Clicks, Paws and Springs: 3D Printing a Mechanism That Steps, Holds and Stores Energy

A ratchet is the first mechanism most people design that has to remember something. A gear turns; a ratchet turns and then refuses to turn back. That small refusal is where the interesting engineering starts, because holding position is not the same problem as transmitting motion. A gear fails loudly — a tooth shears, the drive stops. A ratchet fails quietly. It skips one tooth, then two, and the part that was supposed to stay put slowly walks out of position while everybody blames the load.

This article is about the family of mechanisms that step, hold and store: ratchets and pawls, detents and spring plungers, and the spring-loaded mechanisms that release a fixed amount of stored energy. Three different jobs that share one underlying requirement — the part has to behave the same way on the thousandth cycle as it did on the first.

The previous pieces in this series covered how to lay out motion (mechanisms that move on purpose), how to transmit it (gears and belts) and how to make a part bend on purpose (compliant mechanisms). This one covers the layer above: how to make a mechanism hold, count and release. Every one of those jobs is a stress problem disguised as a geometry problem, and the numbers are unforgiving in ways that are easy to check before you print.

Three mechanisms that are not the same mechanism

The vocabulary gets mixed up constantly, which is why people design a ratchet and then wonder why it will not hold a detent position, or build a spring and wonder why it takes a set after a week. The jobs are different, and so are the numbers that decide whether each one works.

Mechanism	What it does	The number that decides it	How it fails
Ratchet and pawl	Allows rotation one way, blocks it the other	Root bending stress in the tooth	Tooth shears, or the pawl skips a tooth
Detent	Holds a position against a small disturbance	Breakaway force versus the disturbance	Creeps out of position under load or vibration
Spring-loaded release	Stores energy, releases it on command	Spring rate and working length	Takes a permanent set, so it never returns full force

The failure modes are worth reading twice, because only one of them is a breakage. A sheared tooth is obvious and happens once. A detent that creeps and a spring that takes a set both fail gradually, over weeks, in ways that look like the user did something wrong.

Ratchet geometry: the angle is the whole design

A ratchet tooth has two flanks, and they do two different jobs. The drive flank has to take the load and hand it to the pawl without slipping. The ramp flank has to lift the pawl on the return stroke and let it drop back with as little drag as the spring can manage. Get the ratio of those two angles wrong and no amount of print tuning will save the mechanism.

The drive flank is usually specified by the pressure angle, the angle between the flank normal and the direction of the tooth load. A low pressure angle keeps the load pushing the pawl into the tooth, which is self-engaging and stable. As the pressure angle rises, more of the load starts to push the pawl outward along its pivot arm, and past a certain point the pawl simply rides up and over the tooth. In metal that point is around forty-five degrees; printed parts with friction and compliance reach it earlier.

The practical range is narrower than most people expect:

• Drive flank at zero to five degrees from radial gives a tooth that engages hard and barely lifts the pawl. Good for a one-way clutch or a winch. Bad if the mechanism also has to be quiet, because the pawl drops in and slams.
• Drive flank at ten to fifteen degrees trades a little self-engagement for a much gentler engagement. This is where most printed ratchets should live, because the pawl gets a moment to decelerate.
• Ramp flank at twenty-five to forty degrees is the lifting side. Under twenty degrees the pawl drags and the tooth wears. Over about forty degrees the pawl is thrown up too steeply and the mechanism becomes a skip risk.

Note what those three numbers do to each other. A steeper ramp lifts the pawl faster and needs less spring travel, but reverses more load into the pawl pivot. A gentler drive flank engages more reliably but wears the pawl tip faster because the contact area is smaller. There is no free setting; there is only a choice about which failure you can live with.

Ramp flank versus pawl tip: the friction multiplier

The pawl tip is where the design either holds or fails, and it holds or fails because of friction. When the pawl tip slides up the ramp, friction acts against the lift, and the result is an effective ramp angle equal to the real one plus the friction angle rather than the real one alone.

