{"id":6461,"date":"2026-08-14T16:48:00","date_gmt":"2026-08-14T16:48:00","guid":{"rendered":"http:\/\/7haooffice.7thtech.net:82\/?p=6461"},"modified":"2026-09-24T09:20:04","modified_gmt":"2026-09-24T09:20:04","slug":"sand-3d-printing-accuracy-tolerance","status":"publish","type":"post","link":"https:\/\/www.lyafs.com\/fr\/blog\/sand-3d-printing-accuracy-tolerance\/","title":{"rendered":"Pr\u00e9cision et tol\u00e9rances de l'impression 3D au sable : ce que vous pouvez r\u00e9ellement obtenir"},"content":{"rendered":"

Plan on roughly \u00b10.3\u20130.5 mm at the mold feature level from a well-tuned binder jetting sand printer, with the real number moving around depending on geometry, sand type, and binder system. That band describes the mold itself. It does not describe the casting: every alloy shrinks as it solidifies, so a dimensionally perfect mold can still pour a casting that is several millimeters off. Most buyers who ask about sand 3D printing accuracy are really asking about finished parts, and the honest answer is that accuracy planning starts with the mold but ends at the foundry. This article is a principles guide \u2014 where accuracy comes from, how it is verified, and how to spec tolerances you can actually hold \u2014 not a machine spec sheet.<\/p>\n

\"Sand<\/figure>\n

The Direct Answer: What Sand 3D Printing Accuracy Actually Is<\/h2>\n

A binder jetting sand mold holds roughly \u00b10.3\u20130.5 mm on stable, well-oriented features. That is a band, not a fixed number, because three independent things set it: the layer height (typically 0.28\u20130.4 mm on production machines), the binder’s behavior in the sand, and how the feature is oriented in the build. A boss printed standing vertical holds tighter than a shallow slope, and a fine core printed in fine sand holds tighter than the same core in coarse foundry sand. Anyone quoting you a single “accuracy spec” without asking about your part is simplifying past the point of usefulness.<\/p>\n

“Feature level” matters more than “overall accuracy.” A 1,800 mm mold can hold its overall envelope within the same band while a 6 mm hole, a thin rib, or a core print sits at the loose end of it. That is normal, and it is the same behavior you see with conventional no-bake tooling. The machine defines the floor; the geometry and material define where you actually land. If you are new to the process, the layer-by-layer mechanics behind that band are explained in our guide to how sand 3D printing works<\/a>, which walks through recoating, printing, and curing.<\/p>\n

The caveat that gets skipped in most marketing: mold accuracy is not casting accuracy. The mold is the intermediate, not the deliverable. When the metal is poured, the mold wall deforms slightly under ferrostatic pressure, the coating soaks into the surface, and the alloy then shrinks and distorts during solidification. A foundry pouring pump bodies and valve housings does not ask the printer “what tolerance do you hold” \u2014 it asks what the mold will do after the metal arrives. That is why the rest of this article spends as much time on the foundry side of the equation as on the printer.<\/p>\n

Where Sand 3D Printing Accuracy Comes From<\/h2>\n

Accuracy in binder jetting is the sum of machine resolution, material behavior, and orientation \u2014 it is not one spec you can look up. The print head places binder droplets in the XY plane with good repeatability; the recoater lays sand in layers that set the Z resolution; the binder then bleeds through the grain pack and defines where the edge of a feature actually ends up. Each factor can be tuned, and each has a cost attached.<\/p>\n

\"High<\/figure>\n

The table below summarizes the five factors that dominate, what each one does to accuracy, and how a print service controls it. Read it as a control list: every row is something a serious service calibrates and documents.<\/p>\n

