Four-bar linkage: the queen of motion chains

Four links, four pins, and nearly every piece of controlled motion you've ever seen. The wiper that sweeps the windshield, the suspension that keeps the wheel vertical as it travels up and down, the press that multiplies your force at the end of the stroke: underneath all of them is the same chain of four pinned bars. That's no accident. It's the smallest mechanism that still has any kinematic interest — three bars form a rigid triangle that doesn't move — and at the same time the most versatile: with those same four bars, changing only their lengths, you go from a crank that turns full circles to a rocker that barely oscillates. Master the four-bar and you understand where half the mechanism library comes from. In FDM, nearly all of your success rides on the clearances at those four pins.

Four links, one degree of freedom

The four-bar chain is a closed loop: four links joined by four revolute joints — four pins that allow only rotation. One of the links is fixed to the world and acts as the frame; we call it the ground, whatever its actual shape, because it's the bar that doesn't move. Of the other three, the two that connect to ground are the input link and the output link, and the one joining them across the top, without touching ground, is the coupler. Which of the two grounded links ends up being the crank — the one that turns all the way around — and which the rocker isn't yours to choose here: the Grashof rule settles it, and that's the next section.

That there are exactly four joints isn't arbitrary. Count the degrees of freedom: three moving links in the plane have nine, and each revolute joint removes two, since it leaves only rotation free. Three links times three is nine; four joints times two is eight; one is left. One degree of freedom means it's enough to move a single link — turn the input by some angle — for the whole chain to be determined: the coupler and the output each have a single compatible position they must take. That's exactly what you want in a mechanism: one input motion, one predictable output motion, and nothing left undetermined. Add a link and a pair of joints and you'd have two degrees of freedom, a chain you can no longer control with a single input.

Grashof decides whether a link turns or only rocks

Here's what makes the four-bar so rich: the same four bars behave in radically different ways depending on their proportions, and there's a clean rule that predicts which behavior you get. Add the length of the shortest bar and the longest; compare that total to the sum of the other two. If that sum is less than the sum of the other two, the chain is a Grashof chain, and at least one link can turn full circles relative to another. If it's greater, no link turns all the way around: they all only oscillate.

The reason is geometric. For a link to make a full turn it has to be able to pass through every configuration, including the one where it stretches out aligned with its neighbor and the opposite one where it folds back over it. If the shortest bar and the longest together aren't longer than the other two, the loop can always close in those extreme positions and the short bar traces the whole circle. If that sum is too large, the loop jams short of closing and the motion is truncated into an oscillation.

That boundary defines entire families. Given that Grashof holds, which link you fix as ground changes the mechanism: fix the shortest and you have a double crank, where both grounded links turn all the way around; fix one adjacent to the shortest and you have the classic crank-rocker, the one that turns a motor's continuous rotation into an oscillation — the wiper case; fix the one opposite the shortest and you get a double rocker, where the two grounded links only oscillate but the coupler turns all the way around. A single table of four lengths yields four different mechanisms, depending on which one you fix as ground.

The coupler curve and variable mechanical advantage

So far we've looked at the links joined to ground. But the most interesting link is the coupler, the one that floats without touching the frame. As it moves, any point on the coupler — even one that juts out from the bar, off the line joining its two pins — traces a coupler curve: a closed path that's neither a circle nor a straight line, but a figure that can have nearly straight stretches, loops, or cusps. That's the power of the four-bar as a path generator: by choosing the proportions and the coupler point well, you draw the path you need. Approximating a straight line without linear guides, tracing the step of a walking leg, carrying a body along a specific arc while keeping it oriented: all of it comes from shaping a coupler curve.

There's a second virtue, this one on the force side. The mechanical advantage of a four-bar — how much torque comes out for each unit of torque you put in — isn't constant: it varies over the cycle. It's governed by the transmission angle, the angle between the coupler and the output link. When those two approach perpendicularity, the transmission angle is around 90° and the output receives the force most efficiently. When they align, the angle falls toward zero and transmission worsens: almost all of the coupler's force pushes into the output pin instead of turning it. That's why a good design keeps the transmission angle away from the extremes throughout the useful range of travel.

Mechanical advantage shoots toward infinity at a different point: when the input aligns with the coupler, the two in a straight line. There the mechanism multiplies force enormously with very little motion, and toggle presses and clamping jaws live off that effect: they work right at the edge of that alignment, where a modest push becomes a very high clamping force. But that same alignment between input and coupler is the dead point, and what gives a clamp its power is a trap in a mechanism that has to keep turning. It's worth knowing which case you're in.

