How to Define CNC Machining Tolerances That Balance Precision, Manufacturability, and Cost

Quick Answer: CNC machining tolerances should be defined by functional requirement, not by what the machine can theoretically achieve. Standard production CNC machining holds ±0.050–0.100 mm reliably across batches; precision machining achieves ±0.010–0.020 mm with controlled process; tight tolerances of ±0.005 mm or tighter require special conditions (thermal control, CMM inspection, high-end machines) and are not scalable in normal production. The cost relationship is non-linear: tightening from ±0.050 mm to ±0.010 mm can increase machining cost by 2–5×. The correct tolerance is the loosest that meets the functional requirement — not the tightest the machine can produce. The two most common tolerance design failures are over-tolerancing non-critical features (adding cost without benefit) and ignoring tolerance stack-up across assemblies (causing assembly failure even when all individual parts pass inspection).

Why Tolerance Decisions Determine Manufacturing Outcomes

Every tolerance specification on a CNC-machined part is simultaneously a functional claim, a manufacturing constraint, and a cost commitment. Understanding these three consequences simultaneously is what separates tolerance specifications that work in production from those that produce scrap, assembly failures, and cost overruns.

The functional claim: A tolerance specification states that the feature will function correctly anywhere within the specified range. For a bearing seat specified at H7 (e.g., 30 H7 = 30.000 to 30.021 mm), this is a claim that any bore dimension within this range will produce acceptable bearing performance. If the tolerance is set too wide, the dimensional variation spans both acceptable and unacceptable functional outcomes.

The manufacturing constraint: The tolerance must be achievable by the machining process, material, and setup planned for production. A tolerance of ±0.005 mm specified for an aluminum part with a thin wall (1.5 mm) is not achievable in batch production because thermal expansion, cutting force deflection, and part-to-part residual stress variation will produce dimensional variation that approaches or exceeds this range without extraordinary process controls.

The cost commitment: Machining cost scales non-linearly with tolerance tightness. Cycle time increases because tighter tolerances require slower feeds, lighter depths of cut, and additional finishing passes. Scrap rate increases because the acceptable variation window is smaller relative to the process’s natural variation. Inspection cost increases because tighter tolerances require higher-resolution measurement systems and longer inspection time per part.

Realistic CNC Machining Tolerance Capability

The critical distinction is between what a machine can achieve once in a controlled setup and what a process can achieve repeatedly across a production batch. A high-end CNC machining center may demonstrate ±0.003 mm positioning accuracy in a calibration test. In production, with tool wear over a 200-part run, thermal expansion during a 4-hour batch, and part-to-part material variation, the process capability for that same machine may reliably support only ±0.008–0.012 mm on a steel part at typical geometry.

ISO 2768 General Tolerance Standards

When a CNC drawing specifies “ISO 2768-m” (medium class), it establishes default tolerances for all untoleranced dimensions:

• Linear dimensions 0.5–3 mm: ±0.10 mm
• Linear dimensions 3–30 mm: ±0.20 mm
• Linear dimensions 30–120 mm: ±0.30 mm

ISO 2768-f (fine class) applies tighter defaults: approximately half the medium-class values for each range. These general tolerances apply to all features not individually toleranced on the drawing. Specifying ISO 2768-m does not mean all features are held to the same tight standard — it means features without explicit callouts are held to the general tolerance class.

How Functional Requirements Drive Tolerance Specification

Bearing Fits

Rolling bearings require controlled dimensional relationships between the bore and shaft — either interference (press fit, where the shaft is larger than the bore) or clearance (slip fit, where the bore is larger than the shaft). ISO 286 defines the standard fit system:

H7/g6 (sliding fit, common for precision shafts): H7 bore (+0 to +21 µm for 30 mm nominal) with g6 shaft (-9 to -20 µm for 30 mm nominal), producing 9–41 µm clearance. This produces a shaft that slides freely without perceptible play.

