Rigid-Flex PCB Stackup: Flex Core Z-Position and Material Selection

Why "bolting two stackups together" fails

The usual way a rigid-flex stackup gets drawn is this: lay out the rigid section the way you would any multilayer board, drop a flex core into the middle, and window out the prepreg over the flex region. The table is complete, the total thickness adds up, impedance has been calculated. Stackup frozen.

That process skips the one quantity that actually determines how long a rigid-flex board survives: whether the copper inside the flex core sees tension or compression when the board is bent.

A conventional multilayer stackup table is a static bill of materials. Each row is a material and a thickness; get the order right and you are done. A rigid-flex stackup table is not static. The flex section gets bent during assembly, and at that moment every row in the table enters a loaded state. Copper layers farther from the bend neutral axis see more strain. Copper layers sitting on the neutral axis see almost none.

So the real criterion for freezing a rigid-flex stackup is not "is the thickness right, is the impedance right." It is where in Z the flex core's copper lands. That has to be decided in the first step of stackup planning, not after the rigid section is already laid out, because once the flex core's position is fixed, the rigid section's layer count, parity, symmetry and total thickness are all locked with it.

> If you are still deciding whether this design should be rigid-flex at all versus a rigid board plus a separate flex jumper, that is the previous question. See rigid-flex versus flex PCB construction boundaries. This article assumes rigid-flex is already the decision.

Rule 1: the flex core's Z position outranks everything else

Single flex layer: as close to the neutral axis as you can get

If only one flex copper layer has to cross the bend region, the stackup should place that copper as close as possible to the thickness center of the flex section. When the flex section bends, its own neutral axis sits roughly mid-thickness of that section. Modulus differences between coverlay, bonding film and the polyimide base shift it, but it stays near the middle in order of magnitude.

Two consequences follow:

• The coverlays above and below the flex section should be symmetric. Thick on one side and thin on the other pushes the neutral axis toward the thick side, and the copper moves away from it.
• If the flex core is single-sided construction (one polyimide layer, one copper layer), the copper is inherently off-center. Pulling it back toward the middle with symmetric coverlays is the only compensation available at the stackup level.

Two flex layers: symmetry is a requirement, not a preference

With two flex copper layers crossing the bend, one necessarily sits above the neutral axis and one below. During the bend, one is in tension and the other in compression. The only thing you can do here is make the two layers strictly symmetric about the neutral axis: same polyimide thickness, same copper weight, same coverlay construction.

Break the symmetry and the two layers no longer see equal strain. The one that fails first is always the outer one. A single decision in the stackup table — "coverlay A on top, whatever B we have in stock on the bottom" — is enough to reduce the board's bend life to the life of the weaker of the two layers.

Three or more flex layers: go back and change the design

As the number of flex copper layers crossing the bend keeps rising, the outermost layer's distance from the neutral axis necessarily grows, and so does its strain. At this point the question worth asking is: do all of these layers actually have to cross the bend? The common workable alternative is to hold the bend region to one or two flex layers and complete the remaining layer transitions inside the rigid section, rather than bending the whole stack.

SKPCB's rigid-flex capability is up to 20 layers (engineering-confirmed). But that 20 is a total board layer count. It does not mean the bend region can carry 20 flex copper layers. The flex layer count in the bend, the bend radius and the bend-cycle requirement all have to go to engineering review together. You cannot back-calculate them from the total layer ceiling.

Rule 2: bend radius and stackup thickness are interlocked

Bend radius is not something fixed at assembly. It is locked by stackup thickness at the stackup stage.

SKPCB's flex product specification states a bend radius of at least 6× board thickness. "Board thickness" here means the thickness of the flex section, not the total thickness of the rigid-flex board. The relationship runs both ways:

• Mechanical hands you a bend radius, usually dictated by enclosure space, and that puts a ceiling on flex section thickness.
• The flex stackup — layer count, polyimide thickness, coverlay, adhesive or not — determines actual thickness, and therefore the smallest bend radius you can actually build.

The working order should be: get the available bend radius from mechanical, back out a thickness budget for the flex section, then decide flex layer count and materials inside that budget. Doing it the other way around — fill the flex section with every layer the signals want, then ask mechanical whether it bends — usually returns a no.

One more distinction: install-and-stay bending versus dynamic repeated bending. The first is bent once during assembly and then fixed. The second is bent repeatedly over the product's life: clamshells, sliders, live hinges. The two are not in the same league in terms of flex layer count, copper foil type and bend radius. This use condition belongs in the design inputs going to engineering review. It determines feasibility more than any single row in the stackup table.

Rule 3: adhesive versus adhesiveless decides more than bond strength

Flex cores and coverlays come in two material systems, and the difference matters far beyond "does it stick."

Three places where this actually drives the stackup decision:

Thickness budget. The adhesive layer eats directly into the bend radius budget. On designs with a tight bend radius, adhesiveless is often not a performance preference but a thickness necessity.

