Tolerance
How tight should the tolerance be on an acrylic part?
Three tolerance bands, what each one costs, and why a tight limit on acrylic is meaningless until somebody agrees on a reference temperature.
Tolerance
Three tolerance bands, what each one costs, and why a tight limit on acrylic is meaningless until somebody agrees on a reference temperature.
Almost every acrylic part falls into one of three tolerance bands. Putting a part in a tighter band than it needs is the most common way to pay too much.
| Band | Value | Use it when | Relative cost |
|---|---|---|---|
| General fabrication | ±0.3 mm | The part is visually judged, not assembled to anything precise. Display panels, covers, decorative parts. | Baseline |
| Production fit | ±0.2 mm | The part mates with another part or sits in a frame, and the gap is visible or the assembly must not rattle. | +15–25% |
| Precision fit | ±0.1 mm | Holes must align with a fixed metal frame, or a face must seal against a gasket. | +40–80% |
The cost increases are not linear because each band requires a different amount of care: tighter inspection, slower feeds, temperature-stable machining, and often annealing.
This is the part that catches engineers coming from sheet metal, and it is worth doing the arithmetic once.
Acrylic’s coefficient of linear thermal expansion is roughly 7 × 10⁻⁵ per °C. Mild steel is about 1.2 × 10⁻⁵. So acrylic moves about six to seven times as much for the same temperature change.
A 500 mm acrylic panel machined at 20 °C and measured at 30 °C will read about 0.35 mm longer. A 1,000 mm panel will read 0.7 mm longer. Both are larger than the ±0.1 mm tolerance that was specified.
So the question “can you hold ±0.1 mm on a 1 m part?” has no answer until someone says at what temperature. The practical protocol is to agree a reference temperature — usually 20 °C or the machine’s operating temperature — and measure there.
| Part length | ΔT | Change in length | Comparable to |
|---|---|---|---|
| 200 mm | 10 °C | 0.14 mm | ±0.1 mm tolerance, exceeded |
| 500 mm | 10 °C | 0.35 mm | ±0.3 mm tolerance, exceeded |
| 1000 mm | 10 °C | 0.70 mm | ±0.5 mm tolerance, exceeded |
| 2000 mm | 10 °C | 1.40 mm | Beyond any precision claim |
The second reason a part changes size after it leaves the machine is internal stress, and this one is worse because the movement happens over days or weeks.
Cast acrylic sheet carries residual stress from manufacture. Machining removes material from one side, unbalancing it, and the part relaxes toward a new equilibrium. A flat panel can bow; a pocketed part can twist.
Annealing releases the stress before it can move the part. The part is heated below its softening range, held long enough for the polymer chains to relax, then cooled slowly enough that cooling does not reintroduce stress. For parts with a tight flatness or a sealing face, this is not optional.
The sequence matters: rough machine, anneal, finish machine the critical face, then measure. Doing the finish cut before annealing measures the part in a state it will not stay in.
| Part type | Anneal? | Reason |
|---|---|---|
| Thin flat panel, ±0.3 mm | No | Stress movement is smaller than the tolerance |
| Bent guard, 10 mm+ | Yes | Forming stress plus drilling causes cracks and crazing |
| Sealing face, ±0.1 mm flatness | Yes, before finish cut | Otherwise the face bows after delivery |
| Optical window | Yes | Flatness and refractive consistency both depend on it |
What is realistic in practice, assuming annealed material and an agreed reference temperature.
| Feature size | General | Production | Precision |
|---|---|---|---|
| Under 100 mm | ±0.3 mm | ±0.2 mm | ±0.1 mm |
| 100–300 mm | ±0.3 mm | ±0.2 mm | ±0.1 mm |
| 300–600 mm | ±0.5 mm | ±0.2 mm | ±0.15 mm |
| 600–1200 mm | ±0.8 mm | ±0.3 mm | ±0.2 mm |
| Over 1200 mm | ±1.0 mm | ±0.5 mm | By agreement |
Note the widening with size. It is not laziness — it is thermal expansion plus the practical limits of holding a large flexible part rigid while measuring it.
A drawing where every dimension carries the same tolerance is a drawing nobody has thought about. Five categories are worth marking:
1. Mounting hole pattern. If the part bolts to a fixed frame, this is the dimension that decides whether it assembles. Tolerance the pattern, not necessarily the individual holes.
2. Mating edges. The edge that sits against another part or into a groove.
3. Sealing faces. Flatness matters more than size here, and they are different callouts.
4. Optical apertures. Where the image or light path passes, surface quality and flatness both matter.
5. Overall length only where it stacks. If the part is one of several stacked in an assembly, the overall matters. If it hangs on a wall, it does not.
Everything else can be general tolerance. On a typical part this reduces the precision-machined area to 20–30% of the surface, which is where most of the cost saving comes from.
A real example: a display bracket specified at ±0.1 mm on every dimension, including the outer profile and the decorative cut-outs. Nothing about the part required it — it was copied from a machined aluminium part.
Re-specifying to ±0.3 mm general with ±0.1 mm on the two mounting holes cut the price by about 35% and changed nothing about how the part worked. The bracket had been expensive for two years because nobody had asked why the tolerance was there.
The reverse also happens: a part specified at general tolerance where one hole has to align with a metal frame, so every unit needs rework at assembly. The fix is to tolerance the hole pattern and leave the rest alone.
The question to ask on every dimension is: what happens if this is at the edge of its tolerance? If the answer is “nothing”, it does not need a tight tolerance.
Tolerance without a reference temperature. Unmeasurable and unenforceable. The part passes inspection at the supplier and fails at the customer, and both are correct.
Precision fit without annealing. Measures correctly on delivery, moves over the following weeks. The complaint arrives after the invoice is paid.
Measuring a flexible part unsupported. A 1 m panel measured flat on a bench and measured held at one edge will give different readings. Agree the fixturing as well as the tolerance.
On small features and with an agreed reference temperature, sometimes — but it is not a sensible general claim. A 200 mm part changes 0.14 mm for a 10 °C temperature change, so ±0.05 mm across that length is smaller than the thermal noise. On features under 50 mm it becomes achievable with annealing and a temperature-stable environment.
Two reasons: thermal expansion scales with length, and holding a large flexible part rigid enough to measure it accurately gets harder. Neither is a limitation of the machine — they are properties of the material and the measurement.
Almost always, because it requires slower feeds, more inspection, temperature control and often annealing. The exception is when a tight tolerance lets you skip a secondary operation — for example, machining a locating feature precisely enough that a jig is not needed at assembly.
As a separate callout from the dimensional tolerance, with the measurement method named — for example “0.1 mm flatness across the 300 × 200 mm face, measured on a granite plate at 20 °C”. Without the method named, the number is not testable.
A DFM note comes back with the quote, listing the dimensions that are driving cost and what each one would save if relaxed. It is usually the most useful part of the reply.