Flatness Test Granite Surface Plate: How the Reference Plane Works

• 1
  What a Granite Surface Plate Actually Is
• 2
  Why Flatness Matters on CNC Parts
• 3
  Granite Surface Plate Grades and Tolerances
• 4
  Running a Flatness Test on the Plate
• 5
  When a Granite Plate Is Not the Right Tool
• 6
  How the Plate Fits a CNC Quality Chain
• 7
  Flatness Test Granite Surface Plate Procedure
• 8
  Which Grade and Setup Fits the Job
• 9
  Symptoms, Causes, and Fixes

A granite surface plate is a block of natural granite that has been cut, ground, and hand-lapped until one face sits close enough to a true plane to serve as a reference. Inspectors call that face the working surface. Everything else on the plate exists to keep the working surface stiff and stable: the thickness, the ribbing underneath, the mounting points.

The material is usually black granite or pink granite. Black granite is denser and slightly more wear resistant, so it holds up better in a busy inspection room. Pink granite is often chosen for larger plates because it is easier to source in long sections and stays flatter across a long span. Neither choice changes the grade; the lapping does.

Granite is used instead of cast iron for three practical reasons. It does not rust, so a humid shop cannot corrode the reference face. It has low thermal expansion, so a few degrees of room temperature drift moves the surface less. And it is non-magnetic, so a magnetic chuck or a steel part does not pull a burr into the surface.

One point that surprises people new to inspection: the plate is not a measuring instrument by itself. It produces no number. It only gives you a plane that you trust. The instrument on top of it, whether a dial indicator, a height gauge, or a CMM probe, produces the number. If the plate is wrong, every number taken on it is wrong in the same direction.

• 1
  Working surfaceThe lapped face used as the reference plane
• 2
  GradeA flatness tolerance stated in micrometers or microinches
• 3
  Repeat readingHow closely a second sweep repeats the first
• 4
  MountingThree-point or Airy support to avoid twist

Flatness is a geometric tolerance. It controls how far a surface may deviate from a perfect plane, with no datum referenced. On a machined part, flatness usually decides function rather than fit. A hydraulic manifold with a 0.02 mm wave across its sealing face will weep oil under pressure even though every hole is in position. A robot joint flange that is not flat will tilt the bearing race and shorten bearing life.

The tolerance is also hard to inspect honestly. A caliper cannot see flatness, and a CMM reports the plane it fits to the measured points, which hides a bowed surface if you sample too few points. That is why shops keep a granite plate on the floor: the plate gives a physical plane to compare against, and the indicator shows the deviation directly.

In our own machining, flatness callouts show up most on sealing faces, bearing seats, vacuum chucks, optical mounts, and fixture bases. These are parts where the surface does the work. A 4,000 mm maximum processing size helps with the large ones, but the inspection side has to keep up: a plate that is only flat to 20 μm cannot sign off a 5 μm part.

A useful way to think about it: the plate sets the floor of what you can prove. If a drawing asks for 0.01 mm flatness and the plate is good to 0.005 mm, the measurement has meaning. If the plate is good to 0.02 mm, you are guessing. This is why plate grade is a purchasing decision, not a housekeeping detail.

• 1
  Sealing facesWave across the face causes leaks, not misalignment
• 2
  Bearing seatsTilt loads the race unevenly and shortens life
• 3
  Fixture basesA base that rocks re-clamps the part differently each cycle

Grade is a flatness tolerance, not a quality adjective. The common laboratory grades are AA, A, and B, and each one states how far the working surface may deviate from a true plane over the whole area. Grade AA is the tightest and is used for calibration work and master reference plates. Grade A covers most inspection rooms. Grade B is a toolroom plate for setup and marking, not for final buy-off.

Two plates can share a grade and behave differently. A large plate at a given grade is harder to make than a small one, because the tolerance applies over the whole surface. A 1,000 × 1,000 mm plate at Grade A is a more serious object than a 300 × 300 mm plate at the same grade. Always read the tolerance with the plate size.

Repeat reading is the second number worth asking about. A plate can be flat when cold and still not repeat if it rocks on its supports. Before you trust a sweep, run the same path twice and compare the readings. If they differ by more than a fraction of the tolerance, fix the mounting before you blame the plate.