H7/p6 (press fit, common for bearing outer ring): H7 bore with p6 shaft (+15 to +28 µm for 30 mm nominal), producing 6–28 µm interference. The bearing outer ring is pressed into the housing with an interference that prevents rotation in the housing under operating loads.

Using the wrong fit produces immediate failures: a clearance fit where press fit is required allows the bearing to spin in the housing (damaging the housing bore); a press fit where clearance fit is required deforms the bearing ring, reducing bearing life and potentially preventing assembly.

Sliding and Motion Fits

Surfaces that slide against each other require controlled clearance — large enough to prevent seizing under thermal expansion and oil film requirements, small enough to prevent excessive play that causes positioning error or vibration. Typical design clearances:

• Linear slides (precision): 0.010–0.025 mm
• Rotary motion (standard): H7/e8 or H7/f7 depending on speed and load
• Reciprocating motion with lubrication: 0.020–0.050 mm per side

For sliding fits, surface finish is equally important to dimensional tolerance. Ra > 1.6 µm on a sliding surface accelerates wear even within dimensional tolerance.

Sealing Surfaces

Hydraulic, pneumatic, and fluid sealing surfaces fail through two mechanisms: dimensional gaps that exceed the seal’s compression range, and surface roughness that produces leak paths between surface peaks. Tolerance for sealing is therefore two-dimensional:

• Flatness: ≤0.010–0.020 mm for metallic face seals and O-ring groove mating faces
• Surface finish: Ra ≤ 0.8 µm for dynamic seals; Ra ≤ 1.6 µm for static seals

A sealing surface that meets dimensional tolerance but has Ra 3.2 µm will leak. Specifying only a dimensional tolerance for a sealing surface is an incomplete specification.

Assembly Alignment Features

Dowel pins, precision holes, and location bosses that position mating parts relative to each other require positional accuracy — not just size accuracy. Two holes that are each correctly sized but positioned ±0.5 mm from their nominal locations may produce a pattern that does not accept the mating part’s pin pattern at all. Positional accuracy requires GD&T true position callouts, not dimensional tolerances alone.

Why Tight Tolerances Increase Cost Exponentially

The cost of CNC machining does not increase linearly with tolerance tightness — it increases exponentially as tolerance approaches the process capability limit.

Cycle time: To reliably hold ±0.010 mm, the machining process must produce parts where the sum of all variation sources (tool deflection, thermal effects, workpiece fixture compliance, machine positioning error, tool wear) stays within ±0.010 mm. This requires reducing feed rate to reduce cutting force and thermal input, making light finishing passes to bring dimensions to final size under controlled conditions, and waiting between operations for thermal stabilization. A part that machines in 20 minutes at ±0.050 mm may require 40–60 minutes at ±0.010 mm.

Scrap and rework: The process’s natural variation (expressed as the standard deviation σ) does not change because the tolerance is specified tighter. A process with σ = 0.005 mm produces 0.27% out-of-tolerance parts at ±0.015 mm (±3σ). If the tolerance is tightened to ±0.010 mm (±2σ), approximately 4.6% of parts are out of tolerance — a 17× increase in scrap rate. At ±0.007 mm (±1.4σ), approximately 16% of parts are out of tolerance.

Inspection: Standard dimensional inspection with micrometers and digital calipers provides measurement resolution of approximately 0.001 mm — adequate for tolerances down to about ±0.010 mm. Tighter tolerances require CMM inspection, which is approximately 3–5× slower per part and requires controlled temperature (CMM rooms are maintained at 20±0.5°C because steel dimensions change approximately 0.001 mm per 100 mm per 1°C of temperature change).

Secondary operations: When standard CNC turning or milling cannot reliably achieve the required tolerance, secondary operations are added: precision grinding for bore and shaft diameters requiring ±0.003–0.005 mm; honing for cylindricity; lapping for flatness. These operations add cost steps and process risk.

