CNC foam cutting: how the edge actually behaves
Foam is not a soft metal. It compresses, springs back, and melts at the tool tip, so the fixture and the feed rate matter more than the spindle. This page covers the mechanics, the grades that cut well, and the point where CNC foam cutting stops being the right call.

Key takeaways
What happens at the edge in cnc foam cutting
Rigid materials fail by brittle fracture or ductile shear ahead of the cutting edge. Foam does neither cleanly. The cell walls collapse under the rake face first, then tear. That collapse is elastic up to a point, which means the tool has to travel several thousandths of a millimeter further than the finished surface before the material actually separates. Push too slowly and the foam springs back into the tool path.
The practical result is that depth of cut and feed per tooth are coupled in a way they are not in aluminium. A 6 mm two-flute upcut bit running 18,000 rpm at 0.05 mm per tooth removes EPS cleanly. Drop the feed to 0.01 mm per tooth at the same rpm and the same tool rubs instead of cuts. Heat builds, the cell walls soften, and you get a glazed band along the wall.
Heat is the real constraint. Foam conducts almost nothing, roughly 0.03 W/m·K for EPS against about 200 W/m·K for aluminium. Friction heat has nowhere to go except the tool edge and the immediate surface. Once the surface passes its glass transition or melt point, the wall smears closed and no amount of downstream sanding fully opens it again.
Chip evacuation matters less than in metal but not zero. Foam dust is light and travels with the airstream, so a downcut tool on the finishing pass pushes dust back into the kerf and improves the top edge. Rough with upcut for clearance, finish with downcut for a clean shoulder.
- 1Rake anglePositive rake, 10–20°, peels the cell wall instead of pushing it
- 2Flute countTwo flutes for most densities; single flute for soft EPS above 20 mm depth
- 3Edge sharpnessA dull edge triples cutting force in foam, far more than in metal
- 4CoatingUncoated polished carbide; some coatings grab and tear open-cell foam
Which foam grades behave predictably
Not every foam is a routing material. The distinction that matters on the shop floor is cell structure, not density alone. Closed-cell foams such as XPS, EPP and most PU tooling boards hold a defined wall, cut with a clean chip, and repeat within ±0.1 mm on a stable setup. Open-cell foams and loose expanded beads tear at the surface and vary from part to part even with identical programs.
Density drives feed rate more than any other single variable. Below roughly 30 kg/m³ the material has little resistance, so the tool tends to pull into the work and the feed rate needs limiting. Between 30 and 120 kg/m³, which covers most PU tooling board and modelling foam, cutting is predictable and you can hold wall thickness on a 4,000 mm part. Above about 200 kg/m³ the material starts to behave closer to a soft plastic and standard plastic cutting parameters apply.
Humidity changes dimensions. Foam takes on moisture from the air, and a 500 mm EPS block can move 0.3 mm across a humid week. For any part where two foam sections must fit together, machine both in the same session and from the same conditioned stock. Do not cut halves a day apart and expect the joint to close.
Foam also has a grain, of a sort. Blocks are molded or extruded with a direction of cell elongation. Cutting across that direction gives a cleaner wall than cutting with it. On a five-axis setup, we rotate the part in the CAM model rather than accept the tear-out on one face.
- 1Good candidatesXPS, EPS above 25 kg/m³, PU tooling board, EPP, PVC foam
- 2Poor candidatesLoose bead EPS, open-cell acoustic foam, wet or recycled foam
- 3Avoid entirelyFoam with embedded metal inserts or abrasive filler
Workholding is where most foam jobs fail
A foam block weighs almost nothing, so it does not resist the cutting force by mass. Every newton the tool applies has to be reacted by the fixture. On a 4,000 mm part, that force is small but the lever arm is long, and any movement at the far end shows up as a taper along the length.
Vacuum through a bleeder board is the default for flat work. Foam is porous, so vacuum pulls through the part rather than holding it, which is why a sealed bleeder and a skimmed surface matter. For EPS, seal the block edges with tape or a sprayed skin so the vacuum stays under the part instead of bleeding through the top face.
For contoured parts, bond the block to a sacrificial MDF backer with hot-melt or double-sided film and cut through the foam into the backer by 0.5 mm. The bond carries the shear. Tabs, 3 to 5 mm thick, hold the part through the last pass and get cut by hand afterwards.
Never clamp foam directly with a vise or toggle clamp. The jaws crush the cells on contact, and the dent does not spring back. A 0.2 mm crush mark on a 50 mm block stays a 0.2 mm mark. On a finished visual surface that is a scrapped part.
- 1Flat workVacuum bed with sealed bleeder board, tape the block edges
- 2Contoured workBond to MDF backer, cut 0.5 mm into the backer
- 3Through cuts3–5 mm tabs at four points minimum on long parts
- 4NeverDirect vise or toggle clamps on a finished foam face
What tolerance foam can actually hold
The machine is not the limit. Our five-axis centers hold ±0.005 mm in aluminium, and the same linear accuracy applies when the spindle is over foam. What changes is the material response. Foam deflects under the tool during the cut and relaxes after, so the finished surface sits somewhere between the programmed path and the deflected path. On a 100 mm deep wall with a 6 mm tool, that springback is measurable.
For a 100 mm tall feature in 60 kg/m³ PU board, expect ±0.1 mm on the wall position and about ±0.15 mm on the floor. Deep, thin walls are the worst case: a 2 mm wall at 150 mm tall will flex away from the tool during the finishing pass. Either rough both sides and finish symmetrically, or accept a looser tolerance and check with a template rather than a CMM.
