Bartosz Ciechanowski
patreon
On "Mechanical Watch"
🕑 Added 2022-05-04 15:07:41 +0000 UTCI'm not even entirely sure where the idea to write about mechanical watches came from. I think part of it was inspired by drawing the watch in the GPS article, and part came from seeing ads for watchmaking sets.
I myself don't even own any traditional watches, so the topic was very new to me. Thankfully, it was easy to understand the gist of how mechanical watches work just from watching some YouTube videos and I decided to give the topic a go.
The watch movement I present in the article is ETA 2824-2 which seems to be a reasonably popular Swiss movement used in many watches. I'm happy with that choice as it contains some features like the date ring and automatic winding that don't exist in simpler devices, but it also wasn't too packed to make the mechanism overbearing.
When it came to recreating the parts in CAD, I primarily relied on basic drawings found in the movement's technical specification. However, I also bought a replica of this movement on eBay – that turned out to be invaluable. By measuring with calipers I was able to model everything in real units, I could look at sections that weren't visible in the PDF drawings, and having the physical object itself also helped me understand how some of the components work.
Unfortunately, in the process of disassembling the movement, the click spring jumped out and vanished, never to be found again. I ordered another copy of the movement as a functioning backup. This actually ended up being a blessing in disguise since I could dedicate the first one to a more invasive analysis. I had to irreversibly tear apart some of the components to see what's inside and understand how they work, but I could keep the second movement working to test its features. I don't think I'm actually planning to buy a proper mechanical watch, but that second movement I have will serve as a neat reminder of the work I did.
Watchmaking absolutely requires use of magnification – the parts are really tiny. Here's a picture of the pallet fork resting next to a US penny which has a diameter of 0.75 in or around 19 mm:
In a few cases, even using a 7x magnifying glass wasn't enough. I had to resort to taking a picture of the part through that magnifying glass, and then look at the photograph itself, e.g. to count the number of teeth in some gear. I've certainly gained a lot of appreciation for the steady hands and patience that people working on watches must have.
The CAD modeling process took a lot of time – I ended up with almost 80 individual parts. Some complex components, like the mainplate, took longer than others, but even some simple ones, like the stop lever, required a few iterations to make them fit into the final assembly.
Similarly to the article on Internal Combustion Engine, I ended up making models in Shapr3D on my iPad. I still like the app, but I feel that it's quite easy to get disorganized in it. The final design ended up being a total mess, which is perhaps more of a testament to my CAD management abilities:
I originally planned to include a watch case as well, but they're surprisingly much larger than the movement itself, so to make it fit I would have to make all the other parts smaller and harder to see. The idea of having to work on another full featured CAD part definitely played a part in the decision to stick to just the movement.
Modeling the parts is one thing, but making them realistically interact with each other on the website is another. For example, I spent a big chunk of time making sure that the teeth of the escape wheel rub against the jewels in the pallet fork without any gaps, or that the date corrector actually slides up and down in its groove and only changes the date when turned one way. These little details sell the simulation, but they were taxing to make.
For the model rendering I reused a bunch of tech from the Internal Combustion Engine article. I originally wanted to do something more ambitious with more realistic renderings of metal surfaces, but I think I'm happy with the current clean design. The colors work reasonably well together, although not all pairs of hues combine nicely, so it took some iterations to end up with the current state.
One of the hardest parts of writing the article was, as it's often the case, coming up with a convincing progression of topics. It was a real struggle to decide how to introduce the concept of time tracking in the first place. I originally wanted to start with a pendulum clock to present the escapement and the notion that some force has to act on the shafts for a clock to run.
I ultimately ended up abandoning that idea since I convinced myself that I can explain these concepts using the watch itself, but even then I had a hard time deciding if I should start with the balance as the source of actual time tracking and then show how it's used to unlock the escapement and how that escape is powered, or start from the other end by showing how things are powered first and then show how to make them track time. As you may have seen, I opted for the latter approach. I'm reasonably happy with that path, but I'm not 100% sure that I've cracked this explanation.
Comments
Cyril Perrotin
I've find your article by accident as a watch enthusiast myself and the level of detail is amazing. I've shared your article with a friend who works at Tudor (Rolex owned) and his colleagues from the movement team where impressed by your article.
meowmix
I joined your Patreon for this article. You did such a great job with it, well done!
Tommaso Poletti
Fascinating article Bartosz! And WOW for the animations!
Tommaso Poletti
It would be super nice having that code on GitHub so that we can learn math from you :D
Evan Fields
Another excellent article, one of my favorites so far. After reading this I was filled with lust to own a mechanical watch--an entirely new desire for me. Even after sleeping on it for a few weeks the itch was there, so I bought one with a clear face that reveals the internal mechanisms. Having read this article gives me so much appreciation for the incredible complexity therein, and really contributes to the enjoyment of wearing the watch.
Bartosz Ciechanowski
I currently don’t have any writeups on the topic, but I don’t minify JavaScript on my website, so at least the code is easily accessible
David
Just incredible. I love mechanical watches and have followed your blog for a long while now. I was so pleasantly surprised to see these two worlds collide. I'm new to your Patreon. Do you have an article with technical information on how you render your models to the page with physics?
Nikita Lesheno
It's a jaw-dropping article, both on the explanations front and the satisfying attention to detail in the animations. For example, right after "the side-to-side movement of the pallet fork is limited by [...] the pallet fork bridge", I was surprised to see that you simulate the pallet fork actually pushing the escape wheel around when you play with the slider, instead of playing an animation back and forth. I created a mechanical clock animation a while back (https://clock.leshenko.net) but it was way lower quality and things didn't really line up there. Makes me appreciate even more the fact that everything is pixel-perfect in your article.
Shane Wighton
Gotcha. If you ever look at more complicated projects on a desktop, solidworks has a $99/yr fully featured version for makers. Also check out the 3D connexion space navigator for much better camera navigation than a mouse
Nikita Vasilyev
Fantastic work!
Bartosz Ciechanowski
Thanks Shane! I used a bunch of CAD software back in college days, but my recollection is that most of it is quite bloated and expensive. The killer feature for me in Shapr3D on the iPad is the ease of camera movements with hand gestures and the unbeaten comfort of drawing sketches with the pencil. I probably wouldn't do more serious projects on a tablet though.
Shane Wighton
Hey, big fan of your posts. Very information dense and useful presentation. I can tell you put an enormous amount of time into them. Very rare and refreshing to see. I was a bit surprised to hear you modeled all of that on an iPad. Have you tried using desktop CAD for any of your projects (e.g. solidworks, inventor, onshape, etc)?
Nikita Prokopov
Amazing work!
jakob
When I think about it, the approach of starting from the main spring works well because it runs on its own, and is the solution to a familiar problem: "How to store power/get motion". The solution then creates a new problem (runaway motion) which is solved again by the escapement. Starting from the other end would have meant starting with a solution for a problem the reader hadn't seen yet, which is unsatisfying.
Lukasz Rewerenda
Fantastic stuff. My hat is off to you for a low level approach to implementation. I think its worthy of an article itself. Thank you!
Samuel Lough
Starting with the mainspring helped me more than other explanations have in the past. I found I would get the mainspring/balance spring confused but your explanation path was intuitive. great work.