How Material and Geometry Affect Achievable Tolerance

The machine’s positioning accuracy is only one factor in dimensional capability. Material thermal behavior and part geometric stiffness often limit achievable tolerance more than the machine does.

Aluminum

Aluminum’s coefficient of thermal expansion (~23 µm/m·°C) is approximately twice that of steel (~12 µm/m·°C). For a 100 mm aluminum part, a 5°C temperature increase produces 0.0115 mm dimensional change. This means that a part machined at 25°C and inspected at 20°C will measure 0.0058 mm smaller than when machined — not because the machining was wrong, but because of thermal contraction.

For tolerances of ±0.010 mm in aluminum, temperature variation during machining and inspection must be controlled to ±2–3°C to prevent thermal effects from consuming a significant fraction of the tolerance budget. This is routinely achievable in precision machine shops but is not the standard condition. For tolerances of ±0.005 mm in aluminum, temperature control to ±1°C is necessary.

Stainless Steel and Residual Stress

Stainless steel components — particularly those cut from rolled plate or bar — contain internal residual stresses from the forming process. When material is removed by machining, the stress state changes and the part distorts. This distortion can move features by 0.020–0.100 mm or more depending on part geometry, material history, and the amount of material removed. The distortion is not visible during machining — it appears after the part is released from the fixture.

For precision stainless steel parts (tolerances tighter than ±0.020 mm), a rough machining pass followed by stress relief (either thermal annealing or simply allowing the part to relax on the fixture for an extended period) before the finish machining pass is standard practice.

Thin Walls and Long Shafts

Cutting force deflects the workpiece away from the tool during the cut, then the material springs back after the tool passes. For a thin wall (1.5 mm aluminum), the elastic deflection under typical milling forces can be 0.010–0.030 mm — comparable to or exceeding precision tolerance requirements.

For a shaft with L/D ratio (length-to-diameter) above approximately 5:1, the shaft deflects under cutting force, producing a bowed profile instead of a straight cylinder. Above 10:1 L/D, tailstock support (for turning) or steady rest support (for long turning or grinding operations) is required to achieve consistent tolerances.

DFM response to geometric instability: Design minimum wall thickness appropriate to the material and tolerance requirement (1.5–2.0 mm for aluminum at ±0.020 mm; 2.0–3.0 mm for steel). Use staged machining (rough to within 0.3–0.5 mm, then finish) so that deflection during rough cutting doesn’t affect finish dimensions. For long shafts, design support features or accept that precision tolerances require additional process steps.

GD&T and Tolerance Stack-Up

Why Size Tolerance Alone Is Insufficient

A size tolerance specifies that a feature’s physical dimension (diameter, length, width) falls within a range. It says nothing about where that feature is located, how it is oriented relative to other features, or whether its geometry is within a functional shape requirement. Two holes drilled to the correct diameter but 0.5 mm off their nominal position pattern will fail to accept the mating pin pattern regardless of their size accuracy.

GD&T (per ASME Y14.5-2018 or ISO 1101) provides the additional controls that size tolerances cannot:

True position: Controls the location of a feature (hole, boss, slot) relative to a datum reference frame. Expressed as a cylindrical tolerance zone (ø0.050 mm, for example) within which the feature’s axis must lie.

Flatness: Controls the variation of a surface from a perfect plane, regardless of its nominal dimensions. A sealing face specified at flatness 0.010 mm must have all surface points within a 0.010 mm thick parallel plane band.

Concentricity and total runout: Controls the relationship between the axis of a feature and a reference axis. For rotating shafts and bearing seats, runout directly affects vibration, noise, and bearing life.

Perpendicularity and angularity: Controls the angle of a feature relative to a datum. A mounting face that is 0.5° out of perpendicular to its bore produces angular misalignment in assembly.

Tolerance Stack-Up Analysis

When a product consists of multiple machined parts assembled together, the dimensional variation of each part contributes to the variation of critical assembly dimensions. This accumulation is tolerance stack-up.

