Just Wondering....

If it works using the full weight shell on a single string, then there are two quick fixes. Double the drive weight or cut the runtime option in half. This assumes you are not already using the largest reasonable weight shell or the shortest available runtime option.

The Easy Build Clock has the most stable frame of all the wall clocks. I suppose it could sag if the wall screw is too loose. Or if you hang 25lbs on it. :)

Engaging friction is mysterious. When it occurs near the escapement, adding additional drive weight doesn't always help. The teeth just jam together harder. Frame sag induced problems act the same way. Adding additional drive weight makes the frame sag even more. But at least frame sag has a visual clue when adding extra weight.

The 5 year clock has metallic bronze colored PLA, but I don't know if it was labeled "silk".

I had the medium escapement running overnight with a 5 pound weight. It seems reliable, but should have a bit more safety margin, possibly around 7-8 pounds. I will probably do a bit more testing and release the new escapements as an update to the design.

A major upgrade is in progress for the 32 day Easy Build Clock. It gets the new gear tooth profiles throughout, plus alignment tabs added to the frame components, and the new more secure winding key. The prototype is being tested with 6.5 pounds in 32 day mode and is robust with 8 pounds. It was getting close to being released before this Moon Phase side project. Hold off on printing any Easy Build Clock parts for another week or two. The parts kit will also get an update to go with the new changes.

Is your clock stopping because the cord is binding up? Or does single string mean that you are doubling the effective drive power? What the clock is doing when it stops is important for debug. The pendulum should free-swing for at least 10 minutes, so if the clock stops within a minute, then there must not be any power getting to the escapement. The upcoming release improves the gear engaging friction, reducing the frequency of a "dead" escapement. It only slightly improves the cord management by increasing the winding drum from 0.75" to 0.88" in 20 day or shorter modes. It reduces the winding drum to 0.55" only in 32 day mode. I do not have much problem with the braided fishing line listed in every manual. I do sometimes guide the cord when winding so it doesn't all pile up in one spot.

Correct. Izz is the only one that matters since it is the MOI about the rotational axis, assuming of course you have made your arbor your Z axis in CAD. To clarify, in my analysis, I did not reduce the size of the teeth. Only the outer most ring that the teeth are riding on. So no change to the locking and impulse surfaces at all.

I was going to ask you what CAD program you were using, because I notice nearly all of your curved surfaces are significantly tessellated, such that the flats are huge. This is especially troublesome I think in the ratchet. The inner surface that the paws slide on is tessellated, which is likely contributing to the jamming issue. It is most noticeable on the outer surface of the clock face, which is the largest circle. But I can even detect it on the escapement gears.

I use Fusion 360, and aside from Solidworks, is in my opinion the best out there for DIYers. Super cable CAD program with CAM, Analysis, Simulation, Rendering. It does it all. You can use it for free, with some limitations, to try it out.

In Fusion, the STL are exported with zero detectable tessellation on any of my curved surfaces, no matter how large. Round surfaces 3D print perfectly round. Maybe in your CAD program of choice, take a look at your STL Export settings, particularly Chordal Tolerance and Deviation and Angular Tolerance and Deviation.

Here is a pretty good guide by Markforged: How to Create High Quality STL Files for 3D Prints

I use TurboCAD Platinum with the faceter quality set to 100. There are expert modes that allow higher detail, but file sizes start getting huge.

Most gear generator programs approximate smooth curves using hundreds of tiny line segments. I always draw over these with a few arcs that comes within about 0.0001" of the initial shape, which is significantly smaller than the 3D printer resolution. The STL file will end up tessellated anyway. 3MF files might preserve the arcs when it becomes more supported.

I have started simplifying some of the unimportant gear surfaces. The rim is now made up of straight line segments instead of curved lines. There are only a few surfaces like the dial that could benefit from higher resolution.

As you mention, the escapement teeth only have two important surfaces. They can be left intact while making everything else narrower. I did this on the medium sized pallet. Material was removed from the unused surfaces.

The free version of Fusion 360 only applies if you have less than $500 in sales per year, so I don't qualify. It might be worth it even for the annual cost.

On file size, I'm wondering if TurboCAD handles STL differently than Fusion. My higher quality STL's aren't that huge. Totally manageable.

