Acrylic expands at roughly 7×10⁻⁵/K — six times steel. A part toleranced to ±0.05 mm in a temperature-controlled shop can be 0.1 mm out once it leaves. For precision-industry parts, measurement temperature belongs on the drawing before tolerance does.
In precision work the failure is usually environment and mounting, not machining.
A window laid out from the static position gets hit on the first tool change. We set position and size from the machine builder's motion envelope, not from what looks right on the elevation — and that has to be settled before drawings are released, because afterwards it is a new pane.
Internal coolant above 70 bar will crack a 5 mm pane, starting directly in front of the nozzle. That position needs polycarbonate, or 3 mm behind a metal mesh. Thickening acrylic does not fix it.
A finished face pressed onto an aluminium pad picks up marks, and a metal pad also disturbs inductive probes. Acrylic or POM pads, polished on the working face and rounded R1 at the edge, spread the contact stress instead.
An 8 °C swing over 24 hours moves a 600 mm part by 0.34 mm — far beyond the stated tolerance. Precision parts are calibrated at 20 °C, the drawing states the measuring temperature, and an out-of-tolerance reading is corrected for temperature before the part is scrapped.
Precision parts are machined and inspected in the temperature-controlled area, with key dimensions recorded per piece. Capability limits sit on the Optical & precision page.
On a five-axis machine the acrylic part is rarely the precision component — it is the window, the fixture plate or the inspection aid that has to stay put while everything around it moves. That reframes the specification: what matters is stability, not the tightest number on the drawing.
Cast acrylic expands about 70 microns per metre per degree Celsius. On a 600 mm window, a 10-degree shop-to-shop difference is 0.4 mm of movement — more than the tolerance the drawing often demands. So the first question we ask is not "what tolerance" but "measured at what temperature, and compared against what". For anything held to ±0.1 mm we state the measurement temperature on the inspection record, because a number without that is unrepeatable.
Order of operations decides whether a tight part stays tight. Rough machine, leave 1 mm of stock, anneal to release the stress that machining introduced, then finish machine and measure. Skip the anneal and the part will move after the last cut — sometimes 0.15 mm over a week — which is why a first article can pass inspection and a part from the same batch fail it a month later. Every plate we supply for a fixture application gets this sequence.
| Part | Governing requirement | What we hold |
|---|---|---|
| Machine window | Clear of the motion envelope at full travel | Outline matched to the machine model; ±0.5 mm on the opening, no tolerance on clearance |
| Window, coolant side | Seal against 70 bar mist | 8–10 mm cast, annealed, 15 mm minimum from hole to edge, bonded gasket land |
| Vacuum chuck plate | Flatness and repeatable seating | ±0.1 mm over 500 mm after annealing; flatness recorded cold |
| Fixture pad | No witness mark on the workpiece | Soft face or felt overlay, edge broken 0.5 mm |
| Inspection aid | Dimension stable over the shop year | Annealed, 20 mm maximum section, engraved datum marks rather than printed |
For high-pressure coolant there is one more constraint that is easy to miss: acrylic is attackable by some synthetic coolants at a machined edge, so the seal land is specified as a machined face plus a bonded gasket rather than a clamped edge. Where the coolant chemistry is unknown, we run a 24-hour coupon soak before quoting the sheet grade.
Precision acrylic work is priced by machining time and inspection time, not by material. On a typical fixture plate or window: sheet 15–25 percent, machining 30–40 percent, annealing 8–12 percent, inspection and documentation 10–15 percent, packing 5–10 percent. A part that needs two setups costs more than a part with twice the outline and one setup.
The biggest single lever is the number of datums. A design that references everything to one machined face and one edge can be made and checked in a single setup; a design that dimensions from four different faces forces re-clamping and re-datuming, and each re-clamp carries its own error. When we see a drawing with four datum callouts, the first thing we suggest is a simplification — usually the buyer's own metrology gets easier at the same time.
The second lever is flatness versus thickness. A 15 mm plate holds flatness far more easily than a 6 mm sheet of the same footprint, and the cost difference is usually smaller than the scrap risk on the thin option. Where a thin part is genuinely required, we would rather add a machined rib than promise a flatness we cannot hold through shipping.
| Item | What we need | Why it changes the price |
|---|---|---|
| Tolerance map | Which features are critical and which are not | A blanket ±0.1 mm costs more than a map and delivers nothing extra |
| Measurement temperature | The temperature the tolerance is specified at | 70 microns per metre per degree; without it the number is unrepeatable |
| Motion envelope | The machine model or the travel envelope, if it is a window | A window inside the envelope is a crash, not a tolerance issue |
| Coolant and pressure | Fluid type and bar rating | Decides seal land design and whether a soak test is needed |
| Datum scheme | Which face and edge everything is referenced to | One datum means one setup; four datums mean four |
If the part is a replacement for one that drifted, send the inspection record from the original. A dimension that moved after delivery points at annealing, and knowing that saves a sample round.
Key dimensions measured per piece, with temperature and settling time on the record.
On a single part, yes; for a stable batch we quote ±0.1 mm. Beyond that the limit is the material, not the machine: acrylic springs back, moves with temperature and releases stress for 24–48 hours after machining. If ±0.05 mm is genuinely required, the drawing must also fix the measuring temperature, the settling time and the datum — otherwise buyer and supplier each measure a different number.
Polycarbonate takes impact but not solvents: prolonged coolant contact hazes and crazes it, which is harder to live with. Coolant-heavy stations get acrylic on a replacement cycle; impact-risk stations get polycarbonate behind a metal mesh. It is decided per station, not per machine.
Acrylic and POM, usually. Acrylic holds flatness, polishes well and lets you see the workpiece seating, which suits thin sheet. POM resists scratching and damps vibration, which suits heavier cuts. Hole and groove pitch follows the part outline — too dense and the plate loses stiffness, too sparse and thin parts will not hold.
Only by bending, not by machining, and the radius is limited by thickness. A compound curve in acrylic is formed on a heated jig rather than cut, so the practical constraint is a minimum radius of about three times the sheet thickness and a maximum sheet of around 20 mm. Where the geometry cannot be formed, the answer is a flat panel in a machined frame.
Plus or minus 0.2 mm after annealing on a 15 mm cast plate, measured cold on a granite surface. On 6 mm sheet of the same footprint the achievable figure is closer to 0.5 mm, because thin sheet takes a set from handling and from packing. Where tighter flatness is needed on a thin part, a machined rib is more reliable than a promise.
