A machining fixture holds one part while you cut it. An assembly fixture holds several parts while you join them, and that is a harder job, because the parts have tolerances of their own and the fixture has to absorb all of them at once. If the locators are wrong, the clamp distorts a part instead of holding it, or the clearance is too tight for the real dimensions coming off the line, then the fixture does not align anything — it just makes the misalignment consistent. This guide walks through the three things an assembly fixture has to get right — where the parts sit, how they are held, and whether they land in the same place on the hundredth cycle as on the first — and shows how to design each one so the assembly goes together without a fight.
Locators decide everything: the datum comes first
Before you think about clamps, you have to decide how each part finds its position, and that means choosing the datums. The datum is the feature — a face, a bore, a slot, an edge — that the part sits against, and it has to be the same feature the part was machined to in the first place. If the part was located on a bored hole during machining, the assembly fixture should locate on that same bore, not on an outer edge that drifted relative to it. Matching the assembly datum to the machining datum is the single biggest lever you have, because it means the part arrives at assembly already referenced to the same features the fixture will use.
The classic locator scheme is the three-point rule — three points define a plane, two define a line, one defines a point — and it works in assembly the same way it works in machining. A part resting on three points cannot rock; it is fully seated and predictable. The mistake is adding a fourth point to the plane — a fourth pad that looks harmless but now fights the other three, because real parts are not perfectly flat and four points on a slightly cupped surface means the part rocks between whichever three happen to touch. Fewer, well-chosen contact points beat more points every time, because each extra point is a chance for the part to sit two different ways.
The same logic applies to holes and pins. A round pin in a round hole locates the part fully, which is fine until the hole tolerance varies, and then a pin that was a slip fit on one part is an interference fit on the next. That is why the standard practice is a round pin plus a diamond pin — the round pin fixes the position, the diamond pin fixes only the rotation, and together they tolerate the hole-to-hole distance variation that a pair of round pins would jam on. The diamond pin is oriented so its flat sides face the round pin, which lets the second hole breathe in the direction it actually varies while still clocking the part.
Clamps: hold the part without moving it
A clamp's job sounds simple — push the part against the locators and keep it there — but the direction and the force matter more than the clamping power. The clamp has to push the part into the locators, not lift it off them. A clamp that comes down at an angle, or that pushes sideways against a surface that is not a locator, will slide the part off its datum before it ever locks it down, and the part leaves the fixture in a position you never saw. The rule is to clamp perpendicular to the locating face and directly over a locator wherever possible, so the clamping force travels in a straight line into the support and does not bend or tilt anything.
Force is the second half of the problem. You need enough to hold the part through whatever the assembly operation does to it — pressing, driving a fastener, applying adhesive — but too much force distorts a compliant part, and a distorted part measured in the fixture is a lie that corrects itself the moment the clamp comes off. Thin walls, long spans, and cast or sheet-metal parts all spring under clamp load, and the fix is to place the clamp so the load goes into a rigid feature, spread the load over a larger pad, or back the part with a support on the far side so the clamp is squeezing against something solid rather than bending the part. A spring-loaded or adjustable clamp that gives the part somewhere to go, rather than crushing it, is often the difference between a fixture that aligns and one that just bends parts into position.
Clearance and tolerance stack: design for the worst real part
The single most common assembly-fixture failure is designing for the nominal part and then discovering the real parts do not fit. Every dimension on every incoming part has a tolerance, and when you stack several parts in one fixture, those tolerances add up. The fixture has to accept the combination of the largest parts in the stack, not the nominal ones, or the operator will be filing, forcing, and skipping the fixture on the shifts where the parts run big. That means the clearance between the part and the fixture wall, or between the pin and the hole, has to be sized against the worst-case part, not the drawing's nominal value.
Working out that stack is a small arithmetic job, but it is the job that separates a fixture that runs for a year from one that gets abandoned in a week. List every dimension that affects fit — the hole position, the boss height, the wall thickness, the locating feature on each part — and add their tolerances in the direction that makes the assembly tightest. That total is the minimum the fixture has to absorb, and it should be less than the clearance you designed in. If it is not, you have two choices: open the clearance, which costs some of your alignment accuracy, or tighten an incoming tolerance, which costs money upstream. Naming the stack and doing the addition up front is cheaper than finding it on the line, because on the line it arrives as scrap and rework instead of as a number on a page.