Ahh, that's true on free Fusion. I neglected to consider that this is a business for you, lol. I use it strictly for my hobbies, but I easily justify the cost in what it allows me to accomplish. There is no way I could do all the design, analysis, CAM, 3D printing, laser cutting, CNC, etc, etc, without it. I pay for a full seat as it is worth it to me.

On another subject, I suddenly am not getting any notifications on this forum. I checked my settings, nothing changed. Any ideas?

Extra weight is harmless. I just tested mine at 11 pounds. It gets louder and the pendulum amplitude gets bigger.

I forgot to mention, the files are optimized for the classic slicing algorithm. The Arachne engine keeps changing and adding extra gap fill any time the software updates. The classic slicer is much easier to keep clean print lines.

I use brands in the $18-22 price range, Overture, Hatchbox, Eryone, etc. None of them are labeled PLA+. I tried some $9 spools and decided it wasn't worth it. One had grooves along the length so it would underextrude.

btw: The first experiment with 3 pounds stopped running after 2.5 hours. The escapement didn't move when the pendulum was moved back and forth. It started up again before I could identify the stuck gear. I switched it to a 5 pound weight and the pendulum amplitude is much stronger. I expect it to run reliably.

This update needs two copies of pendulum_arm_mid_a to tune the rate. There isn't enough range in the stock arms to get the extra 0.6" of length.

Yes, PLA. I left PETG experiment on the back burner for now.

Now it makes sense lol... I sent some messages to Steve but not sure if he saw them either. Oh well now I know :-)

Now I understand why :-) btw I forgot I had put the 21 day gear in the Easy Build. Reprinted the 7 days with 5lbs. and it's been running fine for a few hours now double stringed.

Also moon clock running strong. All is good in the world lol...

Good to hear. One small issue with the Easy Build clock is a 4:1 ratio in runtime options, but it is very rare for anyone to mention which one they are using when asking questions. :)

I will add labels to the gears before releasing the update to make the selected option more obvious.

I took the clock apart again. I also shortened the string to only the amount needed to hit the floor. There was still a lot of string left. Also I noticed on the ratchet the pinion seemed a tad bent. Could be because when I built that clock I used 3 mm aluminum rods instead of stainless steel rods. I couldn't get the stainless steel at the time. Now that I have lots of it I went and replaced that pinion and also the one in the bottom wind up gear. As that was also aluminum. And somehow I lost one of my springs in the ratchet. SoI replaced all new springs. It appears my EW is very well balanced once I found my old propeller balancer lol... Big RC plane and RC helicopter guy here from another life time ago. Was reading a lot of PLA creeping articles last night. But I'm going to see how it goes with what I did just now. If it still stops, I'm going to print over in PETG CF the bottom gear first since I'd is taking the heaviest pull of weight and right now it's in PLA it might have already started to sag and I just don't see it. Weight shell is currently at 7.455 lbs.

Do a simple test the next time the clock stops. Gently move the pendulum to one side to see if the escapement moves. Excess engaging friction could cause the escapement to stall. It may even be moved slightly backwards so it is not even touching the pallet.

Aluminum can have a lot of different alloys and hardnesses. If you are using material from an RC plane, then it is likely good stuff. It is easy to tell if it is bent.

I am still working on the assumption that the root cause is engaging friction in the involute gear tooth profiles combined with a high inertia escapement. Plus, it appears that PLA may change over time causing it to get stickier. I don't know if PETG is better or worse.

Normally there is enough power reserve to keep running even with a small amount of engaging friction. You can often hear the power fluctuations where the escapement gets quiet. Sometimes it loses a beat and recovers. Other times the clock stops.

My latest experiment is to reduce the escapement size and change the gear tooth profile to a cycloidal modification that I have been experimenting with. I know it is better to only change one thing at a time, but both changes should be improvements.

Reducing the escapement size requires changing gear ratios.

The existing design uses three sets of 54:12 gears with a 25 tooth escapement giving 4556.25bph with a 24.48" pendulum.

The new design uses three sets of 60:12 gears with an 18 tooth escapement giving 4500bph with a 25.09" pendulum. There is enough backlash in the design to squeeze in the extra gear teeth and just enough clearance in the frame for the new slightly larger gears.