Interlayer behavior in the transition zone. The rigid-to-flex boundary is where stress concentrates most on the whole board. An adhesive layer here is a soft interlayer that behaves unlike the materials on either side of it. How that gets handled belongs to transition-zone process engineering — see rigid-flex transition zone manufacturing; this article does not repeat it. What matters at the stackup level: choosing an adhesive system means the transition zone construction has to be confirmed along with it. You cannot pick the material first and figure out the transition later.

Z-axis behavior through thermal cycling. A multilayer rigid-flex board goes through lamination and then assembly reflow, several thermal cycles in total. Adhesive, polyimide and epoxy prepreg do not expand alike in Z. The more layers you add, the more that mismatch accumulates. This is one of the root reasons the rigid-flex layer ceiling sits far below the conventional multilayer ceiling.

Coverlay material selection on its own — polyimide coverlay versus photoimageable coverlay, adhesive versus adhesiveless — has been covered for pure flex designs in flex PCB coverlay and polyimide material selection. The extra consideration in rigid-flex is that the coverlay choice feeds back and constrains how many rigid layers you can build, because the coverlay's thickness and behavior in the flex region determine how the prepreg window edge gets filled and butted in the rigid section.

Three material combinations and their boundaries

Rigid-flex stackups that actually get built fall into three combinations.

Combination A: polyimide flex core + epoxy prepreg rigid section (most common)

Polyimide in the flex section, conventional FR-4 cores and prepreg in the rigid section, prepreg windowed over the flex region.

• Fits: the large majority of install-and-stay designs with moderate layer count and conventional multilayer routing needs in the rigid section.
• Boundary: the Z-axis expansion mismatch between polyimide and epoxy accumulates as layers increase; the window edge over the flex region is sensitive to both resin flow and stress.
• Engineering decision points: which Z position the flex core occupies, and where the window edge sits relative to the start of the bend.

Combination B: all-polyimide construction

The entire stack uses a polyimide system, no epoxy prepreg.

• Fits: designs that need higher temperature capability across the whole board, or that require the rigid section to behave materially like the flex section.
• Boundary: substantially higher cost than Combination A; fewer available part numbers; the rigid section's stiffness comes from polyimide thickness rather than FR-4, which is thickness-inefficient.
• Engineering decision point: does the rigid section genuinely need the same material, or is this just avoiding a material interface?

Combination C: hybrid with local stiffening

The flex core runs through the entire board, and rigidity exists only in the local areas that need mounting and assembly, achieved via stiffeners or local build-up.

• Fits: designs with many bend regions, small scattered rigid areas, and an overall thin-and-light requirement.
• Boundary: the boundary between stiffened and unstiffened areas is itself a stress concentration; overall panel size and flatness control get harder.
• Engineering decision point: is the stiffening built into the stackup or attached downstream? Those are completely different process routes and the design inputs have to say which.

All three combinations need to be evaluated against SKPCB's thickness capability (0.2–6.0 mm) and maximum panel size (standard 600×570 mm). The workable thickness range for rigid-flex is normally narrower than for purely rigid boards and needs stackup-by-stackup confirmation.

Trace width and impedance: the flex section does not run the rigid section's numbers

The easiest thing to miss when freezing the stackup is that trace width capability and impedance conditions are two separate sets of numbers for the flex and rigid sections.

SKPCB's minimum trace/space capability comes in three tiers, identical for inner and outer layers: standard 4/4 mil, advanced 3/3 mil, developmental 2/2 mil. The flex product specification separately states a 2 mil minimum trace. Those are manufacturing capability figures, not bend-region recommendations. Trace width in the bend region should be explicitly relaxed relative to the board minimum, and the bend region should be kept clear of thin traces, vias, pads and any discontinuity. That is a design rule question, not a capability ceiling question.

On impedance: the dielectric construction in the flex section (polyimide, adhesive, coverlay) is nothing like the rigid section (epoxy prepreg, cores). When one controlled-impedance net crosses both, each section has to be calculated separately and given its own trace width. Extending the rigid-section width straight into the flex section is the single most common cause of rigid-flex impedance failures.

> For conventional rigid multilayer stackup planning — layer count, symmetry, thickness budget — see 4-layer versus 8-layer multilayer buildup planning. This article only covers what changes once a flex core enters the stack.

Questions to answer before freezing the stackup

Answer these before the stackup table is frozen and you avoid most of the rework cases. Anything marked subject to engineering review should be submitted together with the stackup file.

Use conditions (the prerequisite that governs everything)

1. Install-and-stay bending or dynamic repeated bending? If dynamic, what order of magnitude of bend cycles and at what frequency?
2. What minimum bend radius does mechanical give you? Does an enclosure or connector constrain the free length of the flex section?
3. Is the bend single-direction or bidirectional? Are there multiple bends in the same flex section?

