Recently, I have had several clocks slow down and eventually stop running after a few weeks of months of operation. The symptoms are the clock has stopped and manually moving the escapement back and forth does not produce any movement in the escapement. Something in the gear train is completely stuck. Wiggling a few gears or winding the clock a few clicks will often allow the clock to start running again. However, it only runs for a short while and stops again with the same symptoms.
All the troublesome clocks were built using silk PLA because I really like the way the gears look. My best guess is friction as the gears move. The gear profiles are correct, or at least the gears in CAD are correct. Errors could creep in if the gears do not match the expected CAD profile. I will try to find or create a test pattern specifically designed to test the gear accuracy. It could require a change to the extrusion multiplier or XY size compensation. I already have elephant foot compensation under control.
The other consideration is friction as the gears slide past each other. Most of the gear movement is a rolling action with no sliding, but there is always some sliding as the teeth are starting to engage. I printed some small squares of material that allow comparing friction between different brands of PLA. The test jig would have a section with normal PLA next to normal PLA and another section with silk PLA next to silk PLA. Sideways pressure is applied and whichever one moved has the lowest friction.
One thing that was apparent was that lightly loaded PLA of any type has low friction, but friction is much higher when pressure is applied before sliding the pieces. This would be roughly comparable to dynamic friction with a light load and static friction with pressure applied.
I tested 12 different spools of PLA consisting of 3 silk PLA, 2 translucent PLA, and 7 normal PLA. The 3 worst performers were silk PLA, followed by 1 of the translucent, a normal PLA, then the other translucent PLA. The best 6 were normal PLA. Most of the normal PLA brands were difficult to distinguish, but the silk PLA samples were obvious that the friction was higher.
The next step was to replace some failing silk PLA gears with normal PLA gears. The clock definitely runs better, but I can tell that there may still be some issues. I will continue looking for better fixes.
A few other suggestions are to use the shorter runtime options where they are available. I have been running my 32 day easy build clock with runtimes of 8 to 10 days. I am also looking at different tooth profiles with lower friction. I have tested some modified cycloid gear profiles, but they were slightly worse. Involute gears seem to have lower friction in a clock. The next test is to compare different pressure angles to see if one is better than the others. I may also try finer pitch gears where they are closer to the escapement so the pinions can have more teeth.
Let me know if there are any other ideas. I would like to get back to having a fully functional clock with a 32 day runtime. I know the clock works, because I have had it working. However, the high static friction needs to be fixed for the clock to be super reliable. I am looking for a fix.