Every design habit on this blog until now has treated the printed part as something that has to resist. It holds a load, transmits a torque, keeps a bolt from loosening, keeps chemical from eating it. This article is about the opposite job: a part whose entire purpose is to move air from one place to another and to do it a predictable number of times. A vacuum cup that has to grab a lens. A bellows that has to push a shutter. A duct that has to deliver cooling to a hotend. A small valve that has to open on a signal.
None of these are exotic projects. All of them are being printed right now, mostly badly, and mostly by people who discovered after the third revision that a printed pneumatic part fails in ways that have nothing to do with the strength lessons we have spent the last month refining. Air does not care about your wall count. It cares about whether a molecule can walk through the polymer, whether a lip can collapse, and whether a 0.3 mm gap between a layer line and a gasket is a leak or a seal.
Air Is a Different Kind of Payload
Start with the thing that makes printed pneumatics unlike printed anything else. A structural part is judged on one number at one moment: does it break, or not. A pneumatic part is judged on a continuous variable every single second: how much flow is crossing the boundary that is not supposed to have flow.
That shifts the whole design problem. A bracket can be 20 percent over-built and nobody notices. A vacuum cup that leaks 20 percent of its pumping capacity may simply never seal, because the pump's free flow has to exceed the leak rate before the cup can pull down at all. Leakage is not a margin you shave. It is a threshold you either cross or do not.
There is also a second, quieter difference. Air is compressible, so the part does not just pass flow, it stores it. Every pocket, every tube, every internal cavity is a spring. That spring is what makes printed bellows and printed diaphragm actuators possible at all, and it is also what makes them slow, mushy and difficult to tune. Volume is a design parameter, not leftover space.
Finally, air is everywhere, and it is buoyant. Unlike a fluid in a closed loop, there is nothing holding air inside a part except geometry and surface quality. A part that is sealed by a lip has no backup. There is no viscosity to slow an escape, no surface tension to bridge a gap. A 50 micron channel that would barely weep water will breathe air freely.
Five Failure Modes, Five Different Fixes
The reason printed pneumatics get abandoned is that all five failure modes look identical from the outside: the part does not hold. Someone then reaches for the strongest filament in the drawer, which fixes none of them. Separating them at the start saves a week.
| Failure | What you observe | Real cause | Correct fix |
|---|---|---|---|
| Permeation | Holds, then drifts up over minutes | Gas dissolving and diffusing through the polymer itself | Wall thickness, barrier material, less surface area |
| Interlayer leak | Leak rate varies with print orientation | Bond planes between layers are diffusion paths | Orientation, temperature, flow — never more walls alone |
| Fit leak | Leaks only at joints and inserts | Uneven flange faces squeezing a gasket unevenly | Gasket groove geometry, flange stiffness, bolt spacing |
| Crush | Works once, then seals worse | Lip or bellows compressed past its elastic range | Limit the stroke, not the force |
| Dead volume | Response is slow and weak but it does hold | Internal cavity acting as a spring that eats the stroke | Reduce internal volume, shorten the path |
Notice that only the first row is really a material problem. The middle three are geometry and process, and the last is arithmetic. This is the same pattern as every other functional-printing topic: the intuitive fix (stiffer material) addresses the smallest slice of the problem.
Permeation: The Leak You Cannot See or Locate
Thermoplastics are not barriers. They are, at best, slow filters. Gas molecules dissolve into the polymer on the high-pressure side, diffuse across, and desorb on the low side. The rate depends on the material, the pressure difference, the wall thickness, and — critically — the total area exposed.
For practical purposes FDM materials sort into a rough order of increasing gas transmission: PPS-CF10 and other semi-crystalline high-performance materials sit at the good end because their crystalline regions are largely impermeable; PA6-CF20 and PA12-CF10 are better than unfilled nylon but still mediocre; PETG and PLA are poor to fair depending on crystallinity; and TPU 95A is the worst of the common materials by a wide margin because its amorphous, low-density structure is exactly what a gas molecule wants to move through.
This is a different problem from the one behind capillary leaks, where liquid finds a path along a surface. Gas does not need a path: it goes through the material itself.
