https://fellerts.no/projects/epoch.html
Fredrik Flornes Ellertsen
Projects V
Kintsugi repair Snus reduction EOG + REM sleep Nixie alarm clock 2048
algorithm Design notes
Contact
Mechanical Watch: Exploded View
In May 2022, someone posted to Hacker News Bartosz Ciechanowski's
blog post explaining how mechanical watch movements work. Since then,
his blog has been my absolute favorite corner of the Internet. His
posts are not just well written and easy to follow, the accompanying
interactive illustrations are magnificent.
The first illustration in his blog post about mechanical watch
movements allows you to "explode" a ticking mechanical movement and
rotate it to inspect its every component from any angle. I owe my
foray into the hobby of watchmaking to Bartosz, but that's not what
this blog post is about. Instead, I want to scratch an itch I've had
for years now: How cool would it be to hold an exploded view of a
real mechanical watch in your hand?
[Bartosz_Ci]
Screenshot of the exploded view render from Bartosz Ciechanowski's
blog post.
Apparently, nobody builds such models
I figured that surely, someone has had this idea and built it before.
On eBay you'll find cubes of resin embedding various random
components from mechanical watches, but they are typically sold as
"steampunk art" and bear little resemblance to the proper assembly of
a mechanical watch movement. Sometimes, you'll find resin castings
showing every component of a movement spread out in a plane like a
buffet---very cool, but not what I'm looking for. Despite my best
efforts, I haven't found anyone who makes what I'm after, and I have
a sneaking suspicion as to why that is.
[watch-buff]
Every component making up the movement of a beautiful 1960's Longines
Flagship wristwatch I worked on recently, laid out "buffet style".
Building an exploded view of a mechanical watch movement is
undoubtedly very fiddly work and requires working knowledge about how
a mechanical watch is assembled. People with that skillset are called
watchmakers. Maker, not "destroyer for the sake of art". I guess it
falls to me, then, to give this project an honest shot.
...how do you build such a model?
Here comes my favorite part: jumping head-first into a project
requiring a set of skills I don't even know exist, let alone possess,
following a process that develops as the project evolves. So how
would one go about building a real-life exploded view of a mechanical
watch movement?
The first thing to consider is what type of mechanical watch movement
to "explode". Although he doesn't explicitly say, I think Bartosz
Ciechanowski's blog post is based on the ETA caliber 2824-2
mechanical watch movement (or one of its many clones, such as the
Chinese PT5000). This is a very robust and popular automatic
(self-winding) mechanical wristwatch movement found in many watches.
It's considered one of the all-time "workhorse" movements of the
industry. I highly encourage the curious reader to check out this
YouTube video of a very skilled and equally witty watchmaker
servicing a Hamilton watch featuring this exact movement. The video
will also reveal how many absolutely miniscule components make up the
ETA-2824, making it less than ideal for my initial prototyping. A
larger and simpler movement would be much better.
Luckily, in the late 1800's to early 1900's, many people carried a
pocketwatch about their person. However, as smaller wristwatches
gained popularity among men, beginning around the first world war
with trench watches, pocket watches lost their value as timepieces.
Many were melted down to reclaim the gold used to case the movement.
Today, you can find gorgeous, hand-engraved movements from the turn
of the 20th century on eBay for next to nothing.
Pocketwatches are the simplest mechanical watch movements out there:
they typically don't have any date complications, aren't
self-winding, and usually don't even have a central seconds
hand---that's a relatively modern feature. There's a reason budding
watchmakers start out practicing on pocketwatch movements.
Now for the hard part: How do you suspend 50-100 tiny components to
form an exploded view of the assembly?
[pocketwatc]
A beautiful early 1900's pocketwatch movement from eBay. $20 for a
piece of horological art.
Layered resin casting
The model I'm planning to build must stand up to being handled, which
means it must be solid. In turn, that probably means that I need to
cast my components in clear epoxy resin. My first idea was to build
the exploded model up, layer by layer, letting the resin cure between
each layer. To save time, I experimented with clear resin that cures
when exposed to UV light. There are several reasons why this did not
work:
1. The resin I purchased ended up with a heavy yellow tint after
curing.
