Most makers discover annealing the same way. A bracket printed in PLA snaps along a layer line, someone on a forum says bake it, and the part comes out noticeably stronger. So the habit forms: anneal everything, it can only help.
It can also destroy a part. Ask anyone who has pulled a dimensionally critical housing out of the oven to find it bowed 2 mm and no longer able to seat its lid. Annealing is not a strength upgrade you apply to a finished print. It is a second thermal process, and like the first one it has a temperature window, a stress direction, and a dimensional cost.
This guide covers what actually happens inside the polymer when you anneal, which materials respond and which barely do, how to plan for the shrinkage instead of discovering it, and the process details that decide whether the part survives the oven at all.
What Annealing Actually Does to an FDM Part
An FDM part is not one solid object. It is a stack of extruded roads, each laid down hot and cooled within seconds while the layer beneath it is already rigid. That rapid, uneven cooling locks two separate problems into the geometry.
The first is frozen-in internal stress. As each road cools it wants to contract, but it is bonded to neighbours that are already cold and stiff, so it cannot. The contraction becomes locked tension inside the part. This is the same mechanism that lifts corners off the bed during a print. What you see at the edges is only the visible portion. The rest stays inside the finished part, waiting.
The second is incomplete interlayer fusion. When a new road lands, the polymer chains at the interface have a fraction of a second to diffuse across the boundary before the material drops below its glass transition temperature (Tg) and freezes in place. They never get far. The result is the well-documented weakness of FDM parts: strong in the XY plane, weak in Z.
Holding the part above its Tg addresses both. Above Tg the polymer chains regain enough mobility to slide, so locked-in stress relaxes. The same mobility lets chains diffuse across layer boundaries and entangle with the layer below, which strengthens the Z axis specifically, the direction that normally fails first.
In semi-crystalline materials there is a third and larger effect: additional crystallisation. Printing cools these polymers so fast that they solidify only partly crystalline. An anneal gives crystals time to nucleate and grow, and crystallinity is what drives stiffness, heat deflection temperature and chemical resistance.
Amorphous vs Semi-Crystalline: The Split That Predicts Everything
This single distinction explains why annealing transforms a nylon part and does almost nothing measurable for a PETG part.
| Polymer class | Typical FDM materials | What annealing buys | Dimensional risk |
|---|---|---|---|
| Semi-crystalline | PLA, PA / Nylon, PPS, PEEK | Large gains: crystallinity increases, so stiffness, heat resistance and creep resistance all rise | High — crystals pack denser than amorphous polymer, so the part contracts |
| Amorphous | PETG, ABS, ASA, PC, TPU | Modest: stress relief, fewer micro-cracks, steadier Z strength | Moderate to low — relaxation rather than crystallisation |
Read the table as a cost-benefit statement. Semi-crystalline materials pay you back the most strength and charge you the most shrinkage. Amorphous materials are cheap on dimensions and pay back relatively little. There is no free strength here, and the materials where annealing earns its keep on strength are exactly the materials where you must compensate for the size change.
The practical consequence for the common shop materials: ABS, ASA and PC are amorphous, so treat annealing as a stress-relief step rather than a strength treatment, and reach for it when you are fighting micro-cracking or delamination rather than chasing raw stiffness. PLA and nylon are where the strength argument is real.
Material-by-Material Annealing Reference
These are starting ranges, not recipes. Oven calibration, part mass and wall thickness all shift the real optimum, which is why a test coupon comes before a batch.
| Material | Approx. Tg | Typical anneal band | Expected outcome |
|---|---|---|---|
| PLA | ~55–60 °C | ~80–110 °C | Large gain in stiffness and heat resistance via crystallisation. High warp risk without support |
| PETG | ~80 °C | ~65–75 °C | Stress relief only, little crystallisation. Modest benefit — often not worth the oven time |
| ABS | ~105 °C | ~90–100 °C | Stress relief, fewer micro-cracks, steadier Z strength. Amorphous, so mainly relaxation |
| ASA | ~100–105 °C | ~90–100 °C | Stress relief while keeping UV and weather stability. Similar to ABS |
| PC | ~147 °C | ~85–100 °C (stress relief); ~130–135 °C for full relaxation | Removes residual shear stress that causes delayed creep and post-machining cracking |
| Nylon PA12 | ~40–50 °C | ~120–150 °C | Largest strength and heat-deflection gain of any FDM polymer. Must be dried first. Significant XY contraction |
| TPU | varies | generally not annealed | No meaningful benefit; the material is already above its useful Tg at room temperature |
Two entries deserve a closer look because they are the ones most often done wrong.
PLA: The High-Reward, High-Risk Case
PLA is the material most people anneal first and the one most likely to come out deformed. Because it is semi-crystalline and its Tg is close to household temperatures, the crystallisation payoff is large — 20–40% gains in stiffness and a substantially raised heat deflection temperature are typical. The same crystallisation is what pulls the part out of shape, especially on large, thin or flat geometry.
The workaround is to anneal inside a mould or with packed support material that constrains the part while it shrinks. Salt, plaster or fine sand packed around the print is a common technique precisely because it holds the geometry without needing a rigid cavity.
PC: Why Stress Relief Matters More Than Strength
Polycarbonate sits in a category of its own because of how high its Tg is and how much residual stress survives printing. Even a chambered print carries internal shear stress of roughly 15 MPa after cooling. That stress does not announce itself. It shows up weeks later as slow creep, or the moment you machine, drill or tap the part and a crack initiates right where the stress was concentrated.