Flex section stackup

1. How many flex copper layers cross the bend region? Can that be reduced to one or two?
2. Is the flex core single-sided or double-sided construction? Has the copper's position relative to the flex section's thickness center been confirmed?
3. Are the top and bottom coverlays symmetric — same material, same thickness, same construction?
4. What is the total flex section thickness? Does it satisfy the ≥6× board thickness relationship against item 2?

Material system

1. Adhesive or adhesiveless flex core? Is the reason thickness budget, cost, or part number availability?
2. Polyimide coverlay or photoimageable coverlay? Adhesive or not?
3. Rigid section via Combination A (epoxy prepreg), Combination B (all-polyimide) or Combination C (local stiffening)?
4. If there is stiffening, is it built into the stackup or attached downstream?

Rigid section and overall board

1. What is the total layer count? Is it within SKPCB's rigid-flex capability of up to 20 layers (specific stackup subject to engineering review)?
2. Is the rigid section stackup symmetric? Has symmetry been rechecked after the flex core was inserted?
3. Nominal board thickness and tolerance? Does it fall within 0.2–6.0 mm?
4. Is the board outline within the standard 600×570 mm panel size?

Electrical

1. Are there controlled-impedance nets crossing from rigid to flex? Has each section been calculated separately for trace width?
2. Has the bend-region trace width been explicitly relaxed relative to the board minimum? Is the bend region clear of vias, pads and copper discontinuities?

Acceptance

1. Acceptance to IPC-6012 Class II or Class III? Do the rigid-flex portions need separately agreed criteria under the IPC-6013 system?
2. Are IATF 16949 process documents required? Are UL-recognized material part numbers required?

For the full statement of producible layer count, trace width, thickness and hole capability, see manufacturing capabilities. For acceptance classes, inspection coverage and traceability, see quality system. The rigid-flex product page and specifications are at rigid-flex PCB, and pure flex at flex PCB.

FAQ

Does the flex core have to sit exactly in the middle of the stackup?

Not the middle of the stackup — the thickness center of the flex section. On a rigid-flex board those are not the same thing. The board's geometric center includes rigid-section material, but bending only happens in the flex section, so the reference is that section's own thickness center. In practice this means making the coverlays and bonding films above and below the flex section symmetric so the flex copper lands near the middle.

The 6× bend radius rule — is that total board thickness or flex section thickness?

Flex section thickness. The total includes rigid-section cores and prepreg, and those materials do not participate in the bend. Using total thickness gives you an oversized bend radius and wastes enclosure space. Conversely, assuming you can calculate from some thinner single layer gives you an undersized radius that exceeds what the material will take. The stackup file should explicitly call out the nominal flex section thickness.

Is adhesiveless always better than adhesive-based?

Not always. It depends on the binding constraint. Adhesiveless is thinner and its interlayer strain distribution is more predictable, which is a clear advantage when the bend radius is tight or the bend is dynamic. But it costs more and has fewer available part numbers. For an install-and-stay design with a generous bend radius and a modest layer count, adhesive-based is entirely adequate and paying more for adhesiveless is not worth it. The criterion is item 2's bend radius and item 1's use condition, not the material's grade.

What is the maximum layer count for rigid-flex?

SKPCB's engineering-confirmed figure is up to 20 layers. That is a total board ceiling, not a guarantee that any given stackup reaches 20. Flex layer count in the bend, bend radius, use conditions and material combination together narrow the actual workable range. Rigid-flex stackups beyond conventional layer counts are subject to engineering review — submit the stackup file together with the use conditions.

Why is the rigid-flex layer ceiling so much lower than conventional multilayer?

Conventional multilayer capability is 40 layers in standard production and 64 layers maximum; rigid-flex stops at 20. The gap comes from material behavior mismatch. A rigid-flex stackup contains polyimide, adhesive (if adhesive-based) and epoxy systems simultaneously, and all three behave differently in Z through lamination and reflow thermal cycles. More layers means more accumulated mismatch across interfaces. A conventional multilayer board is one material system top to bottom and has no such accumulation term.

Can vias go in the bend region?

They should not. A via in the bend region is a discontinuity in both stiffness and geometry: the barrel copper's stiffness differs from the surrounding flex material, so strain concentrates at the barrel wall during the bend. Layer transitions should be completed in the rigid section, leaving the bend region with continuous straight routing only. The same goes for pads, test points, copper discontinuities and trace corners — keep them out of the bend.

How do you handle an impedance net crossing the rigid-flex boundary?

Calculate and specify per section. The flex section's dielectric is polyimide plus coverlay, possibly plus adhesive; the rigid section is epoxy prepreg and cores. The dielectric environments are entirely different, so the same trace width yields different impedance in each. The stackup file should annotate trace width separately for each section for the same net, and specify the transition treatment at the boundary. SKPCB's controlled-impedance capability should be confirmed against the product specification, and the specific implementation of a cross-boundary impedance net is subject to engineering review.