top of page
Search

Flatness Tolerance Machining: How SCPM Holds Planar Surfaces

carystraley
Nov 5
11 min read

Ask a machinist what makes a plate flat and you will get a dozen answers, and most of them skip the part that gets rejected at incoming inspection. Flatness tolerance machining is less about the cutter and more about how the part is held, how stress releases as material comes off, and whether the inspection setup measures what the drawing actually asks for. At Summit City Precision Machining in Fort Wayne, Indiana, we see planar surfaces fail for reasons that have nothing to do with spindle condition. This article explains how we control planar tolerance on CNC milling work, where flatness and parallelism get confused, and what we check before a part ships.


Table of Contents



Quick Takeaways


Key Insight

Explanation

Flatness needs no datum

Flatness is a standalone form control. The zone is two parallel planes the surface must lie between, so the callout does not reference any other feature.

Parallelism needs a datum

Parallelism controls orientation to a datum. Checking it on a granite table with a height gage measures parallelism, not flatness, so the setup must match the callout.

Flatness must sit inside the size tolerance

Flatness refines the size requirement. The eMachineShop reference uses a 0.2 flatness zone as its example, and a looser value could let a part exceed its size limits.

Fixture for the released part

Clamp forces that look harmless during roughing can spring thin sections. The finished surface may bow once the part is released, so plan the finish fixture around the free state.

Separate roughing and finishing when stock removal is heavy

Removing most of the stock, letting the part relax, then taking a light final planar pass is a reliable way to avoid form error that appears only after release.

Measure with enough points

A flatness result from a few touch points is an estimate. Use a dense grid on a leveled height gage setup or a CMM scan that covers the full length and width of the face.


What Flatness Actually Controls


Flatness is the simplest-sounding callout on a print and one of the easiest to misread. Most of the time it controls the form of a single planar surface, and the tolerance zone is two parallel planes that the entire surface must fit between. Nothing else on the drawing changes that zone, which is why flatness is applied directly to the surface and does not reference a datum.


Machined steel plate on a granite surface plate being checked with a dial indicator for flatness tolerance machining.

In practice, a flatness callout is a refinement of the size requirement. The eMachineShop reference notes that flatness must be less than the size tolerance on the same surface. If the flatness zone is looser than the size band, a part can be dimensionally in size yet still exceed its allowable envelope, so it is worth checking on every print review.


Surface flatness on sealing faces and primary datums


Surface flatness shows up most often on sealing faces, where a gasket or O-ring needs even contact, and on surfaces that serve as a primary datum. The gdandtbasics reference explains that surface flatness is evaluated in an unconstrained state, unrelated to other features. The part is judged on its own form, not on how it sits against anything else.


A common mistake is to treat a flatness value as a suggestion for the finish cut. On a sealing face, the fastest way to discover that a flatness callout was ignored is a leak at assembly. Writing the callout to the exact face it controls, rather than to a general note, keeps the shop focused on the surface that actually matters.


Feature of size flatness and the derived median plane


When flatness is applied to a feature of size, such as two opposing parallel faces, it controls the derived median plane instead of each face on its own. That plane is built from the center points of line segments across the feature. The gdandtbasics reference notes this use is uncommon, but it is useful when opposing faces must be controlled symmetrically about each other.


This control also limits the bow or bend that the two faces can carry together. Thin, wide plates are the usual case where this matters. It is most relevant on parts where both faces seat against something, because a single-face flatness callout can pass each face while the plate still curls.


Flatness Versus Parallelism in Practice


Flatness and parallelism get mixed up on the shop floor more than anywhere else in GD&T. Both involve planes, both use dial indicators, and both can cause a part to be rejected for the same visible reason. The difference is the reference. Flatness stands alone, while parallelism is always measured relative to a datum.


Why the granite table test measures the wrong thing


Placing a part on a granite surface plate and sweeping a height gage across the top is a common shortcut. The gdandtbasics reference is direct about it: that setup measures parallelism, not flatness. The plate becomes the reference, so the result only tells you how the surface relates to the plate.


To measure flatness on that same plate, the setup has to level the referenced surface to the inspection plane first. Shims, adjustable supports, or a tunable fixture are the usual tools. Without that leveling step, a flat part can look bad and a warped part can look good, depending on how it happened to rest.


What changes when a datum enters the callout


Parallelism is a relationship. It asks whether one surface runs parallel to a datum surface within a zone bounded by two parallel planes. If the datum is not stable, the parallelism number is not stable either, which is why datum establishment matters as much as the machining itself.