Two consequences follow. First, thin walls leak and thick walls do not, and the relationship is linear in thickness — doubling a 1.5 mm wall to 3 mm halves the permeation rate. That is a much cheaper fix than switching materials. Second, surface area matters as much as thickness, so a large flat panel is the worst possible shape for a vacuum reservoir and a long thin tube is far better than its volume suggests.
For a vacuum cup held for seconds, permeation is almost never the limiting leak. For a part that must hold pressure overnight, it is the only thing that matters, and no amount of print tuning will fix it.
Layer Direction Decides Which Kind of Leak You Get
The single most useful thing to know about printed pneumatics is that a wall is not a wall. It is a stack of flattened beads with deliberate, microscopic gaps and incomplete bonding between them. Those bond planes are the fastest path for gas through the part, and their orientation relative to the pressure gradient decides your leak rate more than any material choice.
If the pressure pushes along the layers — that is, if the part is oriented so the bead lines run parallel to the escaping path — the gas has to cross every bond plane in series, and the leak is small. If pressure pushes across the stack, the gas needs to cross only one bond plane, and it will.
In practice this means one rule that is easy to remember and easy to violate: orient the part so that the pressure direction cuts across layer lines rather than running parallel to them. For a cylindrical vacuum cup, that means printing it upright, with the axis vertical, so the pressure difference across the wall is radial and has to cross many bond planes. Print the same cup on its side and the leak moves to the seam where two layers meet, and no amount of wall thickness recovers it.
The second-order variables are temperature and flow. Hotter extrusion and slightly higher flow both improve bond quality and reduce interlayer leak, which is the opposite of what many people expect — the "cool and fast" instinct that gives crisp corners also gives more permeable walls. For pneumatic parts, deliberately print 5 °C to 10 °C hotter than your cosmetic settings and check for drooping on overhangs rather than chasing surface perfection.
Vacuum Cups: Geometry That Seals Before It Holds
A vacuum cup is not a suction device. It is a seal that happens to be connected to a pump. The pump only has to remove the air inside the enclosed volume once; after that, the cup holds because the external atmosphere presses the workpiece onto the lip. Everything about cup design is therefore about getting a closed loop first and keeping it closed while the load tries to peel it off.
Three numbers decide whether a printed cup works.
- Lip thickness. A printed lip has to be thin enough to conform and thick enough not to tear. Below roughly 0.8 mm the lip folds instead of sealing; above about 2 mm it stops conforming to surface texture and only seals on optically flat parts. A 1.2 mm lip with a 25 degree chamfer at the sealing edge handles most painted metal and printed plastic.
- Lip contact width. Wide lips feel secure and seal worse. A contact band of 0.6 mm to 1.0 mm concentrates the atmospheric load into a line and produces a sealing pressure far above the vacuum level. Wider bands simply divide the same force over more area.
- Internal cavity depth. Shallow cavities reduce dead volume but give no room for the workpiece to be pulled into the cup. On a flat part, 0.5 mm is plenty. On a curved part, the cavity has to be deep enough that the lip can move without bottoming out, or the workpiece lands on the floor of the cup and unseats the lip from the other side.
Material choice follows from the lip, not from the load. A rigid lip on a rigid part gives a line contact with no compliance; a compliant lip forgives surface texture and slight misalignment. This is the one case in functional printing where the flexible material is the right answer: TPU 85A for textured or slightly curved surfaces, TPU 95A for flat and clean ones where you want the lip to resist rolling under shear.
The flange that holds the cup is where the tolerance stack stops being academic. The cup, the flange, the gasket and the manifold all have a tolerance, and a leak appears when they add up in the wrong direction at one point on the ring.
Two attachments matter more than the cup itself. The first is a rigid backing plate behind the cup: a soft cup glued directly to a printed manifold will peel from the edge inward under load, and the peel starts at the point where the cup wall meets the plate. Give the cup a flange and clamp it, or print the cup with a rigid ring bonded around it. The second is a coarse filter or a deliberate labyrinth at the port, because a printed cup will ingest dust and swarf and pass it into the pump.
Bellows: Stroke Comes From Geometry, Force Comes From Area
A printed bellows is a stack of flexible walls and rigid plates, and it is almost always designed wrong in the same way: people make it too short and too thick. Those two choices point in the same direction and both are bad, because a bellows' usable stroke is set by wall length and wall count, while its force is set by effective area — and adding wall thickness improves neither.