2. My puny UV flashlight took forever to cure a pour even 1 mm deep.
3. The seams between layers are very visible.
The first two problems can probably be solved by throwing money at
the problem, but the last issue seems to be a problem with varying
indices of refraction throughout my casting. The resin art community
is well-aware of this problem, and the suggested solution is to pour
the next layer of resin before the previous layer has fully cured.
This supposedly helps the two layers fuse, and should make the
transition between layers much less pronounced. Unfortunately, UV
resin seems to cure from the outside in, meaning I needed to use
"regular" two-part epoxy instead.
To achieve the effect I want, I need around 20 layers for a regular
pocket watch movement. If I were to cast each layer in a transparent
container, adding components and epoxy as the previous layer was
half-cured, I would be doing nothing else for a solid week. Instead,
I figured my best bet was to cast all 20-or-so layers at once, and
stack them together once the epoxy was semi-cured. So, my next
project was to order some casting silicone and cast myself a mold
suited to casting thin disks of clear epoxy. A baking tray studded
with poker chips formed the ideal mold for casting the silicone.
My resin disks solved problems 1 and 2 above, but problem 3 is still
unsolved. In addition, this was a very messy and challenging way to
cast resin: the half-cured disks are floppy, resin runs everywhere
and bubbles get stuck between the layers. I bought a small vacuum
chamber to combat the bubble issue, but escaping gas weaked havoc on
the disks of resin. I learned enough about resin castings to finally
realize that layered casting is not the way to go.
[uv-resin]
1st attempt: embedding some nails in layers of UV resin.
[silicone-m]
2nd attempt: Embedding more nails in disks of clear resin in a
silicone mold...
[silicone-m]
yeah, no.
Suspending each component with fishing line
It took a while to accept that the only hope I had was to cast the
entire model at once. This is difficult because I need to somehow
suspend each tiny component in a way that is robust enough for me to
pour resin all over it and then pull out any air bubbles.
Scarred and burnt by resin's index of refraction, I went looking for
very thin rods of plastic or acrylic that I could cut to length and
glue my components to. Fishing line fits the bill. Specifically,
monofilament nylon leader used in fly fishing. It comes in many
thicknesses, has an index of refraction very similar to epoxy resin,
and is cheap. The only major challenge with using fishing line is
that it "remembers" its spooled shape, so every segment I cut off has
a slight bend to it. I was able to remedy this somewhat by stringing
it repeatedly across the grill pan in my oven and baking it at 150degC
(300degF) for an hour or so. When done, the middle segments are
reasonably straight and significantly stiffer than what I started
with.
The assembly process actually resembles "proper" watchmaking a lot.
Of course, tweezer control and steady hands are important. But
applying tiny drops of CA glue to components with a pinhead is a lot
like applying oil to bearing surfaces and jewels. I just find it
amusing to do it with glue instead of oil---CA glue being the
complete opposite of a lubricant.
Armed with a pair of helping hands and a set of self-closing
tweezers, I found the process of building up the exploded model of a
scrap watch movement to be very satisfying.
[fishing-li]
0.7 mm monofilament fishing line attached to bridge screws.
[crown-whee]
Monofilament fishing line maintains the distance betwen components. A
jig holds everything still while the CA glue hardens.
Resin casting at home
At this point I've tried a number of different epoxy resins claiming
to cure crystal clear. As far as I can tell, they all turn out
appreciably clear. Their main differences are viscosity, cure time
and how much air they trap when mixing. Some resins claim to expel
bubbles when curing. They manage this to some extent, but to get
completely clear castings you either need a vacuum chamber to pull
out all the air from your cast, or a pressure chamber to completely
squash any pockets of air. I went for a vacuum pump because it's a
fun thing to have around (boiling water at room temperature doesn't
get old). Another benefit of the vacuum pump is that you don't need
to leave your cast in the chamber for the entire cure time. My
castings have a lot of voids that love to trap air, and the vacuum
chamber does a good job of pulling out the trapped air.