In metal the friction coefficient is small enough that this term is a rounding error. In printed polymer on printed polymer, static friction commonly runs between 0.15 and 0.4 depending on material, surface finish and whether the part was printed flat or on edge. At 0.3 the friction angle is roughly seventeen degrees, so a ramp drawn at thirty degrees behaves like one drawn at forty-seven. That is the entire margin, and it is why a ratchet that looks perfectly reasonable in CAD skips on the bench.

friction angle   =  arctan( mu )
effective ramp   =  designed ramp angle  +  friction angle

mu = 0.20  ->  friction angle  11.3 deg
mu = 0.30  ->  friction angle  16.7 deg
mu = 0.40  ->  friction angle  21.8 deg

Two ways out, and both are usually better than reducing the ramp angle. The first is to polish or vapour-smooth the ramp so the friction term drops. The second is to treat the pawl as a compliant element rather than a rigid one: let it bend to follow the ramp, so the lift comes from the beam's deflection instead of from sliding contact. That is the same logic used in compliant snap and latch design, and it converts a friction problem into a strain problem, which is measurable.

Root stress: where ratchet teeth actually break

Nobody has ever broken a printed ratchet tooth at the tip. Teeth break at the root, and they break in bending, because the tooth is a small cantilever with the load applied near its top corner. The bending moment at the root is the load multiplied by the height of the contact point above the root, and the stress is that moment divided by the section modulus of the root cross-section.

For a tooth of width w, root thickness t and height h, with a tangential force F at the tip:

section modulus     Z  =  w * t^2 / 6
bending stress      s  =  F * h / Z  =  6 * F * h / ( w * t^2 )

The squared term is the whole story. Doubling the root thickness divides the stress by four; doubling the height doubles the stress. Short and stubby is strong, tall and pointed is weak — and tall and pointed is exactly what people draw, because it looks like a ratchet.

Printed teeth have an extra problem: the root is usually where the layer lines run parallel to the tooth face, so the bending stress arrives perpendicular to the interlayer bond. That is the direction with the weakest strength, the same weakness described in load paths in printed parts. A tooth that survives its first cycle will still fail at the root, just later.

Tooth height h (mm)	Root thickness t (mm)	Relative stress	Practical verdict
2.0	1.6	1.0 (reference)	Sound shape, takes a real load
2.0	1.2	1.8	Usable for light loads only
3.0	1.2	2.7	Marginal; expect root failure
3.0	0.8	6.0	Will not survive a first hard stop
4.0	1.0	7.7	Tooth is decorative

The table is relative, not absolute, but the ratios hold. If you want a ratchet that carries anything, keep the tooth shorter than one and a half times its root thickness and add more teeth instead of making them bigger. More teeth means a smaller step per tooth, which is usually what you wanted anyway.

Why a root fillet is worth more than a bigger tooth

A sharp internal corner concentrates stress, and in printed parts the concentration happens in a region where the bond is already weakest. A fillet of roughly one-quarter to one-third of the root thickness spreads the load over a wider region and moves the peak away from the corner. In practice the fillet buys more than the same material added to the tooth thickness, because it acts on the local peak rather than the average.

The catch is that a fillet at the root changes the ramp geometry just above it. Draw the fillet first, then re-cut the flanks — not the other way around. It is the same ordering discipline that decides whether a boss survives, covered in part shaping.

The pawl is a beam, not a lever

Most printed pawls are drawn as a rigid finger pivoting on a pin. That description is wrong in a way that matters, because a printed finger over fifteen or twenty millimetres is not rigid. It deflects under load, and that deflection is what lets the tip walk off the tooth.

The useful approximation is a cantilever of length L, width w and thickness t with a tip load F:

tip deflection   d  =  4 * F * L^3 / ( E * w * t^3 )
tip stress       s  =  6 * F * L   / ( w * t^2 )

Bending modulus for printed PETG in the plane is around 2.0 GPa, which is where the numbers get sobering. A pawl twenty millimetres long, six wide and three thick, loaded with twenty newtons, deflects about 0.2 mm and sees roughly 22 MPa — survivable, but the deflection is already the same order as the engagement depth. Shorten the pawl to twelve millimetres with the same cross-section and the deflection drops to about 0.04 mm while the stress halves.

