Scott-Russell linkage: exact straight line from a rotation

Almost every mechanism that draws a straight line is cheating. The Watt and Chebyshev linkages trace a curve so flat over a short stretch that you mistake it for a line, but it's an arc, and the moment you move away from the center, the deviation shows. The Scott-Russell doesn't cheat: with three links and the right proportion between them, a point on the mechanism travels along an exact straight line, not an approximate one, while the crank turns. But this isn't free geometric magic. In exchange for that exactness you need either a slider with its guide or a precisely positioned fixed pivot, and the lengths must hold to the millimeter. Everything here is decided by the ratio between the links and by the play in the pivots; fail at either one — the proportion or the pivots — and your exact straight line goes back to being an arc with pretensions.

The kinematics: why the line is exact and not approximate

The Scott-Russell is, at heart, a crank-slider with an output arm sized to cancel the curvature. A crank turns about a fixed pivot and its tip describes an arc. Hanging off that tip is an output link whose other end is forced to move along a straight guide — a slider — or, in the inverted version, constrained by a second pivot. The point you care about — the output — is at the free end of that link.

The key is a single geometric condition, and it helps to name the three points so you don't lose track of it. Call A the end that runs on the slider, B the point where the crank engages the link — the crank's tip — and C the output point. The exact straight line appears when A, B, and C are collinear and the crank is the same length as each half of the link, that is, when the crank length equals |AB| and |BC|, with B at the midpoint of segment AC. Put another way: the crank's fixed pivot stays equidistant from A, B, and C, all three on a circle centered on that pivot.

That equality of distances is exactly what rectifies the arc, and you don't need anything exotic to see it. With A running on the guide and B describing its arc, the crank's fixed pivot is equidistant from A and C at every instant; C ends up trapped on the line perpendicular to the slider guide that passes through that fixed pivot. It's the elementary geometry of the isosceles triangle, not an inversive transformation. That's why it's better not to call it an "inverter": the real inverter, the one that turns an arc into a line by circular inversion, is the Peaucellier-Lipkin, and it relies on different physics. The Scott-Russell pulls its line from a property of equal lengths, and that's why the line is mathematical as long as those lengths are respected. If the ratio between the links drifts, you don't get a worse line: you get a different curve.

That's the fundamental difference from approximate straight lines. The Watt and Chebyshev linkages also depend on specific proportions — their straight stretch only lands where it should if the links are correctly sized — but their error is gradual: if the proportion drifts a little, the curve degrades a little. The Scott-Russell flips that bargain. It gives you the exact line, but it demands a sliding guide or a finely toleranced pivot, and it doesn't forgive a length deviation: the line fails all at once.

Material shrinkage is the enemy here

In most mechanisms you can afford for the links to come out a hair shorter than you drew them: PLA shrinks as it cools, yes, but if every link shrinks by the same amount, the whole mechanism ends up scaled and keeps working. Not in the Scott-Russell, because what matters isn't the absolute length of each link but the ratio between the crank and the output link, and that relationship is only preserved if both links shrink by exactly the same amount.

That almost never happens. A longer link accumulates more absolute shrinkage. A link oriented differently on the bed shrinks differently. And one that crosses a different infill zone cools at a different rate. Any asymmetry moves the ratio between the links, and moving that ratio takes C off the line. The practical defense is to print the two critical links with the same orientation and the same thermal regime — ideally in the same batch, laid down the same way — so that if they do shrink, they shrink together and the proportion holds. And measure the printed links before you assemble: if the crank doesn't already match each half of the output link in the actual printed part, don't expect the mechanism to forgive you.

Print the links so the load runs along the beads

A bar link is slender by definition, and it works in tension and bending along its axis. That makes print direction the deciding factor for reliability. If you print the bar upright, with the layers stacked along its length, every axial stress pulls directly on the bond between layers — the weak plane — and the bar delaminates or snaps along a layer line under a load it would easily have held lying down. Lay it flat on the bed, with its length in the plane of the layers, so the beads, not the bond between layers, carry the load. The full reasoning is in Layer orientation for motion.

The other failure point isn't the bar, it's the pivot eye. There the cross-section narrows around the hole and concentrates stress; if on top of that the layer line or the perimeter seam falls aligned with the load, that's where the crack starts. Thicken the material around each hole and make that wall practically all perimeter; if your slicer lets you, take the seam to the unloaded part of the eye. A Scott-Russell bar that breaks almost always breaks at a pivot eye, not at the center of the link.

