Tolerances that look right on screen and fail in plywood
Tolerances that look right on screen and fail in plywood
I once cut a test joint where the tab was 0.05 mm too wide. It looked perfect in CAD. On screen the fit line was clean. In plywood, the panel would not slide home without a mallet and a prayer. That is the lesson: wood does not behave like steel, and the tolerances for CNC router plywood furniture design have to account for it.
Plywood is not uniform. A sheet labeled 18 mm can measure 17.1 mm to 18.7 mm under EN 315. Even within one batch I have seen 0.3 mm variation across a 2500 × 1250 mm sheet. Humidity matters too. Plywood swells and shrinks with moisture changes, typically up to about 1 mm per meter between dry and wet conditions. A joint that is tight in winter can rattle in summer.
Slots and dados
The rule is simple: measure the actual sheet, then design the slot to the measured thickness plus a small clearance. I start with about 0.1 mm total clearance over the measured thickness for a snug glued fit. If the joint needs to come apart later, I open it to 0.2–0.3 mm. If it has to hold itself together dry, I go slightly under, around 0 to +0.05 mm, and accept that glue is the real fastener.
Dogbone corners are mandatory. A CNC router bit is round, so an internal square corner always has a radius. If you forget the dogbone, the rectangular tab hits the corner radius and the joint jams. I add them to every slot and pocket before exporting the DXF.
Bolt and pin holes
For M6 bolts in plywood, I use 6.6 mm clearance holes as the default. That is the standard normal clearance fit, and it leaves enough room for the bolt to slide through even if the hole is slightly undersized or the sheet has swollen. If I need a tighter alignment, I go to 6.4 mm, but only when the assembly is done indoors and the humidity is stable.
Shelf pins use 5 mm holes for the European system. I keep them on 32 mm centers, 37 mm from the front edge. The fit should be a snug slip fit, not an interference fit. If the pin has to be hammered in, the plywood will compress and eventually crack around the hole.
Press fits in wood
I avoid true press fits in plywood except for small dowels or steel pins. Wood crushes rather than yields cleanly like metal. If a client insists on a press-fit steel pin, I use a 9.97 mm hole for a 10 mm pin in aluminum brackets, but in plywood I would rather use a 9.9 mm hole, a drop of epoxy, and call it good. The wood fibers will compress slightly and the glue does the rest.
The test joint
Before I cut a full sheet, I always cut a small test joint from an offcut. Same bit, same feeds and speeds, same plywood batch. It costs ten minutes and a scrap of material. It saves me from discovering a fit problem after the whole sheet is ruined.
The test joint also reveals whether my CNC machine is cutting true. A worn bit, a loose collet, or a slightly warped sheet can add 0.1–0.2 mm of error. If the test is off, I adjust the slot width in CAD and re-export before committing to production.
When I let it be loose
Some joints are better loose. A van bed frame that bolts into a van floor needs room for the floor to flex, the plywood to swell, and the owner to assemble it without a clamp collection. I would rather use slotted holes and washer plates than chase a perfect 0.05 mm fit that will fail the first time the van drives over a speed bump.
Tolerances in plywood are not about being as tight as possible. They are about being honest with the material. The CAD file can pretend everything is exact. The workshop cannot.
[IMAGE: A dark macro-style render of two plywood pieces meeting at a dado joint. One piece shows an oversized CAD overlay with tight tolerance dimensions in amber. The other shows the real wood with visible gap and compressed fiber at the corner. Dogbone fillets highlighted in amber. Graph paper background. Clean, technical, premium workshop aesthetic. No text. 3:2 aspect ratio.]