Our rule in review is simple: if the drawing does not name a datum, the callout is almost certainly a flatness requirement, and the inspection setup should be built to find form error rather than orientation. Mixing those up is one of the most expensive review comments a shop can receive, because the part passes the wrong test and fails at the customer's receiving dock.


Fixturing and Workholding for Flat Surfaces


Most flatness failures we see start before the cutter touches the part. A blank clamped against a vise jaw, or a fixture plate with a chip under one corner, can spring into a shape under clamp load. The machined surface looks flat on the machine, then shows a bow on the inspection table once the clamp is released.


Clamping force and thin-wall distortion


Thin walls and plates are the most sensitive. Clamp force that is harmless on a thick block can pull a thin section down into the jaws, and the cut face is flat only while the clamp holds. When the part comes out, the surface returns toward its free state.


The practical fix is to spread the clamping load, use more support points than you think you need, and keep clamp pressure as low as the operation allows. Where possible, clamp across the thicker sections and let the thin area sit on support rather than being squeezed into the fixture.


Matching support points to the finished part


Soft jaws and custom fixture plates work best when they support the part the way it will be supported in service. If the part will sit on three feet in the customer's assembly, finishing it on a random set of points can produce a surface that is flat in the fixture but rocks in assembly.


Pro tip: Mark the fixture support points on the setup sheet with the part revision. When the same part is re-run months later, operators can match the original support layout, which removes one of the most common sources of batch-to-batch flatness drift.


CNC Milling Flatness Control


On a CNC mill, flatness comes from three things working together: the cutter, the toolpath, and the amount of stock the finish pass has to remove. A machine can hold excellent positioning and still produce a surface that is out of flat if the cutting conditions push the part around. Good milling strategy is mostly about keeping the finish pass light, stable, and repeatable.


CNC milling machine face-milling a flat aluminum workpiece held in a fixture to control planar surface flatness.

Face mill selection and runout


Face mill runout shows up as a repeating pattern on the surface, a light ripple that a gage can detect even when the surface looks clean. Insert seating, holder condition, and spindle taper cleanliness all affect runout. A face mill that cuts well at roughing can still leave a finish pattern that fails a tight flatness callout.


For critical planar work, verify runout before the finish operation and do not assume that a set of inserts used for roughing is fine for the final pass. A fresh finishing insert in a reliable holder costs far less than a scrap plate and a second setup.


Finish pass depth and stepover


Heavy finish passes deflect thin walls and generate heat, and both move the surface. A light finish cut with a controlled depth of cut tends to hold form better than a single aggressive pass that looks efficient on the cycle time. Stepover on a face finish should be chosen so the scallop height sits well inside the tolerance zone, not just barely within it.


In practice, a planar finish that looks clean under the machine light can fail a CMM grid because the cusp pattern and a slight tilt combine. Checking the finish pattern on a representative sample before running a full lot saves a lot of rework.


Pro tip: Run a sample part through the finish pass and measure it on the same setup you plan to use for production inspection before committing a batch. Most surprises come from changes in holding, not from the cutting parameters.


Inspection Methods That Prove Planar Tolerance


A flatness callout is only as good as the measurement that proves it. Two methods are common on our side of the work: a height gage or indicator on a properly leveled setup, and a coordinate measuring machine that compares many measured points against virtual planes. Each has a different failure mode, and both need to be set up deliberately.


A flatness number tells the shop what to measure, but the setup decides whether the measurement means anything.

Height gage and indicator on a leveled setup


The gdandtbasics reference notes that flatness can be measured with a height gage or indicator run across the surface, but only if the referenced feature is held parallel to the granite table. That leveling step is the part most people skip. The setup needs adjustment so the surface can be brought to the inspection plane, and the reading then reflects form rather than tilt.


Use a grid of readings, not a single line across the middle. A few passes along one axis will miss bow in the perpendicular direction, which is exactly where a thin plate tends to deviate.


CMM point grids and virtual planes


Modern CMMs are faster for this measurement because they can construct virtual planes and compare the measured surface profile against them. The gdandtbasics reference describes it as a 3D measurement, so points have to cover the length and width of the part to confirm the whole surface is in tolerance. A sparse touch pattern can pass a part that is out in the middle.


Our CMM programming covers the scan pattern for each planar face, including points near the boundary, because edge roll-off is easy to miss when the scan stops short of the edge. Report the deviation value and the point count, not just pass or fail, so the next setup can be corrected with real data.


