Gooseneck: a ball-joint chain that holds its shape

You want an arm you bend by hand, let go of, and it stays exactly where you left it—no nut to tighten, no lever to lock, no spring to snap it back. That's a gooseneck: a string of small ball joints held by friction that folds at will and keeps whatever shape you give it. You find them in desk-lamp arms, phone mounts, the steerable necks on a shop light. And here's what almost no one sees until they print their first one: this mechanism doesn't work despite friction, it works because of it. Everything that counts as a flaw in a hinge or a pivot—rubbing, binding, seizing—is the virtue here. Designing a gooseneck means designing just the right amount of seize, repeated link by link, neither too much nor too little.

A chain of ball joints that adds up friction

The kinematics are those of many ball joints in series. Each link is a spherical joint—a ball nesting in a socket—and the ball of one link grows out of the tail of the next: ball, socket, ball, socket, chained together. A single ball joint has three rotational degrees of freedom: two bending directions, which are the ones that curve the chain, plus a twist about the link's own axis, which adds no curve. What matters is that distributed bending. The assembly behaves like a continuous arm, able to take almost any curve, because there are no discrete articulations you feel as you bend it: the angle comes out spread across all the joints, smooth, because each one contributes a little.

What holds that curve against gravity is the contact friction at every ball-socket interface. When you bend the arm and let go, the weight of the overhang creates at each joint a torque that friction has to balance, whichever way it points: with the arm bent upward gravity tends to unfold it, and with the arm horizontal and weight at the tip it tends to collapse it downward. The friction between the ball and its socket opposes both directions, and it only appears because the ball goes in tight and the socket grips it with a real contact pressure. Each joint resists a little. The trick is that this little adds up along the chain, but not evenly: each joint only has to overcome the local torque that falls to it, and that torque grows as you go down toward the anchor. No single ball joint on its own would hold the weight of the whole arm; what holds the assembly up is that each one overcomes exactly its share. That's why a gooseneck "remembers" its position with no locking mechanism at all: the memory is distributed friction, not a ratchet or a brake.

The retention torque sets how many links and how big

The number that governs the mechanism is the retention torque per link: how much moment a joint holds before it starts to give. It has to be greater than the torque gravity demands of the most loaded joint, which is always the one at the base—the one carrying the weight of everything hanging ahead of it multiplied by the distance that weight sits out in overhang. If you size things so the base joint holds, the ones higher up have margin to spare, because each carries less arm and less overhang.

That torque comes from two things: the force with which the ball and socket press together (the contact pressure of the fit) and the radius of the ball (the lever arm with which that force generates moment). A bigger ball gives more retention torque at equal contact pressure, because the friction acts farther from the center of rotation and, on top of that, over more surface area. Watch what you hold constant: if you fix the linear interference—the few tenths the ball oversizes relative to the socket—a bigger ball spreads those tenths over more material and the pressure doesn't simply rise; what sizes the torque is the pressure, not the loose play. Hence the first design decision: big balls and few, or small balls and many. More links give you more flexibility and a finer curve, but they cost you in two currencies. The first is accumulated play: every interface has its slop, and thirty slops in series add up into a looseness you feel as a tremor at the tip—and the ones near the base contribute far more to that tremor than the ones near the tip, because the lever arm of the whole chain ahead of them amplifies it. The second is weight to hold up: more links is more mass hanging off the overhang, which raises the torque the base has to resist, exactly what you're trying to keep from giving. The well-resolved gooseneck is the one with links as large as the curve fineness you need will allow.

Printing the chain pre-assembled in one go (print-in-place)

The gooseneck is the showcase case for assembled printing (print-in-place): the whole chain of balls and sockets prints already built, link inside link, and comes off the bed flexible without a single assembly step. You don't print parts to fit together later; you print the mechanism already working. That's possible because between each ball and its socket you leave a print clearance that keeps the material of one part from welding to the material of its neighbor as it's laid down.

And that's where all the difficulty lives, because that clearance has to satisfy two things at once that pull in opposite directions. It has to be small enough for the joint to grip and give friction—if you overdo the clearance, the chain comes out loose and holds nothing—but large enough that the ball and socket don't end up fused into one rigid block. Fall short and the links weld together and you pull a solid rod off the bed that won't bend: the most common and most frustrating failure of assembled printing, because you can't see it until you try to move the part and it won't move. The margin between "comes out welded" and "comes out loose" is a few tenths, and it's exactly the margin your particular printer can eat up entirely if you don't know it.

Print orientation rules the spheres. A ball is a surface whose slope changes continuously, so printing it in flat layers makes its outline come out stepped, and on the face pointing downward you get overhang the slicer fills as best it can. Orient the chain axis so that stepping lands where it least interferes with the contact that has to give friction, and count on the underside faces of the overhang coming out coarser than the top ones. The clearance that works on the well-oriented half of the sphere may be too scant on the hanging half, where the material droops and leaves gaps. And there's an even more serious orientation factor: repeated bending stresses the bond between layers, which in FDM is always the weak plane. Bend whichever way you bend, the tail of the link—its thinnest section—works in tension, and if the layers sit perpendicular to that tension, they end up delaminating. Orient so the link's neck doesn't flex right against the inter-layer bond. It's the same physics that governs any part that moves, worked out in Layer orientation for motion: here it decides whether the chain comes out articulated or comes out in one block, and whether it lasts or splits open.

