engineering solutions / high precision

High Precision Components Machined to Tolerances You Can Verify

A tolerance on a drawing is a claim about the part. It is not a claim about the process that made it. XS Precision machines high precision components for OEM assemblies where the fit, the seal or the alignment decides whether the product works — and we report what was measured, with which instrument, against which datum.

tolerance bandhalf-scale 50 mm
+0.005-0.005nom ∅ 12.000
A ±0.005 mm band on a 12 mm bore is roughly one tenth of a human hair, measured at 20 °C. The dots are seven parts from one run — inside the band, and clustered rather than spread across it.
the question this page answers

Not “can you hold it” — “can you hold it on part 200”

Any competent shop can produce one part inside ±0.005 mm. An experienced machinist, a sharp tool, a slow finishing pass and a careful setup will do it. The question that actually matters when you place an order is whether the same number comes back on the two hundredth part, on a Tuesday in August, measured by someone else with a different instrument.

That is a different capability, and it is built out of things that never appear on a machine specification sheet: how the part is clamped, how long it soaks before it is measured, whether the coolant is chilled, whether the finishing pass follows a stress-relief step, and whether anyone is plotting the results. Accuracy is a system, not a machine parameter.

±0.005mm
Working limit on critical metal features when the part is small and rigid, measured at 20 °C. Beyond that, the constraint stops being the machine.
20°C
The reference temperature. Steel moves about 11 µm per metre per °C, aluminium about 23 — a 300 mm aluminium part measured at 25 °C reads roughly 34 µm larger.
1.33Cpk
The point where a process has margin. Below 1.00 it is not capable and needs 100% inspection; 1.67 is the usual gate for precision work.
where the microns actually go

Six things eat the tolerance band before the tool does

On parts under 500 mm, these are the ranges that show up on a real shop floor — not machine specifications. Read them as a budget: if your callout is ±0.005 mm, the full band is 10 µm wide, and workholding deflection alone can consume all of it before a single chip is cut. At XS Precision this budget is written down before the first cut, which is why two quotes for the same print can differ by more than the machining itself.

Error budget
Workholding deflection10–50 µm
Taper or chatter on one face. Short tools, low clamping force, vacuum or soft jaws.
Thermal drift5–20 µm
Size walks as the shift goes on. 20–30 min spindle warm-up, chilled coolant, ±2 °C shop.
Tool wear5–20 µm / 100 pcs
Gradual size change through the run. Probing between operations, or a fixed tool-change count.
Chip recut5–15 µm
Poor finish and size scatter. Through-spindle coolant, peck cycles.
Measurement error2–10 µm
Scatter that appears only at inspection. Soak the part, use the right gauge for the feature.
Machine geometry2–5 µm
Roundness and squareness errors. Laser calibration, ballbar checks.
Precision cylindrical grinding of a steel shaft between centres, the grinding wheel throwing a fan of sparks into the splash guard while coolant is sprayed at the contact point
Precision cylindrical grinding: the last pass decides the size, but the steady rest, the coolant temperature and the wheel condition decide whether the next part matches.
tolerance band decides the whole inspection plan

What each tolerance band actually costs you in inspection

The machining is often not the expensive part. As the band narrows, the instrument, the inspection frequency and the paperwork all step up — and that step is where most of the extra cost sits. Knowing which band a callout falls in tells you what the quote has to include. Every high precision CNC machining order at XS Precision is planned this way: the band on the print sets the instrument, the inspection frequency and the first article package, and all three are priced in from the start rather than discovered later.

