How to Check Sheave Alignment

Why this matters

Alignment on a belt drive is not a straightedge touching four points. It is the angle at which the belt enters the groove, and that angle is what saws a sidewall down and throws a belt off a rim. The straightedge is only an instrument for estimating it, and it is a valid instrument under conditions people rarely check: the surfaces it rests on have to be machined, parallel to the grooves, and the same width on both sheaves. Get any of that wrong and you will carefully align to a reading that has nothing to do with where the belt runs. The other half of the problem is that a gap in thousandths means nothing until you divide it by the right baseline, and the right baseline is different for an angular error than for an offset.

Isolate, and mind what moves

Lock out and tag the motor at its disconnect. A wheel holds rotational energy after the power is off, which is stored mechanical energy under 29 CFR 1910.147: verify zero rotation by eye and block the wheel or close and secure the damper before reaching past the guard. Release a spring-loaded tensioner only with the tool the design provides and stay clear of the arm's swing path. Before loosening motor hold-down bolts, support the motor: a motor on inclined slide rails travels when the bolts come loose, and hands do not belong under it or between it and the base. Thin stainless shim stock cuts, so handle it with cut-resistant gloves and deburr any edge you make. If the drive has just been running, let it cool or handle the sheaves with gloves rated for the surface temperature. A laser alignment tool is typically a Class 2 device: do not look into the beam and do not leave it set at eye height across a walkway. Opening an electrical enclosure to reach the disconnect is electrical work under 29 CFR 1910.333(b)(2), proved live-dead-live per NFPA 70E-2021, 120.5, and the guard goes back on before restart under 29 CFR 1910.219 in general industry or 29 CFR 1926.300(b) on a construction site.

Step 0: prove the sheaves are not lying to you

Before any alignment reading, rotate each sheave by hand a full turn, gripping the rim with fingers clear of the groove entry, and watch the rim and face against a fixed pointer.

If the reading changes as the sheave rotates, you have runout, and no alignment is possible until it is resolved. Runout comes from a bent shaft, a sheave pulled up crooked on its taper bushing, a damaged bore or keyway, or a bushing that was tightened unevenly. The maker publishes runout limits; the field disqualifier is simpler, which is that a straightedge gap that moves as you turn the sheave cannot be corrected by moving the motor. A tech who averages that reading and shims to it has aligned to a number that exists at exactly one angular position.

Check soft foot at the same time, because a motor that rocks on its base will land somewhere different every time you tighten it. That has its own article and it is worth reading before you shim anything.

Step 1: pick your reference surface honestly

A straightedge across the two sheaves' outer faces works when those faces are machined and parallel to the grooves. Three cases where it does not:

• Stamped or formed sheaves, where the face is not a machined reference at all.
• Different face widths, which is the common one. A single-groove driver against a three-groove driven sheave has outer faces that are nowhere near the plane of the groove actually in use. Aligning their faces guarantees the belt is misaligned by the difference.
• A projecting bushing flange or hub that holds the straightedge off the face.

Where any of those apply, the reference is the groove in use, not the face. A groove-referencing laser tool that seats in the groove, or a taut string laid in the corresponding groove of each sheave, gives you the plane the belt actually runs in. On a multi-groove drive with an unused groove nearer the motor, reference the groove the belt is in, then check that your correction has not pushed a different groove out.

Step 2: slack the belt before you measure

Take the tension off before taking any reading. On soft or spring-loaded motor mounts, belt pull drags the motor into a position it will not hold once the belt comes off, so a reading taken under tension is a reading of where the belt has pulled the motor to, not where it is bolted. The exception is a rigid, bolted-down base with no slide travel, where the difference is negligible and you can leave it tensioned.

Step 3: read the four points and name the error

Lay the straightedge across both reference surfaces so it touches the larger sheave at two points, and read the gaps at the smaller sheave's two points with feeler stock. What you see names the error and the correction differs for each:

Most real drives have two of these at once, which is why you name them before you correct anything. Correcting an offset by rotating the motor produces a drive that is offset and angular.

Step 4: convert the gap to an angle, against the right baseline

This is the step people skip, and it is where the tolerance actually lives, because belt makers publish alignment tolerance in degrees rather than in thousandths.

For angular error, the baseline is the distance between the two straightedge contact points on that sheave face, which is essentially the sheave's face diameter:

angular error = arctan(gap difference / face diameter)

For offset, the baseline is the center distance, because the belt has that whole span to correct across:

entry angle from offset = arctan(offset / center distance)

That second one carries a consequence worth committing to memory. A 1/16 in offset is not a fixed severity. On a 16 in center distance it is arctan(0.0625 / 16) = 0.22 degrees; on a 60 in center distance the same offset is 0.06 degrees, nearly four times gentler. Any rule of thumb quoted in thousandths of offset with no center distance attached is missing the term that decides whether it matters.

