I have made several of the clocks from this site as well as a number of others. Most of them were printed on a Prusa MK3S, and recently replaced this with a Prusa MK4. I have a few observations about print quality.
One of the big changes with the MK4 is the ability to use Input Shaper (IS), which allows for much faster print speeds. For example, it reduces the print time for the back frame of SP13 from 10 hours to a little over 5. However, when printing the gears (with silk PLA) I got poorer print quality than I used to get with the MK3S, notably on the overhangs underneath the pinions. They have loops which sag downwards. These probably don't affect the functioning of the clock as the overhangs don't engage with other gears, but they look nasty.
When I have seen problems like this before, it is usually the result of the filament not having enough time to cool down. Some people recommend printing more than one part at once, so they total time per layer is higher. I don't usually like this, as you can get some stringing between the parts. I'm not going to bore you with all the different things I tried, but here are a few recommendations:
reduce the temperature by 20 degrees from Prusa's "Generic PLA Silk" settings.
slow down the print. There are multiple ways of doing this: use the non-input shaper profile (basically the same as the MK3S's one); add a modifier over the height range where there are problems and slow it down there; and set the Dynamic Overhang settings in the profile to lower values such as 15%. All of them have about the same effect, at the cost of an increase in print time. For example, on gear 1 of SP13, the non-input shaper profile takes 68 minutes compared to 50 for IS profile.
use a modified fan duct. Most of the problems were on the side of the print furthest from the cooling fan. There are several replacement fan ducts on printables which direct air more evenly around the print, for examplehttps://www.printables.com/model/535705-prusa-mk4-fan-shroud-for-better-tpu-cooling.
The final and first try look like this, before any other cleaning up:
Another source of reduced quality comes from seams. You either use random seam positioning (as Steve recommends) and get zits on the surface, or aligned and get a scar. Both of them show up badly with silk filament. Another option is scarf seams, which overlaps the ends of the seam using variable depth extrusion. Go look it up for the details. This feature is not in PrusaSlicer yet (as on 2.8.0). Instead you can use OrcaSlicer, an experimental slicer that is ultimately derived from PrusaSlicer. To see the difference, compare these:
OrcaSlicer on the left, PrusaSlicer on the right, both before cleaning up and some temperature tuning. It's still not perfect, but a lot nicer. The strings were due to some filament stuck to the nozzle at the start of the print.
Hope this helps!
I'm in agreement that over-extrusion is at the heart of this, with the amount of extrusion depending on the filament. With the gears, I used Geeetech silk as shown above, and then redid them with Hatchbox PLA. Even the best of the Geeetech prints had some issues, for example when you look edge on at the gear teeth, there is sometimes a slightly bulge. The Hatchbox prints were perfect.
For the support column inserts, the problems I reported on myminifactory were with Hatchbox matte grey PLA, and go away entirely with the same Hatchbox regular PLA used for the gears. I checked the diameter of the filament for all three and they are quite accurate. All the filaments were dried recently, to eliminate another possible source of problems.
I think the extrusion multiplier in OrcaSlicer is "Flow ratio" under Filament>Basic Information, and it has a default of 0.98.
In each case there might be a temperature sweet spot that keeps the over-extrusion down, but I reached my limit on running experiments! I have used Hatchbox matte filament before on a project that needed very precise prints, and it was fine. But that was a different color, so perhaps there's something about the matte grey specifically.
Anyway, I'll leave this here for future reference in case anyone else encounters the same problems.