Two parts from the same 3D printing service, delivered in the same week, both failed a ±0.05 mm tolerance on the same feature. The first was an FDM nylon bracket – the hole drift was due to its placement at the edge of the build platform, where the temperature-controlled chamber was colder. The second was an MJF nylon casing, which was within spec immediately after printing, but shifted by another 0.08 mm after vapor smoothing. Two seemingly identical failure reports, two completely different reasons, and "tightening the drawing tolerance" would not be the answer in either case. The tolerance of a 3D printed part is not a number that can be directly applied to a process, but rather the sum of material behavior, build orientation, on-bed position, and post-processing – a design strategy that can reliably sustain mass production will treat these four factors as a single decision.
What does tolerance truly mean in 3D printing?
"3D printing tolerance" is often treated as a uniform number applicable to any scenario. For design teams entering mass production, this framework is the first thing that needs to be discarded. Additive manufacturing is actually a collection of processes – photopolymerization, polymer powder sintering, metal laser melting, material jetting – each with its own layer physics, thermal history, shrinkage behavior, and post-processing chain. A tolerance that is routine for SLA might be challenging for MJF; dimensions consistently achievable with DMLS might be difficult for FDM with the same geometry. A useful way of thinking is to bind tolerance to four things simultaneously: the function of the feature, the process used to print it, its orientation in the build chamber, and the final inspection method. An isolated number is meaningless; a number tied to these four anchors is the basis for predictable parts.
The goal is never to achieve the tightest tolerance on every dimension, but to achieve stable functionality at an acceptable cost. A drawing that marks all dimensions as ±0.02 mm actually says nothing – it will be ignored where the process cannot meet it, but increases inspection burden on features where it's not needed. A drawing that only tightens critical mating surfaces, bearing holes, and datum features, while intentionally loosening others, is what a production line can truly execute.
Why does print size drift?
Dimensional variation in 3D printed parts is rarely attributable to a single cause. Each process family has its dominant sources of drift, and compensating for the wrong reason will make failures seem random – until the true cause is identified.
| Process Family | Dominant Drift Source | How it Manifests |
|---|---|---|
| SLA / DLP / PolyJet | Curing shrinkage, cleaning deformation, post-curing growth | Feature size changes between cleaning and post-curing; large plate warping |
| SLS / MJF (Polymer Powder Bed) | Heat accumulation, cooling shrinkage, bed position gradient | Same feature has different sizes at the center vs. edge of the bed |
| FDM | Bed adhesion, chamber temperature, extrusion width variation | First few layers constrained by platform; long span warping |
| DMLS / SLM (Metal) | In-build thermal expansion, stress relief, heat treatment | Part continuously moves during build, stress relief, and heat treatment |
| Binder Jet Metal | Green body shrinkage during sintering | Overall shrinkage of 15–25% – compensation must be built into CAD |
These are not anomalies, but rather inherent behaviors of the process. The true design task is to determine which dimensions require process compensation, which need verification through pilot samples, and which can be relaxed because they do not primarily govern function.
Material, layer thickness, and part size all affect the outcome
Layer thickness dictates the achievable resolution of a part even before you touch CAD. Thinner layers improve detail and surface fidelity, but build time increases linearly. At approximately 200 µm layer thickness, stepping artifacts on curved surfaces are visibly apparent; at 100 µm, they become barely discernible; below 50 µm, the benefits of further thinning are outweighed by other sources of error. Layer thickness is not a panacea for tolerance, but setting it rigidly without aligning it to the features needing resolution will needlessly waste process margin.
Part size will amplify problems in ways that test coupons cannot reveal. Small features are stable on almost any machine because thermal and positional errors have nowhere to accumulate; a 300 mm long structural component will accumulate thermal gradients, shrinkage mismatch, and platform positional errors along its length. A single hole on that part might measure correctly, but its relative position to another hole could exceed assembly tolerance. This is why large parts require more comprehensive datum planning and more realistic inspection than test coupons – you are not measuring the same thing.
Build orientation is part of the tolerance strategy
Build orientation is often delegated to production as a scheduling detail. In practice, it's a tolerance decision. If circular features are oriented with their axis vertical, their as-printed contour will be completely different from a horizontal orientation – a vertical circle is built layer by layer (visible stair-stepping, tighter radial tolerance), while a horizontal circle is formed by sintering or curing arcs (smoother surface, but overhanging sides may sag). The same sealing flange printed facing up might achieve 0.1 mm flatness, but could warp by 0.3 mm if printed horizontally with a large span. A small angle of 10–15° off the main plane is often the right answer – it distributes surface quality, support burden, and dimensional stability across the entire part, rather than concentrating all three problems along a single axis.
The specific consequence for design teams is that any tolerance-sensitive feature should be brought up for orientation review before quoting is finalized. If hole locations, sliding surfaces, or sealing surface heights are highly dependent on build orientation, this should be identified during design discussions, not after the first batch fails inspection.
