What was my poor man test? If you take a servo motor and let kmotion run a step response, and give it a step size of 80,000 over 3 seconds, the motor will spins many times to the right and then many times to the left. What I did was clamp the servo/HD assembly to the bench, and laid a piece of blue tape on the bench with a straight black line on the tape. Now I added another piece of blue tape to the end of the harmonic drive with another black line. These two black lines are lined up exactly prior to the test. If you spin the motor with 22,000 pulses the Gearhead will turn exactly one revolution. Now if the gearhead had a good deal of backlash when you spin it in the opposite way it would most likely not line up on the black line again. Now with the step response situation you are spinning the motor many times in each direction and it always comes back to the black reference line on the tape on the table. This means it is always find the same start position.