My casting method goes like this: Mix enough of part A and B of your
resin to fill the mold plus 10-15 %. Mix thoroughly for 3 minutes.
Pour the resin into another mixing container to ensure that no
unmixed resin is stuck to the walls of your container. Use a fresh
stirring stick, and mix for another 3 minutes. Pull the whole
container under vacuum (I manage around -0.96 bar) and leave it there
for 30 minutes. Depending on your resin, the froth may very well
overflow your mixing container: cycle between pulling a vacuum and
letting air back in a couple times to pop most of the bubbles.
Now pour the resin into the mold and go through the vacuum process
again. This time, most of the bubbles you're pulling out come from
around the embedment, not from air trapped in the resin, which helps
keep the frothing down.
[froth]
Some resins froth up to several times the liquid volume under vacuum!
Make sure there's enough room to accomodate this.
Prototype #1: lessons learned
My first somewhat promising cast was in a borosilicate class
cylinder. I won't dwell on the looks of this, because I mostly wanted
to experiment and answer a few burning questions with this prototype:
Q: How visible is the fishing line in the final result?
A: Not very! It's visible in the right light, but I don't think
I'm able to do any better.
Q: Does CA glue interfere with the resin in any way?
A: Most resin datasheets say that CA glue can interfere with the
resin's curing process, but I can't tell from this test.
Q: Does casting in a cylinder make sense?
A: Definitely not. Light refracts on the curved cylinder surface
making it difficult to understand the geometry inside, defeating
the purpose of the model. Cast in a cube.
A week or so later, the cylinder shattered. Resin shrinks as it
cures. On to the next prototype!
[first-cast]
First prototype cast. Fishing line all but disappears in the resin.
Prototype #2: promising results
At this point I have settled on a somewhat structured process.
Disassemble and clean
Start with a clean, disassembled movement. For instructions on how to
properly disassemble and clean a watch movement, I highly recomment
Mark Lovick's Watch Repair Course. Assemble the train of wheels with
their bridges and seize them by depositing small amounts of CA glue
on the pinions. Just like with the watchmaker's oil, a sewing needle
with a flat spot is perfect for picking up a small droplet of CA glue
and depositing it where it's needed. Capillary action wicks it
between the components. Less is more when it comes to CA glue: the
bond between surfaces is stronger when no excess is used, and the
cure time is around 30 seconds even without the use of an activator.
I made a small jig for cutting pieces of fishing line accurately and
squarely out of a janky hand press tool.
Assemble trainwheel side
Work starts on the train wheel side of the movement (often called the
watchmaker's side) because that's where most of the complexity lies.
While working here, the movement can rest in a movement holder. I
attach long strands of fishing line to the end of each of the bridge
screws. Self-closing tweezers hold the bridges in place above where
they seat on the mainplate, and the screw/fishing line combo is
threaded through the screw hole in the bridge into the corresponding
holes in the mainplate. More glue binds the bridges to the fishing
line.
Subassemblies such as of the keyless works, motion works, balance
assembly, etc. can be constructed separately and fastened as a unit.
Small, flat components such as cover plates, intermediate wheels and
cap jewels each get their own short studs of fishing line. I dip the
studs in CA glue and place it near the center of mass of the
component I'm working on: the fishing line's flat ends allows it to
stand upright.
Flip and assemble the dial side
Before starting work on the dial side of the movement, it must be
flipped over. I transfer the assembly into the jaws of some tweezers
and apply closing force with a rubber band. The tweezers are clamped
by "helping hands" glued to a piece of cardboard which allows me to
spin my work around. Work continues similarly on the dial side.
Finally, some thin transparent nylon sewing thread attaches the
mainplate to a pegwood stick that allows me to suspend the whole
assembly over a mold, ready for casting resin.
