Good to know, although I don't understand why. Thanks for posting the solution.

PrusaSlicer has a setting to detect thin walls. It may have different names in other slicers. I only use 0.4mm nozzles with Prusa MK3/4 style printers.

Great question. Google "coefficient of thermal expansion" to get a list of changes per degree of temperature change. Plastic is typically about 3x worse than brass or steel. But, as you have noticed, modern houses with heating and air conditioning really only change temperature by a few degrees twice a year. Two of my oldest clocks, SP2 and SP3, used carbon fiber rods which should have very low thermal expansion. It added to the bill of materials and added extra assembly steps. The clock wasn't really any more accurate so I switched back to simple printed pendulum rods. This works well enough as long as the temperature changes are minimal. Maybe the "carbon fiber" archery arrows I used were not actual carbon fiber, but closer to plastic with a few bits of fiber embedded. I never actually tested their expansion rate. There is another material called invar that is very close to zero expansion. It is expensive. Other options include taking advantage of the differences in expansion between brass and steel to build a gridiron pendulum rod. Or use plastic and steel. It would add a bit of complexity.

Gear 8 has a slightly larger outer diameter with the update. The gear teeth end up very close to the bottom support column. The two frame pieces affected are frame_back_lower_b with the support column and frame_front_lower_b that connects to the column. It would probably still work with the old parts. You could build the clock using just the new gears and decide if you want the new frame parts later. A new dial is an optional part if you want different looking globes.

Excellent conversion. The colors look great. And the remontoire makes it really easy to live with. Thanks for sharing.

MyMiniFactory has a messaging system if you don't want to post it here.

The video is very helpful. The entire gear train normally rotates with around 6-8 lbs on the weight shell, which is only 3-4 lbs on each end of the string. The pulley has a small diameter and you are pulling at the larger diameter teeth. It should take significantly less than 1 lb of force on the teeth to get the escapement to rotate. Parts appear to be flexing in the video, so it is likely that you are applying more than 1 lb of force. My guess is that there is extra friction in the gears near the escapement. Torque reduces and speeds increase closer to the escapement. Friction near the escapement becomes exponentially worse than friction near the weight shell. The escapement is barely moving around 15 seconds in the video where it looks like force is being applied to gear 8. You could try debugging the gear train without the pallet. Hang a 4 lb weight directly on the string. The gears should start spinning quickly. If you stop the escapement, it should start spinning quickly. Also, go through the friction related debug steps. Test that all gears spin easily on the arbors. Check all gears for end shake. The gear 4 stack in the center of the clock is a common area for a lack of end shake. The good news is that your clock is incredibly close to being fully functional. Once you find the friction source, it should start working great. I like the reverse colors on the globe.

Interesting. Try cleaning the pendulum bearings in alcohol and let them run dry. Spin them by hand on a piece of 3mm rod and use the two that feel best. PLA seems fairly tolerant of any lubrication types. Use anything you have if you want to try it. 3-in-1 oil, printer rail lubricant, etc.

Hi Al, Thank you for providing lots of information with your post, especially a description of the escapement characteristics and the pendulum free swing test. It is surprising how many people expect debug help by asking "My clock keeps stopping, what should I do?". If the escapement is energetic, then the entire gear train is likely in good shape. This usually narrows it down to the pendulum support bearings, but the free swing test runs for >15 minutes. The bearings in the kit have a dry lube added for rust protection. Sometimes they work slightly better if you remove it using alcohol. Sometimes, a pendulum that slowly degrades over time is an indication of not enough drive weight. 4.5kg should be more than enough to keep SP13 running. One caution is to make sure that the frame adjustments are made to keep it from sagging with the extra weight. All of the things you have done should produce a working clock. My suspicion is that the involute gear tooth profiles could be the issue. I was always under the impression that involute and cycloidal gears have similar efficiency. Both styles have a mostly rolling action with very little friction. There is always some sliding friction as the teeth start to engage and as they disengage. Friction is higher if the engaging occurs before the line of centers because the teeth push into each other. Friction is lower after the line of centers where the teeth are pushing away. Involute gears have much more engagement before the line of centers compared to cycloidal gears. Adding extra drive weight sometimes makes the teeth jam together harder. Some filament brands have higher friction than others and I do not know any good way to test them other than to build a clock. A simple fix is to add a very thin coat of lithium grease to the pinion teeth. You can wipe most of it away, so it is nearly invisible. Only the pinions need to be greased. I have started converting many of my oldest clocks to cycloidal style "Perfect Print Gears". So far, SP2, SP3, SP4, and SP5 have all been upgraded. These were major upgrades because I ended up redrawing the entire frame. The newer clocks should go a bit faster. I am currently converting the SP14 moon phase clock and will start on some of the others next. The biggest difference is that the clocks are more reliable. For example, the 32 day clock used to just barely stay running in 32 day mode with an 11lb weight, and it might randomly stop every few weeks. The new gear profile is extremely stable with a 9lb weight. Sorry for the long winded description of telling you to try lubricating the pinions. Steve

The square base was an early prototype that got heavily modified by the time the official release was made. The mounting hole locations were moved and possibly several other changes.

