Plastic is a thermal insulator and an electrical insulator, and both of those facts are usually treated as background. They are not background. They are the two properties that decide whether a printed part becomes a housing, a bracket, a duct, or a part that quietly fails at month eight because the thing it was holding got hot and the plastic under it crept.
The previous articles in this series covered what a printed part has to carry: load, torque, pressure, motion, grit. This one is about the two loads nobody models — heat and current. Neither one shows up as a force in CAD. Both of them are cumulative, both of them are invisible until they are not, and both of them have a habit of failing at the interface between the plastic and the metal part bolted to it.
What follows is not a lesson in thermodynamics. It is the practical version: how much heat a printed part can actually tolerate, how to build a thermal path out of a material that conducts heat about a thousand times worse than aluminium, how to route a wire through a wall without turning the wall into a fuse, and where the regulatory lines sit if the part is going anywhere near mains voltage or a product listing.
One thing up front, because it is the most common misunderstanding: a printed part is almost never a heat sink. It is a heat path, and the path is only as good as its thinnest, most resistant section.
Two loads that do not appear in a static analysis
A bolt load is a number. A motor's heat output is a number too, but it arrives continuously, and the part's response to it depends on time in a way that static strength never captures. That produces a specific family of failures that all look like "the plastic went soft."
| Failure | What is actually happening | Where you see it |
|---|---|---|
| Boss collapse | Screw preload plus temperature pushes the plastic past its heat deflection point; the boss shortens and the screw goes loose | Motor mounts, driver boards screwed to printed standoffs |
| Creep under constant load | Well below the deflection temperature, but held for hundreds of hours, the part keeps moving | Anything preloaded, anything hanging |
| Thermal cycling fatigue | Plastic expands roughly five to ten times more than the metal bolted to it; every cycle strains the joint | Heated beds, outdoor enclosures, anything that switches on and off |
| Localised scorching or deformation | A small hot component with no spreader; the heat has nowhere to go | Resistors, regulator tabs, heater cartridges, halogen lamps |
| Insulation breakdown | The plastic is not thick enough, or a sharp printed edge has cut through a wire's own insulation | Anywhere a conductor passes through a printed wall |
Every one of those is a design decision, not a printing defect. And every one of them is cheaper to fix in the model than in a reprint of a housing that has already been assembled twice.
The rest of this article splits into three parts that share one idea: heat and current both follow the path of least resistance, and you are the one who decides where that path goes.
What a printed part can actually take: the temperature chain
The number on a filament spool is not the number that matters. What matters is a chain of four temperatures, and the part is limited by the weakest link in that chain under real load, not on a datasheet.
- Glass transition — where the polymer stops behaving like a rigid solid. PLA is around 55 to 60 °C, PETG around 75 to 80 °C — the reason a PETG housing is a much better electronics enclosure than PLA, ABS and ASA around 100 to 105 °C. Above this line, stiffness collapses fast.
- Heat deflection temperature under load — the same transition, measured with a load applied. It is always lower than the dry glass transition, sometimes by twenty degrees, because the load does half the work of softening the part.
- Continuous service temperature — a manufacturer's judgement about long-term creep and ageing, usually well below both of the above. This is the number to design to.
- Melting or degradation — the point of no return. Not a design limit, a safety backstop.
The practical consequence is a margin rule. Design so that the hottest steady-state temperature the part will ever see is at least 20 °C below the heat deflection temperature under load, and preferably 30 °C. A PLA bracket with a heat deflection temperature near 55 °C is therefore a good choice up to about 25 to 35 °C, which covers a lot of indoor electronics and almost no motor mounts.
Margins matter more than absolute numbers because of the second-order effects. A part at 60% of its deflection temperature has already lost a meaningful fraction of its stiffness and will creep under a load it held happily when cold. That is the mechanism behind creep, fatigue and preload — the part is not melting, it is relaxing.
