Flex PCB Neutral Axis: What a Simple Strain Estimate Can Miss

UPDATED October 10, 2026

Discuss your flex PCB project →

A flex PCB strain estimate is useful only when its radius, thickness and neutral-axis assumptions refer to the same cross-section. Using the total thickness in a familiar formula can conceal the location of the copper being evaluated. It can also confuse an inside bend radius with a radius measured at the neutral axis.

This article develops a simple geometric estimate and an original sensitivity example. Use it with the flex PCB design guide to document assumptions, not as a pass/fail rule for a finished circuit.

Define the radius before calculating strain

Consider an initially straight strip bent into a circular arc. Assume pure bending without axial extension at a neutral line, plane cross-sections remaining plane and no slip between layers. Let Rn be the radius at that neutral line, theta the bend angle in radians, and y the signed distance from it toward the outside of the bend.

The neutral-line arc length is Rn times theta. The parallel line at distance y has arc length (Rn + y) times theta. Dividing their length difference by the original neutral-line length gives the engineering strain estimate epsilon = y / Rn. This derivation is a geometric model under the stated assumptions; it does not establish a material’s allowable strain.

If the drawing gives an inside radius Ri and the neutral line is a distance zn from the inside surface, then Rn = Ri + zn. Write both quantities in the calculation record. Substituting Ri directly for Rn changes the model even if the drawing and worksheet both use the letter R.

Know when the midpoint assumption applies

For a homogeneous, symmetric rectangular section in the elementary pure-bending model, the neutral axis lies halfway through thickness t. The outer surface is then t/2 from that axis. If Rn is known, the outer-surface strain estimate is t/(2Rn). If Ri is known instead, the same model gives t/(2Ri + t).

A flex laminate contains different materials and patterned conductors. Its geometric midpoint is not automatically the mechanically appropriate neutral axis. Under an ideal linear-elastic, perfectly bonded layered-beam model with no net axial force, the neutral-axis location can be derived by balancing the axial forces: zn = sum(Ei Ai zi) / sum(Ei Ai). Here Ei is each layer’s applicable axial modulus, Ai its sectional area and zi its centroid measured from the same datum. This is a modeling relationship, not a source of missing material properties.

If those inputs or the bonding assumptions are uncertain, label the neutral-axis position as uncertain. Do not insert a guessed modulus merely to produce a precise-looking result. Real patterned regions, local transitions and nonlinear behavior can require a more appropriate analysis.

Check the copper location, not only the outer surface

The following example is deliberately hypothetical. Assume an inside radius of 5.00 mm, a total section thickness of 0.10 mm and a copper region extending from z = 0.065 to 0.085 mm, measured from the inside surface. Compare two assumed neutral-axis positions without claiming that either represents an actual material stack.

The calculation evaluates the farther copper boundary, not just its centroid. In the second case the copper spans the assumed neutral axis, so one part is in compression and another in tension. The smaller listed magnitude follows from the assumed axis location; it is not evidence that a manufacturing change will create that axis or achieve a particular life.

Both hypothetical cases have the same total thickness and inside radius. Their different estimates show what is lost when a worksheet records only those two inputs. Neither percentage is an acceptance limit, and neither predicts a cycle count.

State what is held constant in a comparison

A comparison of two stackups needs more than two strain percentages. State whether the inside radius, neutral radius, outer envelope or applied bending moment is held constant. These are different boundary conditions. The example above holds the inside radius constant and changes the assumed neutral-axis position. A calculation that instead fixes bending moment would need bending stiffness to determine curvature before using the strain relationship.

This distinction matters when a packaging drawing fixes the available space while a material proposal changes thickness. Keeping the same numerical radius in a spreadsheet may accidentally change the physical constraint. Draw the inside and outside surfaces in the enclosure and identify which surface actually contacts a support. If there is no contact defining the arc, explain how the assumed radius was obtained. A free span need not follow a perfect circular shape merely because a CAD rendering uses one.

For a sensitivity study, change one input at a time where that is meaningful and label the exercise as hypothetical. The purpose is to identify inputs that deserve better information. It is not to select the most favorable combination of assumptions and present it as the manufactured result. If several inputs are linked by the actual construction, preserve those relationships when evaluating a proposed change.