Common planning tolerances for each process
Planning tolerances are typically expressed as a "base value per inch + percentage of nominal length." The table below integrates the most common starting values for major additive manufacturing processes. Treat this as a planning reference, not a specification – specific parts on specific machines with specific post-processing may be significantly tighter or looser than the reference.
| Process | XY (per inch / 25 mm) | Z (per inch / 25 mm) | Length Factor | Directional? |
|---|---|---|---|---|
| SLA / DLP | ±0.05 mm (±0.002 in) | ±0.127 mm (±0.005 in) | +0.1% of nominal length | Yes – Z looser |
| PolyJet | ±0.127 mm (±0.005 in) | ±0.127 mm (±0.005 in) | +0.1% of nominal length | Slightly |
| SLS (Polymer) | ±0.25 mm (±0.010 in) | ±0.25 mm (±0.010 in) | +0.1% of nominal length | Essentially isotropic |
| MJF (Polymer) | ±0.30 mm (±0.012 in) | ±0.30 mm (±0.012 in) | +0.1% of nominal length | Essentially isotropic |
| FDM | ±0.15 mm to ±0.50 mm | ±0.15 mm to ±0.50 mm | Machine-dependent | Yes – first few layers tighter |
| DMLS / SLM (Metal) | ±0.076 mm (±0.003 in) | ±0.152 mm (±0.006 in) | +0.1% of nominal length | Yes – before heat treatment |
Two caveats make this table robust. The metal row refers to the as-printed state; stress relief and heat treatment will further shift dimensions, and critical tolerance metal parts often reserve 0.3–1.0 mm for post-machining on datums and mating surfaces. The polymer powder bed rows (SLS, MJF) describe the "sandblasted but not vapor-smoothed" state; vapor smoothing adds approximately 0.05–0.10 mm to nominal feature sizes (solvent reflow deposition), and this increase must be pre-planned into clearances or subtracted at the CAD stage.
How post-processing changes dimensions
The most common single tolerance error is measuring before post-processing is complete and then treating that number as the final value. Every post-processing step changes something. The magnitudes in the table below are typical values – not every part will experience all of them, but if the post-processing chain includes any of these steps, the nominal features on the CAD must be considered before planning inspection.
| Operation | Dimensional Effect | Typical Magnitude | Design Correspondence |
|---|---|---|---|
| Sandblasting | Uniform material removal from exposed surfaces | 0.05–0.10 mm per side | Add 0.05 mm to critical outer walls |
| Vapor Smoothing (Polymer) | Solvent reflow deposits material on edges and fine features | Small features +0.05–0.10 mm | Subtract 0.08 mm from nominal when specifying vapor smoothing |
| Manual Sanding | Local removal, highly directional | 0.05–0.30 mm per side | Reserve machining allowance for sanded surfaces |
| CNC Post-processing (Metal / Rigid Polymer) | Controlled removal to datum | 0.3–1.0 mm allowance | Datum surfaces printed oversized from nominal |
| Primer + Painting | Cumulative coating on all surfaces | +0.05–0.15 mm per side (two layers) | Subtract coating thickness from mating surfaces |
| Heat Treatment (Metal) | Isotropic shrinkage / expansion, part-dependent | Up to ±0.3% of nominal | Verify with test samples before release |
| Dyeing (SLS) | No geometric effect, but seals surface roughness | Negligible | Plan clearances for either dyed or undyed, not mixed |
Inspection: Match the method to "how much error you can afford"
Inspection costs escalate sharply with precision, and the worst default is "use the most precise one just to be safe." The correct default is to match the inspection method to the consequences of feature failure, plus one level of margin: for dimensions where drift only adds an hour of rework, calipers suffice; for dimensions where drift can trigger a recall, a CMM or CT scan is needed.
| Method | Typical Resolution | Relative Cost | When to Use |
|---|---|---|---|
| Calipers, Micrometers, Go/No-Go Gauges | ±0.02 mm | 1× | General dimensional checks, features with loose fit tolerances |
| Optical / Visual Measurement | ±0.01 mm | 2–4× | Planar features, 2D pattern verification |
| First Article Inspection (FAI) | Instrument-dependent | 3–5× | Production control implementation, batch verification |
| CMM (Coordinate Measuring Machine) | ±0.003 mm | 5–10× | Critical assembly datums, mating features |
| 3D Scanning (Structured Light) | ±0.05 mm within 100 mm | 4–8× | Whole part deviation maps, large organic geometries |
| Industrial CT | ±0.02 mm, including internal features | 15–30× | Internal channels, hidden porosity, high-consequence parts |
Application Cases
Press-fit Bearing Holes on an MJF Casing
A robotics team is making a polymer casing that requires press-fit stainless steel bearings, nominally 8.000 mm. The bearing supplier requires an interference fit of 0.02–0.04 mm – meaning the hole should be between 7.960–7.980 mm. MJF's typical capability of ±0.30 mm for a 25 mm nominal length is an order of magnitude worse than this requirement. The team has two ways to save this design, and the choice depends on volume.
Prototype and early production stages (< 300 units) – The design prints the hole 0.4 mm undersized, then reams it to spec as a secondary operation. A reaming station can consistently achieve 7.975–7.980 mm, and the added cost per part is much less than switching to a metal press-fit kit or changing processes.