Where the heat actually comes from — and how much of it there is
Before designing a thermal path, get an honest estimate of the source. The numbers below are the ones that show up in desktop machines and small appliances, and they span three orders of magnitude, which is exactly why "it gets a bit warm" is not a design input.
| Source | Typical dissipation | What that means for a printed part |
|---|---|---|
| Small signal board, MCU, sensors | Under 1 W | Irrelevant thermally; only creepage and clearance matter |
| Stepper driver or motor driver board | 2 to 10 W | Needs airflow, not conduction; standoffs must be short and vented |
| LED array or high-power LED | 3 to 30 W, concentrated | The single worst case; needs a metal spreader at the source |
| Heater cartridge, hotend, heated bed underside | 30 to 300 W | Keep plastic out of the radiant path entirely, not just out of contact |
| NEMA 17 stepper at moderate load | 5 to 15 W, mostly in the windings | Heat travels through the shaft and the mounting face into your bracket |
| Lithium cell under load | 1 to 5 W per cell | Cell needs air; the printed holder must not insulate it |
Two entries deserve comment because they are the ones people get wrong. A stepper motor is a heat source that arrives through its mounting face — the bracket is not holding a motor, it is holding a motor and a hot plate. And an LED is not a small heat source; a 10 W emitter in a 20 mm package has a power density that will deform any plastic touching it, regardless of the material.
For anything in the 10 W and up range, the honest answer is that the plastic should not be the thermal solution at all. It should be the mechanical structure, and the thermal solution should be metal.
Plastic is a bad conductor, and that is a design constraint not a defect
Thermal conductivity of common printed materials sits around 0.15 to 0.3 W/(m·K). Aluminium is about 200. That is a factor of roughly a thousand, and it has a specific design consequence: a printed heat sink is not a heat sink.
What a printed part can do is bridge a gap, and the bridging calculation is simple enough to do on paper. Thermal resistance of a slab is thickness divided by (conductivity × area). For a 3 mm thick PETG wall, 20 mm square, that is 0.003 / (0.2 × 0.0004) which comes out near 37 K/W. Push 3 W through it and you get a 112 °C rise across the wall.
That single number explains most of the failures in this category. Three watts is nothing. A hundred degrees is a destroyed part. The lesson is not "use thicker plastic" — thicker plastic is worse. The lesson is that the thermal path must be short, wide, and mostly metal.
| Design move | Effect on resistance | When to use it |
|---|---|---|
| Halve the wall thickness | Halves it | Only where structure allows; rarely enough on its own |
| Double the contact area | Halves it | Always worth doing; free in CAD |
| Insert a metal spreader at the source | Removes the bottleneck entirely | Anything above 5 W with a small footprint |
| Add a metal bolt path through the part | Creates a parallel low-resistance route | Motor mounts, board standoffs near heat |
| Add fins | Helps only if there is airflow and the base conducts | Almost never on a printed part alone |
Note the last row. Fins on a printed part are decorative unless three conditions hold at once: the base can spread heat laterally, there is real air movement, and the fin spacing suits that airflow. Printed fins on a poorly conducting base move almost nothing. The honest move is a metal plate with a printed shroud around it.
Air is a component: ducts, shrouds and vents that do not strangle themselves
Once you accept that conduction through plastic is a bad plan, the design problem changes from "how do I conduct heat" to "how do I move air." That is a problem a printed part is genuinely good at, because ducts and shrouds are exactly the kind of hollow, thin-walled geometry that FDM does well.
Three numbers govern almost every printed duct:
- Free area, not opening size. A grille that is 70% material has 30% of the area it appears to have. Pressure drop goes roughly with the square of velocity, so starving a fan by half its area costs far more than half its flow.
- Fan intake needs more room than fan exhaust. A fan pushed against a wall loses a large fraction of its flow to recirculation. Keep at least one fan diameter of clear space upstream.
- Bends cost pressure. Two 45° bends beat one 90° bend, and a sharp internal corner is worse than either. Round every inside corner in a duct, on the inside of the turn specifically.