\n\n\n\n\n\n\n\n\n\n
Factor<\/th>\nEffect on accuracy<\/th>\nHow to control<\/th>\n<\/tr>\n<\/thead>\n
Layer height<\/td>\n0.28\u20130.4 mm steps on near-horizontal faces; Z dimensions land in layer increments, not continuously<\/td>\nPick layer height per part; orient critical faces vertical; accept steps on slopes or finish them<\/td>\n<\/tr>\n
Binder saturation<\/td>\nOver-saturation swells features and blurs edges; under-saturation gives weak, friable walls<\/td>\nTune saturation to the sand and part; calibrate per material; check bleed on holes and thin cores<\/td>\n<\/tr>\n
Sand grain size<\/td>\nCoarse grains pack rougher and blur fine detail; fine grains smooth surfaces but slow venting<\/td>\nMatch grain fineness to feature size and alloy; use finer sand for cores and detail faces<\/td>\n<\/tr>\n
Feature orientation<\/td>\nHorizontal tops show stair-stepping; vertical walls hold XY best; long unsupported spans sag<\/td>\nKeep critical dimensions in-plane; support long spans; orient fine features vertical<\/td>\n<\/tr>\n
Post-processing<\/td>\nCuring, coating, and handling shift dimensions; a mold measured green can move after baking<\/td>\nMeasure after curing; control coating thickness; handle and transport cured molds on flat supports<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/figure>\n

Layer height is the factor buyers ask about first, because it is printed on the brochure. It sets the Z resolution: at 0.4 mm layers, a vertical stack of 25 layers is 10.0 mm; the system cannot resolve a 9.9 mm dimension in Z without changing layer height. At 0.28 mm you gain resolution and lose build speed, which is why the tradeoff between layer height and throughput shows up directly in per-mold cost. For most molds, the geometry is what dictates the choice: large flat copes and drags print fine at 0.4 mm, while a complex core with thin sections earns the finer setting.<\/p>\n

Binder behavior is the factor that separates a calibrated process from an uncalibrated one. The inkjet printhead deposits droplets on demand, and the binder then wicks through the sand grains before curing. That wicking, plus the local saturation level, decides the true edge of every feature \u2014 a hole prints slightly undersized if the binder bleeds inward, slightly oversized if the print is tuned hot. This is also where open material systems matter: if you can change binder and sand independently, you can calibrate for your own alloy and feature mix instead of accepting a locked recipe. Our comparison of open vs. closed material systems<\/a> explains why that freedom shows up in both accuracy and running cost.<\/p>\n

Mold Accuracy vs. Casting Accuracy: Where the Real Tolerance Goes<\/h2>\n

The finished casting is where the buyer’s tolerance actually lives, and the casting’s deviation from nominal is dominated by alloy shrinkage, not by the mold’s \u00b10.3\u20130.5 mm. Typical linear solidification shrinkage for the common casting alloys runs roughly 1\u20132%: gray iron at the low end, steel and ductile iron higher, aluminum alloys in between. On a 400 mm dimension, 2% is 8 mm \u2014 an order of magnitude larger than the mold band. If the foundry does not compensate for that shrink in the CAD before printing, the mold can be perfectly accurate and the casting will still be wrong by several millimeters.<\/p>\n

Section thickness makes it worse. Thin sections freeze first and thick sections feed them, so the part shrinks at different times in different places and distorts instead of shrinking uniformly. Ribs pull, flanges bow, and the parting line gap changes. Because sand is a good insulator, the mold wall thickness and the cores also change the local cooling rate, which shifts where the shrink shows up. This is why two foundries pouring the same alloy can legitimately use different shrink allowances: their gating, risering, and section practices are different, so their compensation differs.<\/p>\n

Pattern makers have compensated for this for a century by building shrink into the pattern. Patternless printing just moves that step into CAD: the shrink allowance, machining allowance, and core-print clearances are applied to the digital model before it is printed, so the mold prints at final-plus-shrink instead of final. That is exactly the workflow described in our patternless sand casting guide<\/a>. The practical consequence is that a print service cannot quote you casting accuracy from the model alone \u2014 it needs your alloy, your section geometry, and ideally your gating layout. The foundry’s shrink factor<\/a> is part of the job, not an afterthought.<\/p>\n

This is also the point where a digital foundry earns its name. Casters moving to printed molds typically start with a trial casting, measure the result, and lock in their compensation factors for that alloy and part family. Pump and valve houses, prototype lines like the one supporting BMW Landshut’s casting work, and aerospace primes qualifying trial castings all run the same loop: print, pour, measure, adjust. The discipline behind that loop is what our article on the digital casting foundry<\/a> covers in more detail.<\/p>\n