Worst-case stack-up: Assumes all tolerances are at their worst limit simultaneously. For an assembly with four gaps each tolerated at ±0.020 mm, worst-case stack-up is ±0.080 mm. If the functional requirement for the total gap is ±0.050 mm, the design fails in worst case.

Root Sum Square (RSS) stack-up: Assumes tolerances are statistically independent and normally distributed. RSS combined variation = √(T₁² + T₂² + T₃² + T₄²). For four tolerances of ±0.020 mm: RSS = √(4 × 0.0004) = ±0.040 mm. This is less conservative but statistically represents approximately 99.7% of assemblies.

Stack-up analysis is essential before finalizing tolerances on any multi-part assembly. If worst-case stack-up exceeds the assembly’s functional requirement, the options are: tighten individual component tolerances (increasing cost), redesign the assembly to reduce the number of tolerance contributors, or accept statistical methods (RSS) after validating process capability.

Process Capability and Production Repeatability

Cp and Cpk

Process capability indices quantify whether a manufacturing process can consistently produce parts within specification:

• Cp = (Upper Spec Limit – Lower Spec Limit) / (6σ). Measures the ratio of tolerance range to process variation, assuming the process is centered.
• Cpk = minimum of [(USL – mean) / 3σ, (mean – LSL) / 3σ]. Adjusts for process centering — a process offset toward one limit has lower Cpk than a centered process with the same variation.

Target values for production:

• General production: Cpk ≥ 1.33 (process variation ≤ 75% of tolerance band, ~64 ppm defect rate for centered process)
• Critical features: Cpk ≥ 1.67 (process variation ≤ 60% of tolerance band, ~0.6 ppm for centered process)

A process with Cpk = 1.0 produces approximately 0.27% defects — 2,700 ppm — which is typically not acceptable for production. A process with Cpk = 0.8 produces approximately 2.5% defects.

The tolerance-capability matching principle: Before specifying a tolerance in production, confirm that the process Cpk for that feature and process combination meets the minimum target. Tolerances set without this confirmation are hopes, not specifications.

Statistical Process Control (SPC)

SPC monitors process output over time using control charts (X-bar and R charts, individuals and moving range charts) to detect process drift before it produces out-of-specification parts. For tight-tolerance production machining:

• Control limits are established from actual process data — not from the specification limits
• Out-of-control signals (trend, shift, single point beyond 3σ) trigger investigation and correction before scrap accumulates
• Tool offset compensation based on in-process measurement at specified intervals prevents dimensional drift from accumulated tool wear

Without SPC on tight-tolerance features, problems are discovered at final inspection after scrap has already been produced.

Secondary Process Compensation: Anodizing and Coating Effects

When CNC-machined parts undergo secondary processes (anodizing, electroplating, PVD/CVD coating, electroless nickel), the coating adds material to all surfaces. This systematically changes critical dimensions:

Anodizing (Type II, 10–25 µm typical): Approximately 50% of the anodize layer thickness grows outward (increasing external dimensions, decreasing internal dimensions). A 10 µm anodize layer increases a shaft diameter by ~10 µm and decreases a bore diameter by ~10 µm. An H7/g6 fit that was assembled successfully before anodizing may have zero clearance or interference after anodizing.

Hard anodizing (Type III, 25–75 µm): More significant dimensional impact — 75 µm hard anodize can produce 37–50 µm dimensional change on each surface, enough to close fit clearances entirely.

Design response: Dimensions that must meet tolerance after anodizing must be machined to pre-anodize dimensions that account for the coating addition. For a 30.000 mm shaft that must finish at 30.000 mm after 15 µm hard anodize (approximately 7.5 µm radial), the machining target before anodizing is 29.985 mm. This pre-machining compensation must be documented in the process plan and drawing, not left to interpretation.