The escapement had two options to fit the existing frame with a 5.7" distance to the pendulum pivot point. I started with the smaller 3.0" option. It has been running for over 90 minutes with 3 pounds of drive weight. It should have around 5 pounds to run reliably. Changes were made from the pallet down to gear 4.

Here are the files for this experiment.

The escapement looks a bit small, so I am working on the 4.8" version to see how much extra drive weight will be required. I will try to post files for this change later today.

Steve

I'm sure Steve will chime in as well but,

the double bounce on the entry pallet occurs on my clock as well. Although it's a tiny loss of energy, I would say that is not your problem stopping the clock. The double bounce I believe is a combination of a bit to much elasticity between the pallet and the escape tooth, coupled with a bit to harsh of a drop, coupled with a low drive torque on the EW which permits the EW to reverse direction upon bounce. The exit side of my escapement doesn't bounce. Because I think the drop is a bit less harsh on that side, and the elasticity between the pallet and the tooth are on opposite faces, which affects the lock geometry and collision dynamics of the left side compared to the right.

The missed beat is a major issue. That shows that there is not enough torque at that instant to drive the EW adequately during the entry of that escape tooth. But amazingly, there seems to be pretty good energy in the EW the rest of the time. Does adding just a little bit more weight, say 1/2 pound more, eliminate the missed beat?

To me, it looks like something is hanging up the EW at that instant. Important question to try and find culprit; does the missed beat happen only once, at the same tooth on the EW, each rotation of the EW? Or does the missed beat occur twice during one full EW rotation? If only once during each EW rotation, then I would rule out the EW pinion, since it would drive the next gear down a bit over 4 times for each EW rotation (54/12). If it happens randomly, then the issue may be further down the going train. If it happens at the same EW tooth each time, I would first claim EW imbalance, or something else weird is going on with the EW or between the pallet and the EW at that spot.

I agree with you sosalty62.

I have seen the double bounce, possibly on the moon phase clock. My interpretation was that the escapement did not come up to speed quick enough to impart energy to the pallet. Too much static friction, or too much weight caused it to start spinning slow. The pallet had mostly moved out of the way by the time the escapement started really moving. The tooth tip barely touched the pallet. The escapement keeps gaining energy until the locking side hits. The pallet has a slight imbalance in operation. Both arms are the same length cosmetically, but this skews the active surface distance by the thickness of the pallet teeth. I may go back to unequal length arms with equal length active surfaces.

In my clock with the double bounce escapement, there was a gear stack pinched by the frame and stealing much of the energy. Fixing the end shake solved most of the double bounce by allowing the escapement to engage better with the pallet. The heavy escapement may still be a problem to be fixed.

The missed beat in the video is concerning. The clock may be slightly out of beat. An unbalanced wheel would change the energy as it turns. It may also be a design issue with too much engagement friction in the gear pairs. A 25:12 escapement to pinion ratio means that the pinions will engage at a different position for every escapement tooth. One of them will be the worst case. I have some experiments with a different gear tooth profile that may reduce this.

That seems like a lot of mismatch that certainly should be corrected. The CAD model is symmetrical in the moon phase clock. Do you think it is caused by a tilted Z axis? The Ender 3 printers are similar to Prusa with the power supply serving as the brace holding the frame vertical. It may also depend on the position of the Y axis rails.

It is a good idea to balance the escapement. Wooden clock builders mount the gears and gently blow on one of them. An unbalance would show up during this test.

Your comment makes me wonder if any of my clocks have unbalanced escapements. I used to distribute the spokes equally and let them randomly hit the escapement teeth. I recently started placing the spokes to land between teeth for better printing. 25 teeth with 5 spokes is balanced. The latest clock I am working on is the 32 day easy build clock with 26 teeth and 5 spokes, so it is slightly unbalanced. However, it is running in 32 day mode with 6.4 pounds of drive weight. The escapement is small but could be balanced better. Thanks for the suggestion to get me thinking about this.