That is why the short, stiff pawl wins: not because it is stronger, but because it stays where the geometry put it. A pawl that deflects is a pawl whose effective engagement angle changes under load, and an engagement that changes under load is a ratchet that skips when the load arrives and works perfectly when you test it by hand.

Pawl pivot: the part that decides the engagement angle

A pawl that swings on a printed pin will develop play, and play at the pivot translates directly into a change in engagement angle at the tip. If the pivot clearance is 0.2 mm and the pawl is fifteen millimetres from pivot to tip, the angular free play is already about 0.8 degrees before anything bends. That is on top of the bending, not instead of it.

Three ways to get the pivot right, in increasing order of effort:

1. Print the pawl and its pivot as one compliant element. No pin, no clearance, no play. The pawl is a shaped beam rooted in the housing. This is the most reliable option on a printer and the least used, because it does not look like a pawl in CAD.
2. Use a long pivot with controlled clearance. The design rules are the same as any printed pivot — see pins, pivots and bushings — with the added rule that the pivot should be longer than the pawl is wide, so rocking plays less of a role than free play.
3. Bush the pivot with a harder insert. Worth it only when the pawl cycles tens of thousands of times, which for a printed part is rarely the actual duty.

Whichever you choose, measure the free play before trusting the engagement. Pinch the pawl against the tooth and see how far the tip can move without the housing flexing. Anything over a tenth of a millimetre is a skip waiting to happen.

Torsion springs: size the wire, not the coil

Printed torsion springs are a genuine option and a genuine trap. They work — a printed coil spring in PETG will hold a pawl against a tooth for thousands of cycles — but only if the wire cross-section is chosen as an engineering number rather than a decoration.

The governing relationship for a round-wire torsion spring is that the bending stress at the inside of the coil is proportional to the moment divided by the wire diameter cubed, with a stress-concentration factor near 1.1 to 1.2 for curvature:

stress  s  =  K * 32 * M / ( pi * d^3 )
rate    k  =  E * d^4 / ( 64 * D * n )     (D = mean coil diameter, n = active turns)

Two consequences follow immediately. Halving the wire diameter multiplies the stress by eight and divides the rate by sixteen — printed coil springs are extremely sensitive to a dimension that is often one extrusion wide. And the number of turns raises the travel without raising the stress, which is why long, many-turn springs are more forgiving than short stiff ones.

In practice a printed torsion spring wants a wire of at least 1.2 mm to survive being printed, which sets a fairly narrow window: enough turns for the travel, a coil diameter large enough that the wire is not over-bent, and an inner diameter that clears whatever it wraps around. Print it with the coil axis in the plane of the bed if you can, so the bending stress does not land across layer lines.

Detents: a different problem with a different number

A detent does not have to resist a full drive load. It only has to resist whatever tends to move the part between operations: a nudge, gravity, vibration, or the spring of a cable. That makes it a far more forgiving mechanism, and also a far more frequently miscalculated one, because the number that matters is not strength but breakaway force.

The detent ball or pin sits in a groove, and the force needed to push it out follows the groove angle, with friction again working against you:

breakaway force  ~  F_spring * tan( groove_half_angle + friction_angle )
retention        ~  F_spring / tan( groove_half_angle + friction_angle )

This is where printed detents usually go wrong: the groove is drawn deep and steep because deep and steep looks secure, and the result is a detent that cannot be moved by hand. A thirty-degree V-groove in a half-compressed 0.8 mm printed leaf will resist several newtons. That is fine for a control knob detent and hopeless for a lid latch that has to open with one finger.