Wow, 0.88 sounds like quite a change to the flow rate. Sounds good as long as it works.

I have considered adding a speaker based "chime", since all other experiments have not sounded good. Traditional chime bars are difficult to find and only sound good when placed inside a resonating sound box. Holding the chimes loosely will create a dull sound. Holding it firmly against a solid object creates a much richer sound. 3D printed clock frames would need some serious experimenting to make anything reasonable sounding. And the mechanism for operating the chimes would easily double or triple the number of gears required. I also tried some wind chime style rods using commonly available components like metal conduit. It followed the best known methods from a few web sites but never sounded rich enough as a clock chime. The next best option might be a simple battery powered chime module. They appear to be available for around US$15 including the clock movement or US$10 for the standalone module. There are two wires coming from the clock movement to trigger the chime unit. Not sure if it is a simple connection or if it needs an electrical pulse to trigger it.

Nice colors and nice extra touches. The gold and black go well together. Thanks for the feedback regarding PETG. I only tried it once and didn't see any friction reduction, so I went back to PLA. This clock is one of the few that does not have standoffs that adjust to an uneven wall. The compact frame doesn't allow them. I use these bearings https://www.aliexpress.us/item/2251801799263630.html I usually test 4-5 bearings by placing them on an arbor and spinning them by hand. The two that spin the longest get used on the pendulum. I cannot remember the last time that this simple selection process was not good enough for at least 15 minutes on the pendulum free swing test. Occasionally, the pendulum gets sluggish after a few months and the clock fails the free swing test but passes after another cleaning in 91% alcohol. They usually run forever after the second cleaning. Maybe this supplier has better than average quality. They are listed as ABEC-1, which is the lowest possible rating other than "unspecified". They have worked really well for me.

I love both of them. The grandkids should love them. The bomb is super cool. Is it just glued around a standard weight shell? It looks like it has very little clearance around the pendulum bob. The ghost bob is also super cool. The updated SP2B design was a good choice for the reversal. The Perfect Print Gears are symmetrical so they can run backwards. This limits the components that needed to be flipped to the obvious asymmetrical parts (pallet, escapement, ratchet, and lower frame with cord hook). The fancy gears in the original SP2 design are asymmetrical and every single one would need to be flipped.

Good to hear that the update is working well. It matches my experience with an increase in stability. I am slowly working my way through updates to several other clocks. So far, each update has taken about a month to fully test and document the changes. The moon phase update is next.

It must be just that one vendor. I have built around 100 fully programmed kits and cannot recall a single Arduino Nano that was defective. There is sometimes a bit of fiddling to randomly switch between the old and new bootloader. My sources include both Amazon and AliExpress. Amazon appears to have mostly 3rd party vendors shipping through Amazon warehouses. I try to stick vendors once I find one that I like. Two Amazon vendors I have used successfully are AITIAO and AITRIP. I have only ordered once from AliExpress and recall having to do the most fiddling. They were possibly a mixed batch using both bootloader types. Amazon has been much more consistent for me. The only true known good supplier might be the official Arduino store. They now have 18 different flavors with "Nano" in the name. The oldest one is the one used in the clocks at US$24.90 and only available with a mini USB port. I think I will stick with the clones to get the USB C port. At some point, I should test the controller with some of the newer Arduino boards.

All the clocks work with 1.8 degree motors. The crazy gear clocks are slightly quieter with 0.9 degree motors, but even 1.8 degree motors are quieter than most commercial battery powered clocks.

Both the front and top profiles look great. That type of filament could be tricky because each gear will look different. You could end up with colors that flow nicely, except one gear looks out of place. Dual color is much more predictable. It really stands out with the extra motion in the crazy gear desk clock.

Some of my clocks use generic purple silk PLA. Then I switched to MatterHackers Blue Raspberry Quantum PLA. It looks awesome, but normal price is $42 for 0.75kg spools. This is what I am using now. Some manufactures also make a tri-color silk PLA, but the colors blended together too much. I prefer dual color for the gears.

The video helps considerably. It shows that the escapement has plenty of energy. It looks like the gears have low friction and should allow the clock to operate properly. The issue might be with the pendulum. Bearings that are properly cleaned of grease should allow the pendulum to free swing for around 10-20 minutes with no additional energy supplied. So why does the pendulum lose enough amplitude for the clock to stop in less than a minute? Run a pendulum free swing test by removing the escapement and measuring how long it takes for the pendulum amplitude to degrade from +/- 5 degrees down to a small amplitude of around +/- 0.1 degrees. It does look like the spacers around the pallet and pendulum are too small. I will look into the design and update them if needed. In the meantime, add extra spacers or print new spacers with the Z height increased to fit. Most slicers allow splitting misc_spacers.stl into individual components so you can isolate the two spacers you need.