≥ ±0.1 mm
Digital calipers
First article plus periodic sampling  ·  Inspection record with actuals
±0.02 – 0.1 mm
Outside micrometers, bore gauges
First article plus event-triggered recheck  ·  Dimensional report with instrument ID
±0.005 – 0.025 mm
CMM or vision system
FAI plus in-process hold points plus final  ·  Full CMM report plus ballooned drawing
< ±0.005 mm
CMM, air gauge or interferometer
100% inspection  ·  CMM report, SPC charts, full FAI package
A bridge-type CMM measuring a machined aluminium housing: the touch-trigger probe with its ruby ball stylus touching the top face of the part clamped on step blocks on a granite surface plate
A bridge CMM with a touch-trigger probe is the reference method for complex geometry: it measures true position and form against a datum structure, not just a length.
the 10:1 rule and the argument zone

If the gauge cannot resolve it, the number on the report is decoration

Digital caliper
Fast, portable, and inadequate below ±0.05 mm.
Resolves0.01 mm (operator-dependent)
Good forNon-critical external lengths, setup checks
Watch outCannot verify true position or form
Micrometer / bore gauge
The workhorse for size on simple features.
Resolves0.001 mm
Good forDiameters, wall thickness, small steps
Watch outMeasures a length, not a geometric relationship
Coordinate measuring machine
Required once position, form or a datum frame is involved.
Resolves0.001 mm and better, at 20 °C
Good forTrue position, flatness, profile, concentricity
Watch outNeeds a stated datum structure to be repeatable

Agree on the decision zone before the parts are made

The gauge-maker’s rule asks for an instrument ten times more accurate than the tolerance; 4:1 test uncertainty ratio is the modern equivalent. ISO 14253-1 takes the next step and treats measurement uncertainty as part of the decision: the specification limit is expanded by the uncertainty, and anything landing in that grey band is arguable — pass at our bench, fail at yours.

This is not a reason to loosen the print. It is a reason to settle the instrument, the datum, the contact condition and the recheck rule while you are still quoting, not after the carton arrives.

Conformance zoneinside the limit by more than U
Uncertainty zonewithin U of the limit — arguable
Non-conformance zoneoutside the limit by more than U
A precision stepped shaft held in a V-block on a granite surface plate, measured with an outside micrometer across a ground diameter, gauge blocks and a dial test indicator beside it
A micrometer across a ground diameter, on a granite plate, with gauge blocks beside it: for a simple size callout this is faster than a CMM and just as reliable.
the sequence that keeps a tight part tight

Rough, relieve, finish, soak, measure

Precision is lost twice: once when the part moves after the cutter leaves it, and once when it is measured at the wrong temperature. Both are avoidable with a fixed order of operations — and neither is fixed by slowing the finishing pass down. XS Precision runs this sequence as a fixed route on tight work, not as a suggestion.

step 1
Rough
Leave stock. Remove the bulk of the material and the bulk of the internal stress in one go.
step 2
Stress relieve
Titanium and nickel alloys release residual stress after cutting; a thin section can move 0.02–0.05 mm.
step 3
Finish
Light, consistent depth of cut with a fresh tool and a spring pass.
step 4
Soak
30–60 minutes at 20 °C before final inspection. No shortcuts.
step 5
Measure
Against the stated datum, with the instrument named on the report.
Machined aluminium and steel parts laid out in rows on a granite surface plate inside a temperature-controlled metrology room, soaking to room temperature before final inspection
Parts soaking on a granite plate inside the temperature-controlled room: a warm part reads big, and the error is the part’s own.
datums and setups

Five axes buy you position. They do not buy you size.

Cutting five faces in one setup removes re-fixturing error, which is often the single largest contributor to a bad position callout. That is a real gain. But size tolerance is governed by the tool, the spindle and the thermal state, so “put it on a five-axis and hold ±0.005” only works for the geometric half of the print.

The other half of the problem is the datum itself. A datum should describe how the part is located in the product, not how it is convenient to hold in a vise. Pick the wrong one and the part passes inspection in the fixture and still sits crooked in the assembly.

One setup, five facesPosition tolerance improves materially; re-fixturing error is removed from the chain.
Size toleranceUnchanged — still tool, spindle and thermal control. Five axes do not sharpen a cutter.
Datum choiceSelect the faces that locate the part in service. A cosmetic edge answers a different question.
Mixed standardsState whether ASME Y14.5 or ISO 1101 governs, and the edition. Mixing them adds ambiguity, not precision.
Drawing vs CADResolve revision conflicts before quoting. The revision that gets measured is the one on the print.
An anodised aluminium optical mounting bracket with a bored through-hole and counterbored mounting holes, bolted to a stainless optical breadboard with a threaded hole grid
An optical mounting bracket on a breadboard: the bore, the mounting holes and the reference face are what actually locate the optics — the outer dimensions do nothing.
same print, different material, different outcome