Groove half-angle	Relative breakaway	Where it belongs
15 deg	Lowest	Lid catches, thumb-operated selectors
30 deg	Medium	Rotary selectors, tool holders
45 deg	High	Anything that must resist being bumped
60 deg	Very high	Rarely correct on a hand-operated part

Two practical notes. First, printed detents loosen over a few hundred cycles, because the groove edges burnish and the friction angle drops; a detent that feels right when new will settle loose. Design the initial breakaway toward the stiff side and expect it to relax. Second, the spring side matters at least as much: the leaf that pushes the ball needs enough travel. A detent with 0.3 mm of spring travel cannot follow a 0.6 mm deep groove, so the ball rides on the flank and the position never settles at all.

Leaf springs: strain per layer, not force

The most common printed spring is a flat leaf, and the number that decides whether it survives is the surface strain at the outer fibre, not the force it delivers. For a cantilever of length L and thickness t deflected by a distance d:

surface strain  e  =  3 * t * d / ( 2 * L^2 )
stress          s  =  E * e

Note that the strain depends on thickness divided by length squared. A leaf 20 mm long and 1.0 mm thick deflected 3 mm sees about 1.1 percent strain. Halve its length and the same deflection sees four and a half percent, which is past the point where PETG yields and well past where it will take a permanent set.

The usable strain budget in printed parts is smaller than in bulk material, because layer boundaries concentrate strain. As a working rule, keep cyclic leaf strain under about 1.5 percent in PETG, 2 percent in polypropylene-like materials, and around 3 percent in TPU where the material is designed to bend. The general strain logic for flexures, including per-material numbers, is in compliant mechanism design.

Spring rate and why a printed spring softens

Spring rate for a printed cantilever follows the standard formula with the printed modulus inserted:

k  =  3 * E * I / L^3     with  I  =  w * t^3 / 12
so  k  =  E * w * t^3 / ( 4 * L^3 )

The cube on thickness means the rate is dominated by a dimension that a 0.4 mm nozzle prints as two or three extrusion widths. Being one extrusion off is not a small error: going from two widths to three changes the rate by a factor of three and a half. That is why a printed spring designed in CAD and a printed spring on the bench are different objects, and why the sensible approach is to print a test strip with three thicknesses and pick the one that feels right.

The softening over time is separate. Printed springs lose force because material creeps under continuous load, the mechanism described in creep, fatigue and preload. A spring held compressed for a month at room temperature will not fully recover when released. If the mechanism must hold force for a long time, design it so the spring spends most of its life near natural length and only deflects when actuated.

Storing energy: the numbers that make it work

A spring-loaded mechanism that releases energy — a latch, a push-push catch, a spring-return lever, a clicker — has an energy budget, and either the spring can supply it or it cannot. The budget is small and easy to calculate:

stored energy   E  =  0.5 * k * d^2        (d = deflection from free length)
surface strain  e  =  3 * t * d / ( 2 * L^2 )

Those two equations together define the design space. For a given leaf the energy goes as the square of the deflection, but the strain goes linearly with deflection. So every doubling of stored energy costs double the strain, and the strain ceiling is what caps the design.

Leaf 20 mm long, 8 mm wide	2 mm deflection	4 mm deflection	6 mm deflection
0.8 mm thick (PETG)	0.13 J, 0.5% strain	0.52 J, 1.0% strain	1.17 J, 1.5% strain
1.2 mm thick (PETG)	0.44 J, 0.7% strain	1.76 J, 1.4% strain	Over strain budget
1.6 mm thick (PETG)	1.04 J, 0.9% strain	Over strain budget	Over strain budget
2.0 mm thick (TPU 95A)	0.50 J, 0.8% strain	2.00 J, 1.5% strain	At limit

The pattern is clear: a thicker spring stores more energy per millimetre of travel with less travel available before the strain budget is spent. A thin spring stores less energy but travels far. A long thin spring with generous travel stores the same energy as a short thick one while running at much lower strain, which is why the good examples of printed spring mechanisms look almost floppy in CAD.

Two more things decide whether the energy comes back. First, the stored energy is only available in one direction of load, so a mechanism that has to push both ways needs two springs or a linkage. Second, the release path has to be stiffer than the spring, or the energy goes into flexing the frame instead of moving the load — the same double-stiffness problem that makes a printed latch feel mushy.

