Parallelogram linkage: keeping orientation while it moves

If you want a platform to rise without tipping the cup it carries, or an articulated arm to push a camera away without it ending up pointed at the ceiling, you don't need a servo correcting the tilt: the geometry alone will do it. The parallelogram linkage is the bar chain that solves this at the root. It's a four-bar with one added constraint — the opposite sides made equal — and from that constraint comes a property that looks like magic and is pure kinematics: the body you move never rotates, no matter how long its travel. It keeps its orientation constant from one end of the motion to the other. The whole subtlety lies in honoring that equal-length condition to the millimeter, and in keeping the four joints from introducing the play that ruins the property.

Why the coupler doesn't rotate

A generic four-bar has one fixed bar (the frame), two rockers or cranks pivoting on it, and a fourth link joining the free ends: the coupler. In the general case, that coupler both translates and rotates at once, and its points trace complicated curves. The parallelogram is the special case where you decide the frame and the coupler measure exactly the same, and the two side bars measure exactly the same as each other too. With those two equalities, the four bars form a geometric parallelogram at every instant: opposite sides stay parallel no matter what.

The consequence is the one you're after, and it's worth pinning down its root cause. What keeps the coupler still is not so much the frame-coupler equality as the fact that the two pivoting bars are equal and stay parallel: they start parallel, they swing through the same angle at the same time, and the coupler, tied to both tips, has no choice but to stay parallel to itself. It translates without rotating. Every point on the coupler — its center, a corner, the spot where you bolt the platform — traces exactly the same circular arc, the same radius, just shifted. This is what's called curvilinear translation: the body moves along a curved path, but its angular orientation is constant. Don't confuse it with straight-line motion: the platform rises along an arc, not a straight line, so it advances a little horizontally as it climbs. What the parallelogram guarantees isn't that the path is straight, but that the orientation doesn't change. A tray that's horizontal at the top stays horizontal at the bottom, even though it swept an arc to get there.

It's still a four-bar

It's worth not losing sight of the fact that a parallelogram is a four-bar that behaves well only as long as it stays a parallelogram, and that good behavior depends on the configuration, not on the construction. It isn't a property the mechanism has built in and keeps no matter what: there are positions where it's lost.

The delicate spot is the singular positions. The singularity arrives when the two cranks become collinear with the frame — all four bars lined up on a single line — and that position need not be at the end of travel: it can fall in the middle of the working range. There the parallelogram passes through an instant where it's undefined: from that position it can keep being a parallelogram or it can cross over and become an anti-parallelogram, the configuration where the side bars cross in an X instead of staying parallel. The moment it crosses, the property is gone: the coupler starts to rotate and your platform tilts. The mechanism gives no warning; at the singularity it has two paths ahead and, with nothing to guide it, it can take the wrong one. That's why a parallelogram that in theory keeps its orientation can, in practice, tip the load: it has passed through the singularity and come out on the wrong side.

Where it's used

The parallelogram is the standard answer whenever you need to move something without letting it tilt. The textbook example is the platform that rises level: a specimen holder, a camera mount, the tray of a robot that picks up a part up high and sets it down below without tipping it. The classic articulated-arm lamp is a parallelogram — often two in series: the head points where you leave it and doesn't nod on its own as you deploy the arm, because each segment keeps its orientation. That said, chaining stages adds up orientation errors, so in a chain of several parallelograms the per-stage tolerance has to be stricter than for a single one. In transport, an arm that mustn't tilt its load — picture moving a full glass — uses the same principle so the contents don't spill as it travels through the arc.

There's a subtler use than leveling: the parallelogram serves to cancel parasitic rotation out of a motion you want to be purely linear. When a mechanism translates a body but adds an unwanted twist to it, chaining it to a parallelogram cancels that twist and leaves the translation clean. You use it not because you care about the height, but because you care that the body doesn't rotate as it travels.

