Growth and shrink are not the same percentage
This is the part that costs people a form. A shop measures springback the obvious way: the form was twelve inches, the part came off at thirteen and a half, so it grew by twelve and a half percent. Then, wanting a finished radius of eighteen, they take twelve and a half percent off eighteen and cut a form at 15.75 inches. The part comes off that form at 17.72, 0.28 of an inch tight, and nobody can see where it went.
The correct operation is division, not subtraction. If the part comes off at 1.125 times the form radius, the form has to be the target divided by 1.125, which is sixteen inches, not 15.75. The size of the error has a tidy closed form: it is the target radius multiplied by the springback squared. At twelve and a half percent on eighteen inches that is 18 times 0.015625, or 0.28 of an inch. At twenty five percent it is more than an inch. The page prints both radii and the part each would produce, so the difference is in front of you rather than in the form. Those are the exact figures it prints with the scaling field set to do not scale, which is the setting that uses the trial number as it stands.
Plies, not thickness
A three quarter inch curved rail made from three quarter inch stock is a steam bending problem. The same rail from eight plies of three thirty-seconds is a gluing problem, and the two behave completely differently, because in a lamination the glue lines are what stop the plies sliding past one another as the stack bends. More glue lines mean less slip, and less slip means less stored energy trying to open the curve back up.
The usual shop assumption is that springback falls with the square of the ply count at a fixed total thickness, and this page will apply it if you ask it to. It is worth being clear that it is an assumption. What is not an assumption is the direction: a stack of many thin plies opens up less than a stack of a few thick ones at the same finished thickness, which is why the answer to a stubborn bend is almost always thinner stock rather than a tighter form. The honest way to use this page is to run one trial at the ply count you actually intend, and then set the scaling field to do not scale.
The blank is thicker than you think
Eight plies at three thirty-seconds is 0.75 inches of finished stock and nothing like 0.75 inches of board. Each ply has to be sawn oversize and taken down, and every cut between two plies throws away a kerf. On the defaults here, nine plies including a spare, at a thirty-second of planing allowance and a 0.045 kerf, need a blank 1.485 inches thick to produce 0.84 inches of ply. Nearly half the board is sawdust and shavings.
That ratio is why bent laminations are usually resawn from one wide board rather than assembled from separate pieces: consecutive plies off the same blank are sequential, so the grain runs continuously around the bend and the finished edge reads as one piece. Losing that continuity is the real cost of running short.
Where this sits
A curved stack needs even pressure over its whole area, which in practice means a bag rather than a row of clamps — the vacuum press pump down calculator covers the pump side and the glue and clamp calculator covers pressure and clamp counts. Flat panel work from the same resaw is the veneer layup calculator, and the stock arithmetic in board feet is the board foot calculator. Sheet metal springs back too, by an entirely different mechanism involving yield and modulus rather than glue line slip, and that is the sheet metal springback calculator.
Questions people ask
How do I measure the radius of a part that came off the form?
Three points and some arithmetic beat a template. Lay the part on a flat surface, stretch a string across the inside of the curve between two points a known distance apart, and measure the deepest gap between the string and the curve. The radius is the chord squared divided by eight times that gap, plus half the gap. A 24 inch chord with a 4 inch sag gives 576 divided by 32, plus 2, which is 20 inches. Do it on the form as well as the part, so both numbers come from the same method.
Why does the form radius come out smaller than the finished radius?
Because the part opens up when the pressure comes off, so the form has to be tighter than the thing you want. That is what overbending means. If your numbers ever produce a form radius larger than the target, either the springback entered is negative, which is not physical for a laminated bend, or the test bend measurements are the wrong way round.
Can I skip the test bend and use a published springback figure?
You can look one up, but it will be a figure for someone else's glue, ply thickness, ply count, clamp time and species, and every one of those moves the answer. This page does not supply a figure for exactly that reason. A test bend is one form, a handful of offcuts and an afternoon, and it produces a number that belongs to your shop. If you must start from a published figure, treat the first real part as the trial and expect to recut the form.
Does the glue matter as much as the ply count?
It matters, and this page cannot tell you by how much. The mechanism is that the glue line has to resist the plies sliding past each other under sustained load, and adhesives differ enormously in how much they creep over hours and years. That is a data sheet property and in some cases not published at all, which is another argument for the trial: your trial includes your glue, held for your clamp time, at your shop temperature, and no calculator has access to any of that.
What ply thickness is safe for a given radius?
Not a question this page answers, and be wary of anything that does answer it confidently. The strain in the outer fibre of one ply is roughly the ply thickness divided by twice the bend radius, and the page prints that ratio, but what strain a particular board of a particular species at a particular moisture content will take before it breaks varies hugely, including between two boards off the same tree. The usual method is to resaw one ply and bend it dry around the form. If it survives, you have your answer for that stock.