Room Temperature Measurement Machine Control: Why 20 °C Decides Your CMM Data

• 1
  Key takeaways
• 2
  How air temperature turns into a measurement error
• 3
  Which materials punish a warm inspection room
• 4
  Where the 20 ± 2 °C rule comes from, and where it breaks
• 5
  Setting up a measurement room that actually holds its setpoint
• 6
  Temperature control level against part and tolerance

A bridge CMM does not measure a part in the abstract. It measures the distance between two points in its own coordinate frame, and that frame is built from a granite table, aluminium or ceramic arms, steel guideways, and glass scales. Every one of those materials has a coefficient of thermal expansion, usually written as CTE and expressed in μm per metre per °C. When the air changes, each component changes length by its own CTE.

The scale is the reference. Most industrial glass scales are calibrated at 20 °C and carry a CTE near 8 μm/m·°C. If the scale warms by 4 °C, a 500 mm travel reads about 16 μm long before the part is even considered. That is roughly three times a ±0.005 mm tolerance. The error does not announce itself; the machine simply reports a slightly different number.

The part adds a second error on top. An aluminium bracket at 25 °C is longer than the same bracket at 20 °C by about 115 μm per metre. A steel shaft over the same span grows only about 55 μm. If you measure both in a warm room and compare the readings to a 20 °C drawing, the aluminium part looks oversized and the steel part looks slightly long, and neither is a machining defect.

This is why the room setpoint on a room temperature measurement machine is not a comfort setting. It is the assumption baked into every reported value. Push the air 5 °C away from 20 °C and you have changed the meaning of the data, even if the machine is perfectly calibrated and the probe is clean.

• 1
  Scale errorReference length changes with scale temperature, roughly 8 μm/m·°C.
• 2
  Structure errorGranite, aluminium and steel move at different rates and bend the frame.
• 3
  Part errorWorkpiece CTE applies to the feature you are checking, not to the whole part.
• 4
  Probe and stylusRuby and tungsten carbide styli also move, though the effect is small on short reach.

Material choice decides how much of the room error lands on your drawing tolerance. Aluminium is the worst common case. With a CTE around 23 μm/m·°C, a 200 mm aluminium housing gains roughly 23 μm for every 5 °C above 20 °C. On a ±0.005 mm bore position that is four times the tolerance, so a warm room can make a good part fail and a bad part pass on different days.

Stainless steel and carbon steel sit near 11 to 12 μm/m·°C, roughly half the aluminium figure. A 200 mm steel feature moves about 11 μm over the same 5 °C. Titanium TC4 is lower again at about 9 μm/m·°C, and Inconel is lower still. Invar and some ceramics are used in metrology fixtures precisely because their CTE is near zero.

Plastics and composites break the rule in a different way. POM, PA and PEEK absorb moisture and change dimension over hours, not minutes. A carbon fibre plate can be near zero along the fibre direction and much higher across it, so a single CTE number does not describe the part. For these materials, temperature control helps, but soaking time and humidity control matter just as much.

The practical split is simple. If the drawing tolerance is tighter than 20 μm and the part is aluminium, you need a real temperature-controlled room. If the part is steel and the tolerance is 50 μm, a stable shop corner with a few degrees of drift may be acceptable, provided you measure the part and the master at the same time.

• 1
  Aluminium 6061About 23 μm/m·°C. Treat 5 °C drift as a 115 μm/m error.
• 2
  Steel 1045About 11 μm/m·°C. Half the aluminium risk on the same length.
• 3
  Stainless 316About 16 μm/m·°C. Between aluminium and carbon steel.
• 4
  TC4 titaniumAbout 9 μm/m·°C. Small thermal error, harder to machine.

The 20 ± 2 °C band is a convention, not a law of physics. ISO 1 defines 20 °C as the reference temperature for geometrical product specification, and most drawings, scales and calibration certificates inherit it. A room held inside that band keeps the uncorrected error small enough that a ±0.005 mm machine can still be trusted on short features. Step outside the band and the machine may still repeat well; it simply reports a length that belongs to a different temperature.

Repeatability and accuracy separate here. Repeatability is the machine hitting the same number on the same part again and again, and it survives a warm room surprisingly well. Accuracy against a 20 °C nominal is what degrades. A shop that only compares parts to each other can tolerate wider drift. A shop that compares parts to a customer drawing cannot.

Feature length changes the calculus. Thermal error scales with the length being measured, so a 30 mm bore position barely moves while a 1,500 mm frame diagonal moves a lot. Two metres of aluminium at 5 °C off nominal is about 230 μm. That is why large gantry machines and large parts are usually the first to get a temperature-controlled room, and why small benchtop work is often fine at 22 °C.

There is a floor on how much control pays. Air conditioning, insulation and a stable foundation cost money and floor space. If the tightest tolerance on the drawing is ±0.05 mm and the longest feature is 50 mm, the thermal error is under 6 μm for aluminium and the control effort is hard to justify. Match the room to the tolerance, not to a habit.

• 1
  Short featuresUnder 50 mm, a few degrees of drift rarely changes the accept or reject call.
• 2
  Long featuresOver 500 mm, every degree matters and gradient control matters more.
• 3
  Tight toleranceBelow ±0.02 mm on aluminium, hold the room at 20 ± 1 °C.
• 4
  Repeat-only workComparing parts to each other tolerates wider swings than comparing to nominal.

Start with the air path, not the thermostat. A split air conditioner mounted on the wall blows cold air across the machine and the part, which creates a local gradient even when the room average looks good. Ducted supply with low-velocity diffusers, or a unit with an inverter compressor that modulates instead of cycling on and off, keeps the air moving gently and the temperature flat. On-off cycling is the most common cause of a 1.5 °C sawtooth that never shows up on a wall thermometer.

Insulate and seal the room before sizing the unit. An inspection room with an uninsulated roller door, a window facing afternoon sun, and a gap under the wall will not hold ±1 °C no matter how large the air conditioner is. Add door sweeps, double-glazed panels or blinds, and a small entry vestibule if people walk in and out. Every opening is a path for humid outside air, and humidity drives rust on granite and steel surfaces.

Soak the part before you measure it. A part that just came off a machining center at 35 °C needs time to reach room temperature, and the soak time depends on mass and material. A thin aluminium bracket can settle in 30 to 60 minutes. A 20 kg steel casting may need several hours. Measuring early is the same as measuring in a warm room: the number is real, but it belongs to a different condition.

Place the machine away from heat sources. Motors, compressors, ovens, forklifts and even a bank of monitors all dump heat into the air. Keep the CMM at least a few metres from any of them and never under a direct diffuser. Record the room temperature and the part temperature at the time of measurement. If a reading is ever questioned, that log is the only way to show the condition under which the number was produced.

• 1
  Use inverter controlModulating compressors hold ±0.5 °C better than on-off units.
• 2
  Diffuse the supply airNever blow cold air directly at the machine or the part.
• 3
  Soak by massSmall aluminium parts 30–60 minutes; heavy steel castings several hours.
• 4
  Log the conditionRecord air and part temperature with every inspection report.