Rapid Prototyping surface texture requirements: achieving target roughness on finished parts

Content Menu

● Introduction

● How Rough Are As-Built Surfaces, Really?

● Process-Specific Causes of Roughness

● In-Process Controls That Actually Move the Needle

● Post-Processing Routes That Scale

● Real Production Examples

● Remaining Limitations and Trade-offs

● Conclusion

● Q&A

 

Introduction

Surface finish on additively manufactured parts has become one of the biggest hurdles when moving from prototype to production. Ten years ago most shops accepted whatever came out of the machine because the part was only for fit checks or marketing photos. Now the same processes are expected to deliver sealing surfaces that hold 10 bar, bearing seats that survive millions of cycles, or internal flow paths with controlled turbulence. The drawings no longer say “as-built acceptable.” They call out Ra 3.2 µm, sometimes Ra 1.6 µm, occasionally Ra 0.8 µm or better, and inspectors enforce it.

The problem is that every additive process leaves its own fingerprint on the surface. Layer lines, partially melted powder, support marks, stair-stepping, balling, dross – all of these push the as-built roughness well beyond what most functional requirements allow. The gap between what the machine naturally gives you and what the drawing demands can be 10–30 µm Ra, and closing that gap in a repeatable, cost-effective way is what separates shops that are still making prototypes from shops shipping thousands of certified parts a month.

This article pulls together practical lessons learned across FDM, SLA, SLS, binder jetting, and metal powder-bed processes. The focus is on what actually works on the shop floor rather than laboratory curiosities.

How Rough Are As-Built Surfaces, Really?

Numbers vary by material and machine generation, but here are typical ranges engineers see in 2024–2025 production environments:

• FDM/FFF (PLA, ABS, PC, ULTEM): vertical walls 6–18 µm Ra, top surfaces 2–8 µm Ra, angled surfaces up to 30 µm Ra because of stair-stepping.
• SLS Nylon 12 (PA 2200/PA 12): 10–20 µm Ra on all orientations, classic orange-peel look.
• SLA/DLP resins (standard and engineering grades): 1–5 µm Ra on upward faces, 8–15 µm Ra where supports touched.
• Metal LPBF (316L, Ti64, IN718, AlSi10Mg): up-skin 4–9 µm Ra, down-skin 12–35 µm Ra, sidewalls 8–15 µm Ra.
• Binder-jetted 316L or 420 stainless before infiltration: 15–25 µm Ra and visibly porous.

These are not worst-case numbers; they are what qualified builds deliver when the machine is running within spec.

Process-Specific Causes of Roughness

Material Extrusion (FDM/FFF)

The two dominant mechanisms are layer stacking geometry and bead-to-bead consistency. On inclined surfaces the staircase height is approximately layer height × (1 – cos θ), where θ is the angle from horizontal. At 45° a 0.25 mm layer already creates 18 µm peak-to-valley steps before you even consider over- or under-extrusion.

Vertical walls show grooves from the nozzle path and small ridges where adjacent beads overlap. Nozzle temperature, speed, cooling rate, and extrusion multiplier all move the roughness a few microns.

Laser Powder Bed Fusion (Metal)

Melt-pool instability dominates. On down-facing surfaces the powder bed conducts heat poorly, so the melt pool becomes deeper and wider than intended. Surface tension pulls liquid metal downward, forming dross and attached spheres. On sidewalls the scan lines themselves leave ridges, typically 5–10 µm high for 40–60 µm layer thickness.

Contour passes help, but if the contour energy is too low you get lack-of-fusion; too high and you get keyholing and spatter.

Polymer Powder Bed (SLS)

The surface is essentially a collection of particle necks. Roughness scales with powder particle size and degree of sintering. Newer fine powders (D50 ≈ 30 µm instead of 60 µm) and better chamber temperature uniformity have brought typical values down from 18 µm Ra to around 9–12 µm Ra on recent machines.

In-Process Controls That Actually Move the Needle

Build Orientation

The single cheapest improvement. In metal LPBF, making a critical surface up-skin instead of down-skin can cut required post-processing time by 80 %. In FDM, rotating a part so a sealing face becomes a top layer instead of a 30° incline can drop roughness from 25 µm to 3 µm without changing a single parameter.

Modern build-prep software (Materialise Magics, Autodesk Netfabb, nTopology) now predicts Sa/Ra for every triangle and suggests orientations that minimize roughness on flagged faces.

Layer Thickness and Exposure Settings

Dropping layer thickness from 50 µm to 30 µm in LPBF typically reduces down-skin Ra by 4–8 µm, but build time goes up roughly 40–60 %. In FDM, moving from 0.3 mm to 0.12 mm layers often halves staircase height on 45° surfaces.