How Accuracy Is Verified: From First Article to Trial Casting<\/h2>\n

Verification is a three-layer process, and each layer answers a different question. A coordinate measuring machine (CMM) touches a handful of datum features with high precision and tells you exactly where those points sit relative to nominal. It is accurate, slow, and sample-based \u2014 right for checking a few critical dimensions on a first article, wrong for understanding the whole mold at once.<\/p>\n

Structured-light scanning is the layer that catches what a CMM misses. A scanner projects a pattern onto the mold, captures a full-field point cloud, and compares it to the CAD model, producing a color map of deviation across the entire surface. A good industrial scanner resolves comfortably finer than the tolerances you are checking, so it tells you where the mold bowed, where a core shifted, and where the band actually sits \u2014 not just at three datum points, but everywhere. For complex cores where the internal geometry cannot be touched, computed tomography gives the same full-field comparison inside the part.<\/p>\n

The third layer is the one that matters most commercially: the casting trial. Scanning the mold verifies the mold; pouring and measuring the casting verifies the process. The typical workflow runs like this: print a feature artifact or the first mold, cure and measure it against nominal, adjust the shrink and compensation factors, pour a trial casting, measure the casting, and repeat until the loop closes. Once the factors are locked, production molds print to the same recipe with only spot checks.<\/p>\n

A print service worth using will do the first two layers for you and hand over the dimensional report. That report, plus a casting trial at your foundry, is what turns a printed mold from “close enough” into a documented, repeatable process. If you want the mold-and-core work handled by people who run this loop daily, our sand mold and core 3D printing service<\/a> takes parts from CAD through printed tooling with measurement included.<\/p>\n

When Accuracy Doesn’t Matter<\/h2>\n

A sand mold is a consumable. It is poured once and broken out of the casting; it never wears, never warps through reuse, and never needs to hold a tolerance on a surface the metal never sees. That simple fact changes the accuracy conversation completely. The runner system, risers, vents, and the far side of the parting line carry no dimensional requirement beyond “works,” and spending the tolerance budget on them is pure waste.<\/p>\n

More importantly, many of the features that need real precision in the finished part are machined after casting anyway. Bores, sealing faces, mating flanges, and locating surfaces get cut in the machine shop; the casting only needs to arrive with enough stock, consistently. Once a feature is going to be machined, the mold tolerance on it loosens to “stock present everywhere it is needed,” which is a far easier job than holding the finished dimension in sand. Designers who understand this stop asking for machining tolerances on a consumable mold and start asking for reliable stock.<\/p>\n

Compare that with a machined pattern, and the economics become obvious. A CNC-machined wooden or resin pattern can hold \u00b10.1 mm or better \u2014 but it costs on the order of \u20ac3,000\u201315,000, takes 4\u20138 weeks to make, and only earns that precision because it is reused thousands of times. A printed mold is single-use and accurate to \u00b10.3\u20130.5 mm, which is plenty for a casting that is going to be machined anyway. The traditional vs. patternless comparison on this site<\/a> walks through exactly where that tradeoff lands for different production volumes, and our sand 3D printing cost guide<\/a> shows how the tolerance decision feeds the per-part number.<\/p>\n

None of this argues that accuracy is irrelevant \u2014 it argues that accuracy should be bought where it pays. Hold tight tolerances on the features the casting must keep green, add stock and let the machine shop own the tight ones, and let the mold be a mold. That is how a foundry spends its tolerance budget without overspending on tooling.<\/p>\n

Practical Guidance for Designers and Buyers<\/h2>\n

Start with the features that will be machined. Mark every bore, face, and seat that gets cut after casting and add machining stock per the foundry’s convention \u2014 typically a few millimeters per side \u2014 before the model ever reaches the printer. Then check the features that must stay green: those are the only ones that actually test the mold’s accuracy, and they are the ones to dimension carefully and verify on the first article.<\/p>\n

Orient critical features deliberately. Keep dimensions that matter in the XY plane, where the print head defines the edge; avoid relying on Z accuracy across many layers; place fine details on vertical walls rather than horizontal tops; and support long spans so they do not sag under their own weight before curing. Orientation is free accuracy \u2014 it costs nothing to rotate the model, and it buys the tight end of the band on the features you care about.<\/p>\n