Push-push catches: the sequence, not the spring

A push-push catch — press once to latch, press again to release — is the classic printed spring mechanism, and it usually fails for a reason that has nothing to do with spring force: the heart-shaped cam track is not deep enough, so the pin jumps across it.

The track works because the pin follows a closed path: in along one flank, into the notch, out along the other. Two things break it. A shallow notch lets the pin slip past the detent on the way out, so the mechanism never holds. A track that is too shallow relative to the pin's lateral slop lets the pin jump diagonally at the junction, so the mechanism skips a state.

The rule that survives testing: make the track at least twice as deep as the pin's lateral free play including pivot clearance, and give the notch an over-centre entry of a few degrees so the pin is committed to the branch it entered. In printed parts that usually means a 3 mm deep track for a 1.5 mm pin — deeper than it looks like it needs to be.

Clockwork, and why the escapement is the honest test

An escapement is the purest form of the mechanism this article is about: a train of gears that would run away, held back by a pallet that releases one tooth at a time. Printed escapements do work, and they teach the lesson above in a single object.

Three conditions have to hold together. The driving torque has to be large enough to overcome friction between the escape wheel and the pallet but small enough that the pallet can hold it back. The pallet has to drop far enough to clear one tooth and no further. And the frame has to be stiff enough that the arbor spacing does not change under load, because the release depth is measured in tenths of a millimetre.

The third condition is what fails first. An escapement on a printed frame is a mechanism whose critical dimension is set by a part that flexes, which is the definition of a mechanism that will run on the first day and stop a week later. If you build one, build it on a frame at least 6 mm thick with metal inserts at the arbors, and accept that the gears are the easy part.

Materials: the honest table

Ratchet and spring mechanisms load material in a way that exposes every weakness at once: repeated bending, contact wear, continuous stress, and sometimes heat. The material choice is not the usual strength ranking.

Material	Where it wins	Where it lets you down
PETG	Best all-round spring and pawl material; high strain before yield, low friction against itself	Creeps under long-term load; teeth wear if not burnished
PETG-CF	Stiffer pawls and teeth with less deflection; excellent wear resistance	Lower strain to break; poor choice for the spring itself; abrasive on nozzles
PA6-CF	Highest fatigue resistance for teeth that cycle thousands of times	Hygroscopic — needs drying or the parts come out porous and weak
PA12-CF	Better dimensional stability than PA6-CF, good toughness	Less stiff, more expensive, still needs dry filament
TPU 95A	Silent ratchets, low-force detents, anything that must not click loudly	Too soft for engaged teeth; the pawl deforms instead of holding
TPU 95A-HS	Same low noise, printable at usable speed	Still a poor load-bearing tooth material

The unusual entry is TPU, which is almost never right for the tooth and often right for the pawl spring. A flexible pawl biasing against a rigid PETG tooth gives a quiet ratchet with a low engagement force, and the pawl's flexibility absorbs the impact energy that would otherwise be spent breaking the tooth tip. If you want a ratchet you cannot hear, that is the combination. The hardness reasoning behind it is in choosing flexible filament by hardness.

Two process notes that are not optional. Print ratchet teeth with the tooth flank running along the layer direction where possible, so the root bending stress lands in-plane rather than across the interlayer bond. And keep the wall count high around the tooth root: three to four perimeters at the root do more than any amount of infill behind it, because the stress never reaches the core — the same argument as in shell count and infill, applied where it matters most.

Print-in-place ratchets: what really decides freedom

A ratchet printed as one assembly is a genuinely good idea and a genuinely unreliable one. When it works, it works for years. When it fuses, it fuses at the pawl tip, and the pawl is the one feature with no room to be freed by hand.