What to watch when printing it in FDM

Here theory meets your printer, and theory is demanding. The whole property — the coupler doesn't rotate — rests on an exact geometric equality: the two pivoting bars measure the same. Any difference between them introduces rotation, but not a constant drift that grows with each step: it turns the parallelogram into a generic four-bar whose coupler orientation error depends on position, going to zero in some configurations and reaching a maximum in others along the arc. How much that length error translates into degrees of tilt depends on the bar length and the spacing between pivots, not on a fixed rule; but to give you a sense, one degree of tilt over a 200 mm tray is already about 3.5 mm of drop at the end, so margins that seem small show up. The good news is that in FDM you control this better than in almost any other process, because the distance that matters is the one between pin centers, and that one you set yourself in the model. Model the two bars with the same center-to-center dimension and, better still, make them from the same piece or print them in the same orientation, so any systematic process error affects both equally and cancels out.

The second front is the play in the four pins. A parallelogram has four joints, and the play of each one adds up. Here you're not designing a pivot that just has to spin freely: you're designing a chain whose geometry depends on those four points being where they belong. Too much play and the parallelogram stops being a perfect parallelogram at every instant, so the platform nods within the play even when the bars measure exactly the same. The play you choose for these pins is a trade-off: enough to turn without seizing, the minimum to keep it from wobbling. That number comes from calibrating your machine, not from a table; it's worked out for you in Tolerances for moving parts. And watch for another effect of printed play: FDM pins have high friction and stick-slip, and if the friction isn't equal on both sides, one side drags the other and rotates the coupler even when the lengths are perfect.

The third is the print orientation of the bars. A parallelogram bar works in bending in the plane of the mechanism: when the platform carries a load, each bar tends to bend within that plane. Print the bar lying flat, with the plane of the mechanism parallel to the bed, so the beads run along the bar's length and the bending tension falls on long, continuous fibers, not on the bond between layers, which is the weak plane. A slender bar printed on edge, with the layers perpendicular to the bending, delaminates where you least expect it. The full reasoning for how orientation decides strength is in Layer orientation for motion.

There's a fourth factor that's neither play nor delamination, and that on long loaded trays tends to dominate over the other two: the bending stiffness of the bars. Even if the center-to-center lengths are perfect and the pins have no play, a slender bar flexes elastically under load, the pin centers shift, and the coupler nods. It's a reversible nod — it disappears when you remove the load — but it's real, and the only defense is section: bars that are taller in the bending plane, not more material spread into infill.

The three failure modes

A parallelogram fails in three ways, and it's worth recognizing them because each is fixed differently. The first is loss of parallelism: the platform stops staying level and tilts along the travel. The cause is geometric — side bars of different length — or play — accumulated clearance in the pins. Tell them apart by the signature: if the tilt is repeatable and a function of position (zero at some points of the arc, maximum at others, but always the same for a given position), suspect the lengths; if there's hysteresis — the platform ends up higher or lower depending on which direction you approach from — suspect the play.

The second is collapse at the singularity: the mechanism passes through an aligned position, crosses over to the anti-parallelogram, and from there the coupler rotates and the load tips. It's a catastrophic failure, not a progressive one: either you've forbidden it with a stop or you're going to suffer it. The third is pin wear: with cycling, the surfaces of the holes polish and enlarge, the play grows, and a wobble appears that wasn't there at first. It's the loss of parallelism that arrives with use, which is why it pays to start with the tightest play the motion will tolerate: it gives you margin as the part ages.

The summary is short. What guarantees constant orientation is two things, only two: that the crank lengths are truly equal and that the play is controlled. Everything else — the choice of bars, the distribution of motion — is built on top of those two. And if your travel flirts with the aligned position, add a third non-negotiable condition: a stop that decides for the mechanism which side of the singularity you want to live on. Before sizing those four pins, stop by Tolerances for moving parts: the number you take from there is the one that separates a platform that stays level from one that nods on every cycle.