Friday, July 29, 2022

New clock experiments, part 2

For the next round of work on the new clock, I changed the configuration to have a frame which is shorter overall, and with the escapement set off to one side.

There is nothing holding the end of the side piece in place so it tends to splay out a little, but will do for this phase. (The illustration shows one version of the design. I used different minute wheels and weight drums in the experiments.)

For the first test, I made a minute wheel with a winding drum built into it, and set the clock going. With a weight of 560g, it ran for a few minutes and then stalled. Raising the weight to 680g, it ran until the weight hit the floor. The weights were not chosen with any care: I use a water bottle, and it's just a result of how much water I put in it.

Theory says that for a 50mm drum, you should unwind the string at 15.7 cm per hour if the motion is continuous. This means that for a typical 1.5 m drop, you would get a run time of around 9.6 hours. It's not enough for a practical clock, but will do while I am experimenting. I also added a weight drum on a separate arbor with 3:1 gearing.

(Yes, I know the string is tangled round the gear. It's the only video I took before disassembling it and I noticed it too late. It wasn't tangled when I did the tests.)

The 3:1 gearing triple the run time to 28.7 hours. However, I didn't get this arrangement to run reliably. It would usually stall after 10-15 minutes, though I did have one run lasting an hour. The weight I was using was 1340g (the heaviest I could get with a water bottle), and this probably explains it. You would expect something more like three times 680g, that is about 2kg, would be needed. I think the frame may also have been starting to distort slightly. Initially when I tried using the 3:1 gearing, I added a 20 tooth gear on the minute arbor, such that both it and the minute wheel were held onto the arbor with set screws. I could not get them to hold tightly enough and one or other would slip. In the end I glued the two gears together, and if I use this in future, I'll print them combined.

Finally, I put together a rough version a weight drum with a ratchet. I thought this would be straightforward, but ended up going through several variants. Looking at other clocks, I see three main styles:

  • the drum (which in all of these is within the diameter of the minute wheel) has a ratchet on on end, and there are pawls freely pivoted on the minute wheel. Gravity makes them drop into place. A lot of wooden clocks use this.
  • the drum has a ratchet on the end and there are sprung pawls attached rigidly to the minute wheel.
  • the ratchet is inside the wheel and the pawl are pushed outwards by spring. Used in Steve Peterson's SP5 (on a separate arbot).
The three styles are illustrated here:

(Credits: Jacque Favre Clock One, TheGoofy design on Thingiverse, Steve Peterson SP5).

I played around with the gravity approach for a bit, but found it hard to get the pawls to drop into place at the right time, and settled on using sprung pawls. They work at any position at any orientation and work well given that you can print thin springy plastic.

With a 50mm drum, I measured a weight drop of 8cm in 32 minutes, or about 15 cm/hour. There is something puzzling here as some other measurement showed quite different values. I also tried a drum 25mm in diameter. As expected, this halved the drop per hour. You would expect to have to increase the weight, but I got away with the same 680g, albeit with a rather weak tick. It would probably have stalled if I let it go on for longer.

I put the weight drum directly on the minute arbor for the tests I've just described. It is not a good way of doing things, as you really need the arbor with the weight to be supported at each end. That doesn't work well with some configurations of the hour train and works even less well if you are going to bring out a seconds arbor. As I mentioned before, a separate weight drum arbor with a 3:1 reduction didn't run without stalling, and so I decided to try 1:1 gearing. There should not be any problems with this, and indeed it worked fine. So this gives me the configuration I want to use for the next version, with the option that the 1:1 gearing could be changed.

Incidentally for some of these prints, I used a 0.6mm nozzle with an Arachne-based slicer (PrusaSlicer 2.5.0 alpha). Allegedly this gives as good precision as a 0.4mm nozzle. See Thomas Sanladerer's video for details. It seems to work well, though I still have a little bit of tuning to reduce stringing and blobbing, and it is more prone to producing elephant's foot. For now, it's good for faster prototyping, and I'll stick with 0.4mm when I want better accuracy.

