Home› Technical Notes› Layer Thickness vs. Surface Finish in DLP

Layer Thickness vs. Surface Finish in DLP Jewelry Casting: An Engineering Guide

By NEWAY 3D Printing•Published 2026-11-21•Technical Notes

Surface finish is the first thing a jeweler judges on a cast piece, and in DLP wax printing it is decided long before the investment flask goes into the furnace. Three variables dominate that decision: layer thickness in Z, pixel pitch in XY, and the exposure dose used to cure each layer. They are not independent. Changing one without adjusting the others is the most common reason a printer that produces acceptable patterns at 50 µm suddenly delivers rough or dimensionally loose patterns at 25 µm.

This note works through what each variable controls, how the errors add up, and how to calibrate a DLP platform so that a thinner layer actually buys a smoother casting instead of only a longer build.

What layer thickness actually controls

Layer thickness is the vertical sampling interval of the model. It does not smooth a surface; it decides how closely the printed surface approximates the intended one. Every inclined or curved face is rebuilt as a stack of prisms, and the visible result is a staircase whose step height equals the layer thickness and whose tread width depends on the local slope angle.

For a face inclined at angle θ from horizontal, the horizontal offset between consecutive layers is:

Δx = t / tan(θ)

where t is the layer thickness. At 50 µm and a 30° overhang, Δx is about 87 µm. At 25 µm it drops to 43 µm. That is the entire geometric argument for thin layers: the staircase does not disappear, it halves. Anyone promising a "smooth" surface from layer thickness alone is describing a smaller staircase, not the absence of one.

The staircase effect, quantified

The arithmetic matters because it explains why some models benefit from 25 µm and others do not. Near-vertical walls have a large tan(θ), so Δx shrinks and the staircase becomes invisible even at 50 µm. Near-horizontal surfaces have a small tan(θ), so Δx grows and the steps become obvious. A ring shank, a domed signet face and a filigree surface are three completely different problems at the same layer thickness.

In practice, the roughness contribution from staircasing scales roughly linearly with t for a given slope. Moving from 100 µm to 50 µm halves the step height; moving from 50 µm to 25 µm halves it again. The gain never reaches zero, and the cost in build time is strictly proportional to the reciprocal of t. A 25 µm build takes twice as long as a 50 µm build of the same model, and four times as long as a 100 µm build.

Pixel pitch sets the horizontal floor

Layer thickness is only the Z term. The XY resolution is fixed by the projection system, and no layer thickness can recover detail that the pixel grid cannot address. A 4K DLP projector covering a 192×108 mm build area gives roughly 47 µm pixels; the same resolution over a smaller field gives smaller pixels and finer edge definition. This is why two printers with identical layer thickness settings can produce visibly different surface quality.

The practical consequence is that below a certain layer thickness, further reduction buys very little. If the pixel pitch is 47 µm, dropping from 50 µm to 25 µm layers halves the staircase in Z but leaves the XY contour error unchanged. For surfaces dominated by horizontal curvature rather than vertical slope, the pixel pitch is the binding constraint, and time spent at 25 µm is largely wasted.

Peel force: why thin layers are harder, not easier

There is a widespread assumption that thinner layers are mechanically gentler because each layer is a smaller increment. The opposite is often true on conventional release films. A thinner layer means a larger number of separation events for the same build height, and each event still has to break the same cured interface. With a standard film, the accumulated peel load per unit of build height rises as layer thickness falls, which shows up as more frequent support failures on tall, thin parts and as visible layer lines from elastic recovery of the part between layers.

Oxygen-permeable release films change this calculus. A persistent inhibition layer at the interface keeps the cured part from bonding to the film, so the separation force stays low regardless of layer count. That is the condition under which 25 µm and even 10 µm layers become practical for production rather than a laboratory demonstration: the geometry improves, and the peel penalty that normally accompanies it does not appear.

Exposure dose must follow layer thickness

Photopolymer cure is governed by absorbed energy, not by time. The dose delivered to a layer is irradiance multiplied by exposure time, and the depth to which that dose penetrates is described by the Beer–Lambert relation. Halving the layer thickness does not halve the required exposure, because the resin still needs enough energy to reach the previous layer and form a bond, and because oxygen inhibition consumes a fixed portion of the dose at the free surface.

The failure modes are asymmetric. Under-exposure at thin layers produces weak interlayer bonding and delamination that only appears after burnout, when the pattern has already been invested. Over-exposure produces over-cure: the cured region spreads laterally beyond the intended pixel boundary, closing fine gaps, filling slots and growing external dimensions. On a ring, that shows up as a size that will not fit the mandrel after casting. The correct procedure is to hold the dose per unit thickness roughly constant and then fine-tune with a dimensional test coupon rather than a visual one.

Choosing layer thickness per model class

A single layer thickness for every job is a compromise that serves nobody. The following starting points are conservative and assume a well-calibrated 4K DLP platform with an oxygen-permeable film.

Model classLayer thicknessRationale
Heavy signet and mens ring bodies50 µmLarge flat and gently curved faces; staircase is not the dominant defect
Thin shanks, prongs, fine settings25 µmSlope-dominated surfaces; support stability matters more than speed
Filigree, micro-pave, texture25 µm or belowFeature size approaches pixel pitch; need maximum Z sampling
Master models for rubber moulds25 µmSurface is transferred to every production copy, so errors multiply
Large sculptural or hollow forms50–100 µmSurface curvature is low; build time dominates the economics

These are starting points, not rules. The correct value for a specific model family is established by printing a dimensional coupon and a surface coupon at two candidate thicknesses, casting both, and measuring. Visual inspection under a bench lamp is not a measurement.

A calibration routine that holds

Calibration drifts because resin changes. Viscosity shifts with temperature, pigment settles, and a partially used bottle behaves differently from a fresh one. A routine that survives production pressure has four steps and takes under an hour.

First, establish the dose curve. Print a stepped exposure matrix at the target layer thickness and identify the window where interlayer bonding is complete but lateral growth is still within tolerance. Record the window, not a single number.

Second, measure dimensions, not appearance. Print a coupon with a known outside diameter, a known slot width and a known pin gauge, then measure all three after post-cure. Over-cure shows up as a diameter that runs large and a slot that runs narrow. Adjust exposure in small increments until both are inside tolerance.

Third, verify the Z term. Print a tall step block and measure total height against nominal. A consistent negative error indicates compression from over-driven first layers or insufficient support; a positive error points to over-cure at the base.

Fourth, re-check after any resin batch change. The dose window is a property of the resin lot as much as the machine. A new lot that prints visibly darker or lighter needs a fresh coupon, not an assumption that the old recipe still applies.

Bottom line

Layer thickness is a sampling decision, not a smoothing control. It sets the vertical step height; pixel pitch sets the horizontal floor; exposure dose decides whether the part is dimensionally correct. Thinner layers only pay off when the release mechanism keeps peel force low and the dose is recalibrated for the new thickness. Get those three aligned and 25 µm produces castings that need less finishing. Skip one and the same setting produces patterns that look acceptable in resin and fail after burnout.

If you are evaluating a DLP platform for jewelry casting, the useful question is not the minimum layer thickness on the datasheet. It is whether the machine holds dimensional tolerance and surface quality at the layer thickness you actually intend to run, with the resin you actually intend to cast.

Ready to discuss your project?

Tell us what you need — our engineering team replies within 24 hours.

Or contact us directly

💬
WhatsApp / WeChat
▶
YouTube
🌐
DLP / LCD 3D printers & casting resins for jewelry and dentistry. Reply within 24 hours.