CNC Milling

Choosing the Right Milling Bit

Shape, diameter, blade length, and flute count all determine what you can cut and how fast. A mill bit looks like a drill bit but costs ~€60 versus ~€2, because it has cutting edges along its sides, not just the tip. Treat them accordingly.

Flute types each have a trade-off: upcut pulls chips out (great for through-cuts, can tear the top surface); downcut gives a cleaner top but chips stay in the pocket; compression combines both and is the best choice for plywood. This week we used a 6 mm flat end mill for drilling, profiling, and pocketing, and a 6 mm ball end mill for 3-axis finishing passes.

The ceramic spindle at rest. At 1600 RPM this is what does the actual cutting, treat the bit with the same respect as the machine.

Feeds, Speeds & Key Parameters

Three things to nail before hitting Run: cutting speed / feed rate (too slow burns the material; too fast snaps the bit), spindle RPM (always verify on the controller, some machines ignore the G-code value), and cut depth per pass (rule: rough depth = ½ × bit diameter, so 15 mm stock with a 6 mm bit = 3.1 mm per pass across 5 passes).

Never pause mid-cut to change bits unless the program is explicitly designed for a tool change. Plan your operation order and don't deviate.

Parameters — Tongue Drum

Stock: 1220 × 1217 × 15 mm plywood. All values loaded from the tool library rather than typed manually.

• Tool: 6 mm flat end mill — drilling and profiling
• Spindle RPM: 1600
• Feed rate (drill / plunge / cut): 2600 mm/min
• Clearance plane: stock max + 20 mm Z
• Total cut depth: 15.3 mm | Depth per pass: 3.1 mm (5 passes)
• Cut direction: climb — lateral movement only
• Entry / exit: none / none
• Bridges / tabs: outside profile cuts only

Setup Checklist

• Open RhinoCAM → Mill plugin → load tool library (30 tools).
• Review tool specs: length, diameter, RPM, cut rate, plunge rate. Cut rate is most critical.
• Set up layers: profile, screws, pocketing, engrave — colour-coded.
• Define stock: align origin to 0, 0, 0. Use preset stock size; don't modify the alignment.
• Post-process: use ShopBot OK / Raptor depending on machine.

RhinoCAM simulation — the parallel finishing toolpath over the stock. Always run the simulation and check the bottom of the bed before exporting G-code.

2-Axis Operations

Screw operation first — 15 mm clearance, cut parameter 1. Material must be secured before any profiling starts.

Profiling: select lines → 2-axis profiling → Load from Tool. Set clearance plane manually. Cut direction: climb, inside/outside for closed curves. Outside cuts require bridges/tabs in Advanced Cut Parameters, leave everything else blank. Cut levels: 15.3 mm total, 3.1 mm per pass.

Pocketing: same tool and most parameters. Default leaves a 0.6 mm stock offset, always set it to 0 in Global Parameters → Cut Parameters. If you move the object, right-click → Regenerate Setup Strategies.

3-Axis Operations (STL workflow)

• Insert STL, place below Z = 0, confirm inside stock volume.
• Don't place too low, more material above = more roughing time.
• Horizontal Roughing: 6 mm flat, 75% stepover, 0.6 mm stock.
• Parallel Finishing: 6 mm ball, 0 stock, quality 25.
• Holes in the mesh need a capping surface, otherwise the finishing pass plunges through.
• Disable spiral paths; verify engage path in simulation.

How Tongue Drums Actually Work

A tongue drum is a pitched idiophone, meaning the material itself vibrates to produce sound, with no strings or membranes involved. Each tongue is a cantilever beam: fixed at one end (where it connects to the body) and free at the other. When you strike the free end, the tongue flexes and oscillates at its natural resonant frequency. That frequency becomes the pitch you hear.

The physics is the same as a diving board: a longer, thinner board flexes more slowly (lower frequency, lower pitch); a shorter, stiffer one snaps back faster (higher frequency, higher pitch). For a tongue drum, you control pitch by changing tongue length and tongue width. Length is the dominant variable, doubling the length drops the pitch by roughly two octaves.

Tuning: The Role of Tongue Length

The resonant frequency of a free-clamped beam follows: f ∝ (thickness × width) / length². That squared relationship on length is what makes it so sensitive, a small change in length has a large effect on pitch. In practice:

• Shorter tongue → higher pitch. The lowest note (G2) needs the longest tongue; the highest (C3) needs the shortest.
• Width affects both pitch and volume, a wider tongue moves more air and resonates louder, but also stiffens it, nudging the frequency up slightly.
• Thickness is fixed by the stock (15 mm plywood), so length and width are the only two tuning levers available in the design.