Step 5: correct in an order that does not undo itself

1. Vertical angular first, with shims. Shimming tilts the motor, which changes everything downstream, so it goes first. Common machinery practice is to keep the stack to a small number of clean, deburred, full-size shims per foot rather than a pile of thin ones, because a thick stack behaves like a spring and reintroduces soft foot.
2. Horizontal angular next, by rotating the motor on its base.
3. Axial offset last, by moving a sheave along its shaft. This is the only correction that does not disturb the other two. If the sheave has to move outboard to line up, stop and think first: moving a sheave further from its bearing increases the overhung load, and it is usually cheaper to move the other sheave inboard.
4. Re-tension, then re-check alignment. Tensioning pulls the motor, and on slide-rail mounts it can pull it out of the position you just set.

Worked example: a drive that took three passes

A blower drive on a rigid base with slide rails. Center distance 20.0 in. Both sheaves have machined outer faces of the same width, driver face diameter 4.5 in, driven 11.0 in, so the straightedge across the faces is a legitimate reference here.

Step 0. Both sheaves rotated by hand a full turn against a fixed pointer, gaps steady within the feeler stock's own resolution. No runout, so the readings mean something.

First reading, belt slacked. Straightedge touching the driven sheave at both points. At the driver: front gap 0.000 in, rear gap 0.040 in.

Naming it. One point touching and the other open is angular in the horizontal plane. The angle is arctan(0.040 / 4.5) = 0.51 degrees. Belt makers commonly publish something on the order of half a degree for classical V-belt drives, and considerably tighter for synchronous drives where tooth entry is unforgiving; the number to work to is the one on your maker's data sheet. At 0.51 degrees this drive is sitting on the limit, which is exactly the condition that produces a belt that wears one sidewall and lasts most of a season, so nobody connects the two.

Correcting it. Motor rotated on its base until both driver points read the same. New reading: front 0.030 in, rear 0.030 in.

Second error, exposed by fixing the first. Equal gaps at both points is pure axial offset of 0.030 in. Its entry angle is arctan(0.030 / 20.0) = 0.086 degrees, which on this center distance is comfortably inside tolerance. Worth noting anyway, because the same 0.030 in on a compact drive with an 8 in center distance would be arctan(0.030 / 8.0) = 0.21 degrees, close to half a typical classical V-belt allowance spent on an offset nobody would have bothered with.

It was corrected anyway, by moving the driver sheave inboard on its shaft rather than moving the driven sheave outboard, which would have added overhung load to the blower bearings. Reading after: 0.010 in offset, both points equal.

Third pass, after tensioning. Belt tensioned to the maker's new-belt deflection force, then the straightedge went back on: front 0.025 in, rear 0.025 in. Tensioning had pulled the motor about 0.015 in along its rails. Still inside tolerance at arctan(0.025 / 20.0) = 0.072 degrees, and now it is a measured number rather than an assumption.

What skipping the third pass looks like. The drive leaves commissioning with a documented 0.010 in offset and actually runs at 0.025 in. That is fine here. On a slide-rail motor with more travel and a tighter center distance it is the difference between inside and outside the belt maker's allowance, and the evidence arrives months later as a belt worn on one sidewall that everyone reads as a bad belt.

How to verify it held

• Re-check after the run-in re-tension. Tension drops during run-in, the motor is adjusted again, and alignment moves again. The alignment that counts is the one after the last adjustment anyone makes.
• Read the belt at the next visit. A belt worn on one sidewall and not the other is residual offset or angular error. Even wear on both sidewalls is normal service. The full wear-pattern read has its own article.
• Watch where the belt tracks in the groove. With the guard closed and the drive running, viewed from outside the plane of rotation, the belt should sit steady. A belt that walks toward one flange, or that climbs at start and settles once running, is telling you the entry angle is wrong even when the straightedge said it was not, and that usually means your reference surface was not the groove.

References

• 29 CFR 1910.147 for mechanical isolation and stored rotational energy; 29 CFR 1910.333(b)(2) and NFPA 70E-2021, 120.5 for enclosure work and live-dead-live proving
• 29 CFR 1910.219 (general industry) and 29 CFR 1926.300(b) (construction) for guarding of belts, pulleys and sheaves
• Belt and sheave manufacturer engineering data for alignment tolerance in degrees by belt type, sheave runout limits, and bushing installation procedure
• See related: Why Soft Foot Defeats an Alignment; How to Tension a Belt Without Guessing; What Over-Tensioning Does to the Bearings Either Side