Thermal drift is the second limit. Foam has a high coefficient of thermal expansion for its stiffness, and a shop that swings 5 °C across a shift will move a 1,000 mm foam part by roughly 0.3 mm. That alone can consume the tolerance budget on a large pattern. Measure parts in the same room they were cut, and let the block reach room temperature before the finishing pass.
If your drawing carries a ±0.05 mm callout on a foam part, the honest answer is that we will hit it on a short rigid feature and miss it on anything long or thin. Say which dimensions actually function and we will hold those.
- 1Short rigid features±0.05 mm achievable under 50 mm with a stable setup
- 2General walls±0.1 mm is the working figure on PU tooling board
- 3Long thin walls±0.25 mm or looser; plan for a fit check
- 4Large patternsAdd 0.3 mm per metre for thermal movement
Toolpath strategy that keeps the surface intact
Rough with a larger tool and a constant engagement. Adaptive or trochoidal paths keep radial engagement around 10 to 15 percent of the tool diameter, which keeps the heat per unit length low and the tool load steady. In foam, the benefit is not tool life, it is surface consistency: a steady load produces a uniform wall, while a varying load produces bands where the foam was pushed rather than cut.
Finish with a single continuous pass wherever the geometry allows. Stopping mid-surface leaves a witness mark that will not sand out evenly, because the foam under the stop point is compressed more than the surrounding material. On long parts, ramp in rather than plunging, and keep the lead-in outside the finished surface.
Stepover for finishing sits between 5 and 10 percent of the tool diameter for a visual surface and up to 30 percent where the part will be coated. A 6 mm ball nose at 0.4 mm stepover gives a surface that reads smooth to the eye on PU board. The cusp height at that stepover is under 0.01 mm, which is below what the material can hold anyway.
Climb cutting on the finishing pass gives a cleaner edge on closed-cell foam. Conventional cutting tends to lift the top fibers and leave a fuzzy line on the shoulder. On open-cell foam the difference is smaller and tear-out dominates regardless of direction.
- 1RoughingAdaptive path, 10–15% radial engagement, upcut tool
- 2FinishingClimb cut, single continuous pass, 5–10% stepover
- 3Lead-inRamp or arc entry outside the finished surface
- 4Depth per passUp to 1× tool diameter in foam, watch heat not load
CNC foam cutting versus molding and hot-wire
Use this to pick a process before you commit to tooling.
| Method | Best for | Tolerance | Tooling lead time |
|---|---|---|---|
| CNC routing (3-axis) | Flat and shallow contoured panels | ±0.1 mm typical | None |
| CNC routing (5-axis) | Deep contoured and undercut shapes | ±0.15 mm typical | None |
| Hot-wire cutting | Tapered and ruled EPS forms | ±0.5 mm on a good setup | None |
| Vacuum forming over a foam mold | Thin shells and quick panels | Follows the mold, ±0.2 mm | Days, not weeks |
| Injection molding | Above roughly 10,000 parts a year | ±0.05 mm on foam-grade resin | Weeks plus mold cost |
| Hand shaping | One-off artistic forms | No repeatable figure | None |
When to route foam and when to walk away
Route foam when the part is a pattern, a mold blank, a fit-check model or a low-volume shell, and when you can live with ±0.1 mm and no structural load. Go to hot-wire for ruled EPS forms, and to molding once annual volume passes roughly 10,000 pieces. If the part must carry load or hold a sealing face, foam is the wrong material and no amount of toolpath tuning changes that.
Questions we get on foam routing
Can you hold ±0.005 mm on a foam part?
No, and any shop that says yes is quoting a machine spec rather than a material result. ±0.005 mm is what our five-axis centers hold in aluminium.
Foam deflects under the cutting edge and relaxes afterwards, so ±0.1 mm is a realistic working figure on PU tooling board and ±0.25 mm on soft grades.
Which foam should I specify for a vacuum-forming mold?
PU tooling board in the 60 to 120 kg/m³ range. It cuts clean, holds a screw thread for inserts, and survives the heat cycles of short vacuum-forming runs.
EPS is cheaper but dents and degrades after a few pulls. Use EPS for a one-off shape, not a production mold.
Does foam need a finishing pass at all?
It does if the surface is visible or will be coated. A roughing-only surface on foam has visible scallops and a compressed skin that paint will not sit on evenly.
Light sanding at 320 grit after a proper finishing pass is usually enough. Heavy sanding on a rough surface just re-opens torn cells.
How big a foam part can you machine?
Our largest travel is 4,000 × 400 × 150 mm on one setup. Larger foam assemblies can be cut in sections and bonded.
For anything over 2,000 mm, plan the joint positions before you send the model. Foam joints are more visible than metal joints.
Will cutting fluid damage the foam?
Yes, most of them. Solvent-based fluids dissolve EPS and XPS on contact, and water-based fluids swell the cells.
We cut foam dry with air blast for chip clearance and a high-efficiency filter on the extraction.
What CAD data do you need for a foam pattern?
A STEP or IGES solid is ideal. Surfaces-only models are workable if the surfaces are trimmed and closed, but open edges create gaps in the finish pass.
Send the model with the pull direction marked if it is a pattern for molding. It changes the toolpath strategy.
Send us your foam part and the dimensions that matter
Upload a STEP file and we will return a quote, a DFM note flagging the features that cannot hold tolerance in foam, and a recommended grade within 12 hours.
12-hour quoteNo minimum order quantity100% inspection before shipment