Getting from a target frequency to a tongue dimension requires working backwards through that formula. I used reference values from existing tongue drum designs and adjusted proportionally for 15 mm plywood, then verified in simulation before cutting.

Optimising for Volume: The Body as Resonator

The tongue produces the pitch, but the body cavity amplifies it. The enclosed air inside the drum acts as a Helmholtz resonator, the box has a natural resonant frequency determined by its internal volume and the size of the sound hole opening. When the body's resonant frequency is close to the tongue's fundamental, the two reinforce each other and the instrument is significantly louder.

This created a real design tension: a larger body gives more resonance and volume, but the CNC bed constrains total panel size, and the finger joints need enough material to be structurally sound. Three competing constraints:

• Body volume — larger = louder and richer sustain, but increases material use and cut time.
• Sound hole size — the 220 × 100 mm ellipse was sized to tune the Helmholtz frequency toward the mid-range of the pentatonic scale (roughly A2), so the body amplifies the most-used notes most strongly.
• Finger joint depth — joints that are too shallow pull apart; too deep and they eat into the tongue root area, destabilising the tuning of the lower notes whose tongues sit closest to the edges.

The CNC Constraint: Kerf and Tongue Root

The 6 mm bit introduces kerf — material removed by the cut itself. For the tongue cuts, this means each slot is 6 mm wide regardless of how thin you draw it in CAD. That kerf width affects the effective tongue width (slightly narrower than designed) and the stiffness at the root.

The root is the most critical dimension: this is where the tongue is clamped to the body, and where stress concentrates when the tongue flexes. Too thin a root and the tongue cracks over time. Too wide and the tongue can't flex freely, killing sustain. I added a small radius at the root of each tongue cut, both to reduce stress concentration and because the dog-bone fillet that the round bit requires there actually functions acoustically, giving the tongue a cleaner pivot point.

In short: the dog-bone, usually a pure fabrication workaround, became an acoustic feature.

Why CNC?

The frame has both flat engraved pattern (vines, botanical linework) and raised 3D sculpted relief (flowers, depth in the ornamental borders). That mix makes it a natural 3-axis CNC job, laser can't give depth, and 3D printing wouldn't have the warmth of wood grain that connects back to nani's hand-drawn work.

CNC Strategy

• Horizontal Roughing: 6 mm flat, 75% stepover, 0.6 mm stock. Clears the bulk fast.
• Parallel Finishing: 6 mm ball, 0 stock, quality 25. Follows surface contours to reveal the botanical relief.
• Outer frame silhouette: separate 2.5D profiling operation, cut outside the line, bridges to hold in place until the final pass.
• Any holes in the mesh need a capping surface so the finishing pass doesn't plunge through.

The wood was enormous, measure twice

The stock sheet was 1220 × 1217 mm, bigger than I expected to be handling. Before you can fixture anything you have to physically lift, position, and align it on the bed. We measured twice before screwing anything down. At this scale, an offset origin wastes both time and material with no recovery path.

Dog-bone fillets — the internal corner problem

A round bit physically cannot cut a sharp 90° internal corner, it leaves a curved radius equal to its own diameter. For the tongue drum's finger joints, this meant tabs wouldn't seat flush without clearance. The fix is the dog-bone fillet: a small circular cutout at each internal corner that gives the mating piece room to slide in. It looks odd in the CAD but disappears once assembled. You can't design around physics, you design with it.

Pocketing left a skin of material

First pocketing pass left ~0.6 mm at the floor. The default stock offset in RhinoCAM pocketing is 0.6 mm and it's easy to miss because it feels intentional. Fix: Global Parameters → Cut Parameters → set stock to 0. Worth checking every single time you set up a pocket operation.

Toolpath didn't update after moving geometry

Repositioned the object in Rhino and the simulation still showed the old paths. RhinoCAM doesn't auto-update, right-click → Regenerate Setup Strategies every time geometry moves. Caught in simulation, which is the right place to catch it.

RPM not reading from G-code on the Raptor

The Raptor can silently ignore the spindle RPM in the G-code file and hold whatever was last set manually. Always verify the RPM on the Raptor control panel before running, the software value and the actual spindle speed are not always the same thing.