Designing Planar Tolerances That Hold Up


Good planar tolerances start in design, not in the shop. A callout that asks for a flatness zone tighter than the process can hold repeatably becomes a tolerance the shop chases on every lot. A realistic flatness value tied to function is far more useful than a number chosen out of habit.


Setting flatness below the size tolerance


The eMachineShop reference illustrates the size relationship with a flatness zone of 0.2 between two parallel planes, and notes that a looser flatness value could let the overall size exceed its limits. The practical rule is to set flatness so that form error never consumes the entire size band. Checking that the flatness zone fits inside the size zone is a quick step that catches a lot of print problems early.


Symmetric control of opposing faces


If two faces must mate or seal together, a flatness callout on each face may not be enough. A derived median plane control can address the combined bow, which matters on flanges and clamp plates. Designers sometimes add a parallelism callout on top, but that adds a datum and a second inspection, so it should only appear when function requires it.


Be specific about which surfaces are functional. A note that says all surfaces must be flat gives the shop no way to prioritize fixturing or inspection time, and it usually leads to over-inspection of cosmetic faces.


Choosing a Method for Tight Planar Tolerances


There is no single right way to hit a tight planar callout. The right method depends on part thickness, material stability, the number of setups, and whether the surface is functional or cosmetic. The table below compares three approaches we weigh against each other during process planning.


Coordinate measuring machine probing a machined part on granite to verify parallelism and flatness callouts.

Approach

Where It Works

Where It Fails

Single-setup face milling with a dedicated finish pass

Thick, stable blanks with moderate flatness callouts and short cycle needs

Thin plates or parts with clamping distortion, where the finish reflects the clamp rather than the free state

Rough, relax, then finish in a separate operation

Parts with heavy stock removal or heat-sensitive material, where stress release changes the form

Parts with tight datum relationships that must be re-established after each operation without careful fixturing

Wire EDM for thin features and fine planar edges

Small, thin, or delicate sections where cutting forces would distort the part

High-volume planar faces where cycle time and per-part cost outweigh the benefit of low cutting force


In practice, the first option is the default when the blank is stable and the callout is moderate. When the part is thin or the form error is driven by clamping, the two-stage approach usually costs less than repeated scrap. Wire EDM is a specialized choice, and we reach for it only when geometry or cutting force makes conventional milling unreliable.


The wrong choice is the one made by habit. Whichever method you pick, the inspection plan has to match it, and a sample run has to confirm the planar result before the batch starts.


Frequently Asked Questions


Does flatness tolerance need a datum?


No. Flatness is applied to an individual surface, and its tolerance zone is two parallel planes the surface must lie between. The zone is unconstrained and related only to itself. Parallelism is different because it always references a datum.


Can I check flatness by placing the part on a granite plate?


Not by simply placing it down and running a height gage. That approach measures parallelism relative to the plate. To evaluate flatness, the referenced surface has to be leveled to the inspection plane first, and enough points are needed to cover the full surface.


How tight should flatness be compared with the size tolerance?


Flatness should be less than the size tolerance on the same surface. If the flatness zone is wide enough that form error can consume the size band, a part can be within its size limits on the print yet still fail to fit. Check that relationship during print review.


Why does a part pass flatness on the machine but fail at inspection?


Most often, the clamping load or the removal of material changed the form after release. Thin sections are the usual culprits. Re-measure after the part has been released and relaxed, and consider a separate finish operation for parts where stress release is a factor.


What is the difference between flatness and parallelism on a sealing face?


Flatness controls the form of the face by itself, which determines how evenly it contacts a mating surface. Parallelism controls how the face is oriented relative to a datum, which matters when the sealing face has to sit square to another feature. Some sealing faces need flatness only, and some need both.


Can 5-axis milling hold planar tolerances on multi-face parts?


Yes. Multi-face setups can reduce the number of times a part is re-fixtured, which helps keep faces related to each other. The planar result still depends on clamping, finish strategy, and inspection, so a 5-axis machine does not remove the need for careful fixturing on thin or distortion-prone parts.


What planar tolerance has caused you the most trouble on your parts, and did the fix come from setup, machining, or inspection?


References


Comments


 Precision Machined Components - Fort Wayne, IN                                                      

Join our Email List

  • facebook
  • youtube

©2020 by Summit City Precision Machining Inc. SCPM. • ISO 9001:2015 Certified

bottom of page