Talk to the print service about feature limits before you finalize the model. Ask for the minimum wall thickness, the smallest hole or core the binder system holds reliably at your sand choice, and the empirical band their calibration shows for your kind of geometry. A serious service runs calibration coupons and can show you one; a service that quotes one number with no questions is telling you everything you need to know. And because accuracy is a process property, not a machine property, it is worth comparing how different vendors calibrate before you choose \u2014 our sand 3D printer buyer’s guide<\/a> covers what to ask, and the current ranking of machines by build volume, value, and accuracy<\/a> gives a vendor-neutral starting point.<\/p>\n

Define your datums the way the foundry will measure them. If you dimension from the same reference faces the CMM and the machine shop use, you avoid tolerance stacks and arguments later. And match the spec to the need: if a feature is machined, do not spec the mold at machining tolerance; if a surface is cosmetic, do not spec it at all. The machines themselves are sized for real jobs \u2014 from a J1800 handling a medium mold to a J4000Pro printing multi-ton sections \u2014 and the full sand 3D printer product line<\/a> is the right place to check build envelopes against your part sizes.<\/p>\n

If you are spec’ing a mold for a specific alloy and part family, the fastest way to get a realistic tolerance plan is to talk to people who quote them daily. Request a formal quotation from the 3DPTEK team<\/a> and include your alloy, part size, and which features are machined afterward \u2014 you will get an honest accuracy band plus a recommendation on layer height and sand selection, not a brochure number.<\/p>\n

\"Traditional<\/figure>\n

Written by the 3DPTEK engineering team \u2014 the same team that operates our own foundries and sand 3D printers in daily production. About 3DPTEK<\/a>.<\/em><\/p>\n

Frequently Asked Questions<\/h2>\n

How accurate is 3D sand printing?<\/h3>\n

At the mold feature level, roughly \u00b10.3\u20130.5 mm, depending on geometry, sand type, and binder system. Layer height of 0.28\u20130.4 mm sets the Z resolution, and binder bleed sets the true edge of features in XY. That band describes the mold \u2014 the finished casting adds alloy shrinkage on top, so it is looser.<\/p>\n

What tolerance can I expect on a 3D printed sand mold?<\/h3>\n

Expect \u00b10.3\u20130.5 mm on stable, well-oriented features \u2014 vertical walls, bosses, and core prints \u2014 and looser results on thin sections, shallow slopes, and long unsupported spans. The foundry’s shrink compensation dominates the final casting, so the mold tolerance is only one input into the total accuracy picture.<\/p>\n

Is it as accurate as a CNC-machined pattern?<\/h3>\n

No. A machined pattern can hold \u00b10.1 mm or better, but it costs roughly \u20ac3,000\u201315,000 and takes 4\u20138 weeks, and it earns that precision through thousands of reuses. A printed mold is single-use and accurate to \u00b10.3\u20130.5 mm \u2014 close enough for a casting that will be machined after pouring.<\/p>\n

What surface finish can I expect?<\/h3>\n

The mold surface replicates the sand grain pack, so finish depends mainly on grain size: finer sand gives smoother surfaces, coarser sand gives rougher ones. Expect a texture comparable to a good green sand or no-bake mold. Foundry coatings and washes smooth the surface before pouring and improve the casting’s as-cast finish.<\/p>\n

How accurate will my finished casting be?<\/h3>\n

That is set by alloy shrinkage \u2014 typically roughly 1\u20132% linear \u2014 plus section thickness, gating, and the foundry’s practice. The mold’s \u00b10.3\u20130.5 mm is a small part of the total. With correct shrink compensation and a trial casting, printed molds meet the same casting tolerance classes as conventional tooling.<\/p>\n

How is accuracy verified?<\/h3>\n

In three layers: a CMM checks datum points precisely, structured-light scanning compares a full-field point cloud to CAD and catches warpage, and a casting trial measures the actual poured part. The workflow is print, cure, measure, adjust compensation, pour, and re-measure until the loop closes.<\/p>\n\n\n