As a reminder, even though MoT's is a free site, we do rely on two methods lights on. One is a referral generator that creates (mostly) unobtrusive links. This depends on the site being whitelisted in any ad-blocking software you use. If you could please whitelist this site, we would be ever thankful. (We promise there are no pop-ups, pop-unders or other annoying distractions.)
The other method is donations from members and visitors. These are completely voluntary and there is no obligation. Any donations go directly towards site operations and are greatly appreciated. Click here to donate today.
You may be familiar with my enthusiasm for the mechanical aspect of horology from the Haute Horlogerie Discussion thread. Last summer I decided to immerse myself deeper into that field by designing, fabricating, and building my own clock from scratch. I had little to no previous experience with mechanical design or fabrication, so I anticipated a process fraught with setbacks and trial-and-error. As you can tell by the fact that we're now a good ways into 2020, I wasn't far off the mark. My goal was simply to build a clock that could function as such on a day to day basis, ideally having at least 24 hours of power reserve and telling time to within +/-15 minutes per day. My only design restriction was my refusal to use potentially troublesome springs, ruling out the use of a mainspring or a sprung balance.
This post describes how each component of the clock evolved and how the project ultimately came to failure. Pictures are at the end.
Not having experience with or access to metalworking equipment, I decided to produce the clock with a 3D printer. In addition to allowing me to produce fairly complex parts, it afforded me the luxury of rapid prototyping, essential for an inexperienced designer like myself. In fact, producing parts was so easy that I found myself getting too lax in the design process and making silly mistakes. After some research I chose the QIDI X-Plus 3D printer, chosen primarily for its outstanding price per cubic inch. It's about the largest hobbyist printer available. I'll admit I was slightly dubious at first, but I quickly warmed up to it. It hasn't given me anywhere near as much trouble as I expected. I decided to print in PLA, partly because that's what was included with the printer, but mostly because it's the most versatile and widely available type of filament. The downside to 3D printing is its relative lack of precision, ensuring that there would always be some minor errors in the mechanism and forcing me to design on a very large scale.
I designed the clock in Autodesk Inventor, a sketch-based application for "product design and engineering." Prior to undertaking the project I had some experience with Autodesk Maya but none with Inventor. The learning curve wasn't very steep, and while I'm far from professional with it I've been able to produce the parts required for the project without too much trouble. It also provided me with an invaluable tool: a spur gear generator. All the gears in the clock were produced with this tool using parameters such as the desired ratio and the distance between centers.
The parts I couldn't make myself I had to buy from outside sources, mostly Amazon. These parts were: the stainless steel rods that carry the gears; the small bearings for the gears, which were originally designed for high-performance RC cars; the large bearing for the pendulum; the chain holding the driving weight; the driving weight itself and the weight used in the pendulum bob; and the small spring used in the winding mechanism. All of these materials operated as they should without trouble. I was surprised, however, by the extremely close tolerances between the rods and the bearings. Some bearings simply wouldn't fit on certain rods. To give some scale, a small Sharpie mark I made on one of the rods significantly increased the friction between it and the bearings.
The Design Process
Because of my inexperience with mechanical design, the design process was highly iterative. I started by breaking the project into two stages, first simply designing a stripped down version of the clock (basically just a running seconds hand), then designing the rest of the clock around it. It took many iterations to get the stripped down movement working, but after that the rest of the mechanism was largely dictated by its structure. Unfortunately, this turned out to be a crippling fault.
The Overall Structure
Each part of the mechanism is detailed below, but I thought I should give an overview of the final design here. The movement is in a very unusual shape; if a usual movement can be thought of as linear, mine forms more of a cross, with the gear train forming a straight line from the back to the front with the hours and minutes extending from the sides. As it turned out, this was an unfortunate choice, but it was made for its simplicity. It also aligned with the aesthetic design I was shooting for. The focus of the project wasn't so much on telling time as on keeping time, which I chose to highlight by placing the seconds indicator in the center of the movement/dial and moving the hours and minutes to the edges. The gear train itself is also quite strange; instead of being laid out on a plane and sandwiched between two plates it's placed entirely on two parallel bars, necessitating a constant center distance for all the gears. This distance is dictated by the escapement, which happens to require a depth of 82.435mm.