Three rules make a printed bellows work.
Never let the wall do the moving. The wall is a hinge. It should be thin, and it should bend rather than stretch. A wall of 0.4 mm to 0.6 mm with a smoothly varying profile (a shallow S, not a sharp V) will bend elastically for tens of thousands of cycles in TPU 95A. A 1.2 mm wall of the same material will fight you with high stiffness and then crack at the inner corner where the strain concentrates.
Add stroke with plates, not with wall length. Each convolution of the bellows buys you roughly the same stroke, and the pressure rating stays nearly constant no matter how many convolutions you stack. Adding convolutions is nearly free. Lengthening a single wall is not: it increases strain per unit of stroke and produces buckling.
Bound the stroke mechanically. This is the single most common omission. A bellows without an internal stop will be over-travelled by its own actuator, and once a thin TPU wall has been stretched past about 20 percent it does not return to the same length. The fix is a rigid internal post or an external collar that bottoms out before the wall reaches its elastic limit. It costs one feature and doubles the life of the part.
One more thing: printed bellows are pressure-limited by the plate, not the wall. A convolution plate that is 1.5 mm thick will dome under 0.2 bar and the bellows will simply get shorter instead of pushing. If the bellows has to do work, the plates need ribs or a domed profile — this is the same argument as the ribs and gussets work, applied to a plate that carries pressure rather than a bracket that carries a load.
Where Print Orientation Meets Pneumatics
For most printed parts orientation is a strength decision. For pneumatic parts it is also a leak decision, and the two do not always agree.
| Part | Best orientation for strength | Best orientation for sealing | What to pick |
|---|---|---|---|
| Cylindrical vacuum cup | Upright, axis vertical | Upright, axis vertical | Both agree — print upright |
| Flat diaphragm | Lay flat, flexing in-plane | Lay flat, pressure across layers | Both agree, but add a concentric rib pattern |
| Bellows | Upright, compression along layers | Upright, pressure radial | Upright, with thin plates ribbed instead of thickened |
| Valve body with a seat | Seat face parallel to build plate | Seat face parallel to build plate | Both agree — the seat becomes a single top surface |
| Long duct or manifold | Lay flat, pressure along layers | Upright, pressure across layers | Split and orient the halves, then bond along a non-critical seam |
The last row is the interesting one, because it is the only case where the two objectives genuinely conflict. A duct printed flat as a single closed volume has its layer planes running the length of the duct, which is exactly the direction you do not want. Printing it upright with a hundred millimetres of unsupported overhang is not an option either. The compromise that works is to split the duct along its mid-plane, print both halves with the flat faces down, and join them — with the split line placed where a leak is least harmful, ideally in a low-pressure region or outside the sealing boundary entirely.
That reasoning is the same one behind detachable joints and tolerance design, but the stakes are different: a structural joint that is a little loose still carries load, while a pneumatic joint that is a little loose simply does not seal.
Valves and Seats: Small Features, Large Consequences
Printed valves are where most people discover that "good enough" geometry is not. A valve seat is a line, not a surface, and its whole job is to be the one place where the two halves of the part touch with enough pressure to stop flow.
Three practical rules.
Print the seat as a top surface, not a side wall. Ironing gives you a flat, slightly densified face that seals against a rubber disc with very little force. A vertical seat face carries layer lines as ridges, and those ridges are leak paths that no amount of closing force will bridge.
Seal on a soft disc, not on plastic. A printed seat against a printed poppet gives plastic-on-plastic contact, which has the same creep and the same dimensional scatter as any other fit. A 1 mm disc punched from TPU 95A or a slice of silicone tubing conforms, tolerates misalignment, and is trivially replaceable. This is direct reuse of the gasket and groove logic: put compliance in the cheap consumable, not in the part.
Never rely on a printed spring to close a valve. Compliant printed springs work, but their force is low and their stiffness changes with time and temperature. Use a real spring or gravity, and use the printed feature only to guide.
If the compliant parts of the valve are what interest you, the compliant mechanism article covers the geometry of features that bend on purpose — the same hinge logic a bellows wall depends on.