Material choice for ducts is about temperature, not stiffness. A duct sitting downstream of a 250 °C hotend sees radiant heat even if it never touches anything hot, and PLA will sag simply from being near it. For anything in a printer's own hot zone, ABS, ASA or a glass-filled grade is the sane starting point, and the ASA and ABS grades handle both the temperature and the fatigue of constant minor flexing. Where a duct has to be stiff and dimensionally stable near a heat source, ABS-CF or ASA-CF are the upgrades; the fiber does not conduct appreciably more heat, it just stops the part from creeping out of shape.
And the duct has to be printable without supports inside it, because supports inside a duct are a permanent airflow restriction you cannot remove. Design the cross-section so it is a bridgeable shape, or split it on a plane you can glue and seal — see sealing printed parts for the joint details that keep a split duct from leaking air at the seam.
Mounting anything that gets hot: the boss problem
Nearly all thermal failures in printed assemblies happen at a bolted joint, and they happen for a mechanical reason: a screw joint's whole job is to maintain preload, and preload is the first thing heat takes away.
The plastic in the joint is doing two jobs at once. It is being compressed by the screw, and it is being heated by conduction from whatever the screw holds. Both reduce its stiffness. Under a boss with 15 W of heat behind it, the sequence is reliable: the plastic softens, the boss shortens by a few hundredths of a millimetre, preload drops, the joint loses clamping force, and the hot component starts moving. That movement is the failure — not a crack, just a slow unmooring.
Four moves fix it, and they are cumulative:
- Put metal between the heat and the plastic. A metal washer or a small metal plate under the screw head adds stiffness and spreads the clamp load. It also gives the heat a shorter route to the screw than through the plastic.
- Shorten the standoff and vent it. A long solid printed standoff around a hot screw is an insulated column holding heat. Shorter, with an opening on one side, runs cooler.
- Use the screw as a thermal path on purpose. A steel screw through the part is a conductor of roughly 15 W/(m·K) — fifty times better than the plastic around it. A metal-to-metal stack from the hot component to a metal chassis plate, with the plastic only locating it, is the correct architecture.
- Do not let the boss be the only thing holding a preload. Where possible, arrange the joint so the load path is metal-to-metal and the printed part is in the position of a spacer, not a spring. The joint design principles behind motor mounts and rail brackets apply directly.
For the parts that hold electronics, the geometry of the standoff matters as much as its height. A printed standoff with a hole straight through it, a heat-set insert at the top, and a wall thickness of about twice the insert diameter is close to the best a printed part can do. Insert depth, boss wall thickness and installation temperature are covered in detail in threads in plastic, and the temperature point is the one that bites in a heated assembly: inserts installed into a boss that is already warm sit looser than the same insert in a cold boss.
Heat has to go somewhere: spreading before it goes anywhere
The single most effective move in this whole article is also the cheapest: get the heat out of the small area before it touches plastic. A 10 W source on a 20 mm square is 2.5 kW/m²; spread the same 10 W over a 100 mm square and it is 100 kW/m² — a twenty-five-fold reduction in flux, which is the difference between scorching a part and warming it.
In practice that means a metal interface plate between the hot component and the printed structure, with thermal paste or a gap pad at the interface. The printed part then sees a large, cooler surface rather than a small, hotter one. That plate can be as simple as an aluminium sheet with printed locating features around it — the plastic positions, the metal conducts.
Where a plate is not practical, air is the fallback, and vented housings work if the vents are designed as a circuit rather than as holes. Air enters low, crosses the hot component, and exits high; the vents on the intake side should be larger than the exhaust side so the flow does not stall. A printed housing with vents only at the top is a thermos, not a chimney.
Radiation deserves one sentence, because it is the mechanism people forget with heater cartridges and hot beds. Radiant heat travels through air without heating it. A printed part does not have to touch a 250 °C object to reach its own deflection temperature; it only has to see it. That is why a heat shield is a separate part with a reflective surface, and why "it is not touching anything" is not a thermal argument.
