A coordinate measuring machine (CMM) is sized first by envelope, then by the ISO 10360-2 accuracy class needed for the tightest GD&T callout on the part, and only afterwards by probe type, software and shop-floor fit [S1][S2].
Selection is dominated by four numbers: the X, Y, Z measuring range, the permissible linear error expressed as E = A + B·L/1,000, the maximum part mass, and the operating environment (temperature band, vibration class) [S2][S5].
Envelope sizing: the "twice the part" rule and where it fails
The classic Hexagon and Canadian Metalworking guideline says to choose a CMM whose X, Y and Z measuring ranges are at least twice the width, length and height of the largest part you intend to measure [S1][S2]. The factor of two is not arbitrary: it leaves clearance for fixtures, probe extensions, stylus reach around features, and rerunning a measurement strategy without collisions, which would otherwise consume 20-40% of the nominal volume in real shops [S1][S2].
Where the rule breaks is on long, thin parts (aerospace skins, wind-tower shells, rail car panels). For these, gantry and horizontal-arm CMMs become the only realistic option because bridge frames would need travel in the 10-20 m range along one axis, a domain where published work area envelopes reach up to 20 m × 6 m × 4 m for full-scale airplane-wing and ship-propeller measurement [S3]. A rule of thumb for stretched parts: spec the long axis at 1.5× the part length and the perpendicular axes at 1.2×, then add fixture clearance, instead of doubling everything [S1][S5].
Accuracy and ISO 10360-2: matching the class to the tolerance
CMM linear accuracy is verified under ISO 10360-2 by measuring five certified lengths, three times each, in nine different positions and directions, producing 135 measurements that must all fall inside the manufacturer-declared error envelope E = A + B·L/1,000, where E is in microns and L is the measured length in millimetres [S2].
The selection logic is mechanical: pick the ISO 10360-2 specification that consumes at most one-quarter of your tightest dimensional tolerance (the 4:1 measurement-system ratio still used by most OEM print drawings), which leaves headroom for fixture repeatability, thermal drift and probe-form errors [S2]. For free-form surfaces, blades, and airfoils, ISO 10360-5 form performance (PFT, PFU) becomes the binding spec, not the linear term [S2].
Structure selection: bridge, gantry, horizontal-arm, portable
Bridge CMMs are the workhorse for parts up to a few thousand pounds and volumes from roughly 500 × 700 × 400 mm to 1,200 × 2,000 × 1,000 mm, with a moving bridge on a granite base that gives thermal mass and vibration damping in a temperature-controlled room [S5]. Gantry CMMs split the moving bridge and the fixed rail along the floor, opening up envelopes from 2 m up to 20 m on the long axis for aerospace skins, body-in-white and large castings, at the cost of substantial floor space, foundation work and stricter 20 ± 0.5 °C room specs [S3][S5]. Horizontal-arm CMMs trade a smaller footprint for articulated access, making them the standard pick for engine blocks, transmission housings and complex castings where internal bores and undercuts must be probed without re-fixturing [S5].
Portable CMMs (articulated arms, tracker-based systems) sit outside the fixed-frame envelope and serve large or immovable parts such as turbine casings, on-ship propeller hubs and body-shop jigs, but they inherit the operator's stability as part of the measurement chain, so ISO 10360-12 governs their acceptance test rather than ISO 10360-2 [S4]. For related large-component selection logic, see climbing formwork criteria for height and load and power trowel sizing for confined sites, where similar envelope-versus-rigidity trade-offs apply.
Probe, scanner and software gates
Once the envelope and accuracy class are fixed, probe and scanner choice is gated by feature access and point density. Touch-trigger probes remain the default for tight-tolerance prismatic features and datum reference frames, while scanning probes and non-contact laser/optical heads are required when the inspection routine needs thousands of points on a free-form surface, an airfoil, or a sheet-metal panel [S4][S5]. A practical decision rule: choose touch probing when the print is dominated by ±0.01-0.05 mm linear and GD&T callouts on <50 features per setup; switch to scanning when the surface area exceeds roughly 0.1 m² per setup or the form tolerance is below 0.05 mm [S2][S4].
Software gates then determine what the CMM can actually do with the data: CAD-based off-line programming, GD&T per ASME Y14.5 / ISO 1101, point-cloud comparison against nominal CAD, and statistical process control (SPC) export. Without native CAD import and direct DMIS / I++ interface support, a CMM turns into a manual comparator, which negates most of the throughput advantage over hand gauges [S1][S4]. For reference on how tolerance loops cascade into downstream equipment choice, the PID controller selection logic for HVAC loops uses a similar spec-down, verify-up hierarchy.
Who should NOT buy a bridge CMM, and when to outsource
A bridge CMM is the wrong answer when (a) the part is larger than roughly 1.2 × 1.8 × 0.8 m, where gantry or horizontal-arm geometries win on access and stiffness; (b) the part is heavy, fixed, or ships out for service (large turbines, on-site jigs, marine propellers), where a portable arm or laser tracker is the only practical CMM; (c) annual measurement volume is below roughly 200 part-hours, where the ROI of a climate-controlled CMM room rarely beats a contract inspection lab plus a portable arm [S1][S5].
For very low-volume or one-off inspection, contract labs running 1.5-3.0 µm class bridge CMMs and large gantries will spec, program and report faster than a first-time buyer can commission its own machine, and they remove the need to maintain a 20 ± 0.5 °C, <0.5 °C/h drift environment, which is a hidden cost on the order of 10-20% of the machine price over a 10-year life for a poorly planned install [S2][S5].
Shortlist logic and a decision matrix
A shortlist for a typical precision-machining shop evaluating CMMs in 2026 looks like this, ranked by the four binding criteria above:
1. Bridge CMM, ~1.5 µm + 3 µm/m class, 700 × 1,000 × 600 mm envelope, touch + scanning probe, CAD direct import: the default pick for job shops measuring dies, moulds, aerospace fittings and medical parts with ±5-25 µm tolerances, installed in a 20 ± 0.5 °C room on an isolated foundation [S1][S2][S5].
2. Gantry CMM, ~3-5 µm + 4-5 µm/m class, 3-10 m X-axis, fixed-rail design: the only sensible pick for body-in-white, aircraft skins, large castings and wind-tower shells where the bridge frame would otherwise be over 4 m tall and unstable [S3][S5].
3. Horizontal-arm CMM, ~5-8 µm class, 1.5-2.5 m reach: the right pick for engine blocks, transmission housings and complex castings where internal bores and undercuts dominate the routine and a vertical bridge would need awkward probe angles [S5].
4. Portable arm / laser tracker CMM, ISO 10360-12 class, 1.5-4.5 m reach: the only option for large, fixed or service-position parts (turbine casings, ship propellers, on-site jigs), and the standard backup for any shop running mixed-scale work [S4].
Across these four, the binding axes are envelope (X·Y·Z m³), ISO 10360-2 E constant in µm, part mass in kg, and installed environment class (20 ± 0.5 °C room vs shop-floor with enclosure). The next step for a serious buyer is to request the OEM's ISO 10360-2 test certificate with five lengths, three repetitions, nine positions, and to verify the B constant explicitly, because the B term dominates accuracy above roughly 100 mm of measured length and is where most low-cost bridges hide their real performance [S2].
Component reference pages worth checking: contour measuring machine, vision measuring machine, and linear guide.