Contour and Skin Strategies

Most metal machines now allow separate energy for contours and hatch. A single contour pass at slightly higher power followed by a second “skin” pass at lower power can smooth sidewalls from 12 µm Ra to 6–8 µm Ra with almost no time penalty.

In FDM, the ironing function (PrusaSlicer, Bambu Studio, Simplify3D) makes the nozzle sweep top layers again with 5–15 % flow. Top surfaces routinely reach 1–3 µm Ra on engineering polymers.

Post-Processing Routes That Scale

Mass Finishing (Vibratory, Tumbling, Drag)

Still the workhorse for external surfaces. Ceramic or plastic media in a vibratory bowl takes nylon SLS parts from 15 µm to 3–5 µm Ra in 6–12 hours. High-energy centrifugal systems do the same on metal in 1–3 hours.

Abrasive Flow Machining (AFM) / Extrude Hone

The only reliable way to finish internal channels smaller than 5 mm diameter. Typical results on LPBF cooling passages: from 20–30 µm Ra down to 2–4 µm Ra in one or two passes.

Chemical and Vapor Smoothing

Acetone vapor smoothing on ABS (now done in sealed commercial chambers) drops Ra below 1 µm in 20–60 minutes. Newer polyjet and MJF parts respond well to AMT’s Vapour Smoothing 2.0 process.

Electropolishing and Chemical Accelerated Polishing

Standard for Ti64 medical and aerospace parts. Removes 20–50 µm uniformly and leaves mirror finishes (Ra 0.2–0.5 µm) even on complex lattice structures.

Laser and Abrasive Jet Polishing

Laser polishing is finally moving out of labs. Fraunhofer ILT and similar systems now achieve Ra < 0.8 µm on IN718 turbine components in minutes per part.

Real Production Examples

GE Additive’s Atlas project (large-format LPBF): critical flange faces are oriented as up-skin, receive double contour + skin scans, then light bead blast → Ra 4–6 µm, no CNC required.

Boeing 787 ducting in Nylon 12: airflow surfaces oriented upward, parts receive 8-hour vibratory cycle with ceramic media followed by dye penetrant inspection → consistent Ra 3–4 µm.

Orthopedic knee implants (Ti64 LPBF): HIP → CNC on bearing surfaces → electropolishing → final Ra 0.25 µm on articular faces, 1–2 µm on textured bone-ingrowth zones.

Formula Student teams printing carbon-fiber PEEK end-use suspension parts: 0.15 mm layers + ironing + light vapor smoothing → sidewall Ra 2–3 µm, strong enough for track use.

Remaining Limitations and Trade-offs

Thinner layers and extra contours cost build time. Aggressive mass finishing can round small radii or erode thin walls below tolerance. Chemical processes sometimes attack near-surface microstructure (especially in high-strength aluminum alloys). Internal features below ≈ 2 mm remain almost impossible to polish mechanically.

Conclusion

Getting additive parts to the required surface roughness is no longer a black art. The recipe is straightforward: start with aggressive orientation optimization and proven parameter sets that push as-built quality as far as possible, then apply the minimum post-processing needed to cross the finish line. Shops that document their qualified combinations of orientation, parameters, and post-process steps for each material and feature type are the ones consistently hitting Ra callouts on production drawings without heroic effort.

The machines keep improving – 20 µm layers are becoming common in metal, fine powders in polymer, variable spot size optics – so the gap continues to shrink. But for the foreseeable future, achieving target roughness remains an engineering discipline, not something the printer does by itself. Master the controls and post-process chain for your specific process and material, and additive ceases to be “rapid prototyping” and becomes genuine manufacturing.

Q&A

1. Which costs less in production: printing with 30 µm layers or 60 µm layers plus extra post-processing?
   Usually 60 µm layers plus a short mass-finishing cycle. Build time dominates cost at scale.
2. Can you eliminate all support marks on SLA parts without manual sanding?
   Almost – use minimal supports, orient marks on non-critical faces, then light bead blast or resin-based chemical smoothing.
3. Is there a reliable way to predict final roughness after vibratory finishing?
   Yes. Most media suppliers provide removal-rate curves; combine with pre-measurement and you can hit ±1 µm Ra repeatably.
4. Why do some LPBF sidewalls look shiny and others matte with identical parameters?
   Minor bed temperature drift or oxygen pickup changes oxidation and melt flow. Modern machines with better chamber control have largely solved this.
5. When is it cheaper to hybrid manufacture (print + CNC) instead of pure additive + post-process?
   Whenever a few critical faces need Ra < 2 µm and are accessible to a tool – molds, turbine blade roots, bearing housings.