Current is the other invisible load: routing wires through printed walls
The electrical side of a printed part is mostly about one thing: a hole in a printed wall is not a safe place for a wire to pass. The printed material is an insulator, but the edges of a printed hole are a line of hard, sometimes sharp layer ends, and a vibrating or repeatedly flexed wire will abrade its own insulation against them.
The standard fix is a printed grommet or a strain-relief feature rather than a bare hole. Three geometries actually work:
- A radiused horn. The hole's edge is a full round, at least the wire's own diameter in radius, so the wire bends over a curve instead of a corner.
- A separate grommet. A TPU part with a lip that snaps into the hole; TPU 95A is the right hardness for this — hard enough to hold its shape, soft enough to deform around the wire. Softer grades like TPU 85A are better for very small wires but harder to print as a thin lip.
- A slotted clamp with a zip tie. The wire is anchored so that any pull is taken by the clamp, not by the terminal. This is the one that matters most, because most wire failures are pull failures, not abrasion.
Two rules sit underneath all three. First, the wire path inside the housing must not be able to touch anything that moves or heats. That sounds obvious and is routinely violated when a housing is designed without the harness in the model — model the wire bundle as a swept solid, at its actual diameter plus a few millimetres, and check it. Second, a conductor should never be asked to follow a printed feature with a radius smaller than about ten times the wire diameter; below that, the copper work-hardens and the strand eventually breaks inside intact insulation, which is the hardest fault to find.
Creepage, clearance and the mains line
If the part carries mains voltage or sits near it, several of the following numbers become mandatory rather than advisory, and which standard applies depends on the product and the market. The design rules, however, are consistent across them.
| Concept | What it measures | Why 3D printing makes it harder |
|---|---|---|
| Clearance | Shortest distance through air between conductors | Printed surfaces are rougher and layer lines can host condensation |
| Creepage | Shortest distance along the surface between conductors | A printed surface is porous at the layer interfaces; tracking happens along it |
| Tracking | Carbonised path forming along a surface after a fault | Contaminated or damp printed surfaces track at lower voltage |
| Flammability rating | How the material behaves when ignited | Most filaments carry no rating at all; a few do |
The practical rules that follow, in order of how often they are violated: put a physical barrier rather than a slot between mains and low voltage, so that creepage is limited by the barrier's height and not by a milled-out channel; keep the mains section at the edge of the board and the low-voltage section at the other, with no printed feature bridging them; never rely on a printed rib as the only separation, because a rib printed vertically is a series of layer boundaries exactly along the tracking path.
Material choice matters here in a way it does not elsewhere. Standard PLA and PETG have no flammability rating and should not be the enclosure for mains. Where a part really is near mains or near a heat source with a fault current possible, the honest options are a rated material or a metal enclosure with printed internal structure. Our own range includes a flame-retardant ABS for exactly this use: ABS-FR is the sensible baseline for printed housings that need to hold current and heat at the same time, and it keeps the mechanical behaviour of ABS while adding a rating. On the resin side the equivalent is a flame-retardant photopolymer, which is covered alongside the other engineering resins in the material guides.
Whatever the material, the surface matters too. A printed enclosure that will hold mains should have its internal surfaces sealed or smoothed where creepage is tight, because layer porosity and fingerprints are both tracking accelerants. Chemical resistance and stress cracking is the companion reading for anything that will meet solvents, cleaners or conformal coatings.
