Understand Unilateral and Bilateral Tolerances in Machining

• 1
  What unilateral and bilateral tolerances actually control
• 2
  How unilateral tolerances steer the cut in one direction
• 3
  Why bilateral tolerances stay the default on most prints
• 4
  Choosing between unilateral and bilateral tolerances
• 5
  What each callout costs at the machine
• 6
  Unilateral vs bilateral: where each one fits

A nominal dimension is a target, not a part. The tolerance turns that target into a legal size band, and the machinist cuts somewhere inside it. Unilateral and bilateral tolerances describe two different shapes of that band.

Bilateral means the allowance is split evenly around nominal, written as ±. A 20 mm shaft with ±0.05 mm may finish anywhere from 19.95 mm to 20.05 mm. Unilateral means the allowance sits on one side only, written as +0.10/0 or −0.05/−0.15. Same nominal, different legal window.

That difference is not cosmetic. It changes which tool offsets the operator dials in, how the first-article inspection reads, and whether a mating part still fits after plating. Read the callout before you read the nominal number.

One habit saves trouble: convert every callout to an upper and lower limit before you release the drawing. A window from 19.95 mm to 20.05 mm is unambiguous. A ± sign next to a stacked tolerance is not, especially when two departments read the same print.

• 1
  BilateralEqual allowance above and below nominal; the ± symbol.
• 2
  UnilateralAllowance on one side only; both limits share a sign.
• 3
  Limit formWriting both limits removes all ambiguity.

Unilateral tolerances exist because the fit, not the dimension, is what matters. A bore that must accept a press-fit bearing has one useful direction of error. Oversize may still assemble. Undersize scraps the housing. So the band opens upward and stops at nominal: Ø30 +0.021/0.

The machinist reads this as a target at the top of the window. On a bore, that usually means boring to the high limit and checking with a plug gauge. On a shaft, the logic flips: the band opens downward, and the operator aims at the low limit so the part never comes in oversized.

This is also where plating and coating live. Electroless nickel adds roughly 0.005–0.025 mm per surface depending on bath and time. Anodizing grows the part by a fraction of the oxide layer. If you specify a symmetric ± band on a surface that will be coated, half your parts may land out of spec after finishing. A unilateral band with the growth allowance built in keeps the plated part inside the window.

Unilateral is not automatically harder to machine. It is harder to machine when the single-sided band is tighter than the process capability. A +0.010/0 band on a Ø200 mm bore is a different job than the same band on a Ø20 mm bore. Thermal growth, tool wear, and machine geometry all scale with size.

• 1
  Bores and holesOpen the band upward so the part still assembles.
• 2
  Shafts and pinsOpen the band downward so the part still enters.
• 3
  Coated surfacesAdd the coating thickness into one side of the band.

Bilateral is the default because it is symmetric, and symmetric bands are easier to plan around. The machinist targets nominal, and the wear allowance is split both ways. A ±0.05 mm band on a turned diameter gives the operator 0.10 mm of total room, centered on the number they expect to see on the DRO.

Statistical process control also prefers symmetry. When the band is centered, a capable process drifts to either side with equal risk, and control charts read straight. Shift the band to one side and the process mean has to be re-centered, which usually means a deliberate offset in the CAM program or the tool offset table.

Cost follows the same logic. A centered band lets us run standard depths of cut and normal tool-wear compensation. A one-sided band often forces a tighter target, more frequent in-process checks, and sometimes a finishing pass that would not otherwise be needed.

There is a boundary. Bilateral makes no sense where the function is directional. If a shoulder must seat against a mating face, or a slot must clear a tab, the error direction is defined by the assembly. Symmetry just wastes half your allowance on a direction that never helps.

• 1
  Centered processTarget nominal; tool wear splits both ways.
• 2
  SPC friendlyControl limits sit symmetrically around the mean.
• 3
  Not for fitsDirectional functions need a directional band.

Start with the function. Ask what the part does at that surface. If it slides, presses, seals, or seats, the error direction is already decided by the mating part. If it only needs to be close to nominal with no contact logic, a symmetric band is fine.

Then check the process. Our general machining tolerance is ±0.005 mm, and that number is a shop capability, not a promise on every feature. Small features, deep pockets, thin walls, and long bores all move the achievable band. A unilateral callout tighter than the process can hold will come back as a deviation request or a higher price.

Then check the finish. Fine finishes at Ra 0.2–0.8 μm and high finishes at Ra 0.8–1.6 μm are produced with specific tool paths. A very tight unilateral band may require a separate finishing operation, which adds a setup and a fixture touch.

Finally, check the inspection plan. Unilateral bands are usually verified with go/no-go gauges at one limit. Bilateral bands are verified with calipers or a CMM at both limits. The drawing should make that choice obvious, because the inspector will follow what is written, not what you meant.

• 1
  Function firstContact and fit decide the direction.
• 2
  Process secondMatch the band width to what the machine holds.
• 3
  Inspection thirdWrite the band so one gauge can verify it.

Tolerance width drives cost more than tolerance type. A ±0.05 mm band and a +0.10/0 band of the same total width are usually close in price. What changes is where the operator aims and how often they stop to measure.

A one-sided band can add checks. If the target sits at the limit, the operator has no room to drift, so they measure more often. On a 200-part run, that is real time. On a one-off prototype, it is a few extra minutes.

Coating is the expensive trap. If a symmetric band is written on a surface that will be anodized or nickel-plated, the shop either masks the feature, machines undersize, or asks you to revise the drawing. All three add cost or delay. Putting the coating allowance into a unilateral band on the print avoids the conversation.

There is a limit to how far you can push either type. Below our ±0.005 mm capability, the shop needs a specific process plan, and the drawing should say which features carry the tight band. Broad tight bands across a whole part are a common cause of quotes that look high for no visible reason.

• 1
  Width mattersTotal band width drives cost more than direction.
• 2
  Checks add upLimit-targeted bands need more in-process gauging.
• 3
  Coating trapSymmetric bands on coated surfaces cause rework.