The same caution applies to pneumatic cylinders. A printed cylinder-and-piston pair at hobby pressures will work at 0.5 bar and be marginal at 1 bar, because the seal has to handle both pressure and sliding without a metal bore to reference. If the cylinder has to move something reliably, the bore is the part worth buying.
Ducts, Manifolds and the Arithmetic of Pressure Drop
In a duct, geometry dominates material to an almost embarrassing degree. Pressure drop in a duct scales roughly with the length, inversely with the fourth power of the diameter for a smooth round bore, and is multiplied by every bend, every step and every internal ridge. For printed ducts, the "smooth bore" assumption is the one that has to be checked, because a printed bore carries a helical texture that adds measurable resistance.
| Design choice | Relative effect on pressure drop | Comment |
|---|---|---|
| Double the bore diameter | Falls by roughly 16x | The single biggest lever available |
| Halve the duct length | Halves | Second biggest, and usually free |
| Replace two 90° elbows with one gentle bend | Falls noticeably | Use two 45° rather than one 90° |
| Add a sharp internal step or ledge | Rises measurably | Common where two bodies meet |
| Switch from vertical to horizontal bore | Worse texture, higher drop | Vertical bore prints smoother |
| Increase wall thickness | No effect on flow | Only affects leak and stiffness |
If the duct has to be joined to a housing, the enclosure and chamber guidance on managing airflow and heat is the natural next read, because a duct that works on the bench and starves inside a closed box is a chamber problem, not a duct problem.
Two design habits follow. First, run the duct as a vertical bore whenever the geometry allows: vertically printed round bores come out measurably smoother because the bead is laid into the trough of the previous layer instead of bridging across it. Second, put every transition as far from the working end as possible — the pressure drop right at the outlet is what starves the nozzle or the ejector.
For the parts that feed a printer, this arithmetic is not academic. A fan duct with too small a throat and a 90 degree turn will deliver visibly less cooling to the far side of the nozzle, and it will show up as asymmetric overhangs rather than as a fault anyone can identify. The orientation and support rules change accordingly: the duct that looks best in the slicer preview is often the one that starves the part on the printer.
Materials for Air: An Honest Table
Ranked by how well they hold air and how well they tolerate the pressure, not by how strong they are in a tensile test.
| Material | Air tightness | Best pneumatic job | The catch |
|---|---|---|---|
| PPS-CF10 | Best of the FDM family | Valve bodies, long-term pressurized housings | Hot, expensive, needs a capable printer |
| PA6-GF25 | Good | Manifolds and structural pneumatic bodies | Moisture sensitive before and after printing |
| PA6-CF20 | Good | Cups with rigid backs, cylinder bodies | Dimensional change with moisture |
| PET-GF15 | Fair to good | Ducts and moderate-pressure bodies | Lower HDT than nylon, more brittle |
| PA12-CF10 | Fair | Complex manifolds needing accuracy | Lower stiffness, more permeable than PA6 |
| PETG | Fair | Low-pressure ducts, cosmetic housings | Seals acceptably only at 3 mm and above |
| ABS / ASA | Fair | Warm-air ducts, enclosures | Layer bonding must be solved first |
| PLA | Fair, then poor | One-off test cups and test ducts | Creeps under sustained pressure even at room temperature |
| TPU 95A | Poor as a wall, excellent as a seal | Cup lips, valve discs, bellows | Do not use it as the pressure boundary |
| TPU 85A | Poor as a wall, best as a seal | Textured-surface cups, soft bellows | Hard to print thin, tears at sharp corners |
The row that surprises people is PLA. It is often recommended for pneumatic test parts because it prints easily and looks airtight. It is not a bad test platform — but "looks airtight" is the trap, because PLA creeps under sustained pressure. A PLA part that holds 0.5 bar on the first day will be leaking on the tenth, and the change is slow enough that it gets blamed on the pump.
Testing Without Instruments, in Twenty Minutes
You do not need a gauge to know whether a printed pneumatic part works. Five checks, in order, will tell you where it fails.
- The squeeze test. Cap the outlet, squeeze the part, hold for ten seconds, release. A slow spring-back means the part holds pressure; an instant return means air is getting out somewhere. This is crude and it localizes nothing, but it separates "holds" from "does not hold" in two seconds.
