Flex PCB or Rigid-Flex? Deciding the Construction Boundary

"It has to bend" is an observation, not a construction decision

The most common documentation gap in flex and rigid-flex enquiries is not a missing Gerber. It is a requirement that reads "this section has to bend." Engineering review has to come back with questions: does it bend once during assembly and then stay put, or does it flex repeatedly across the product's service life? Do components sit on either side of the bend? What bend radius does the enclosure actually allow? Without answers, the construction cannot be classified and the quote cannot proceed.

The reason is that flexibility is not a checkbox. It is the entry point to a whole set of construction consequences. Choose pure flex and you get the thinnest overall build and the finest trace capability, but every location that needs mechanical support, connector insertion force, or a heavier component has to be solved with an added stiffener. Choose rigid-flex and the rigid zones are formed inside the board itself, so components, connectors and screw holes have native support. The price is one step back on minimum trace width, a higher overall thickness floor, and a more granular review cycle.

So this article does not explain what a flex circuit is and does not list target industries. It answers one question: given the bend duty and the assembly constraints, does this board belong to pure flex or rigid-flex?

What each page publishes, and which numbers must not cross over

SKPCB splits flexible construction into two deployed capability lines. The published parameter sets are deliberately asymmetric, and that asymmetry matters more than the individual numbers because it decides what you are allowed to cite on a drawing:

The blanks are not omissions. They are facts that must be passed through to the design input as-is. The most frequent engineering error is exactly this cross-over: treating the rigid-flex figure of ≥100,000 cycles as a life commitment for a pure flex circuit, or applying the flex-page rule of ≥6× board thickness to the flex arm of a rigid-flex board. Neither citation holds, because the two constructions do not share flex-layer count, copper weight, coverlay or transition-zone conditions. Where the other page is silent, the drawing says "confirm in engineering review." It does not say a number you derived yourself.

Note one counter-intuitive direction: pure flex has the finer minimum trace width at 2 mil, against 3 mil for rigid-flex, consistent with the 2/2 mil floor in the site-wide capabilities. If your flexible section carries very dense routing, moving to rigid-flex does not loosen the trace rule. It tightens it by one step.

Criterion 1: formed once, or cycling for the life of the product

This carries the most weight, and it is the only criterion you can settle from the requirement itself.

Formed once (static bend). The part is bent into position during assembly and then does not move relative to the housing. Typical cases are folding a flexible run into an enclosure cavity or routing it around a structural member. Both constructions can serve this; the decision passes to the remaining three criteria.

Cycling over service life (dynamic bend). Clamshells, slides, rotating joints, reciprocating heads or sensor arms. Here the question is not whether it can bend, but whether the flex-layer count, copper type, coverlay and neutral-axis position can support the stated number of actuations. The rigid-flex page publishes an explicit ≥100,000 cycles figure; the flex page publishes no equivalent. That does not mean pure flex cannot handle dynamic bending — in practice dynamic applications use pure flex extensively. It means you cannot cite a life figure straight from the page. Dynamic life goes through engineering review against the actual layer count, copper weight and bend radius.

Either way, dynamic bending requires three items written into the design input: the minimum bend radius the enclosure permits, the target actuation count over service life, and the bend direction including whether any torsion is present. "Must flex repeatedly" is not reviewable.

Criterion 2: how many rigid zones carry components or load

This is the most reliable cost divide, and the test is blunt: count the areas that need mechanical support.

• Zero to one. A connector at one end, routing elsewhere, or press-fit contacts only. Pure flex with local stiffeners is usually sufficient and wins on both thickness and cost.
• Two, simple geometry. A small populated area at each end with a flexible run between them. This is the boundary zone; both are viable, and the decision falls to criteria 3 and 4 plus the assembly burden of the stiffeners.
• Three or more, or any zone needing multilayer routing, impedance control or screw loading. Classify as rigid-flex. At this point a rigid zone is no longer just "a patch supporting some parts" — it has to carry real multilayer interconnect within that area. The published rigid-zone range of 2–20 layers exists precisely for this, whereas pure flex offers 1–8 layers across the entire board, with no way to add layers locally.

The decisive question is not component count but whether those zones need independently defined layer counts and interconnect density. Layer count is a whole-board property on pure flex; rigid-flex lets rigid and flex sections be defined separately. The moment the requirement says "this area needs multiple layers, that link only needs two," the construction has already been pushed toward rigid-flex.

Criterion 3: connector count and assembly steps

This criterion prices electrical reliability and line cost at the same time.

In a pure flex solution, every junction with a rigid carrier usually means a connector or a hot-bar/solder operation. Connectors are failure sites in their own right: insertion stress, contact resistance, fretting wear under vibration, mis-insertion during assembly. Eliminating those sites is the entire value proposition of rigid-flex — the rigid-flex page lists "connector elimination" and "reduced assembly steps" among its product features.

Use a substitution ratio: if rigid-flex removes two or more board-to-board connectors along with their assembly and inspection steps, the higher construction cost is usually an acceptable trade. If it removes a single low-cost connector while forcing the board from 1–8 layer pure flex up to a rigid-flex build with a 2–20 layer rigid zone, the trade rarely pays.

One direction gets overlooked: a connector can be replaced on the line; a transition zone inside a rigid-flex board cannot. Under vibration and shock a connector is a liability; in a serviceable product a connector is an asset. That leads directly to criterion 4.

Criterion 4: rework and the scrap unit

This one sets the post-production cost of ownership, and it is the criterion most often skipped early.