Because of my refusal to use springs, a sprung balance was out of the question. The only alternative was, of course, a pendulum escapement. I chose an anchor escapement, so called because the pallets form the shape of an anchor, called the "recoil" escapement. The recoil escapement is the very simplest pendulum escapement, but it has some pretty serious drawbacks. With every oscillation it pushes the entire gear train backwards, hence its name. This results in a lot more wear and less accuracy. A "deadbeat" escapement doesn't suffer from this problem, hence its name, but it's a good deal more complicated in its design. To be honest, what with the poor materials and relative imprecision of my mechanism, I don't think a deadbeat escapement would have made much of a difference. I decided to simply accept the faults of the recoil escapement. I designed mine to the British Horological Institute's specification.
The pendulum is just as important as the escapement. Using simple equations from physics I calculated that the pendulum should be about 690mm long, but that's just a theoretical value. There still needs to be a way of adjusting the pendulum's length for fine regulation. First I tried using a threaded steel rod, the idea being that I could thread the bob up and down to adjust the length. It turned out, however, that the rod was much too heavy to serve as a pendulum. I replaced it with a wooden dowel and instead of threads used set screws to attach the bob to the pendulum and the pendulum to the escapement, allowing small adjustments in length.
For the pendulum bob I initially used a 200g scale calibration weight (which I bought off Amazon) but later decided to replace it with a 100g weight so that it wouldn't require as much energy to push the pendulum.
The power for the clock actually turned out to be the hardest part, or at least the one that required the most iterations. My refusal to use springs necessitated the use of a driving weight like in a grandfather or cuckoo clock. My first idea was to attach the weight to a string wrapped around a drum, but as soon as I tried it in practice it became obvious that it was a terrible idea. The string wouldn't wind properly and liked to tangle and fall off the drum. The system also would have also required a key to wind the clock. It went into the scrap bin very quickly.
My next idea was to replace the string with a chain. Instead of using a regular straight chain I used a jack chain from the hardware store. The barrel that I designed for it used spikes on either side of the chain to "embrace" it. This arrangement actually worked quite well, but I wasn't happy with how the jack chain looked.
Finally, I bought a proper clock chain from Timesavers.com. I really should have done it at the beginning to save myself the hassle of designing two different barrels. Actually, the clock chain barrel was a good deal harder to design than the jack chain barrel, mostly because the clock chain was so much finer and the barrel had to skewer the links instead of embracing them. For a long time I had trouble with the "spikes" of the barrel not lining up with the links of the chain. After the fact, I realized that I could have made my life a lot easier with just a bit of forethought. In CAD, the circle forming the outside of the barrel is defined by a diameter, and the distances between the teeth protruding from the circle are defined by angles. Given the diameter it's pretty easy to calculate the angle corresponding to the required distance between the teeth (in my case, 8mm) using the formula θr=a, where θ is the angle, measured in radians, r is the radius of the circle, and a is the length of the arc on the circle encompassed by the angle. However, with an arbitrary diameter (and therefore circumference), this results in the barrel requiring a non-integral number of teeth. This means that as long as I chose arbitrary round numbers for the diameter of the circle, the distances between the teeth were never quite right. I should have made sure that the circumference of the circle was a multiple of 8mm, ensuring a whole number of teeth. Live and learn.
For the driving weight I initially used a 500g scale calibration weight like the one used for the pendulum bob. The ~250:1 reduction between the barrel and escape wheel, however, reduced the power to almost nothing. I switched to a 1000g, then ultimately to a 3000g weight.
After I built the first prototype I quickly realized that a rigid frame was essential to prevent the driving weight from warping the entire movement and throwing the gears out of mesh. For the subsequent prototypes I simply used a sturdy rectangular frame supported by four corner pieces securely screwed into place.