Temperature matters more than most people expect. Granite moves little, but it is not immune. A plate sitting under a skylight in July and a plate in a controlled room at 20 °C will not read the same on a hot part. Let the plate, the part, and the instrument sit together long enough to reach the same temperature before you record anything.

• 1
  Grade AAMaster and calibration work, tightest tolerance
• 2
  Grade AGeneral inspection and final buy-off
• 3
  Grade BToolroom setup, marking, rough checks

The classic method is the sweep. Clean the plate with a lint-free wipe and a mild cleaner, then let it dry fully. Wipe the part and remove any burr with a fine stone. Any chip under the part is a false high point, and at 0.005 mm it takes almost nothing to create one.

Set the part on the plate and mark the path you will follow. Zero the indicator at one end, then move the part or the indicator across the surface in overlapping lines. Record the maximum and minimum readings. The difference is the flatness deviation for that path. Repeat on a second direction at 90° to catch a bow that the first pass missed.

For a plate that needs to be graded rather than used, the check goes the other way. A repeat reading with a known flat straight edge, or a full grid map with an electronic level, shows how the plate itself deviates. That work belongs in a calibration cycle, not in a quick daily check.

A few practical rules cut most of the error out. Keep the indicator tip clean and the stand rigid. Do not lean on the plate while reading. Do not measure a part straight off a warm machine; let it sit. And write down the room temperature next to the numbers. Six months later, that note explains a disagreement that would otherwise look like a bad part.

• 1
  Clean both surfacesA chip under the part reads as a high point
• 2
  Overlap the pathsMissed strips hide local waves
• 3
  Run two directionsA bow along one axis can vanish across it
• 4
  Record temperatureExplains drift when a part is rechecked later

A granite plate is a reference plane, and it can only measure what sits on it. Very large parts that cannot be placed on a plate need a different approach, such as a laser tracker, a portable arm, or a level-based grid map on the machine itself. A 2 m frame casting is not going onto an inspection plate in one piece.

Very small features are another limit. Flatness on a 3 mm sealing land is better checked optically or with a profilometer trace, because the indicator contact and the part edge interact. The plate is still the setup surface, but it is not producing the number.

Shape also matters. A plate check measures deviation from a plane, not roundness, cylindricity, or position. Those need their own instruments and their own datums. Engineers sometimes send a print with a flatness callout that is really controlling a fit, and the plate will show the surface is fine while the assembly still binds.

Finally, the plate does not remove the need for a calibrated instrument. A plate with a current calibration sticker and an uncalibrated indicator is a calibrated plate doing nothing. Keep both in the same calibration cycle, or the record will not survive an audit.

• 1
  Too large to placeUse a portable arm, tracker, or level grid
• 2
  Too small to contactUse optical or profilometer methods
• 3
  Wrong characteristicRoundness and position need their own datums

On the machining side, flatness is created before it is measured. Fixture rigidity, clamp sequence, and tool wear all push a face out of plane. On a five-axis part, a thin floor or a long unsupported wall can spring after unclamping, so the flatness you measure on the plate may be worse than what the machine produced. That is a fixturing problem, not a measurement problem.

We run 127 high-precision CNC machines across three wholly-owned plants, including 16 simultaneous 5-axis machining centers and 16 mill-turn centers. Parts are inspected before shipment, with raw material checks, in-process monitoring, and final inspection. Measurement reports are available on request, which is where the granite plate earns its place in the record.

Materials change the picture. Aluminium 6061 and 7075 move more with heat than steel, so a plate check right after machining can read high. Titanium TC4 and 17-4PH stainless spring back differently after clamping. For plastics such as POM or PEEK, flatness is often a stress-relief question rather than a machining question.

The plate also supports the front end of a project. When a customer sends a print with an unclear flatness callout, we can measure the first article on the plate and show the actual deviation instead of arguing about the drawing. That is usually faster than a tolerance debate.

• 1
  ClampingOver-tight clamps bow a thin part and it stays bowed
• 2
  Stress reliefMachined surfaces move after material is removed
• 3
  Thermal soakAluminium reads high when measured hot