A resin printer does not reproduce the CAD boundary directly. It projects energy into a reactive liquid, separates a partially cured layer, washes the part and finishes polymerisation in post-cure. Every stage can move the final surface.
Dimensional accuracy is the result of optics + resin photochemistry + mechanics + support stiffness + wash + post-cure + measurement method. Tune compensation only after those variables are stable. If exposure, resin temperature or cure cycle is still moving, a scale factor merely hides process drift.
Vat photopolymerisation works because the resin receives enough radiant exposure to gel/cross-link. A widely used first-order description is the Jacobs working curve, which relates cure depth to incident exposure through the resin's penetration depth and critical exposure.
The equation explains why “more exposure” does not simply make the same geometry stronger. It changes how far polymerisation extends beyond the intended boundary. Pigment, filler, wavelength, photoinitiator and the printer's optical spectrum all affect the relationship.
In XY, edge position depends on the projected image, optical blur, pixel pitch, anti-aliasing and lateral cure beyond the nominal pixel boundary. In Z, layer thickness, penetration into adjacent resin, build-platform calibration, release deformation and accumulated motion error matter. A single “printer accuracy” number hides these different mechanisms.
Increasing exposure can enlarge pins, walls and bosses while narrowing holes and slots. The effect is feature-size dependent; a universal -0.05 mm offset is rarely correct across 0.5 mm slots and 50 mm housings. Calibrate positive and negative features separately and include a range of sizes.
Early layers often receive much higher exposure for build-plate adhesion. That creates the familiar “elephant foot” and can distort dimensions close to the platform. If a critical datum must be printed directly on the plate, use machine-specific early-layer compensation or sacrifice geometry designed to be removed later. Otherwise, move the precision feature away from the over-exposed region.
Each layer separation applies load through the support tree and the green part. Large cross-sectional area, long lever arms and insufficient supports can shift surfaces even when exposure is correct. A deviation that looks like shrinkage may actually be elastic/plastic deformation during peel or tilt release.
For critical parts, compare alternative orientations and inspect the deviation field. If error correlates with support direction or section area, do not “fix” it with global XY scaling.
Post-curing increases cross-link density and can densify the polymer network. Formlabs explicitly notes minor shrinkage during cure and builds compensation into its material settings. The general lesson is that dimensions should be qualified after the exact cure schedule used for production.
Measure only after the part has returned to a stable temperature. A warm freshly cured part can mislead high-resolution measurements.
Resin viscosity, refill behaviour and reaction kinetics vary with temperature. If the printer heats the vat, define a soak condition before the first exposure. If it does not, qualify the process over the real ambient range or control the room. A compensation factor derived at 18°C may not reproduce at 28°C.
Use one plate containing:
Place duplicates across the build area. Plate-position differences reveal optical uniformity or mechanics that one central cube would miss.
Once the process is stable, derive correction factors from repeated data, not one print. Then reprint the artefact and confirm the residual error. A compensation that improves the mean but doubles the variation is not an improvement in process capability.
For production work, collect enough measurements to estimate mean, standard deviation and drift. A simple control chart can reveal resin-lot changes, LCD ageing, film wear or cure-unit variation long before a customer dimension fails.
If you are attempting to qualify ±0.05 mm features with a measurement method that repeats only to ±0.04 mm, the printer may not be the limiting system. Use a gauge suitable for the feature and run a basic repeatability/reproducibility check where the tolerance is tight.
Keep nominal geometry nominal where possible and store process compensation in the print profile, slicer or controlled manufacturing variant. If the CAD is altered to make one printer produce the right hole, the model becomes wrong for machining, moulding and any future printer.
Sources and standards checked at the review date.