Key design action: Tolerance is not solved at the printing stage, but at the system design stage – determining which part bears the precision. MJF is an excellent casing process, but making it bear bearing hole tolerance is using a tool for something it's not good at. Shift precision to reaming or press-fitting metal inserts – let each process do what it does best.
Large DMLS Bracket Grows 0.4 mm After Stress Relief
A 280 mm long aerospace-grade titanium bracket was printed at nominal size, measured within ±0.15 mm on the build platform, but grew by 0.35–0.45 mm along its longest axis after stress relief. The team initially assumed that the as-printed dimensions would carry through to after heat treatment; they subsequently added witness samples to each build, empirically measured the shrinkage coefficient, and applied a -0.0015 scale factor in CAD so that nominal dimensions would be in spec after heat treatment. Since then, each production batch is verified with test samples before parts are released, ensuring the samples fall within the pre-verified shrinkage band.
SLA Dental Model Inspected Before Post-Curing
A dental lab used SLA to print full arch dental models, measured the inter-cuspal distance immediately after cleaning, and then sent them to clinics for use. The crowns from the clinics didn't fit. The models were measured before post-curing, when the resin was still in a "green body" state and dimensions were unstable. Post-curing shifted the arch by approximately 0.1 mm – not a large absolute value, but enough to cause crowns to be rejected. The solution was a process change: the inspection point was moved to after post-curing, and a 30-minute cooling window was added before measurement. The clinic's first-time fit rate returned to over 95%.
Practical Tolerance Classification Table
One decision that reduces disputes more than any other is to clearly define three levels before drawing release. The default values in the table below apply to most general engineering polymer parts; metal and precision optical parts typically require one tighter level.
| Tolerance Level | Use Case | Typical Value (Polymer AM) | Inspection Method |
|---|---|---|---|
| General (Loose) | Non-critical outer walls, cosmetic features | ±0.5 mm or ±0.5% (whichever is greater) | Calipers, sampling |
| Functional | Clearance fits, fastener pockets, mounting holes | ±0.2 mm | Calipers or go/no-go gauges, 100% inspection |
| Precision | Sealing surfaces, bearing holes, press fits, datum features | ±0.05 mm (typically achieved by secondary operations) | CMM or FAI, recorded |
Do / Don't Comparison
| Do | Don't |
|---|---|
| Bind tolerance to function, process, orientation, and inspection | Apply the same tolerance to the entire drawing |
| Specify whether inspection is done before or after post-processing | Assume as-printed dimensions equal final dimensions |
| Reserve machining allowance for critical metal surfaces | Expect DMLS to achieve ±0.05 mm without secondary operations |
| Include witness samples for every batch of heat-treated metal | Use calibration from several months ago |
| Flag orientation-sensitive features to the build operator before quoting | Allow automatic orientation when there are sealing surfaces |
| Transfer precision to parts that are already machined if it benefits the system | Force every feature to meet spec at the printing stage |
| Select inspection method based on failure consequence | Default to "most precise just in case" |
Common Errors and How to Avoid Them
| Error | Why it Fails | How to Avoid |
|---|---|---|
| Same tolerance for the entire drawing | Either ignored where unachievable or wastes inspection costs where unnecessary | Categorize features into General / Functional / Precision |
| Locking in tolerances before selecting the process | Process capability may not meet it; cost skyrockets or parts are scrapped | Select process first, then define achievable tolerances for that process |
| Inspecting before post-curing, heat treatment, or vapor smoothing | Dimensions will still shift after inspection; reported numbers are meaningless | Define inspection point as "after final post-processing" |
| Ignoring orientation's impact on critical features | Sealing flange warping, circular feature drift | Incorporate orientation review into the design release process |
| Drawing lacks datum structure | Cannot diagnose cause of drift when parts are out of spec | Clearly mark primary, secondary, and tertiary datums |
| Treating SLS and MJF as equivalent for precision features | MJF has higher density, lower porosity; different post-processing shifts | Confirm inspection state by process (sandblasted vs. vapor smoothed) |
Pre-release Checklist
Go through this during design review – five minutes can prevent failures that typically surface three weeks later during assembly.
- Every critical mating dimension has a clear tolerance, not inherited from the title block default
- Primary, secondary, and tertiary datums are marked on the drawing
- Process and acceptable alternative processes are specified on the drawing
- Orientation-sensitive features are flagged to the build operator
- Every tolerance annotation specifies the inspection state (as-printed / sandblasted / vapor smoothed / post-heat treatment)
- Post-processing offsets (coating thickness, vapor smoothing increment, machining allowance) are reflected in the nominal dimensions
- Inspection method is chosen based on failure consequence, not lab default
- First article inspection plan is recorded, metal heat-treated parts include witness sample requirement
Design Takeaways
The tolerance performance of a 3D printed part is not achieved by writing ±0.05 mm on a drawing, but by creating a consistent decision chain that links design intent, process selection, build orientation, post-processing, and inspection. Categorizing features into three levels, letting each process work within its capabilities, locking down inspection points only after the full post-processing chain is defined, and reserving space for secondary operations for features that cannot be reliably achieved by the machine – this is the path of least friction to stable mass-produced parts.
0 comments