Clown Dice Carrier: A Study in Tolerances
Around November 20th, I set out to answer a pretty simple question: What does 0.01" actually mean in my workshop?
In theory, specifying a 0.01" tolerance is straightforward. In practice, it's... different. My CNC machine is great, but it aint that great. Wood moves in unpredictable ways. And when you're combining CNC-machined materials with 3D-printed materials, those tiny numbers start to mean very different things depending on which tool made them.
So I designed this project specifically to learn my ecosystem's... we'll call them nuances, yeah?
Why a Clown Head?
Two reasons.
First, a good friend of mine runs a group of lampooners known as the "Clowns" of nopixel, who've become an endless source of inspiration for my projects. It's just a great theme.
But more importantly: this shape is a nightmare to work with. The clown head is all unconventional curves and pointy edges. It creates problems at every stage of the pipeline. My thinking was simple like, if I can make this work as a compact, usable item, then anything else will be easy.
Turns out I was right.
The Design
This is a double-sided 8-dice carrier with embedded magnets for closure. Each side holds 4 dice in precision-machined pockets, with 3D-printed inserts that snap into place (We'll get to why later).
Key features:
- Double-sided storage for 8 dice total
- Magnetic closure with embedded magnets
- Custom clown face CNC-carved into the exterior
- Precision pockets machined to tight tolerances
- 3D-printed inserts for the dice compartments
The Technical Stuff
I did this in Vectric Aspire and Fusion360, but mostly in Aspire.
- 13 interior toolpaths for the pocket machining sequence
- 7 dice-related operations for the actual dice (next project)
- 3 profile cuts at different angles (quarter-round and 60-degree)
My machine runs well at around 19000 RPM/100 IPM for interior work on walnut and 14,000 RPM for the outside profiles, with feed rates tuned for clean cuts in hardwood. I tried to mill this out of oak but my pass depth of .02 was just too much at those feeds and speeds. I elected to not work with oak on this project because I really didn't feel like fundamentally changing my toolspaths just because oak is a brutal material to mill.
It took 10 iterations to dial everything in. Ten versions of adjusting pocket depths, tweaking magnet hole diameters, and refining the fit between CNC-cut wood and 3D-printed plastic. Each iteration taught me something new about how my specific machines interpret "precision."
Materials
The prototype in the photos is poplar—great for testing because it machines cleanly and shows mistakes clearly. The production run of 20 will be in walnut, which is what this design was ultimately meant for.
Lessons Learned
Most people make square boxes. Mine was all S-curves and T-splines. Taking on a project like this was equal parts challenging, rewarding, and educational.
The Perfect Fit Isn't Perfect
Pine and walnut behave completely differently when it comes to tolerances. With pine, ±0.05" gets you a tight fit. I suspect because it's pithy and expands after milling. I cut four pieces with the same settings and got reproducible results every time.
Walnut is another story. ±0.02" is more than enough for a good lid fit, with just a hint of play that nobody would notice. I made one piece with sleeves cut at 0.015" and the fit was incredible until I realized it created suction. The lid wouldn't come off without unwanted effort.
That's when I learned the counterintuitive truth: you actually need an air gap. A mathematically perfect fit creates a vacuum seal. The perfect fit isn't perfect—it's a trap.
Here's the paradigm I'd encourage everyone to adopt: take a small piece of your skillset that you'd like to improve, and explode it into its own project. Don't just practice tolerances on a test piece—design something real that forces you to get it right.
Now that this is done, anything in this class of item will be straightforward to produce.
If you'd like to support my work, check out my store or find me on Patreon.