Quick Recap ⚙️

Hey there, fellow watch-tinkerers! I’m still deep in big-whiteboard territory—sketching every wheel, pivot, and bridge for my future movement while learning the nuances of a freshly built 4-axis CNC setup. No chips have flown—yet—but today I pinned down the machining playbook for two fully-sintered carbide micro-cutters. Each cutter will sculpt module 0.13 teeth on both the wheel and its mating pinion. Laying rock-solid foundations now should make first-cut day feel almost calm. 🙂

Explore the interactive toolpath playground →

Today’s Deep Dive: Simulating the 3-Stage Cutter Strategy

Both cutters share the same three-step journey; only the final edge profile changes—one hugs the broader wheel involute, the other traces the slimmer pinion flank:

1. Stage 1 – Coarse Roughing
   Planned tool: ⅛″ (3.175 mm) CVD round-end mill from Harvey Tool
   Workpiece: Ø 6 mm solid-carbide blank
   Simulated spindle: 30 000 RPM | Radial DOC: 8 µm | Axial step-down: 60 µm
   

   This first pass slims the 6 mm blank to about 2.2 mm in the cutting zone. Fusion 360’s adaptive roughing, paired with an 8 % synthetic-coolant mix, keeps simulated spindle load under 40 % and temperatures steady.

2. Stage 2 – Detail Roughing
   Planned tool: 1 mm CVD square-end mill
   Simulated spindle: 35 000 RPM | Radial DOC: 4 µm | Axial step-down: 40 µm
   

   This pass shapes the shank relief and rough involute flanks to ±0.02 mm, leaving 30 µm for finishing. A mist-plus-air blast in the simulation prevents heat spikes while the A-axis keeps the blank rolling smoothly.

3. Stage 3 – Edge Finishing
   Planned tool: the same 1 mm bit, side-milling with its flute edge
   Simulated spindle: 38 000 RPM | Radial DOC: 1 µm | Axial step-up: 5 µm
   

   Pure climb cuts deliver a predicted mirror finish of Ra ≤ 0.15 µm. Wheel and pinion curves are saved as separate “trace” paths, swapped in at post time.

Where Fusion 360 Fell Short—and the DIY Detour 🛠️

Fusion handled bulk removal nicely, but it simply wouldn’t allow multipass refinement when using multi-axis strategies like Swarf or Flow. Rather than compromise, I built a home-grown toolchain: TypeScript + THREE.js for 3-D visualization, plenty of trig to unwrap toolpaths, and ChatGPT to sanity-check the math. The script now exports blend-free G-code with helical lead-outs—something Fusion can’t (yet) do—so every line is under my control.

Target metrology: OD 2.000 ± 0.005 mm, tooth height 0.300 ± 0.003 mm—numbers I’ll verify in situ once the cutters arrive. Imagine neon toolpaths swirling like a galaxy instead of metal chips and you’ll have the idea. 🚀

Next Steps 🔧

• Receive & loupe-inspect the Harvey Tool CVD mills (20× check on every flute).
• Clock run-out to < 3 µm in the ER-20 collet system.
• Dry-run Stage 1 G-code above a foam dummy to confirm clearances.
• Tweak coolant-nozzle positions for uninterrupted chip evacuation.

Sign-off & Call-to-Action

Thanks for following along! If you’ve broken-in CVD tooling or have hard-won tips for ultra-fine gear cutting, I’d love to hear them—drop a reaction below or reach out on Nostr at @bitcoinwatchmaker. Your insights help turn these plans into perfectly cut teeth. 🙌

OH MY GOD. Fusion is great... but it can be painful.

All I wanted was a simple G-code pass — just take the CVD-coated mill bit and slowly shape it, one layer at a time.

But nope. Even after paying for the manufacturing extension, Fusion won't do it. Swarf? No multi-pass. Rotary parallel? Same story. Always goes deep. Always skips the careful steps I need.

What is this? this is a tool to cut the gear teeth which are super specific, delicate, so im having to mill fully sintered carbide, with diamond coated (CVD) tools, freaking carbide is what cuts everything else... this according to chatgpt is the limit of what can be done in CNC... not bad for a newbie haha

So... time for the long route. Built a Vite app. Hooked it up with THREE.js. Loads an STL and spits out the G-code I actually want 💻✍️

Normally I cant opensource things due to the NIHS rules being implemented everywhere, this I can if anyone wants it let me know

Big milestone today—the script now spins up wheels, staffs, and a full mainspring barrel in one click, then drops rubies into custom holders like it’s dealing cards at Monte Carlo.