The dead point: when the geometry jams

The dead point (toggle) has its flip side. When the input link aligns with the coupler — the two in a straight line, stretched out or folded — the force the coupler transmits to the output passes exactly through the output pivot. A force pointing straight at the center of rotation produces no torque: zero lever arm. At that instant the mechanism doesn't know which way to keep going, and pushing harder won't unstick it; it only squeezes the pins. It's the position where the crank-rocker reverses the direction of its oscillation, and it's where a poorly designed mechanism gets stuck.

Dead points aren't a defect to be eliminated; they're inherent to the geometry, and you have to design to get past them. There are three usual solutions. One, give the mechanism inertia so it crosses the dead point by momentum, the way an engine's flywheel carries it past the piston's top dead center. Two, phase-offset a pair of four-bars in parallel, so that when one is at its dead point the other is far from its own and pulls the pair through. Three, and often the most interesting: use it. A well-placed dead point is a self-locking position, and that's where jaws that close on their own come from.

But that self-locking isn't automatic, and this is where many designs go wrong. A toggle clamp doesn't lock by "forcing its way back through the point of infinite force." It locks because the input-coupler line crosses slightly past the dead point and comes to rest against a mechanical stop. Past that point, the reaction of the clamped part generates a torque that pushes the mechanism even harder against the stop, not back toward opening. The stop is what holds the position; without it, the geometry has nothing to rest against and there's no stable lock. If you print an over-center clamp and forget the stop, it won't stay closed on its own: it opens.

The printed four-bar lives and dies by its pins

All that elegant kinematics assumes perfect joints. In FDM you don't have them: you have four printed pins, and every one has clearance. This is where the four-bar goes from theory to craft, because the clearance of the pins directly governs whether the path you drew is followed or not.

The central problem is accumulated backlash, the dead play the mechanism takes up before it transmits motion. Each pin has a gap between shaft and hole — the radial play it needs to turn without seizing — and that gap lets the link move a little before it drags the next one along. With a single pin you barely notice. But the four-bar has four in series, and the play accumulates along the chain: the output carries the error of all four at once. The exact amount isn't a simple arithmetic sum — it depends on the instantaneous geometry, and near the dead point a small clearance produces a disproportionate output error — but the direction is always the same: more pins, more error. A coupler point offset from the link amplifies it further by lever arm, so a few tenths of clearance at each eye can turn into millimeters of imprecision in exactly the path you wanted precise.

And the clearance that ruins precision isn't the one you drew. The printed hole tends to come out narrower and the shaft thicker than their nominal dimensions — from the hole shrinking and from over-extrusion on the shaft's outline — so without compensation the real gap isn't the one in the model. That mismatch between nominal and printed is the subject of Tolerances for moving parts, and that's where you get the figure to size each pivot with.

The temptation, then, is to tighten the clearances as far as they'll go. That's a mistake: with too little clearance the pin seizes, the mechanism jams or advances in jerks, and every turn abrades the plastic. You have to give each joint just enough gap to turn freely and not one tenth more, because every extra tenth is backlash. The answer isn't to choose between precise and loose: it's to calibrate your printer with a tolerance coupon and size the four pivots with the real sliding gap, not with a number from a table.

Orient the layers and reinforce the pivot eyes

There's still the other FDM failure, the one that gives no warning until something cracks: delamination. A printed part is strong along the beads and weak between layers. A four-bar link works mostly in tension and compression between its two pins — it pulls and pushes along the bar — so the rule is to orient each link lying flat in the layer plane, with the line joining its two pivot eyes running along the beads. That way the load on the bar's body travels through the strong material and doesn't pull layers apart. Print a link on edge, with the layers perpendicular to the load line, and the bar splits along a layer plane at the first serious stress. The full reasoning for how orientation decides strength is in Layer orientation for motion.

The pivot eyes are the other fragile spot, and laying the link flat doesn't fully solve them. The pin hole takes a radial load in every direction in the plane, so even though the bar's body works through the strong material, the eye's outline always has a weak component between layers. On top of that, the hole concentrates stress around its perimeter, and a thin-walled eye cracks there, from the same hoop stress that splits any loaded hole. Give them wall thickness: a generous ring of material around each hole, made of continuous perimeters and not of infill, which is the only thing that truly resists tension in a circle. On low-cycle pivots, plastic against plastic holds up. But if the mechanism will turn many thousands of times, the printed pin wears: the rubbing files the hole, the gap grows, and the backlash you worked so hard to eliminate comes back, now by wear. For those high-cycle pivots, seat a metal bushing in the eye and let metal turn against metal; the plastic only carries the bushing. How to seat that hardware without cracking the eye is covered in Embedded hardware: magnets, bearings, and inserts.

Know the four-bar thoroughly and half the chapter on linkages falls into place: the crank-rocker, the double rocker, the toggle press, and nearly all custom-designed paths are variations on these four bars and these four pins. The first number you need for any of them to print successfully is a well-calibrated pivot gap — exactly what Tolerances for moving parts covers.