DFM Checklist for Tolerance Optimization

Feature tolerance assignment:

• Identify all features that directly affect function (bearing seats, sealing surfaces, alignment pins, critical fits) and apply functional tolerances to these
• Apply standard tolerances (ISO 2768-m or similar) to all non-critical features (cosmetic surfaces, clearance holes, general dimensions)
• Do not apply the same tolerance level uniformly to all features

Geometry for machinability:

• Internal corners should have radius ≥ tool radius (typically R0.5–2.0 mm minimum); zero-radius internal corners require secondary EDM operations
• Wall thickness ≥ 1.5 mm for aluminum, ≥ 2.0 mm for steel at precision tolerance requirements
• Avoid deep narrow pockets (depth > 4× width) for features requiring tight dimensional control

Datum strategy:

• Primary datum should be a functional surface that contacts the mating part in assembly
• Datum surfaces should be stable (accessible, non-flexible, machined in the same setup as the features they reference where possible)
• GD&T callouts reference datums; ensure datum selection is consistent between design, machining fixture, and CMM inspection setup

Assembly tolerance:

• Perform worst-case or RSS stack-up analysis on all critical assembly clearances before finalizing component tolerances
• Verify that the planned process Cpk for each tolerance can meet the specification with adequate margin (Cpk ≥ 1.33 minimum)

Secondary process compensation:

• Identify all features where anodizing, plating, or coating adds material to critical surfaces
• Calculate dimensional change from coating thickness and apply pre-machining compensation
• Define whether the tolerance applies before or after coating on the drawing

Key Takeaways

• The correct tolerance is the loosest that meets the functional requirement: tighter tolerance adds cost exponentially (2–5× from ±0.050 mm to ±0.010 mm) without performance benefit on non-functional features.
• Standard CNC machining holds ±0.050–0.100 mm reliably in production: precision machining achieves ±0.010–0.025 mm with controlled process; tolerances tighter than ±0.005 mm require special conditions not scalable in normal batch production.
• Tolerance stack-up analysis is mandatory for multi-part assemblies: parts that each individually pass inspection can produce assembly failures if the combined worst-case variation exceeds the functional gap requirement.
• Material behavior limits tolerance more than machine capability in many cases: aluminum thermal expansion at ±5°C temperature variation produces 0.011 mm dimensional change per 100 mm — exceeding a ±0.005 mm tolerance budget without any machining error.
• GD&T controls functional relationships that size tolerance cannot: true position, flatness, concentricity, and perpendicularity are required for features where location, orientation, or form — not just size — determines assembly or performance success.
• Process capability (Cpk) must be verified before specifying production tolerances: a tolerance specified without confirming Cpk ≥ 1.33 for that feature and process is a prediction of yield loss, not a manufacturing specification.
• For OEM procurement teams: drawings that specify tight tolerances on all features uniformly — rather than applying tight tolerances selectively to functional features — produce quotes with hidden cost premiums across every feature, not just the ones that need precision. A DFM tolerance review before issuing drawings for quote consistently reduces machining cost by 15–40% without changing functional performance, by identifying and relaxing non-critical tolerances.

Frequently Asked Questions

What is the standard tolerance for CNC machining?

The practical standard production tolerance for CNC machining is ±0.050–0.100 mm, which aligns with ISO 2768-m (medium class) for general linear dimensions. This tolerance range is achievable across most materials, geometries, and CNC machining setups without special process controls, and provides reliable production stability with economical inspection. Precision machining at ±0.010–0.025 mm is achievable with controlled cutting parameters, stable fixturing, and appropriate tooling, but increases machining cost by approximately 1.5–3×. Tolerances tighter than ±0.005 mm require temperature-controlled environments, CMM inspection, and multiple finishing passes or secondary operations — these are achievable in specialized shops but not scalable to normal batch production economics.

Why do tight CNC machining tolerances cost so much more?