A number of the clock's components need to be securely fastened together. For example, every wheel is attached to a pinion, and the barrel is composed of four separate parts. Some parts only need to be held together by the friction of tapered pins pressed into holes, but others need to be more firmly attached or simply don't have enough space for such pins. For those pieces I used Loctite plastics glue or Super Glue to bond them together.
The only exception is the corner pieces holding together the front and back of the frame. These are aligned by square pegs and held together with screws, allowing me to easily disassemble and reassemble the frame. In order to give the screws sufficient purchase, I printed the corners with a lot of extra outer layers (instead of the usual hollow support structure).
All of the systems that the user interacts with (that is, those responsible for setting and winding the clock) work on the same principle. The barrel is fitted with a ratchet, allowing it to turn in one direction but not the other. This allows the driving weight to exert its force on the mechanism in the proper direction and to be easily reset when it ultimately reaches the floor. The ratchet is retained by a medium-sized compression spring.
Actually, my system is nearly identical to the way the winding pinion and clutch mesh together in the keyless works of a watch (pictured below). When you turn the crown clockwise they mesh and wind the watch, but when you turn it counterclockwise the crown spins freely and you can hear a ticking noise. That's the sound of the angled teeth of the winding pinion slipping over those of the clutch.
Though I didn't complete the hands of the clock, I did have (in my opinion) a rather clever plan for a simple setting mechanism. The hands would be connected to shafts pressed against the gears of the mechanism by springs similar to the one used in the barrel. Where they met, I would attach small silicone rings, which would stick together, carrying the hands with the gears. To set the hands one would simply push the gears backwards toward the wall, separating them from the hands and allowing the hands to be adjusted appropriately. I devised this unorthodox system because my printer can't fabricate ratchets fine enough to allow the hands to be set with an acceptably fine degree of precision.
In the end, my poor design choices added up to an impasse. I completed a running seconds module that could run indefinitely, but after that I was stymied. Not that the clock couldn't be completed—the way forward is obvious. But the final product would be cumbersome, ugly, and embarrassingly inelegant. The picture below approximates what a final clock would have looked like.
The main problem is the fundamental structure of the clock. While the running seconds module looks fairly elegant in itself, there is no way of conveniently expanding from it. Its constant gear depthing imposes serious limitations, as does the shape of its frame. In order to add any kind of functionality it would have to evolve into a weird "Double-Stuff" variation on the traditional planar structure of a clock, with all the gears sandwiched between two plates. I didn't see any point in pursuing that path, so I was forced to simply pronounce the project a failure.
As I mentioned in the introduction, I wasn't expecting this to be an easy project. Despite its ultimate failure I had a lot of fun, and learned a good deal to boot. Additionally, I've gained a new respect and appreciation for the extraordinary haute horlogerie pieces I enthuse over. For your part, I hope you've found this post at least somewhat interesting. I know didn't go into too much detail about the actual design and construction process, so feel free to ask questions. Following is a collection of images from over the course of the project.
The first prototype, which could only run for about 30 seconds. The very very first prototype lacked the robust white frame and was held together only by two thin "struts."
A profile view, featuring the jack chain barrel. Note how **** the gears are. This is because the bearings they ride on were crooked in their seats—the problem was ultimately solved by providing shoulders from them (the bearings) to sit on.
The second prototype (in truly appalling lighting). Though it looks very similar to the final prototype (see below), it's actually quite different. Among other things, the gears still haven't been stabilized and the frame is only held together with friction.
The fourth and final prototype. A fairly elegant thing, if I do say so myself. Pity it's no good for telling time. Its longest trial run lasted 36 hours, and I only stopped it because it was too loud to sleep.
The final barrel and ratchet mechanism. Compare the ratchet to the one out of a keyless works, pictured above.
The 3kg lead driving weight that powers the clock. Also visible is the straight chain that replaced the jack chain of the first prototype.
A close-up view of the left side of the escapement.
The right side.
A profile view. Note the much better spacing of the gears.