Materials snap into place too—brass for wheels, polished steel for staffs, deep-red ruby jewels, and a flash of blued steel where it counts—so the CAD preview already looks heat-blued and bench-ready.

The real magic? A tidy joint system that lets me dial each rotation until the whole train lines up in picture-perfect symmetry. I also reworked the clearances between wheels so the bridges drop in clean and pick up those sweet 0.07 mm chamfers without a fuss. Watching it lock in is pure watch-nerd bliss.

Bonus win: after some head-scratching I finally found the sweet spot for the Bitcoin 10-minute complication—a pinion driving a wheel that meshes with an inner gear. Woah, can’t wait to see that block-time tick! ₿

Next up: diving into milling fully-sintered carbide bits shaped to match each tooth profile—turning pixels into cutting edges one micron at a time. Stay tuned! 🛠️⌚️

Script also adds the cuts so the wheel and pinion can be joined

Kept pushing the script forward — this thing’s growing fast.

Settled on three jewel types. For the center and third wheel staffs, I’m using olive ring jewels with 2 oil cups on the bottom (dial side), and olive/bombe ring jewels on top (plate side).

Fourth wheel gets the same: olive with 2 oil cups. For the escapement, I’m going with one oil cup olive jewel on the bottom and a straight ring jewel with oil cup facing the dial.

Also added a jewel holder to the mix. Makes future servicing easier 🔧 and gives me room to explore materials and maybe some heat bluing 🔵

Olive ring jewel with 2 oil cups

Olive/Bombe ring jewels

Straight ring jewel with oil cup

Rushed ahead a bit... then realized the numbers weren’t lining up. Even the tiniest error matters here.

So I paused. Took a step back. Recalculated everything 🔍

Wrote a script — with a little help from AI 🤖 — to find the exact gear ratios I need. No guessing. Just clean math.

Then came the drawing script, built to match those ratios and follow NIHS standards to the dot 📐

Now it feels solid. Precise. On track.

I’ll keep building on this — adding the Bitcoin complication ⛓️, jewel positions 💎, and initial bridges. One step at a time.

After weeks of learning, mistakes, and exciting breakthroughs — the lathe is finally good to go! 🛠️

I had to remake the backplate three times due to different layouts. The lathe's travel is limited, so I had to get creative. But now it’s dialed in and solid.

⚠️ Manufacturer Woes (Mucho NO recommendo)

To be honest — the manufacturer did a terrible job on this lathe. It came with:

• Insane vibrations
• Super loud noise during operation
• Broken and unconfigurable software
• Default tool change macro that literally caused a collision 😤

That mess sent me down the rabbit hole of rewiring, reprogramming, and rebuilding much of the system from scratch. Painful… but I now know the machine inside out!

🔩 Current Tool Setup (4 Tools)

1. HSS Cutting Tool — hand-ground by me at a 45° angle. My first toolmaking experiment!
2. Micro 100 Grooving Tool — fantastic for tight grooves and channeling.
3. Cutout Tool — great for partying shapes with high precision.
4. Drill Fixture — holds fixed drills securely for consistent center holes.

⚙️ Motor: The Real Challenge

The motor was the trickiest part. I used a Teknic ClearPath Integrated Servo Motor, paired with a power supply, analog send unit, and 22 AWG cables. I’ve never worked with electronics before — so learning to use a multimeter, solder, crimp, and protect everything properly was a major win.

Big thanks to Alexis — couldn’t have done this without his help! 🙏

🌀 Spindle & Drive

I went with a Power Twist V-belt. Pulleys are almost 1:1, and even though there’s some drop in RPM (from 2520 to 2320), it feels solid and precise so far.

✅ Past Weeks Recap

• Polished the keyless works, winding mechanism, bridges, and barrel arbor
• Aligned the Elara 4th axis to within 0.01°
• 3D printed the case, revised leg length to avoid scratches
• Built a new brass fixture for 3.0mm stock (no more Loctite hacks!)
• Achieved 2μm precision in XYZ on the Elara mill
• Got replies from Incabloc and updated the balance staff accordingly
• Started grinding my own lathe tools from HSS bar stock
• Published all blog updates via Nostr — auto-summarized with ChatGPT

🔮 What’s Next?