My printer is good. I print very complex mechanical parts all of the time and never have any issues with alignment, fit, tolerance, etc. I printed 4 EW's on the moon phase project, and every one of them was imbalanced. I could not figure out the source of the imbalance, but I have some ideas. For one, I have not been using "Random" for my seam position. I use "Aligned", which puts the seam in one particular spot around the ring. It could be the accumulation of a tiny bits of extra PLA in that one spot, added up over all of the layers. Or it could just be the nature of 3D printing; minor variations in over/under extrusion during a print, etc. I will need to print an EW with "Random" and compare to rule out seam position as the culprit.

As far as checking the mass balance of the EW model itself prior to printing, I could do a CG analysis on the model to see what the computer says about it (to make sure CG is on exact rotational axis in your model. But I don't have the STEP file of the EW and can't do that analysis with the STL. Your CAD program should have a mass analysis command. Fusion does it easily. Maybe check CG on EW model and report back on your findings. If you would like me to do that for you, I will need the STEP for the EW.

Below is a link to a video I made when I was testing and balancing my EW. The setup shown has the pallet removed and a 0.9 lb weight hung single string. That was the minimum weight needed to have the EW stop due to the imbalance, depending on the rotational position of the EW. By iteratively reducing the weight this way, the imbalance can be made to show up and be detectable, otherwise the wheel would just spin up despite the imbalance. At minimum weight, you can also see that, once the wheel picks up a tiny bit of momentum, it is able to overcome the tiny imbalance, along with dynamic friction kicking in, and the wheel spins up.

With that setup, you can see that the red side of the EW is approximately where the "heavy" side of the EW is. Watch as I slowly rotate the EW to get the imbalance to rise up and over the side of the wheel, enough such that the 0.9 lb drive weight can overcome it. Until it comes back around to the bottom, at which point, the torque on the EW is not enough to "lift" the imbalance back up and over the top. Repeat and repeat to verify. With that issue verified on the wall, I removed the EW and performed the balancing procedure. All super good after that. Clock runs like a top right now. Now, I can literally blow lightly with my breath from a foot away and cause the EW to rotate.

None of the other gears need this level of attention. They have so much torque compared to any gear imbalance that it is in the noise. But for the EW, that is a whole different story. That sucker is super sensitive to reliable performance, at least on my clock it is.

Link to EW tests

That is very interesting. And a bit worrying. The large moon phase clock escapement is balanced by symmetry. I draw one tooth and array it 25 times. One spoke is drawn and arrayed 5 times.

I use TurboCAD Platinum 2018. It only has a 2D analysis modeler. The escapement does show that all parts are 2D balanced, as expected. Fusion 360 didn't even exist when I started clock design. I don't qualify for the "free" version, so I will probably stick with what I know how to use.

So, how can the CAD model be balanced, but the printed part is not? One guess was Z axis alignment, but I will take that back. Z axis alignment could cause wobble, but it should still be balanced, Could the print cooling fan be skewing the weight distribution by blowing from one side? Or maybe the fan is changing the center hole position? Could dual and tri-color PLA be skewing the weight distribution?

The standard methodology for clock design is to make the escapement small and lightweight. It looks so good when it can be such a large part of the design, so I usually make it large. Maybe it should be made smaller on new designs. I will look into this for the moon phase clock. Changing the gear ratios between the minute hand and the escapement from 4.5:1 to 5:1 would allow an 18 tooth escapement with 4500bph. The pendulum would need to be 0.5" longer. The escapement could have a 3" diameter with a 7.5 tooth span, or a 4.9" diameter with a 6.5 tooth span. The existing escapement is 5.6".

PETG is worth a try. I also have some on order that will be here Wednesday.

My suspicion is that involute gears have more engaging friction than cycloidal gears. This is where a new tooth starts to engage. If it engages before the line of centers, it pushes into the other gear and friction can be higher. Any engagement after the line of centers is pushing away with low friction. This effect can be seen in the ratchet that often sticks in one direction but easily slides in the other direction. A thin layer of lithium grease solves the problem in the ratchet.

I believe the same engaging friction issue is happening with the involute gear tooth profile used in many of the clocks. A brass gear with steel pinions does not dig in like softer PLA. That is why I have been trying new gear tooth profiles. If it works, then I will populate the gears to some of the other clocks. The initial results look promising.

The pendulum free-swing test will quickly determine if the bearings are good. I have never seen less than 98% of the cheap bearings pass the free-swing test. Ceramic bearings probably work better but should not be required.