Make the mold, prepare and pour resin
Speaking of the mold: I bought some 2 mm thick 20x30 cm acrylic
sheets. Foamcore or wood could work as well, but I don't want the
vacuum pump to pull air from the mold into the casting. I cut them
into 7x10 cm rectangles and lined them with "epoxy mold tape" (fancy
packing tape that epoxy resin does not bond to). Then I used some
Tec7 construction adhesive to form a cube.
Prototype #2 shows a lot of promise! It's far from perfect, though.
Here are some of the things I'll improve for prototype #3 in no
particular order:
* I struggled with attaching subassemblies to the mainplate because
I need to precisely control the distance between the subassembly
and the target surface: too snug and the assembly ends up
crooked, too far away and the glue doesn't adhere properly. A
proper lab jack (tiny scissor lift) might solve this, so one is
on the way.
* The resin shrunk a lot while curing, most likely because it
overheated. I'll focus on proper airflow for my next cast.
* People noted that it's difficult to see between the components,
so I'll "explode" the next model even more.
* The hands are set to an invalid time. 10:10 is the way to go.
[prototype-]
[prototype-]
[prototype-]
[prototype-]
Protoype #3: nailing the technique
The third iteration incorporated the scissor lift lab jack into the
process which, together with helping hands, allowed for much greater
precision in bringing components together true and square. I'm also
committing another watchmaking sin by placing a small magnet in the
jaws of my self-closing tweezers to gently hold tiny screws. This
allows me to lower a screw into its target drop of CA glue and simply
lift the tweezers once the glue has set. This is a much more reliable
method than trying to hold onto a sub-millimeter screw with
self-closing tweezers whose gripping force I have little control
over. Magnets are sinful in this case because magnetism can cause all
sorts of timing issues in mechanical watch movements, but that's
obviously not an issue in my case.
I also started using a CA accelerator to speed up the assembly,
because waiting for glue to cure is painful. Spraying the accelerator
onto my work would create a huge mess, so instead I spray some into a
lidded container and use tweezers to pick up droplets of the stuff to
deposit accurately onto the glue joints. This stuff is too volatile
and runny to be picked up by a needle. For me, the most effective use
of the CA accelerant is to first place a drop of CA glue on one of
the two mating surfaces, then dip the other surface in accelerant and
quickly bringing them into contact. I believe this works well because
the joint cures along the interface instead of curing from the
outside-in as is the case when spraying accelerant after the joint
has glue in it.
You may see that the balance wheel is hanging from the hairspring in
this casting. The balance wheel marks the end of the train of wheels
and releases tiny amounts of energy from the mainspring 18,000 times
per hour. The balance assembly is the heart of the movement and is
also the most delicate component, and I want to highlight that by
stretching the hairspring to show its form. These components are not
glued in place, and the balance wheel hangs freely from its spring,
meaning I have to cast the whole assembly upside-down to achieve this
effect.
[prototype-]
[prototype-]
[prototype-]
[prototype-]
Protype #4: ETA-2824
At this point I figured I was ready to tackle the final boss of this
project: the ETA 2824 wristwatch movement that we've all seen in
Bartosz Ciechanowski's blog post. Well, I'll tackle the Chinese
PT5000 clone movement instead, because I can't justify spending EUR300
on a genuine movement only to ensure it will never run again.
[pt5000-vs-]
Comparing my PT5000 to Bartosz's render.
This movement arrived from China in good working order---I was
actually suprised at its performance out of the box. It ran with good
amplitude and little positional variance between horizontal and
vertical positions. I became less and less impressed as I
disassembled the movement prior to cleaning, though, because it was
absolutely drenched in oil. Many of the bridges had sharp burrs that
broke off during cleaning. Nothing was broken though. All in all,
these movements really need a proper service before putting them to
service. That's not an issue here though.
I was worried that the much smaller components would be challenging
to work with, but I found the process to be essentially the same as
for larger pocket watch movements. My 0.7 mm nylon fishing wire still
fits through most of the screw holes and the lab jack makes alignment
a breeze.
