Finger Joint Box Generator
Generates a kerf-compensated box as downloadable SVG — and shows why narrower fingers give more glue area, not less.
Fingers per edge
Glue area against finger width
Narrower fingers give more glue area, not less
Each finger protrudes into the mating panel by the material thickness, so it contributes two side faces of t × t — and those sides are the strong glue surfaces, where the finger's end face is end grain like a butt joint. Total side area is 2n·t², and since the edge length L equals 2nw, that comes out to exactly (t/w) × a butt joint's area. At a finger width equal to the thickness, the joint matches a butt joint in area but in good glue surface; at 3× thickness it's a third. The area argument has no optimum — it favours narrower fingers all the way down.
So "two to three times the thickness" is a compromise, not an optimum, and it helps to know which way each term pulls. Glue area wants the fingers narrow. Three things want them wide: a finger much thinner than the sheet snaps while you assemble it, a finger only a few kerfs across is mostly cut edge, and every extra segment is more cutting. The received rule sits where those meet — which is why it's always quoted as a range rather than a number.
The segment count must be odd or a corner ends in an open notch. One panel takes the even segment indices and the other the odd ones, so an edge begins with a tab at index 0 and ends with a tab only when the last index — s−1 — is also even, which happens exactly when s is odd. This is a hard geometric constraint rather than a preference, and it's the thing hand-drawn finger joints get wrong. Each edge here is rounded up to the next odd count, which is why the three axes often end up with slightly different finger widths.
And inside versus outside dimensions is a 2t error, far bigger than any kerf. A 100 mm box in 3 mm material is 94 mm inside; in 12 mm ply it's 76. People agonise over a 0.2 mm kerf and then lose 6 mm to a definition. Decide which one actually matters for what's going in the box, and set the selector above accordingly.
- The kerf shift is half the kerf less a quarter of the clearance, not half of it. A tab and its slot share a boundary, so one unit of shift moves both and buys two units of clearance. Using the natural-looking half gives exactly double the fit you asked for.
- A global kerf offset won't fix the joints. It corrects the outside of each panel and moves slot and tab the same direction, leaving every joint as loose as it started — the commonest way a generated box rattles despite the kerf being "handled".
- Measure the sheet, not the label. Nominal 3 mm plywood runs 2.6–3.1 mm, a spread larger than the entire kerf compensation. The finger depth is set by that number.
- Cut one corner first. Two small panels with a single joint settle the kerf, the fit and the thickness together in a minute — the only test that covers all three at once.
How to use
- Say whether your dimensions are inside or outside — it is a 2t difference.
- Measure the sheet thickness and the kerf rather than trusting the label.
- Cut one corner as a test before committing the whole sheet.
- Download the SVG and check it opens at real-world millimetres.
Frequently asked questions
How wide should finger joints be?
The usual answer is two to three times the material thickness, and it is a compromise rather than an optimum. Glue area actually favours narrower fingers all the way down, because each finger contributes two side faces of thickness squared and the total works out to exactly t/w times a butt joint area. Three things push the other way: thin fingers snap during assembly, fingers only a few kerfs across are mostly cut edge, and every extra segment is more cutting time.
Do more fingers make a stronger joint?
In glue area, yes — and this surprises people, who often assume chunky fingers are stronger. Side glue area is 2n times the thickness squared, so it rises directly with the number of fingers. Halving the finger width doubles the glue area. What you lose is finger robustness during assembly and cutting time, which is why the practical answer lands in a range rather than at the narrowest cuttable finger.
Why must the number of segments be odd?
Because one panel takes the even segment indices and the other takes the odd ones. The edge therefore begins with a tab at index zero, and ends with a tab only when the last index is also even — which happens exactly when the total count is odd. With an even count, one end of every edge is a notch with nothing filling it, and the box corner has a visible open slot. It is a hard geometric constraint, not a style choice.
Why do my three edges have different finger widths?
Because each edge is rounded up independently to the next odd segment count, and three different edge lengths rarely divide into odd counts at the same width. A 100 mm edge and an 80 mm edge at a nominal 7.5 mm finger land on different counts, so their actual finger widths differ slightly. That is correct and invisible on the finished box, since fingers only ever have to match the edge they share.
What is the difference between inside and outside box dimensions?
Twice the material thickness on every axis, which is far bigger than any kerf. A 100 mm box in 3 mm material is 94 mm inside; in 12 mm plywood it is only 76. People agonise over a 0.2 mm kerf and then lose 6 mm to a definition. Decide which measurement actually matters for whatever is going in the box, and set it explicitly rather than assuming.
How much kerf compensation does a finger joint need?
Half the kerf less a quarter of the wanted clearance, applied at every tab boundary — not half the clearance, which is the natural-looking answer and gives exactly double the fit you asked for. A tab and its slot share a boundary, so one unit of shift moves both and buys two units of clearance. The cut gap works out as twice the kerf minus four times the shift.
Why does my box still rattle after enabling kerf compensation?
Almost always because a single global offset was applied to the whole file. That corrects the outside dimension of each panel, which is useful, but it moves the slot and the tab the same direction — so the joint ends up exactly as loose as it started. Joints need the slot drawn narrow and the tab drawn wide, which is two offsets in opposite directions rather than one.
How thick should the material be for a finger jointed box?
Three to six millimetres suits most small boxes, and the choice matters more than it looks because the finger depth equals the thickness. Thicker material gives deeper fingers and more glue area per finger, but also loses more internal volume — twice the thickness on every axis — and demands wider fingers to stay cuttable. Three millimetre plywood is the usual default for a reason.
Should I measure the sheet or trust the label?
Measure, in several places. Nominal 3 mm birch plywood runs roughly 2.6 to 3.1 mm in practice, a spread larger than the entire kerf compensation, and it varies within a single sheet as well as between them. The finger depth in a generated box is set directly by that number, so a slot cut for the label rather than the sheet is loose or impossible depending on which board you picked up.
What fit should a laser cut box have?
Snug for a glued box — tight enough to hold itself square while the glue sets, loose enough to assemble by hand. A press fit with about 0.05 mm of interference holds without glue in plywood or MDF but will crack acrylic, which does not compress. Use a slip fit of around 0.1 mm for anything you intend to take apart, and for acrylic regardless.
Can I cut this box on a CNC router instead of a laser?
Yes, with two adjustments. Set the kerf to your cutter diameter rather than a laser kerf, since the same compensation arithmetic applies to any tool that removes width. And remember that a round cutter leaves a radius in every inside corner, which a laser does not — the fingers will not seat fully unless you add dogbone or T-bone relief cuts at the corners, which this generator does not produce.
How do I test the fit without wasting a sheet?
Cut one corner: two small panels with a single joint between them. That settles the kerf, the fit and the actual material thickness together in about a minute, and it is the only test that covers all three at once. A joint that seats correctly on the test piece will seat correctly on the box, in a way that no set of measured numbers quite guarantees on its own.
🔒 This tool runs entirely in your browser. Nothing you enter is uploaded, logged, or stored.