down position and held there with a fairly heavy spring to keep the cable from oscillating and inducing nonlinear oscillation in the control system. Displacement of the preload spring would be the means for detecting overload condition.

I suggest the implementation be preceeded with a software simulation of the system.

This is a neat project - kinda wish I was working on it!

Bruce