With pure flex plus connectors, the failing unit is separable: replace the flex, keep the motherboard. Rigid-flex bonds them into one part, so damage to any rigid or flex zone makes the whole board the scrap unit. For a board with a 20-layer rigid zone, thickness near the published 4.0 mm ceiling, and assembly already completed, that scrap cost is not a rounding error.

Conversely, the applications the rigid-flex page lists — medical implant devices, aerospace, defence electronics — value exactly the property that there is no connector to work loose. They would rather accept whole-board scrap than field failure. So criterion 4 is not a standalone score. Weigh it against the consequence of failure: when field-failure cost far exceeds material cost, take rigid-flex; otherwise keep a replaceable boundary.

Putting the four criteria together

These are a sequence of filters, not an equally weighted scorecard:

1. Criterion 1 first. If the bend is dynamic with a stated cycle requirement, quantify radius, target cycles and bend direction before anything else, whichever side you land on. Review cannot start without them.
2. Then criterion 2. Three or more load-bearing zones, or any zone needing local multilayer or impedance control, classifies as rigid-flex. The remaining criteria will not change that.
3. Criteria 3 and 4 settle the boundary zone. With two load-bearing areas and a static bend, decide on how many connectors get eliminated and whether the scrap unit can stay separable.
4. When an input is missing, do not guess the construction. Filling the gap costs less than reworking a wrong construction choice.

Worth stating: hybrid configurations exist. A rigid-flex board can still carry external connectors, and a pure flex circuit can take several stiffeners. The four filters assign the primary construction; detailed configuration is still settled in project review.

Inputs to have ready before review

Following the four criteria, review needs this set — not a note saying "must bend":

• Bend duty: static one-time forming or dynamic cycling; for dynamic, the target actuation count over service life.
• Bend geometry: the minimum radius the enclosure allows, bend direction, any torsion, and the length of the flexible run.
• Zone breakdown: how many areas carry components or load, where they sit, the layer count each one needs, and whether impedance control applies (impedance capability per capabilities: ±5%, ±10% standard).
• Connector strategy: how many board-to-board connectors the current concept uses, and which of them the construction is allowed to replace.
• Rework requirements: whether whole-board scrap is acceptable, and the severity of a field failure.
• Mechanical and acceptance: outline and slot requirements (see manufacturing tolerance), surface finish requirements (see surface finish), acceptance class (IPC-A-600 Class 2/3, see quality assurance). Acceptance specific to flexible construction is confirmed in the project documents against the applicable requirements of IPC-2223, IPC-6013 and IPC-4204.

Material-level choices — adhesive versus adhesiveless PI, rolled annealed versus electrodeposited copper, coverlay and stiffener selection — sit outside this article and are covered separately in flex material selection. Stack-up and transition-zone detail is in the rigid-flex design guide. Application-side selection is in the flex PCB application guide. Once the inputs are complete, go straight to quote and engineering review.

FAQ

For a static bend, which construction is cheaper?

On bare board alone, pure flex usually is: 1–8 layers, PI only. But count stiffener attachment, connector material, and the assembly and inspection steps as well. If rigid-flex removes two or more connectors and their operations, total cost can invert. When it removes only one low-cost connector, pure flex normally still wins.

The rigid-flex page states ≥100,000 bend cycles. Can I quote that figure for pure flex?

No. ≥100,000 cycles belongs to the rigid-flex page; the flex page publishes no equivalent. The two constructions differ in flex-layer count, copper weight, coverlay and transition structure, so the figures are not interchangeable. Dynamic life for pure flex is confirmed in engineering review against the actual stack-up and bend radius.

Can the flex-page rule of ≥6× board thickness be applied to the flex arm of a rigid-flex board?

No. That is the published figure on the flex PCB page. The rigid-flex page publishes no bend-radius rule; the allowable radius for its flexible section is confirmed in review against flex-layer count, copper weight and the transition structure at both ends.

My flexible run needs very fine traces. Does that argue for rigid-flex?

The opposite. Pure flex holds 2 mil minimum trace width against 3 mil for rigid-flex. Fine-line demand in the flexible section is not a reason to select rigid-flex. What should drive you there is the layer count and support requirement of the rigid zones.

How do I decide when overall thickness is capped?

The rigid-flex page publishes 0.3–4.0 mm; the flex page publishes no board-thickness range. If the space budget sits clearly below 0.3 mm, the published rigid-flex window no longer applies and the design should go the pure flex route, with achievable thickness confirmed in engineering review.

Only one section needs to bend and the rest is ordinary rigid board. Can it be built as one part?

That is the defining case for rigid-flex. Decide on criteria 2 and 3: whether the rigid zones need independently defined layer counts and interconnect density, and how many connectors get eliminated. If a rigid zone is just a small mounting patch with no multilayer interconnect, pure flex with stiffeners is usually more economical.

Self-check: (1) Core question does not overlap this site's rigid-flex-pcb-design-guide (stack-up and bend-radius rules), flex-pcb-application-guide (application selection) or flex-pcb-coverlay-polyimide-material-selection (material chain); this article performs construction classification only, without expanding stack-up, material grades or industry lists. (2) All figures verified against live flex-pcb / rigid-flex-pcb / capabilities page text; unpublished items are stated as "not published; confirm in engineering review," with no cross-page substitution. (3) No invented bend-life curves, radius conversion factors, stiffener thickness specs, PI elongation values, lamination parameters, yields, lead times, customer cases or "industry average" figures. (4) IPC-2223, IPC-6013, IPC-4204 and IPC-A-600 cited by standard name only, no clause values restated. (5) Eight internal links, all en locale, all verified present in out/; no /industries/* links. (6) Keyword usage natural, no stuffing. (7) All six FAQ answers are supported directly by the four criteria in the body and match the body's wording.

---