Generic rail and carriage: T, slot, or tongue

A linear guide is the simplest promise in mechanics: a body that can only travel back and forth along a straight line, and nothing else. The rail sets the direction, the carriage rides it, and between the two there should be no freedom left but that one translation. The hard part isn't the idea—trapping the carriage in every degree of freedom but one—it's pulling that off in FDM without seizing, without nodding, and without wearing out on the first few passes. And all of that comes down to the profile you choose and the tenths of clearance you leave on each contact face.

One freedom, ideally trapped

A guide rail aims to leave translation along the axis alive and close off the other five degrees of freedom through surface contact. That's the intent. In practice, with the real clearance FDM demands, the carriage isn't perfectly trapped: rotation about the rail's own axis and pitch don't vanish—they're bounded by the ratio of clearance to contact length. An ideal carriage neither twists nor lifts; a printed one does both a little, and how much that little is, you decide with the geometry. Keep it in mind from the start, because half a page down that residual pitch is one of the guide's three failure modes.

Within that kinematics, the two profile families that truly guide—the T-rail and the slot-and-tongue—differ in how they spread the load and in whether the carriage embraces the rail from outside or rides trapped inside a slot. The T-rail puts the profile's wing above the carriage tabs, so any attempt to pull the carriage off the rail vertically runs into those wings: it resists lift well because there's material exactly where the force wants to pry the joint open. The open slot—a channel with a tongue running in it—guides cleanly side to side but does nothing against lift: load it upward and the tongue pulls right out. For it to retain in the vertical, it has to close over the tongue, and at that point it's no longer an open slot but a trapped dovetail- or T-type joint. In other words: the slot case that does take a vertical pull is, strictly speaking, the one that turns it into a T. Don't choose by habit—choose by where the force will arrive. A guide that only loads in its plane is happy with an open slot; one that pulls outward needs the wing of a T to hold it.

The long contact: blessing and trap

The argument for a linear guide over a plain pin is that it spreads the load over a broad surface: less pressure per unit area, less wear, less angular play. But that same breadth is what turns a small error into a big failure. On a short pin, a tenth of a millimeter of deviation hardly shows; on a hundred-millimeter rail, that tenth accumulates over the travel and the carriage finds a spot where the contact pinches and binds.

That's why fitting a long guide isn't reasoned the same way as fitting a short pivot. The extended contact amplifies any departure from flatness, any waviness in the walls, any thermal drift of the profile: what would be spare clearance in a short joint is the only defense against seizing in a long one. A slightly looser fit on a long guide isn't sloppiness, it's physics: you give the system room to absorb the rail's inevitable imperfections without their turning into friction that grows pass after pass. The spare play is the price you pay for the carriage not jamming mid-stroke.

A starting point for clearance

The whole article turns on "tenths of clearance," so it helps to have a starting number, even if your calibration overrides it. For a face that slides in FDM, a good starting point is on the order of 0.2 to 0.4 mm per face, with the high end of the range for the longest rails. It's not a sacred value: it depends on nozzle diameter, on your machine's flow calibration, and above all on the elephant foot of the first few layers, which fattens the base of the rail and eats clearance right where the carriage passes closest to the bed.

That's why the model's nominal clearance and the part's real clearance almost never match. The nozzle doesn't lay down infinitely thin lines: a gap drawn at 0.3 mm can come out at 0.15 mm if the inner wall fattens toward the channel, and vice versa. The operational consequence is simple: compute clearance from the measured part, not from the nominal dimension. Print a short stretch of the profile, measure it with a feeler gauge or caliper, and adjust the model from that measurement before you commit to the full hundred millimeters.

Lay it on the bed, along the sliding axis

This is where FDM imposes its judgment on the geometry. A linear guide is, by definition, long along its axis of motion, and that axis has to lie flat on the bed, in the X/Y plane. The reason is twofold. First: printing the rail standing up, with its length growing in Z, multiplies the time, fills the part with beads running across the slide, and leaves the contact surfaces at the mercy of the roughness between layers. Second, and more important: the sliding direction should follow the beads, not cross them, so that the walls that rub are continuous and smooth in the direction of motion.