The material decides as much as the clearance

The whole behavior of the mechanism—friction, fusion, wear, delamination—hangs on the material, and it's worth choosing with this in front of you. PLA prints assembled cleanly, separates the parts well, and gives a dry, stable friction, but it's stiff and brittle: the link's neck breaks without warning and tolerates being bent many times poorly. PETG is tougher and stands up better to repeated flexing, but it's sticky to print and tends to fuse assembled prints more than PLA does, so the "welded or loose" boundary narrows on you and you need a bit more clearance for the same grip. TPU bends without breaking, but its elasticity costs it retention: a TPU chain springs back to its place instead of staying where you left it, exactly the opposite of what you're after. For a gooseneck that retains, the sensible balance is usually well-calibrated PETG, with PLA as an option if you bend it rarely and the stiffness doesn't scare you.

Calibrating deliberate friction

This is where the gooseneck parts ways with nearly everything else that moves. In a pivot, a hinge, or a guide, you chase the clearance that lets it turn freely without binding; the goal is to eliminate rubbing. In the gooseneck you chase the opposite: a fit tight on purpose, a controlled interference that generates the friction that holds the shape. You're not calibrating a sliding clearance, you're calibrating a repeatable grip.

That changes which table you start from. The clearance catalog that separates "turns freely" from "slides without play" in Tolerances for moving parts describes the generous side of the fit; the gooseneck lives at the other extreme, in the territory of the light grip, related to the pressure logic detailed in Press-fits that hold—only here the interference doesn't fix two parts forever, it lets them move with resistance. You want the grip just tight enough that the joint won't give under load, but not so tight that the chain becomes impossible to bend by hand or that the balls dig in and rub until they seize. It's a narrow band, and the only honest way to find it is to print samples with stepped interferences, bend them and let go, and keep the one that holds the overhang without needing both hands to curve it.

Calibrate fine, because the useful range is narrow and the chain multiplies the error: an interference a touch loose per link is invisible in a single joint, but summed over thirty it gives an arm that gives way; a touch too tight is a joint that's hard to bend, and thirty joints like that no one moves by hand.

What it's for and how it breaks

The gooseneck is the answer when you need something to bend by hand and stay where you leave it with no locking mechanism at all: desk-lamp arms, phone and camera mounts, steerable coolant nozzles, aimable air nozzles, any element you reposition often and want to fix without touching a screw. If what you want is exactly that—position without locking, continuous adjustment, one-handed handling—there's no cleaner solution. If instead you need the position to hold a serious load or to not move even if you wanted it to, this isn't your mechanism: the friction that makes it comfortable is also its load ceiling.

It's worth being clear about its four failure modes, because they're characteristic. The first is that the chain gives under load: if the retention torque per link came up short, the arm gives way on its own, slowly, as soon as you hang something off the tip or the chain's own weight gets the better of the base. You fix it by raising the interference or enlarging the balls, not by adding links, which makes the problem worse by adding weight. The second is the fusion of links in assembled printing: clearance too scant, and the chain comes out welded into a rigid block; it's a calibration failure, not a design one, and you prevent it with the short sample before the long run. The third is the breakage of the link's neck, and it goes hand in hand with the first fix: every time you raise the interference to gain retention, you also raise the force needed to bend by hand, and that force concentrates at the link's tail, its thinnest section. Overtighten—especially in PLA—and the chain becomes so stiff that the neck breaks before it bends. There's a real tension between "it retains" and "it doesn't snap," and it lives right at that section. The fourth is treacherous because it arrives with time: the loss of retention through wear. Every time you reposition the arm, the surfaces of the ball and socket rub and polish, and FDM—with its stepped finish—wears faster than a machined part. After many cycles, the grip that held the shape loosens, and the gooseneck that held firm starts to sag under its own weight.

This invites you to start from the tight side of the useful range—let the new gooseneck run a bit stiffer than comfortable, counting on use to loosen it to just right—but with one condition: that the short sample has already confirmed that at that interference it doesn't fuse and that the neck survives the bending. If the fusion boundary sits right up against the tight side, you've no margin to go stiffer, and forcing it takes you from one failure to another. Wear, besides, doesn't always loosen cleanly: it can leave burrs that bind before they release. Calibrate toward the firm side, yes, but only within the range you've already confirmed articulates.

If your problem was the opposite—the joint turning loose and free instead of retaining—the reasoning inverts completely, and you'll find it in Tolerances for moving parts: the same ball-socket pair, calibrated to remove friction instead of to manufacture it.