The material table, ranked by the job
No single filament wins this category. The right choice depends on which of the two loads dominates, and the table below is ordered by that question rather than by strength.
| Material | Deflection temp (approx.) | Where it fits | Where it does not |
|---|---|---|---|
| PLA+ | 55 to 60 °C | Indoor fixtures, cold electronics, anything below 40 °C | Motor mounts, near heat, anything outdoors |
| PETG | 75 to 80 °C | Housings, ducts away from the hot zone, chemical exposure | Mains enclosures, sustained loads when warm |
| ABS | 98 to 105 °C | Warm housings, ducts near hotends, general heat resistance | High sustained load, precision fits |
| ASA | 98 to 105 °C | Same as ABS plus outdoor UV and thermal cycling | Same as ABS |
| ABS-FR | Around 95 °C | Electronics enclosures needing a flammability rating | Structural load paths |
| ABS-CF / ASA-CF | 100 to 110 °C | Stiff warm brackets, board frames, ducts that must not warp | Anything needing impact toughness |
| PA6-CF | Around 200 °C | Hot, loaded, structurally critical parts | Wet environments, flexing joints |
| PPS-CF | Above 250 °C | Near hotends and heaters; the only print-friendly option in the truly hot zone | Cost, brittleness, print difficulty |
| PET-CF / PET-GF | Around 150 °C | The middle ground: hot, stiff, cheap enough for housings | Impact, repeated flexing |
| TPU 95A | 60 to 80 °C | Grommets, strain relief, seals, vibration isolation | Anything structural or hot |
Two notes on that table. PPS-CF is the material that actually solves the hot-zone problem — a duct or shroud within a few centimetres of a heater cartridge is PPS-CF territory, and PPS-CF10 prints far more politely than its numbers suggest. But it is brittle, and it is the wrong choice for anything that takes a knock. And on the other end, moisture matters as much as temperature for the engineering grades: a wet spool of PA6-CF prints weak parts that look fine, which is the failure mode set out in moisture by material. Dry the spool before blaming the design.
For the parts that only need to sit still and be stiff, the mid-range grades are better value than the top end. PA12-CF10 and PET-CF17 are the two that most often land in the sweet spot for brackets and housings, and PA612-CF15 is the choice when dimensional stability over time matters more than peak strength. If the question is whether to keep ABS at all, the case for ASA covers when the swap is worth making.
Ten checks before the part goes near anything hot
A short list, in the order that catches the expensive mistakes first.
- Write down the steady-state temperature and the peak. Not a guess — measure it, or take it from the component's own dissipation and the area it has to shed it from.
- Confirm the part is at least 20 °C below the material's heat deflection temperature at thickness, under load.
- Identify every bolted joint within 30 mm of the heat source. Each one needs a metal interface or a shorter, vented standoff.
- Check whether any single component exceeds 5 W in under 400 mm² of footprint. If so, it gets a metal spreader.
- Verify the thermal path is metal end to end, and that the plastic is only locating, not conducting.
- Model the wiring as a solid and route it with radii of at least ten wire diameters.
- Put a real relief feature — grommet, radiused horn, or clamp — at every wall penetration. No bare holes.
- If mains is anywhere in the assembly, put a physical barrier between it and low voltage, and pick a rated material.
- Confirm the vents form an inlet-low, exhaust-high circuit, and that no printed support remains inside a duct.
- Run the part at its worst case for an hour and then re-check the torque on every screw. Anything that moved has told you where the margin is not.
That final check is the one worth making a habit. Thermal problems announce themselves as a loss of preload long before they show up as a crack, and a torque wrench after one hour of running costs a few minutes and saves an entire redesign.
Structure, then spread, then move air
The whole subject compresses into an order of operations. Get the structure right first, because a part that does not hold the load is not a thermal problem yet. Then spread the heat at the source, because flux is what destroys plastic and area is the cheapest lever you have. Then move the air, because a printed duct is the one thermal component FDM is genuinely excellent at.
What makes the two invisible loads difficult is that they reward exactly the opposite instinct from strength design. Strength design is about putting material where the load is. Heat design is about removing material from the path and putting metal in its place. Current design is about making every path longer than the straight line and every edge rounder than the corner.
None of that is exotic. It is a set of arithmetic checks and geometric habits that take ten minutes at the start of a design and prevent the class of failures that only appear after a part has been in service for a season — the boss that went loose, the wire that rubbed through, the duct that sagged toward the heater. Those parts all looked right when they came off the plate, which is precisely why the checks belong in the model.