My embarrassingly large collection of failed, rejected, and deprecated parts. Girard looks on with a mixture of outrage and disgust as the pile grows.
- The Sultan of SoWhat
- Posts: 2594
- Joined: Wed Dec 19, 2018 8:52 pm
- Has liked: 2035 times
- Been liked: 3279 times
- Posts: 307
- Joined: Tue Dec 18, 2018 5:57 pm
- Location: Missouri
- Has liked: 321 times
- Been liked: 870 times
Also if you pick this up again and need any help with Inventor, I've been using it every day of the week for the last ten years for my job, so I'd be happy to help if I can!
Sent from my moto g(6) using Tapatalk
Anyone with an education or interest in engineering would be fascinated by multiple aspects of you project, including design and execution. Others would be interested in what makes you tick, no pun intended. Obviously this isn't something one enters into lightly and I suspect that is why you use the term "failure," while the rest of us don't know where to begin with even a response!
Thank you so much for sharing the highs and the lows, I really enjoyed the read. What an impressive undertaking.
Sent from my LG-M322 using Tapatalk
- Posts: 2364
- Joined: Fri Dec 21, 2018 12:43 pm
- Location: Munich & SF Bay Area
- Has liked: 4503 times
- Been liked: 3084 times
I think it looks great and you learned a lot so that is a win. People just don’t do this kind of thing, so just that you took a crack at it should be seen as a major achievement.
In thinking of how I would do this, my first thought was to make something on a much larger scale, possibly water driven. I am not sure why that is where my mind goes, but that is where it went. I might go super low tech and make a graduated hour glass out of a couple of 5 gallon jugs. Neither of those would be mechanical in the way you were thinking, but mechanical in other ways. Also, very much less practical and couldn’t be hung on a wall. OK, I could see the giant hour glass installed on the wall of a museum of modern art or something. They love **** like that.
At any rate, great job on the solid attempt and for documenting your adventure and sharing it with us.
That's quite a generous offer! Though I do hope to make another stab at it, it won't be for a long time. I'd like to learn more about clocks and how they're generally constructed, and possibly look into better materials. I really do appreciate your offer though, and who knows? When the time comes, I might take you up on it!
It took some doing, but I think you should find a video here.
Probably more recently than you think! The 3D printing community seems to take some interest in clockmaking, with many designers making their mechanisms available gratis on websites like Thingiverse.com. Check out projects like this one or this one. The most impressive mechanism I've seen is this one, which even features a tourbillon!
You'd also be interested in the story of Michael Blayney, an anesthetist living on the Isle of Man. He struck up a friendship with George Daniels and ultimately decided he wanted to make a horologium of his own. Daniels challenged him to produce a marine chronometer, so he bought a lathe and, after five years, produced a truly high-grade mechanism. Check out this extremely interesting interview.
Finally, though he didn't design his own mechanism, I highly recommend this series by a YouTuber named Chris, in which he fabricates a small pendulum clock from scratch. He's an extremely admirable craftsman, and produces very high quality content. Here's a video that gives an overview of the process from beginning to end.
Perhaps you'd enjoy this episode of a show called "James May's Man Lab," in which Mr. May and his colleagues design and build a water clock capable of telling the hour and minute! It's light entertainment, really, but entertaining all the same.Sporkboy wrote: ↑Wed Jan 15, 2020 5:42 amIn thinking of how I would do this, my first thought was to make something on a much larger scale, possibly water driven. I am not sure why that is where my mind goes, but that is where it went. I might go super low tech and make a graduated hour glass out of a couple of 5 gallon jugs. Neither of those would be mechanical in the way you were thinking, but mechanical in other ways. Also, very much less practical and couldn’t be hung on a wall. OK, I could see the giant hour glass installed on the wall of a museum of modern art or something. They love **** like that.
- Site Admin
- Posts: 2157
- Joined: Tue Dec 18, 2018 11:27 am
- Location: Minneapolis, MN
- Has liked: 1390 times
- Been liked: 5833 times