Work through the force balance behind a layered model

The weighted neutral-axis expression follows from a specific idealization. At a given curvature, axial strain varies linearly with distance from the neutral axis. For a linearly elastic layer, axial stress is its modulus multiplied by that strain. Multiplying stress by sectional area gives an axial force contribution. Requiring the sum of these contributions to be zero produces the weighted-centroid expression already introduced. The modulus must describe the direction of extension represented by the model.

Take an original dimensionless screening example with two idealized layers. Give both layers the same sectional area. Place their centroids at 0.02 mm and 0.08 mm from the same datum, and assume the second layer has three times the axial modulus of the first. The weighted neutral position is (1 × 0.02 + 3 × 0.08) / (1 + 3) = 0.065 mm. An unweighted average of the two centroids would instead be 0.050 mm. These are invented layer properties, not values for copper, polyimide or adhesive.

The example illustrates the role of stiffness weighting without claiming a practical stackup. Real layer areas must be represented consistently, and a chosen beam section may not capture a patterned circuit’s behavior across its full width. A layer with missing copper cannot be treated as a continuous sheet simply because the material list says copper. Ask the analyst to explain the representation of the actual pattern and the limits of any equivalent uniform layer.

Track the copper boundaries and the bend direction

After selecting a justified neutral-axis model, evaluate both surfaces of each copper region of interest. The distance to a copper centroid alone understates the farthest distance within a finite-thickness conductor. Preserve the sign of strain during calculation, even when the summary table reports only a maximum magnitude. That sign records whether the modeled boundary is on the tensile or compressive side of the bend.

If the circuit bends in the opposite direction, define the geometry again rather than reusing an inside-surface datum without explanation. The surface that was inside may now be outside. Use a cross-section sketch to map the same physical copper boundary into the new coordinate description. A symmetric-looking total thickness does not prove that the copper occupies symmetric positions in the actual section.

For repeated motion, record the strain history assumed for that material location. Moving from flat to a curved state is different from reversing between opposite curvatures, even if one extreme has the same strain magnitude. The elementary estimate can describe those assumed geometric states, but it supplies no fatigue curve or allowable cycle count. Do not convert the calculated strain range into life without an applicable material and construction model supported by evidence.

Identify where the elementary model stops answering the question

The arc model describes a uniform region under its stated assumptions. It does not automatically resolve what happens at a stiffener edge, a cutout, a pad or a local change in width. Mark these features on the route and ask whether the calculated section represents them. Reporting one number for the entire circuit can conceal that the mechanically important location was never part of the calculation.

Similarly, the force-balance expression assumes the layers share the modeled deformation through their bonding. If the design contains intentionally unbonded layers, its interpretation needs review. A fully bonded composite section and several layers able to move relative to one another are different mechanical models. Do not obtain a favorable neutral-axis location by applying the bonded formula to a construction whose load transfer has not been established.

Keep the modeling output attached to a decision. It may justify comparing candidate constructions, identifying a copper boundary for closer study or requesting better dimensional information. To support a product release, connect that analysis to the actual construction, installation geometry and relevant verification plan. This preserves the useful insight of a simple calculation while keeping its scope clear to the next engineer who reads the worksheet.

Keep model output separate from qualification

Minco’s design guide identifies construction, plating, bend geometry and static versus dynamic use among the inputs to forming reliability. Its discussion supports reviewing the whole application rather than treating a thickness ratio as a complete reliability assessment. Minco Flex Circuits Design Guide, Forming section, page 22.

The bend-radius guide helps distinguish installation bending from repeated motion. The bend-radius screening calculator is an early design aid; its result does not supply a neutral-axis model or fatigue qualification for a particular laminated section.

Before release, retain the regional cross-section, radius datum, copper boundaries, material assumptions and the loading conditions omitted from the estimate. Identify where the installed path departs from a smooth circular arc. Then agree representative verification with the responsible design and manufacturing teams. Request a flex construction and bend review with that record so that the calculation and proposed build refer to the same region.

Explore the flex PCB guide and tools →