Tight tolerances increase cost through four simultaneous mechanisms. First, cycle time increases because tighter tolerances require slower feed rates (to reduce cutting force and thermal input), lighter depths of cut, and additional finishing passes — a part at ±0.010 mm may take twice the machining time of the same part at ±0.050 mm. Second, scrap rate increases exponentially: if a process has natural variation of σ = 0.005 mm, tightening the tolerance from ±0.015 mm (±3σ) to ±0.008 mm (±1.6σ) increases the percentage of out-of-tolerance parts from 0.3% to approximately 11%. Third, inspection cost increases because tight tolerances require CMM measurement (3–5× slower per part than manual measurement) in temperature-controlled environments. Fourth, secondary operations like precision grinding or honing may be needed if standard CNC machining cannot achieve the required tolerance reliably, adding cost steps.

How do you choose the right CNC machining tolerance?

Tolerance selection should follow a three-step process. First, identify which features directly affect product function — bearing fits, sealing surfaces, alignment pins, precision clearances — and define tolerances for these features based on the functional requirement (fit class, seal gap, allowable misalignment). Second, apply standard tolerances (ISO 2768-m or similar) to all non-critical features: cosmetic surfaces, clearance holes, general structural dimensions. Third, verify that the planned machining process has sufficient capability (Cpk ≥ 1.33) to produce the functional tolerances reliably across the production batch. The correct tolerance is always the loosest one that meets the functional requirement — not the tightest the machine can achieve. Applying tight tolerances to non-functional features increases cost without improving performance.

What is tolerance stack-up and why does it matter in CNC assemblies?

Tolerance stack-up is the accumulation of dimensional variation across multiple parts in an assembly. Each component has its own dimensional tolerance, and when parts are assembled, these tolerances combine — in the worst case, all tolerances are simultaneously at their worst limit in the same direction. For an assembly with four components each contributing ±0.020 mm to a critical gap, the worst-case stack-up is ±0.080 mm. If the functional requirement for the gap is ±0.050 mm, every part in the assembly will pass individual inspection, but approximately 5–15% of assembled products will fail to meet the functional requirement. Stack-up analysis must be performed before finalizing component tolerances, because the solution — tightening component tolerances, redesigning to reduce the number of contributors, or using selective assembly — must be determined at design stage, not after production has started.

How does anodizing affect CNC machined dimensions?

Anodizing adds material to all exposed surfaces: approximately 50% of the anodize layer thickness grows outward from the original surface, and 50% grows inward (converts base metal). For standard Type II anodize (10–20 µm total thickness), the effective dimensional change is approximately 5–10 µm per surface. For Type III hard anodize (25–75 µm total thickness), the dimensional change is 12–37 µm per surface. This means a shaft-bore fit that was correctly dimensioned for H7/g6 clearance before anodizing may have near-zero clearance after hard anodizing, preventing assembly. The design response is to machine to pre-anodize target dimensions that account for the coating addition — for example, a 30.000 mm shaft that must fit at 30.000 mm after 20 µm hard anodize is machined to 29.980 mm before anodizing. This compensation must be specified in the process plan and drawing, not assumed or left to the machining supplier’s judgment.

Written by the RPS engineering team with 15+ years of precision CNC machining experience producing aluminum, stainless steel, titanium, and engineering plastic components with tolerance requirements from ISO 2768 general to ±0.003 mm precision for aerospace, medical, automotive, and industrial OEM manufacturing. Technical references: ASME Y14.5-2018 (Geometric Dimensioning and Tolerancing), ISO 286 (Shaft and Hole Fit System), ISO 2768 (General Tolerances), Machinery’s Handbook (Fits and Tolerances), Chase K.W. and Parkinson A.R. (Mechanical Assemblies — Tolerance Analysis).

Sourcing Precision CNC Machined Components?

At RPS, we provide DFM tolerance review on all new part programs — identifying over-toleranced features, verifying process Cpk on critical dimensions before production commitment, and providing CMM first-article inspection reports with measured vs nominal data on all precision orders.