Friday, July 22, 2022

Designing a new clock

It's about a year since I started making 3D printed clocks, with Steve Peterson's SP5 clock, and I decided it was time to try designing a clock of my own. I have modified some of the clocks I've made in small ways (adjusting fit) and in larger ones (replacing the motion work in the Swingtime clock). The only complete design I have done was the William Strutt epicyclic design, and even them much of it came from a published diagram. Now it's time to do something from the ground up. I don't know if I will carry this all the way through to a reliable design; it may turn out I don't have the skills or the patience to do so.

The starting point is a basic Graham escapement design with a rather conventional wheel train and motion train. I have a few ideas for ways I would like to refine it over time. The gear tooth counts are taken from this design on Thingiverse, itself remixed from a design by Thingiverse user TheGoofy. Having said that I wanted to design the clock myself, it might seem inconsistent to take the gearing from an existing design, but there are only so many possibilities which give you the right ratios and a number of teeth that you can manufacture. I like this wheel train as it allows you to add a seconds hand easily. Assuming a pendulum with a 2 second period (so about 1 metre long), you get this:

  • Escape wheel: 30 teeth, pinion 30 teeth.
  • Seconds wheel: 60 teeth, pinion 9 teeth.
  • Intermediate wheel: 72 teeth, pinion 10 teeth.
  • Minute wheel: 75 teeth, 16 teeth.
  • Reduction wheel (minutes to hours): 64 teeth, pinion 20 teeth.
  • Hour wheel: 60 teeth.
You have some choice about the gearing from the minute wheel to the weight drum. My initial design has an extra 20 tooth wheel on the minute arbor engaging with a 60 tooth wheel on the weight drum for a 3:1 reduction. If the weight drum has a diameter of 50mm, then it means that for each 3 hour rotation, the weight drops pi times 50mm, so that in 19 hours the weight drops by 1 metre. The pendulum length is something I will reconsider later.

I did the design in Fusion 360, using the standard add-in for generating the gears. The add-in only generates involute gears. I feel sure that I read something which said that cycloidal gears are better in the train from the weight to the escapement where each step increases speed and reduces torque, and involutes are better in the train from the minute wheel to the hour wheel, with the opposite characteristics. However, I could not find the reference again and there are many places which say to use involutes unless the teeth are so small they become fragile. This reference concisely summarizes the debate. If I wanted cycloidal gears there are tools like this one which output a DXF or SVG. I used a 14.5 degree pitch angle and a small amount of backlash to make the spacing between the teeth better. All of the pairs of gears are design to have their centers the same distance apart (60mm) so that I can stack several on the same arbor if I choose, leading to various different modules from 1.33 to 1.5. For the escape wheel and anchor, I used a parametric design a created a few months ago in Fusion 360, based on Jacque Favre's tutorial.

(Side note: what I actually did for the gear was use the add-in, and then project the profile into a new sketch. I could then project the pinion profile into a new sketch. This makes combining them a little easier.)

Here is a picture of a prototype, with a couple of minor parts missing:

The purpose of the prototype is to check everything seems to fit together. The weight arrangement is unfinished. It has no ratchet to allow for rewinding. The frame is very flimsy as I designed to print quickly. There are no bearings or attempt to reduce friction in this version, and the anchor is designed so I can experiment with different positions for the pendulum.

I printed part of this: the frame and the wheel train from the escapement through to the minute wheel, modified to include a weight drum. Originally I wanted to try this with various weights to get an idea of what weight I might need in the final version, but I had cut corners in the frame. So in the end it only ran for a few seconds with a weight hooked up. However, by applying force to the minute wheel by hand I was able to see it roughly working:

The beat isn't right and it takes quite a lot of force to make it run, but it gives an idea that things are generally correct.

More, hopefully, to follow.