A programmed anneal — a prolonged soak around 130–135 °C followed by a slow, controlled cooldown of a few degrees per hour — relaxes that residual stress to roughly 2 MPa and measurably raises impact strength. The result is not a stronger-looking part but a stable one: dimensionally repeatable across months, and safe to machine without stress-release cracking. For anyone holding a dimension across a season, that stability is the whole point.
Which brings up the parts that make PC printing possible in the first place. PC demands a nozzle in the 260–300 °C range, a bed at 100–120 °C and a chamber that stays genuinely hot. A standard all-metal hotend is the baseline requirement — a tungsten carbide nozzle delivers the wear and thermal stability that high-temperature PC blends demand. At those chamber temperatures, bed adhesion stops being about glue and starts being about retaining heat, which is why a Cold PEI build plate that releases cleanly once cool works better than fighting a stuck part hot. And because PC is aggressively hygroscopic, filament that has absorbed moisture will boil at the nozzle and leave porous, weak walls that no anneal can repair: dry it, and if the shop runs several materials, a filament drying cabinet keeps every spool ready rather than drying one roll at a time.
Planning for Shrinkage Instead of Discovering It
The single most common cause of a failed annealing attempt is not temperature. It is doing the anneal on a part whose tolerances never had room for it.
Before anything goes in the oven, run this check:
- Decide whether the part needs to be dimensionally accurate or strong. If both, you need a compensation step, not a shortcut.
- Print a test coupon in the same material and wall thickness, anneal it on the intended schedule, and measure it. Do not extrapolate from a previous different part.
- Scale the model in CAD by the measured factor before printing the real part. Compensation belongs in the model, not in post-processing.
- Orient critical faces. Flat, mating and sealing surfaces should face down on the bed where they are most accurate, and should be supported during the anneal so they stay flat.
- Constrain thin geometry. Packed support material, a mould, or a fixture stops thin sections from curling during the soak.
Because annealing tends to be roughly uniform in the print plane and different through the thickness, isotropic scaling alone is an approximation. For tight-tolerance assemblies, measuring at least two axes on the coupon is worth the extra ten minutes.
Time, Ramp and Cooldown: The Part Everyone Skips
Temperature gets all the attention, but the schedule around it decides whether the part survives. Three rules cover almost every failure.
Ramp Slowly
Dropping a cold part into a hot oven recreates exactly the thermal gradient you are trying to remove. Bring the part up with the oven from cold wherever possible. A fast ramp anneals the skin and leaves a stressed core, which is worse than not annealing at all because it looks successful.
Soak by Mass, Not by Clock
A thin bracket equalises in well under an hour. A thick, dense block needs considerably longer for the core to reach temperature. Soaking for a fixed time regardless of part mass is how people end up with parts that are annealed on the outside only.
Cool Slower Than You Heated
Pulling a part out hot and letting it cool unevenly on a bench is the most common cause of a warped annealed part. Cool it inside the oven, supported, and let the whole cavity come down together. Where the application is stress-critical, a controlled cooldown on the order of a few degrees per hour is what converts a good anneal into a stable part.
A forced-air convection oven with a large cavity and even temperature distribution is the right tool. The print must not sit close to the heating element. Domestic kitchen ovens cycle too widely on temperature and, for anything you intend to eat off again, are a poor idea on principle.
When Not to Anneal
Annealing is the wrong move more often than most guides admit.
- Precision fits and assemblies. Anything that must mate with a machining tolerance, thread into a captured nut, or seat a bearing is a poor candidate unless you have characterised and compensated the shrinkage.
- Large, thin, flat parts. These warp most readily and gain least from the process. It is generally better to accept as-printed strength or to change geometry.
- Parts where toughness matters more than stiffness. Crystallisation can raise stiffness while reducing impact resistance. A part that ends up stiffer and more brittle is not always an improvement.
- PETG, in most cases. The amorphous structure means little crystallisation to gain. You spend oven time and dimensional risk for modest stress relief.
- Any part that was not dried before printing. Annealing a moisture-loaded part bakes the defects in rather than removing them.
Where a part genuinely needs more heat resistance than an as-printed FDM part can deliver and the geometry will not survive an anneal, the honest answer is usually to change material rather than process. Moving up to an engineering polymer with an intrinsically higher Tg — a stabilised ASA filament for UV-exposed outdoor work, or a fibre-reinforced grade such as PETG-CF or ABS-CF where stiffness and creep resistance are the driver — solves the problem at the design stage instead of in the oven.
The Takeaway
Annealing is a stress-relief and, for semi-crystalline polymers, a crystallinity-increase process. It strengthens the Z axis of parts that fail between layers, and it raises heat resistance in PLA and nylon meaningfully. It does not make an amorphous part dramatically stronger, and it always charges a dimensional price that must be paid in CAD rather than discovered after the fact.
The workflow that works: know the Tg, decide whether the part needs strength or accuracy, print a coupon of the same material and mass, measure the shrink, compensate in the model, pack or fixture thin geometry, ramp slowly, soak by mass, and cool inside the oven. Done that way, annealing is a repeatable engineering step. Done as an afterthought on a part with no tolerance budget, it is just a way to ruin a good print faster.
If you are working in PC or a high-temperature blend and want the hardware side sorted before you start, our ABS filament and engineering-grade range, the ASA filament for outdoor stability, and accessories like the automatic filament feeder for keeping multiple materials fed cleanly will get a shop set up for materials that actually require a real thermal plan.