| Design decision | What goes wrong | What to do instead |
|---|---|---|
| Locate on a machined datum | Outer edges drift from the bore; part lands crooked | Use the same datum the machining used |
| Three-point plane seating | A fourth pad makes a real part rock | Three points only; drop the extra pad |
| Round + diamond pin | Two round pins jam on hole spacing variation | Diamond pin oriented flat toward the round pin |
| Clamp perpendicular over a locator | Angled clamp slides the part off its datum | Push straight into the support, no side load |
| Clearance for nominal parts | Big parts do not fit; operator forces the fixture | Size clearance to the worst-case tolerance stack |
| Back support under the clamp | Compliant part bends and reads false in the fixture | Clamp against a rigid feature or add a support |
Repeatability: prove the hundredth cycle matches the first
A fixture that aligns a part once is easy to build. A fixture that aligns it the same way on the thousandth cycle, after the locators have worn and the clamps have settled and three operators with three different techniques have each run it for a shift, is a different engineering problem. Repeatability is not about the first assembly being right; it is about the last assembly being indistinguishable from the first, and that only happens if the fixture is designed to be self-locating — the part can only go in one way, the clamp can only close to one stop, and the operator has no choices to make that could vary the result.
The practical way to buy repeatability is to remove operator judgment from the loop. A fixture with a keyway so the part cannot be inserted backwards, a clamp with a hard stop so it always closes to the same position, and a locating pin that wears against a replaceable bushing rather than against the raw plate — each of these turns a "might do it wrong" into a "cannot do it wrong." Hardened, replaceable contact points matter here, because an aluminum or mild-steel locator wears, and a worn locator is a fixture that drifts a little more every day without anyone noticing. When the contact surfaces are hardened inserts you can swap, the drift never has a chance to accumulate.
Verifying repeatability is a measurement exercise, not a belief. Run a small number of parts through the fixture — not one, but a spread that covers the tolerance range you expect — and measure the same feature on each assembled result. If the spread of those measurements is tighter than your requirement and stays there across a couple of operators, the fixture is repeatable. If the spread is wide, or it changes when a different operator runs it, the fixture is handing you a hint about where the variation lives, and that is worth more than a fixture that happens to measure well once. The fixture is only as repeatable as its worst contact point, so check the locators and clamps as carefully as you check the parts.
Designing the fixture so it can actually be built
An assembly fixture is a machined part in its own right, and it lives and dies by the same rules as the parts it holds. The locator bores have to be positioned to the same accuracy you are trying to achieve in the assembly, or the fixture quietly adds its own error on top of the part tolerances. That is why the fixture plate should be machined, not welded up from hand-aligned stock, and why the locating features should be put in on the same machine in a single setup where possible — the less the fixture's own features drift relative to each other, the less error it contributes to every assembly that runs through it.
The material choice follows the duty. An aluminum fixture plate is light and easy to machine, and it is fine for low-load or prototype assemblies where the clamp forces are modest and the volumes are short. But where the fixture will run thousands of cycles, or where the clamps bear hard against the locators, steel or a hardened insert on the contact points is the right call, because the wear lands on the fixture instead of on the parts. The cost difference is small against a fixture that has to be rebuilt because its aluminum locators wore oval in six months. Design the contact points as replaceable hardened inserts, and the fixture becomes something you service rather than something you scrap.
Finally, think about how the part comes out as much as how it goes in. A fixture that locates and clamps perfectly but leaves no room to reach the fastener, the adhesive line, or the inspection point is a fixture the operator has to fight. Leave the access — clearance for the tool, visibility for the check, room for the assembled part to lift out without scraping a fresh surface — and the fixture does its job without adding a second, slower, error-prone job on top of the first one.
- Locate on the machining datum — reference the same feature the part was cut to.
- Use three points on the seating plane — a fourth pad makes a real part rock.
- Pair a round pin with a diamond pin — fix position with one, rotation with the other.
- Clamp perpendicular, directly over a locator — straight-line force, no side load.
- Size clearance to the worst-case stack — design for the biggest real part, not the nominal.
- Add back support under the clamp — squeeze against something rigid, not through a thin wall.
- Make contact points hardened and replaceable — wear the insert, not the plate.
- Prove repeatability with a measured spread — run a tolerance spread, not one good part.
"An assembly fixture fails for the same two reasons every time: it locates on the wrong feature, or it was designed for the nominal part that never actually arrives. Match your locators to the machining datums, add up the tolerance stack and clear for the worst case, and make the contact points replaceable so the fixture cannot drift. Do those three things and the fixture stops being a source of variation and starts being the thing that holds your alignment exactly where you put it."
An assembly fixture is a small machine with one job, and like any machine it is only as good as its weakest contact. Get the datums matched to the machining, the clamps pushing straight into support, the clearance sized to the real parts, and the contact points built to be serviced, and the fixture will align the hundredth assembly as well as the first. Skip any one of those and the fixture does not hold your tolerances — it just holds your mistakes in place.
If you are building an assembly fixture for a multi-part product and want the locators, clamps, and clearances worked out before the first plate is cut, send the assembly over and we will lay out the fixture and machine it to the same tolerances the parts were made to. See how we handle fixture and jig machining and multi-setup alignment, or get a quote on the fixture.