I don't remember seeing this before. Instead of baluster/bulbous pulley or
capstan, two angled faces with a thin ridge in the centre on the drive
pinion. So the belt must ride on the ridge but why does it not hunt about?
There is a limiting flange either side of the angled flats and the width of
the rubber band is greater than the distance between a flange and the
central ridge, but still why no riding about?. Then what I thought was crude
construction is probably part of the dynamics. The capstan is 2 discs fitted
together but one is 49.94mm diameter and the other 50.04mm so this step
change must make sure the band rides over the ridge but biased to one side,
lightly touching one angled face, but not touching a limiting flange. When I
get the mechanism working again I will try viewing with a xenon strobe light
.. Anyone any input on this ? more or less prone to changes in rubber
restitution/stretch etc, more/less likely to come of when PTO is engaged,
greater/less accuracy in speed of drive with varying driven-side back torque
etc.