- The soapy water sweep. Pressurize to your working pressure, brush on a thin film of dish soap solution, and watch. Bubbles are unambiguous. Sweep the flange first, then the seams, then the ports, then the walls. Interlayer leaks produce a fine foam along a line; fit leaks produce bubbles at a joint; permeation produces nothing at all, which is itself the answer.
- The tissue test for vacuum. For a vacuum cup, hold it down on a glass plate, pull vacuum, and release. The cup should stay attached for several seconds after you stop pumping. If it drops instantly, the leak is at the lip or the flange; if it drops slowly, the leak is in the wall.
- The stroke count. For a bellows, cycle it by hand one hundred times and then measure its free length and its sealing force again. Any measurable change means it is being over-travelled, and the stop needs to move.
- The overnight hold. Fill the part, cap it, and leave it for twelve hours at the temperature it will see in service. A drop of a few percent is permeation and is acceptable. A drop to zero means a real leak, and it will be at a joint, because a wall leak big enough to empty the part in a night would have shown up in step two.
That last check is worth the twelve hours. Every pneumatic part that has failed for me has failed overnight, and in every case the cause was a joint rather than a wall — usually a flange that was not stiff enough to keep even pressure on its gasket.
Where Printed Pneumatics Stop Being the Right Answer
Being honest about the boundary is more useful than another list of successes. Printed pneumatics are the right answer for cups on light parts, bellows for short strokes, ducts and manifolds at low pressure, and valves that open a few times a day. They are the wrong answer for anything on this list.
- Continuous pressure above about 2 bar. Walls, plates and joints all become marginal at once, and the failure is sudden rather than gradual.
- Any part that must hold vacuum for weeks. Permeation through the polymer sets a floor on leak rate that no geometry can beat.
- Sealing a toxic, flammable or oxygen-sensitive gas. The consequence of a leak is not a loss of performance.
- Sliding seals in a printed bore. Without a metal bore to reference, wear and ovality close the clearance and the piston seizes.
- High cycle counts on a thin flexible wall. A printed bellows printed in TPU is a good 10,000-cycle part and a poor 500,000-cycle part.
- Anything in a food or medical path. The layer texture is a bacterial refuge that cannot be cleaned reliably.
What This Article Changes About the Rest of the Series
Everything in the last month of design writing assumed the printed part was working against a force. Air works differently: it works against your surface finish, your bond planes, your wall thickness and your volume, and it does so continuously rather than at one critical moment.
That makes pneumatic parts a good final test of the whole design toolkit, because almost every method we have built up applies and almost every one of them has to be adjusted. The tolerance rules still apply, but the acceptable gap is smaller. The creep and preload rules still apply, but now they decide leak rate instead of clamping force. The chemical resistance rules still apply, but the failure is a slow material change rather than an obvious crack.
The one genuinely new idea is that a printed pneumatic part is a volume with a surface problem. Almost every fix — thicker walls, fewer joints, larger bores, smaller internal cavities, softer seals, mechanically bounded strokes — is a statement about surface area, wall thickness or stored volume. Get those three right and the material choice becomes a second-order decision, which is exactly the opposite of how most people approach it.
Ten Rules, In Order of Payoff
- Orient so the pressure cuts across layer lines, not along them. This alone decides whether the part leaks.
- Print 5 °C to 10 °C hotter than your cosmetic settings, and check overhangs rather than surface finish.
- Make the lip a thin, compliant ring with a narrow contact band; put the compliance in TPU, not in the backing plate.
- Never glue a soft cup directly to a rigid manifold. Print a flange and clamp it.
- Build bellows stroke from convolutions, not from wall length, and always add a mechanical stop.
- Make bellows plates stiff with ribs instead of thick; a domed plate is a bellows that does not push.
- Seat valves on a top surface, seal against a soft disc, and never depend on a printed spring to close.
- Enlarge the bore before you do anything else to a duct. Diameter beats every other variable by an order of magnitude.
- Take long-hold parts off PLA even for testing, because creep under continuous pressure is invisible on day one.
- Test joints first and walls last. Almost every pneumatic failure is a flange, not a wall.
If you are building the duct side of this rather than the vacuum side, the cooling and flow material in cooling and orientation covers the printer-specific version of the same arithmetic. If you are building the seal side, start from the gasket geometry piece and work backwards into the cup.