Buckling and out-of-plane loads: two ways to leave the line

When the output link works in compression — depending on where the mechanism is in its travel — a long, thin bar buckles before it breaks in tension. Buckling gives no warning: the bar holds, holds, and suddenly bows all at once and the output leaves the line. Buckling is governed by the section's axis of least inertia, so the defense is to add depth precisely along that weak axis. The problem is that, in a bar laid down and printed flat, the weak axis is usually the thickness in Z — the layer direction — and reinforcing there worsens the fiber orientation you just worked to get right. Identify the axis of least inertia, reinforce it, and if it coincides with the layer direction, accept that you're choosing between buckling stiffness and axial strength: give the bar exactly the section it needs not to buckle and don't trim it any thinner than the load allows.

There's a second, more subtle mode, specific to printing the mechanism lying down. The whole Scott-Russell works in a plane; printed flat on the bed, that working plane coincides with the plane of the layers, and any load out of it — the output link's own weight, the lateral torque the slider feeds in — bends the assembly in Z, right where the part is weakest between layers. The whole mechanism pitches, shifting C off its line without any bar having changed length. If the application loads out of plane, stiffen the assembly in Z or guide it laterally; it's not enough for the links to be correctly sized in their own plane.

The pivots and the slider: this is where the line escapes

You've nailed the proportion and oriented the bars well. If the pivots wobble, the line goes anyway. The Scott-Russell amplifies joint play toward the output: two or three tenths of slop in the intermediate pivot don't stay at the pivot, they propagate along the output link and turn into a visible deviation from the line and into backlash — the dead play when you reverse direction — every time the mechanism changes direction. It's worth putting the figures in their place: in a pivot printed print-in-place (printed already assembled), two tenths of slop aren't a crude defect but the floor of what's achievable — between elephant's foot and the hole coming out narrow, you normally land at two to four actual tenths. That's why each joint wants the minimum slop that still lets it turn freely, not the generous gap of a crude hinge. Tune it pivot by pivot: the gap just enough to turn without seizing and without wobbling. If they go print-in-place, calibrate that gap on a coupon first; if they take a pin, give it a clean sliding fit. How to reason that gap per side, and why the printed hole comes out narrower than you drew it, you'll find in Tolerances for moving parts.

If your version has a slider — and the canonical form of the Scott-Russell does — that guide adds its own problem: friction and slop at once. If the slider runs loose, its lateral play moves the point that was supposed to run straight and ruins the exactness you bought by nailing the proportions; if it runs tight, friction raises the input force, can cause stick-slip — that jerky advance that leaves the output shaky — and, with plastic against plastic, wears and opens up the fit over time. You want the slider with the least lateral play compatible with running smoothly, smooth walls in the sliding direction — orient the guide so the layers don't form steps transverse to the motion — and, if you can, a generous contact length that spreads the support and reduces the pitching. The inverted version, which swaps the slider for a second pivot, saves you the guide's friction in exchange for another joint whose play you also have to mind: you choose which error you'd rather manage.

When this mechanism is worth it

The Scott-Russell earns its place when you need exact rectilinear motion from a rotation and you don't want — or can't — fit a long rail. A carriage on a linear guide gives you the line, but it takes up the whole stroke as a straight rail; the Scott-Russell generates it with links that pivot in a far more compact volume, starting from a rotation. It's the natural choice for measuring or plotting instruments, where the trajectory's straightness is the function, and for any rotation-to-translation conversion that has to be precise and not merely "roughly straight." Just remember that the line is exact only for C and only over the central stretch of the turn: how you drive the crank and how you manage the variable torque near the dead points is part of the design, not a detail you can leave for later.

What it is not is a forgiving mechanism. If your application is happy with an approximate line over a short stretch and you value robustness against manufacturing imprecision, a Watt or Chebyshev linkage will give you fewer headaches: no guide, no critical proportions, no amplification of the play. The Scott-Russell is for when exactness is non-negotiable and you're willing to pay for it with calibration. When you size it, come back to Tolerances for moving parts: in this mechanism, more than in any other, the slop you choose at each pivot is the straightness you're going to get.