Which materials hold a tight number and which ones walk

A tolerance that is routine on 6061-T6 can be a struggle on 316L or Ti-6Al-4V. If the material is still open, this table is worth more than another decimal place on the drawing. Tell us the material and the quantity and we will tell you which callouts on a High Precision Components print are realistic for that substrate.

materialbehaviour under the cutterwhat it means for your print
Aluminium 6061-T6Holds size well and moves little after cutting. The easiest substrate for tight work.Best choice when the print allows it
Stainless 316LWork-hardens and pushes the tool, so size drifts faster through a run.Plan tool changes, expect more inspection
Titanium Ti-6Al-4VReleases residual stress after cutting; thin sections can move 0.02–0.05 mm.Stress relief is not optional
Inconel / nickel alloysHeat at the cut, fast tool wear, post-cut movement.Budget the inspection, not just the cutting
Engineering plasticsMove with temperature and humidity, and relax after machining.Needs its own strategy and its own soak
stack-up is a question about the assembly path

Two parts can both be in tolerance and still not fit

This is the failure that costs the most, because every individual report says pass. A plate has its holes within specification. The bracket has its holes within specification. The bolts still will not go through, because the two hole patterns were located from different features.

A stack-up that means anything follows the path through which parts locate one another — not a column of every dimension on the print. Worst-case analysis is right where failure is unacceptable or nothing can be adjusted. Root-sum-square is only honest when there is real process data behind the spread, and when the contributors are genuinely independent.

Plate hole pattern, located from the datum edge±0.05 mm
Bracket hole pattern, located from its own base face±0.05 mm
Position of the bracket base against the plate datum±0.08 mm
Hole-to-bolt clearance consumed by both patterns−0.03 mm
Accumulated at the interface±0.15 mm

Nothing here is out of tolerance on its own sheet. The accumulated ±0.15 mm at the interface is what decides whether the bolts enter freely — and it is visible on the drawing long before it is visible on the line. This is the review XS Precision runs on high precision work before a quote goes out.

Finished precision turned shafts with multiple ground diameter steps, chamfered ends and centre holes laid out on a bench with snap gauges and a steel rule
Ground shafts with sharp shoulders and centre holes: the shoulders are the features that locate the bearing, so they are the ones that carry the tolerance.
a sample that fits is not a capable process

What the capability index is actually telling you

Process capability compares the output of a stable process with the specification limits. That word carries the weight: computed on an unstable process, Cpk is a number that will not predict the defect rate and will not survive a customer audit. Ask what data sits behind the claim, which machines were sampled, and whether the process was in control.

Cpk < 1.00Not capable.  ·  100% inspection plus corrective action before production continues.
Cpk 1.00–1.33Marginally capable.  ·  Higher sampling frequency and SPC monitoring required.
Cpk 1.33–1.67Capable with margin.  ·  Normal control plan, control charts kept live.
Cpk ≥ 1.67Comfortable for precision work.  ·  Suitable for tight bands and reduced sampling by agreement.
the most expensive habit in precision sourcing

Blanket ±0.01 mm callouts buy nothing and cost everywhere

When every dimension on a print carries the same tight callout, a supplier has to quote the whole part at the tightest level, even though most of those dimensions do not touch another part. Put a general tolerance note in the title block — Unless otherwise specified, tolerances per ISO 2768-m — and spend the tight numbers where they do work.

Bearing bores and shaft fits — the fit decides the function.
Sealing faces and gasket lands — flatness, not just size.
Dowel and locating features — these set the assembly path.
Optical and sensor mounts — alignment is a position callout.
Thread engagement and preload surfaces.
Outer panel dimensions that nothing mates with.
Cosmetic edges, fillets and non-functional radii.
Clearance holes and access pockets.
Overall envelope dimensions used only for packaging.
Anything whose tolerance opening by 50% changes nothing.
what leaves with the parts

The report is part of the deliverable

Blank ruled inspection forms and a machined aluminium bracket on a steel desk, with a rubber stamp, a pen and the matching part laid on top of the paperwork
A ballooned drawing, the material certificate and the dimensional report: the three sheets that turn a measurement into evidence.