Some components needed special care, in particular the balance shock
springs that protect the delicate pivots of the balance staff from
shocks. These aren't normally found on older pocket watch movements
which is why a broken balance staff is one of the most common failure
modes of pocketwatches. Shock springs are fragile and are, in my
opinion, among the most difficult components to handle when servicing
a watch. Of course, I want to also explode the balance assembly, so I
needed a way to suspend the shock springs above the capstone jewel.
By laying the spring down flat on silicone and placing a drop of CA
glue on it, the surface tension of the glue fills in the inner disk.
Once cured, the spring with its hard and transparent interior can be
lifted off the silicone. This is how the luminous material (lume) is
applied to the hands of a watch---another watchmaker's trick.
The assembly process was luckily uneventful, and I finished the build
off with a black dial and a random set of hands from eBay.
Unfortunately, the casting process went completely south. Only after
I had created a mold, mixed, degassed, and poured resin, did I
realize that my mold was just barely too tall to fit comfortably in
the vacuum chamber. With some hasty modifications to the scaffolding
that supported the exploded model, I was able to pull a reasonable
vacuum on it, but it left the exploded model crooked. To add insult
to injury, the resin seems to have dissolved the paint on the date
indicator ring, which left milky streaks throughout the casting.
All in all, I spent roughly 18 hours stripping, cleaning and
assembling the exploded view of the PT5000. With improved technique I
might get this down below 15 hours, but it's very tedious work, and
rushing means I'll knock something off and have to redo work. Good to
know, because I will be doing this again until I get it right.
[pt5000-wha]
The PT5000 disassembles nicely into a steampunk whale.
[pt5000-ass]
Installing the barrel bridge assembly.
[pt5000-ass]
Assembly complete!
[pt5000-fin]
Final casting. The paint on the date ring, dial and hands dissolved.
Also, the model is crooked.
End game: the wristwatch
Prototype #4 has a couple flaws which I'd like to fix before calling
this project done. The process of assembling the model is more or
less nailed down: I just need to find a way to seal any painted
surface before casting, as well as try to remember that my vacuum
chamber has limited volume. I bought another PT5000, this time in a
complete watch with case and a metal band, and started experimenting
with sealing the painted surfaces. Here's what didn't work:
* CA glue dissolves paint just as readily as epoxy resin. It might
work on simple paint-jobs like the chapter indices and the hands,
but the date ring has crisp lines (the numbers) which I need to
preserve.
* UV curable CA glue didn't cure on top of the paint. No idea why.
* Same with UV curable epoxy resin.
Clear spray lacquer from the hardware store worked well and didn't
dissolve the paint. It did turn yellow as you'll see in the final
casting, but I'm okay with that. I also built a better jig to cut the
fishing line into equal lengths with square ends. It's a bit janky
but worked very well.
There's not much more to say. I'll stop blabbering and make way for
an image series.
[fishing-li]
Janky but effective cutting jig for fishing line.
[painting]
Painting the dial, hands and calendar ring...
[ww-assy-1]
The watchmaker's side of the movement is assembled and fastened to
the case.
[ww-assy-2]
Our watch is held by a stiff cardboard tube (okay, a toilet roll
core) to keep the finished half suspended mid-air while work
continues on the dial-side.
[ww-assy-3]
Aligning the date ring...
[ww-assy-6]
Done! I wish I could keep it in this form, but it's too fragile.
Final result
[final-prom]
[final-prom]
[final-prom]
[final-prom]
[final-prom]
[final-prom]
I don't have the tools or knowledge required to sand this down to a
perfect mirror finish, but that's okay. After 2.5 years of noodling
around with this, I've achieved what I set out to do, and I'm pretty
happy with the result, even if it is difficult to photograph. Figured
I'd write this post to get some honest feedback, maybe even some
tips, in case I decide to continue working on this... I mean, more
interesting wristwatch movements do exist!
Bartosz, if you are reading this, contact me and I'll send you the
final casting. This project would never have happened without your
blog post.
Fredrik Flornes Ellertsen, 2016-2025