In the previous article we made the case that geometric tolerance matters more than dimensional tolerance in robotics. This article is about the manufacturing consequence of that fact.
If what matters is where features sit relative to each other, then the number of times a part gets unclamped is not a scheduling detail. It’s the tolerance budget.
The problem with three setups
Take a typical joint housing: a bearing bore on the front face, a mounting spigot on the back, a cable gland threaded hole on the side, and a bolt circle around the front.
On a 3-axis machine that’s three setups, minimum:
- Clamp on the back, machine the front face, bore and bolt circle.
- Flip, clamp on the front, machine the back spigot.
- Put it in a vise or angle plate, machine the side hole.
Every flip costs time — but that’s the small part. The real cost is that each flip establishes a new datum, and each new datum adds error between the features cut before and after it.
| Setups | Realistic error between features on different faces |
|---|---|
| 1 | governed by the machine — typically under 0.02 mm |
| 2 | re-clamping adds roughly 0.02–0.04 mm |
| 3+ | 0.05 mm and up, and it varies part to part |
That last row is the one that hurts. Not because the average is bad, but because the variation is. An assembly built from parts that each vary a little will sometimes bind and sometimes not — which is exactly what “repeatability problems” look like from the outside.
What one setup changes
Put the same part on a 5-axis machine and rotate the part instead of re-clamping it. The bearing bore, the spigot and the side hole are all cut in one coordinate system.
Three things improve at once:
- Accuracy between features — no new datum, so no accumulated re-clamping error.
- Time — one clamping, one proving-out, no intermediate handling or waiting.
- Consistency across a batch — the thing robotics actually buys. Every part comes off the same setup in the same coordinate system.
For a part where a bore must be concentric with a spigot on the opposite face, this isn’t a nicety. It’s often the only way to hold the callout without going to grinding.
When it genuinely pays
Five-axis earns its rate in these situations:
- Features on three or more faces that must be true to each other. Concentricity, perpendicularity, true position across datums.
- Angled features. A port or boss at 30° off the main axis is awkward or impossible in three axes without a special fixture.
- Deep cavities with draft. Tilting the cutter keeps a short, rigid tool engaged instead of hanging a long one into a corner.
- Complex structural parts. Robot links and wrist bodies with lightening pockets on several faces.
- Low volume. When you’re making 20, a dedicated multi-op fixture doesn’t make sense, so one setup on a 5-axis is cheaper than three setups plus fixture design.
When it doesn’t
This is the part worth being blunt about, because a lot of money gets wasted here.
- Flat plates with pockets on one face. A 3-axis does this at a fraction of the hourly rate. Five-axis adds nothing.
- Simple brackets. Features on two faces where the tolerance between them is ±0.1 mm — re-clamping error is irrelevant at that number.
- Long production runs. Once you’re making 500, a dedicated fixture amortised over the run wins on cost per part, and the fixture itself guarantees the consistency.
- Parts that are mostly round. That’s a turning or mill-turn conversation, not a 5-axis one.
The rule we keep repeating: the question is never “what’s the best machine for this part” but “what’s the simplest process that can hold these tolerances?” Sometimes that’s the 5-axis. Most of the time it isn’t, and a supplier who says so is saving you money rather than upselling you.
The full argument, including why the most capable machine is often the wrong one, is in Which CNC machine does your part need?.
3+2 versus full 5-axis
Worth knowing the difference, because the prices differ and most robot parts only need one of them.
- 3+2 (positional): the rotary axes position the part at an angle, then lock. Cutting happens with three axes. Handles angled holes, angled faces, and multi-face work where each feature is individually simple.
- Full simultaneous 5-axis: all five axes move at once. Needed for continuous sculpted surfaces — impellers, turbine blades, complex organic shapes.
Almost every robot structural part is a 3+2 job. You need the part rotated, not the cutter dancing. Knowing this saves real money, because shops sometimes quote simultaneous 5-axis work when positional would do.
What to put on the drawing
If you want a supplier to quote one-setup work honestly, tell them three things:
- Which feature is the datum. A, B, C in that order.
- Which relationships actually matter. “Bore Ø30 concentric with spigot Ø25 within 0.02” is a complete instruction. “±0.01 on all dimensions” is not.
- That the relationship matters more than the individual sizes. This is the sentence that lets us choose process instead of just following decimals.
Frequently asked questions
Is 5-axis always more accurate than three setups on a 3-axis?
For relationships between features on different faces, yes — because it removes the re-clamping error entirely. For a single feature machined in one setup on either machine, there’s no accuracy difference; there’s only a higher hourly rate.
Can you hold concentricity across a long part on a 5-axis?
Within the machine’s working envelope, generally yes — and considerably better than with multiple setups. Beyond the envelope, the part needs to move regardless, and then it becomes a fixture and datum question again.
Does one-setup machining cost more per hour but less per part?
Often, yes. The hourly rate is higher, but you remove two clampings, two handlings and a proving-out, and you scrap less. On complex parts the total usually comes out lower. On simple parts it doesn’t, which is why the honest answer is always part-specific.
We only need ten parts. Is fixture design even worth it?
Usually not. That’s precisely where one-setup 5-axis beats the “cheaper” 3-axis route: you get fixture-grade consistency without paying for a fixture you’ll use once.
Where we fit
We run true 5-axis alongside 3- and 4-axis in Shenzhen, and we put robot parts on whichever one the geometry actually justifies. If your part is a 3+2 job we’ll quote it as one, not as simultaneous 5-axis — and if a 3-axis can hold your tolerances we’ll say that, even though it’s the cheaper quote.
Send the STEP and the PDF and we’ll tell you what your part genuinely needs.