Test coupons and calibration

The most expensive way to discover that a clearance is wrong is to print the whole part first. Six hours on the bed, the bearing won't seat in its bore, and you've learned exactly one thing about a single fit. A coupon learns it in fifteen minutes for a few cents of filament — a small, deliberately cheap print whose only job is to measure what your machine actually does, not the manufacturer's, not the forum's. Print it, read the result, write the number down, then print the real part knowing the answer instead of guessing it.

A coupon measures what your printer does, not what you drew

The dimension you type on screen isn't the one that comes off the bed, and the offset isn't the same on any two machines. Extrusion width eats into your curves, cooling shrinkage pulls the material in, and your slicer's calibrated flow nudges it up or down a few tenths. The filament adds its own share. PETG oozes and spreads wider than PLA; humidity, color, and brand each shift the figure on their own. The upshot is that two printers running the same file with the same settings give a tight fit on one machine and a loose one on the other. There's no universal number to copy; there's only the one your setup produces, and a coupon is the cheapest honest way to get it.

Three of them cover almost everything you design:

• The tolerance tower measures how much a hole closes up and how much a shaft fattens on your machine. It's a series of calibrated holes tested against a known rod — an 8.0 mm rod against holes of 7.9, 8.0, 8.1, 8.2, and 8.3 mm, say — each labeled with its nominal clearance. You try the rod in every hole. The holes at or below the shaft diameter press in or won't go at all, and as you climb you find the first one that holds fast, the one that slides with no play, and the one that spins free. Those three holes give you your three real clearances.
• The clearance coupon does the same for one specific fit you're already designing: the actual joint geometry, repeated with the clearance stepped, so you never extrapolate from one diameter to another.
• The break coupon measures strength, not fit. A short bar that you pull or bend by hand until it gives, printed in the same orientation as the final part, shows you where it splits — almost always along a layer line, the weak plane between beads. Pulling by hand won't give you a reproducible force, but it gives you what matters: the failure mode and the comparison between orientations.

The discipline is the same in all three: isolate one variable, print a strip of values around it, and pick the one that works. You're not printing a part — you're printing a measurement.

Print the coupon in the final part's orientation

A coupon only tells you the truth if it's printed the way the real part will be, because the offset you're measuring depends on orientation as much as on material. A hole printed upright comes out reasonably round: each layer is a closed ring resting on the one below. The same hole laid down prints with its upper half as an overhang, and near the crown of the circle the hole wall gets so close to horizontal that each bead hangs unsupported over the last. The roof of the hole comes out collapsed and rough, and the circle deforms into an oval. The effective clearance of that horizontal hole has nothing to do with that of the upright hole.

The same goes for breaking. Strength between layers runs about 40–80% of what the material offers along the bead. Where you land in that range depends heavily on material and temperature — well-printed PLA reaches the high end; poorly tempered PETG or ABS drops below 40%. So a coupon printed standing up splits along a layer line under far less load than the same coupon laid down, where the beads run in the direction of the stress. Measure the bar in one orientation and build the part in another, and the number you wrote down describes a part you never printed.

Watch where the fit falls within the coupon as well. Elephant's foot widens the first few layers, so the mouth of a hole nearest the bed comes out narrower than the rest and gives you a falsely tight clearance. Measure in a region away from the first layer, or compensate for elephant's foot in the slicer before printing the coupon.

And only use it across similar diameters. Bead error is absolute: a half-bead biting into the gap measures the same in a 2 mm hole as in a 20 mm one. But its relative weight, and the effect of curvature, change with size, so a tower of 6 mm shafts doesn't guarantee the same gap on a 20 mm pivot. Measure near the size you're going to use.

Measure once, write the number down, reuse it

The point of a coupon is to turn a trial-and-error fit into a known constant. Once the tolerance tower has told you that your real sliding gap is 0.12 mm per side in PLA, that number stops being an unknown: you apply it directly to all your pivots, hinges, and carriages without printing a single coupon again. Same with the overhang limit, the longest bridge distance that holds without sagging, or the nozzle temperature that gives clean layers: measure them once and spend them a hundred times.

These numbers barely move as long as you don't change filament type, nozzle, or machine, and don't seriously alter your flow calibration. They do drift, though — when filament absorbs humidity between sessions, when a different batch comes in, or as the nozzle wears — which is why it's worth rechecking them if something starts coming out odd. Until then, they're yours. Write them on a slip taped to the side of the printer or drop them in a notes file; one calibrated number you trust is worth more than a hundred reprints of the real part. That's exactly what Real printed clearances asks you to capture and keep, and what Iterate and measure turns into a repeatable method. The coupon is how you produce those numbers in the first place.

Iterating a coupon is cheap; reprinting the part isn't

The economics are simple. A tolerance coupon prints in fifteen minutes and burns a few grams of filament. The real part might be six hours and half a spool. Nail the fit on the first try because the coupon already gave you the number, and you print the part once. Guess it, and every failed attempt is the whole part again: the cost of the error scales with the size of the part.

Even when the coupon itself needs a second try — because the first strip didn't cover the good range, or because you want to fine-tune between two columns — it's still cheap. Iterating for fifteen minutes until you nail the number is the investment; reprinting for six hours until you stumble onto it is the waste. Coupons look like a detour when you're itching to print the real part. They're the opposite — the fifteen minutes that keep you from burning six hours on a part that was never going to fit.

With the number in hand, the next step is turning measurement into a habit. Iterate and measure gives you the method to pin down any fit in a few runs; Real printed clearances tells you what gap to expect before you even print the first coupon.