• Finish the balance staff with final jewel specs
• Cut, test, and compare Micro 100 vs HSS tools
• Design and fabricate the elusive crown gears
• Improve pulley ratios and optimize RPM at the spindle
• Figure out how to secure the movement inside the case
• Start polishing and finishing parts for real assembly!

Super excited for what’s coming! Let’s make some beautiful, tiny mechanical art 🌀✨

Website Updates: Reactions, Permalinks, and Smart Summaries ⚡

Some exciting updates to the site this week!

• 🔁 Reactions: You can now react to blog posts — quick way to share a vibe or show support!
• 🔗 Permalinks: Each post now has a clean permalink. Sharing specific moments just got easier.
• 🤖 Smart Summaries + Nostr: We’ve added an automated system that uses ChatGPT to summarize each blog post and auto-publish it to Nostr. No more double posting!

Bit by bit, the site is getting more connected and streamlined. Watchmaking meets automation ✨

Trying Out Micro 100 Tools 🛠️

I've noticed that HSS tools wear out quite fast, especially at the micro scale. So, I’m giving Micro 100 a go — excited to test their high-quality carbide tooling for turning.

Here’s what I’m trying:

• BT-6 (Box Tool): Ideal for creating accurate shoulders and flat faces in confined areas. Great for tight workspaces and forming box-like geometries.
• PF5-050150 (Profiling Tool): A fine profiling tool with a 0.5mm radius, perfect for tiny contours and detailed turning operations.
• GR-018002 (Grooving Tool): Designed for making ultra-fine grooves — this one's got a 0.018" width. Precise, clean groove cuts!
• T-100 (Cutoff Tool): A solid carbide tool for parting off tiny components. Rigid and sharp — ideal for micro-scale cutoff tasks.

Hopefully, these will hold up better than HSS and bring cleaner finishes to the lathe work. Time (and testing) will tell!

Sharpening, Turning, and Chasing Precision

This session was about slowing down and listening to the work.

I manually sharpened my HSS tools using both India and Arkansas stones. No machines. Just me, the edge, and the feeling of steel against stone.

To confirm cutting direction, I studied the old masters — especially T&T&T’s YouTube videos. Watching their moves, the angles, the confidence — that gave me clues I couldn’t learn from CAD or cam simulators.

With my newly sharpened tools, I cut half of a balance staff. The finish? Beautiful. The dimensions? Still oversized — I left too much material. But that was the plan: test control, feel resistance, and watch the chips curl cleanly.

What’s next? I need to dial in G54 more precisely, and set tighter offsets for tool 2+ on my gang-style lathe. That’s the only way to achieve the repeatability and control I’m after.

Every small improvement feels like a step toward mastery — and I’m still happy to fail along the way.

Learning, Trying, Failing… but Happy

This week, I hit an unexpected behavior while setting up my gang-style tool system with Mach4 and Fusion 360.

Here’s the layout:

• Tool 1 is positioned manually and sets G54.
• Tool 2 and beyond use XYZ offsets relative to Tool 1.
• Fusion 360 doesn’t emit Y moves for tool changes in lathe ops — it assumes only XZ movement.

So far, so good... until I discovered this 🤯:

After a tool change using M6, running G0 Y0 still uses the previous tool’s Y offset!

What I mean is this: (M6 macro)

This happens because M6 internally calls setCurrentTool, but Mach4 doesn't apply the Y-offset until after the macro finishes. That means if you issue any move (like G0 Y0) immediately after M6, you’re still running with the old offset. Yikes.

My solution: I created a custom macro: M200.

So then, I modified Fusion 360’s Mach4 Turning Postprocessor to call M200 immediately after each M6 command.

how you may be wondering, well like this: (this is in the mach4-turning postprocessor downloaded from Autodesk website)

And M200 looks like this:

It does three things:

• Moves Z up 5mm
• Moves Y to 0 (now under the correct tool offset)
• Restores Z

Now every tool change is reliable and precise — no more misaligned cuts or offset confusion. It’s a small fix, but one that makes the difference between frustration and confidence on the lathe.

Mach4's quirks taught me a lot this week — and solving them made me enjoy the process even more.