What actually determines whether the assembly frees, in order of importance:

1. No horizontal overhang at the engagement face. The pawl tip and tooth tip must not be printed as a bridge over a gap, or the two will touch by sagging. Redesign the tooth so the critical face is vertical.
2. A gap larger than your printer's horizontal tolerance, placed where it matters. A 0.2 mm gap that works between two vertical walls does not work where the seam lands. Give the pawl tip 0.35 mm and use 0.2 mm everywhere else rather than a flat 0.25 mm everywhere.
3. First layer squish under control. A flat, slightly over-squished first layer joins every feature that touches the plate. If the pawl and housing both reach the bed, they are joined at the plate before they are joined anywhere else. Lift the pawl 0.2 mm above the bed and support it, or redesign so only the housing touches.
4. Seam placement. A seam landing inside the pawl gap adds a blob of exactly the size of the clearance. Set the seam to a corner or to the back of the part, and check it in the preview rather than assuming.
5. Slow the first ten layers. Most fusions happen in the first few millimetres, when the part is still cooling unevenly and the nozzle is dragging.

The escape hatch matters as much as the design: design a freeing slot. A 2 mm access hole behind the pawl, or a place to insert a blade, turns a failed print into thirty seconds of work instead of a reprint. The same principle applies to any print-in-place joint, and the pivot and clearance rules in printed sliding surfaces and manufacturing tolerances apply unchanged.

Testing: five things to check before you trust it

A ratchet that feels right on the bench is not a verified mechanism. Five checks catch nearly every failure this article describes, and all of them take less than an afternoon.

1. The skip test. Load the mechanism to its design load and cycle it fifty times by hand, slowly. A ratchet that skips once will skip again. If it skips, the cause is almost always pawl deflection or pivot free play, not tooth geometry.
2. The hold test. Load it in the blocked direction at twice the expected stall load and leave it for an hour. Look for the pawl creeping up the drive flank. If it creeps, the drive flank angle is too shallow for the friction you actually have.
3. The set test. Deflect the spring to its working travel, hold for a day, release, and measure the free length. Any permanent set means you are over the strain budget and the spring will keep losing force.
4. The noise test. Cycle it twenty times with your ear near it. A healthy ratchet clicks once per tooth. A double click means the pawl is bouncing, which means the spring is too stiff or the pawl too light, and a double click is a skip waiting to happen at speed.
5. The temperature test. Warm the part to the hottest it will see in service and repeat the hold test. Printed polymer loses modulus with temperature, and a mechanism that holds at room temperature can walk at fifty degrees — relevant for anything near a printer, a motor, or a vehicle.

If those five pass, the mechanism is probably sound. If any one fails, the fix is in the geometry, not in the slicer, and it is cheaper to find it now than in a customer's hands.

Rules, ordered by how often they matter

1. Short stubby teeth. Keep tooth height under one and a half times root thickness. This single ratio decides more failures than everything else combined.
2. Add a root fillet before making the tooth bigger. It acts on the peak stress where it actually concentrates.
3. Count the friction angle. Printed-on-printed friction eats fifteen to twenty degrees of your ramp margin. Design as if the ramp were steeper than you drew it.
4. Short pawl, long pivot. A pawl's job is to stay put, and deflection is what moves it.
5. Strain, not force, sizes a spring. Keep cyclic surface strain under 1.5 percent in PETG and you will not lose a spring to a set.
6. Stiffer than the spring, or the frame eats the energy. The release path must be the stiff path.
7. Detents are breakaway-force problems. Match the groove angle to the hand that operates it, not to how secure it looks.
8. Give print-in-place mechanisms a freeing slot. Assume one in five will fuse and design the recovery.
9. Test with load, time and temperature. Hand feel proves nothing about a mechanism whose failure mode is creep.
10. Know when to use a bought spring. For anything that cycles continuously or holds force for months, a stainless compression spring and a printed seat will outperform any printed coil, and that is not a failure of the printer.

Closing

The reason these mechanisms are worth building at all is that they change what a printed part can be. A bracket holds something; a ratchet remembers where it was. The step from one to the other is not a step in print quality, it is a step in thinking — and every one of the numbers above is a way of thinking about a part that has to behave, rather than a part that has to exist.

Start smaller than you want to. One tooth, one pawl, one leaf spring, printed in PETG, tested with a weight and a watch. The mechanism you build after that will work the first time, and the ones after that will work for years.