Blog · 2026-08-15 · Linear
A linear axis is four parts that must agree with each other: a guide, a drive, a coupling and a frame. Get the alignment wrong and it binds, stalls and wears out. Here's how to build one that runs smoothly for years.
Most first linear axes bind. The screw turns, the carriage moves partway, and then everything tightens up — or it runs beautifully in the middle of the travel and jams near the ends. The cause is almost never the parts; it's that the guide and the drive disagree about where "straight" is, and something has to give. Understanding the division of labour between components is what prevents that, so let's start there.
Each part of a linear axis has exactly one job, and problems come from parts doing each other's jobs:
The guide (rails and bearings) defines the path and carries every load that isn't along the axis — weight, side forces, moments from an overhanging tool. The drive (screw or belt) provides motion along that path and should carry only axial force. The coupler joins motor to screw while absorbing the inevitable small misalignment between them. The frame holds all of it rigidly enough that loads don't deflect the geometry.
Three options cover almost everything hobby builders need.
Round rail and linear bearings (LM8UU on 8 mm rod) are the cheapest and most forgiving of rough alignment, since the bearings can rotate slightly around the rod. They're also the least rigid and noisiest, and the recirculating balls wear grooves into softer rod over time. Fine for light, low-precision axes.
Profile rails and blocks (MGN9, MGN12, MGN15) are the modern default and the biggest single upgrade available. They're stiff in every direction, take moment loads well, run smoothly and quietly, and cost far less than they used to. The catch is that they demand a flat, straight mounting surface — a profile rail bolted to a bowed extrusion becomes a bowed rail, and its precision is wasted.
V-wheels on extrusion (the OpenBuilds approach) are cheap, tolerant, adjustable via eccentric nuts, and quiet. They're less rigid than profile rails and the wheels wear, but for a large, low-force gantry they're a sensible economical choice.
Whatever you choose, two guide points spaced as far apart as practical beat two close together. A carriage riding on bearings 40 mm apart will rock; the same bearings 150 mm apart will not. Spacing is free rigidity.
For a screw axis, the parts are: motor, coupler, screw, nut and a bearing at the far end (optional but valuable on longer screws).
The coupler deserves more thought than it gets. Motor shaft and screw will never be perfectly collinear — that's not a workmanship failure, it's reality with printed and machined parts. A rigid coupler transmits that misalignment straight into the motor bearing and the screw as a bending load, which is exactly what wears out motor bearings and makes an axis rumble. A flexible coupler — helical beam or jaw/spider type — accommodates a few tenths of a millimetre of misalignment while staying torsionally stiff. Use a flexible coupler unless you have a specific reason not to. Spider couplers have slightly less backlash than beam couplers and damp vibration better; beam couplers tolerate more misalignment.
The nut mounting is the other half of the same problem. The nut should be located firmly in the axis of travel but allowed to float slightly perpendicular to it, so it can follow the screw's small runout instead of fighting the rails. Printed nut blocks with slightly oversized bolt holes, or an anti-backlash nut with a compliant mount, achieve this. Bolting a nut down absolutely rigidly and hoping everything is perfectly aligned is the most common cause of binding.
Assemble in an order that lets each part find its own position rather than being forced:
1. Mount and align the rails first, with nothing else attached. Bolt them down, then slide the carriage by hand along the full travel. It should feel identical everywhere — no tight spots, no changes in effort. If two parallel rails are slightly non-parallel, loosen one and let the carriage set its spacing before retightening, working along the length.
2. Fit the nut block to the carriage loosely. Don't tighten it yet.
3. Fit the screw through the nut and into the motor coupler, motor mount bolts loose. Let the screw find its own natural position.
4. Run the carriage to one end of travel, tighten the motor mount there, run to the other end, tighten the nut block there. This is the key step: tightening each component at the position where it's least constrained lets the assembly settle into agreement rather than being forced into one builder's idea of straight.
5. Test by hand, then by motor. Turn the coupler by hand through the full travel; effort should be smooth and constant. Then run it under power, slowly, and listen. A rhythmic sound once per screw revolution indicates a bent screw; effort that rises toward one end indicates residual misalignment.
With the mechanism sorted, the motor has to actually drive it. Compute the required torque from the load and lead using the Lead Screw Calculator, then double it — guide friction, acceleration and the general untidiness of real assemblies all eat margin that a clean calculation doesn't show. If you're driving with a stepper, check the required speed against the torque-speed reality in the Stepper Motor Calculator, and set the driver's current limit properly using the Vref procedure. An axis that binds slightly and a stepper whose current is set too low produce identical symptoms — missed steps — so it's worth eliminating the mechanical cause first.
An axis needs to know where it is. Steppers are open-loop, so position is only meaningful relative to a known reference, which means a homing switch at one end and a homing routine at startup. Mechanical microswitches are cheap and repeatable to a few hundredths of a millimetre if approached slowly and consistently from the same direction; optical and magnetic endstops avoid contact wear.
Add a second switch (or a firmware soft limit) at the far end. A carriage driven into a hard stop at speed either skips steps, bends something, or both — and the failure usually happens during the first test of new code, when you least want it. Sensorless homing via a TMC driver's StallGuard is an elegant alternative that eliminates the switch entirely, though it needs tuning to trigger reliably without false positives.
For short screws (under about 200 mm) driven at modest speed, no — the motor bearing and nut are enough. For longer or faster screws, a supported end bearing significantly reduces whip and wobble. It must be a bearing that allows axial float on at least one end, or thermal expansion will preload the screw.
Almost always torque falling off with stepper speed rather than a mechanical fault — verify by turning the axis by hand for smoothness first. If the mechanism is smooth, the fix is a higher supply voltage, lower speed, or a finer lead.
For one-directional lifting, backlash is irrelevant. For positioning that reverses, an anti-backlash nut brings a typical T8 assembly under about 0.05 mm — and approaching every target from the same direction in software removes the effect entirely at the cost of a little extra travel.
When an axis doesn't run sweetly, work through it in this order — each step isolates one component so you're never guessing about two things at once.
Disconnect the drive entirely (undo the coupler or unclamp the belt) and push the carriage by hand along the full travel. If it's rough or tight anywhere, the problem is the guide: rails not parallel, not flat, or bearings preloaded by an over-tight mount. Fix this first, because nothing downstream can compensate for it.
If the guide is smooth, reconnect the screw but leave the nut block loose. Turn the screw by hand. Smooth means the screw and its bearings are fine. A rhythmic tight spot once per revolution means a bent screw — check it by rolling it on a flat surface.
Then tighten the nut block and try again. If binding appears only now, the nut and the rails disagree about the path: loosen, run the carriage to the far end of travel, and retighten there so the nut can find its natural position.
Finally reattach the motor. Binding that appears only at this stage points at coupler misalignment or a motor mount that isn't square. Effort that rises steadily toward one end of travel almost always means the screw isn't parallel to the rails — which is a mounting problem, not a parts problem, and no amount of extra motor torque will fix it.
Rails guide, screws drive, couplers forgive, and everything gets tightened where it's least constrained. Size the drive with the Lead Screw Calculator, and if you're not yet committed to a screw, compare the alternatives in Lead Screws vs Ball Screws vs Belts.