But laying the profile flat brings two problems to light. The first is the overhanging undercuts. The roof of a T-slot, the wing that traps the carriage, the step the tongue slides under—all are overhangs, material the slicer has to stretch over air. If you leave them at a right angle, they come out drooping, sagging, with dangling strands that narrow the gap right where the carriage has to pass. The fix isn't supports—which would leave a rough contact surface on precisely the face that's meant to slide—but a chamfer: tilt the undercut roof to a printable angle so each layer rests on the one below and the overhang builds itself. The 45° from vertical is the limit, not the target: at exactly that angle the overhang comes out at its worst, with droop and roughness. For a face that also defines the contact clearance it pays to be more conservative—tilt it to 30°–40° from vertical, or resolve the roofs into a diamond shape—because the less overhang, the smoother the face comes out. A T-profile well designed for FDM has chamfered wings not for looks but because that's how it prints clean and support-free, with the contact faces coming out smooth off the machine.

The load split decides which profile holds

Choosing the profile is choosing where the force can pull without the guide giving way. A T-rail has material where it matters most for resisting vertical separation: the wings go to work the moment something tries to lift the carriage, and the wider they are, the better they spread that tension. An open slot, by contrast, is meant to guide laterally and accept load within its plane; ask it to resist lift and, unless it closes over the top, it will let the tongue escape on the first vertical load. The day you close it over the top so it holds, remember you've already built a T.

Always reason from the expected load: if the guide only positions in its plane, an open slot is enough; if it carries weight that pulls outward, you need the wing of a T. And don't confuse these guide profiles with the T-nut, which sounds similar but does the opposite: it isn't a carriage that rides the rail, it's a fastener that slides along a profile's channel to its spot and stays put there. The channel's wings trap it in Z while the screw pulls, so it does resist lift, but its job is not to move once tightened, not to slide. It's the backbone of rail-type modular systems, where several accessories share one profile; it's an anchoring mechanism, not a motion one, which is why it falls outside this guide's choice.

The three ways a guide stops guiding

A linear guide fails in three known ways, and they're worth naming because each has its design antidote. The first is seizing from accumulated friction: on long rails, the sum of small deviations finds a spot where the contact pinches and the carriage binds mid-stroke. You fight it with the slightly more open fit already discussed and with contact surfaces that are smooth in the sliding direction—hence the print orientation.

The second is pitch: if the clearance is large, the carriage doesn't just slide, it also twists a little inside the rail, wobbles, and the guide loses precision. Here the remedy pulls the opposite way from seizing, which is why a guide is always a compromise: you need enough clearance not to bind, but not so much that the carriage nods. The lever for winning both at once is the contact length. There's no torque that "straightens" the carriage; it's pure geometry: the pitch angle is, in essence, the arctangent of the total clearance divided by the contact length. Spread the same play over more rail and the possible angle shrinks. So if your carriage wobbles, before tightening the clearance—which would lead you to seizing—lengthen it.

The third is wear of the contact surfaces with continued use. Each pass shaves a little off the faces that rub, and in a printed plastic that means clearance grows over time: what started sliding snug ends up with play. The dominant factor in wear is contact pressure and material—a PLA abrades sooner than a PETG, and a POM slides better than both—so the first lever is to spread the load over broad surfaces to lower the pressure per unit area. Orientation helps in the background: a face that slides with the beads tends to wear a touch slower than one that crosses them, because it isn't tearing out the valleys between bead and bead at every pass. And lubrication is the cheapest mitigation of all: a dry PTFE grease on the contact faces drops friction and wear at a stroke, and often rescues a fit that was rubbing too much without having to reprint anything. That choice of face orientation is the same one that governs the strength and friction of any moving part, and it's worked out in Layer orientation for motion.

Design the guide, in short, working backward from its three failures: calibrated clearance—and one notch more open the longer the rail—so it won't seize; a long carriage so it won't pitch; contact faces oriented with the bead, and a touch of dry lubricant, to delay wear. The profile—T or slot—is chosen by the load direction; smoothness and service life are chosen by the clearance, the orientation, and a rail that comes off the bed straight.