A number without a method is not evidence. For high precision components the packet that ships with the order is what lets your quality team accept the batch, and what lets you trace a problem back to a machine, a tool or a material lot months later. The first article report is the one that settles the method; everything after it is confirmation.

Dimensional reportActuals per balloon, with the instrument and its calibration ID.
Material certificateGrade and heat/lot, traceable to the mill.
FAI packageFirst article inspection to AS9102 or your own format on request.
SPC dataControl charts for the critical characteristics on production runs.
TraceabilityPart-to-heat-lot and part-to-machine records kept for the agreed period.
last kilometre

A 5 µm part that arrives with a dent is a scrap part

Ground and lapped surfaces, sharp shoulders and thin walls do not survive loose packing. Precision parts ship separated in die-cut foam, sealed against corrosion, with the inspection report travelling with the carton rather than in a separate email.

It is a small line item and it protects everything upstream: the measurement, the thermal soak, the stress relief and the finishing pass all get undone by one impact in transit.

Individual cavities or dividers — parts never touch each other.
Anti-rust packaging with desiccant for sea freight.
Ground and lapped faces protected from contact.
Dimensional report and material certificate packed in the carton.
Carton marked with part number, revision and quantity.
Ground steel shafts and a machined aluminium block held in die-cut foam cavities inside a sealed anti-rust vacuum bag on a packing bench, with a desiccant sachet and anti-rust paper beside them
Precision parts held in die-cut foam cavities and sealed in anti-rust pack-out bags: nothing moves inside the carton.
frequently asked

Questions engineers ask before they send the drawing

On small, rigid metal features measured at 20 °C, ±0.005 mm is a realistic working limit, and it requires temperature control, in-process probing and slower finishing passes to repeat across a run. Tighter than that is possible on single features but it becomes a metrology project. A long, thin or large part at ±0.005 mm is usually not a machining problem at all — it is a thermal and fixturing problem.

Usually, but not in a straight line. Moving from ±0.05 mm to ±0.025 mm is mostly process discipline. Moving from ±0.010 mm to ±0.005 mm adds temperature control, probing, slower feeds and a much higher inspection frequency. In practice the larger cost driver is the inspection, not the cutting.

Ask for the inspection method and the report before you order, not after. For a simple size callout a calibrated micrometer is faster and just as reliable as a CMM. For position, flatness or anything referenced to a datum, you need a CMM and a defined datum structure. If a callout needs a CMM, say so at the quote stage so the inspection time is inside the price.

For position, often yes: cutting five faces in one setup removes re-fixturing error, which is frequently the largest single error source. For size, no. Size tolerance depends on the tool, the spindle and the thermal state, and those do not change because the part is on a five-axis machine.

That pattern points at tool wear or thermal drift. The machine was cold at the start and settled in, or the tool developed a wear land across the run. In-process probing between operations, or a fixed tool-change count, catches both before a part goes out of tolerance.

Titanium Ti-6Al-4V and nickel alloys such as Inconel generate heat, wear tools quickly and release residual stress after cutting, so a thin section can move 0.02–0.05 mm. Stainless 316 work-hardens and pushes the tool, so size drifts faster. Aluminium 6061-T6 is much easier. Plastics move with temperature and humidity and need their own soak and handling.

Yes, and we will flag it. Missing datums are the most common cause of measurement disputes: two shops measure the same feature from different references and reach different conclusions. We will propose a datum reference frame that reflects how the part locates in your assembly and ask you to confirm it before the first cut.

send the drawing

Tell us which callouts actually matter

Upload the drawing and we will come back with a manufacturability read: which tolerances are buildable as drawn, which ones move the price and why, what we would measure them with, and which datum structure we propose to work to. That review happens before quotation, not after. Send it to XS Precision and you get an engineer’s answer, not a price list.

STEP / STP / IGES / PDF / DWG / DXF  ·  engineering review before quotation  ·  response within 24 hours

Request a quote today!