Rigid-Flex PCB Design: Stack-Up, Materials and Transition Zones

When a product needs a board-to-board connection that survives repeated movement, the choice is usually between a cable assembly, a flex circuit with connectors at both ends, and a rigid-flex board where the flexible layers are part of the same laminated stack as the rigid sections. The third option is the most expensive and the most capable, and it is the one that removes the connector from the reliability budget altogether.

Rigid-flex PCB design is not simply a matter of drawing a flex outline inside a rigid layout. The materials change, the stack-up becomes asymmetric by necessity, the fabrication sequence grows, and the mechanical behaviour of the finished part depends on details that never appear in a schematic. What follows walks through the decisions that determine whether the board survives the field.

What Rigid-Flex Offers That a Flex Cable Cannot

A flexible circuit with stiffeners can carry a modest number of signals between two connectors, and for many products that is enough. The limit appears when the interconnect needs hundreds of nets, when the rigid sections carry fine pitch ball grid array devices, or when the same assembly must provide electromagnetic shielding and controlled impedance like any high density interconnect build. A rigid-flex PCB design lets the flexible layers leave the rigid stack as part of the same lamination, so the copper that carries the signal is continuous from driver to receiver. There is no solder joint and no connector contact in the path.

The advantage carries obligations. Everything that used to be hidden inside a connector housing becomes part of the mechanical specification of the board. Bend radius, the expected number of flex cycles, the direction of the bend relative to the conductor routing and the stiffness of the coverlay all become design parameters that the fabricator must be told about. Assemblies that fail in the field usually fail at the transition between the rigid and flexible regions, not in the middle of a straight flex run.

Stack-Up and the Flex Core

The flexible core sits in the middle of the stack and is normally a polyimide film with rolled annealed copper, chosen because rolled copper tolerates repeated bending far better than electrodeposited foil. Rigid sections build up on both sides of that core. Because the flex core has to be free of rigid material where it bends, the outer layers are opened with a window, and the resulting stack is deliberately unbalanced in the bend region. That imbalance is not a defect, but it must be understood before the artwork is released.

Common configurations include three flex layers embedded in an eight or ten layer rigid stack, a balance familiar from multilayer board design, and odd layer counts are not unusual when the flexible portion is built from a single sided core to maximise flexibility. The designer also decides which nets belong on the flex layers. High speed pairs kept on the same flex core benefit from continuous reference planes, while power distribution is usually left in the rigid section where copper area is available.

Materials, Coverlay and Adhesives

Adhesiveless polyimide laminates have largely replaced acrylic bonded constructions in demanding products because they thin the dielectric, improve dimensional stability and remove a layer that creeps under thermal cycling. Coverlay protects the flex conductors, and its thickness, which follows the same logic as the board thickness guide, sets how far the neutral axis sits from the copper, which directly influences the strain the copper sees at a given bend radius. Photo imageable coverlay gives tighter openings around pads, while laminated coverlay is cheaper and more robust in abrasion.

Surface finish matters more than on a conventional board. Hard gold over nickel on the flex tail withstands repeated insertion into a zero insertion force connector, while immersion finishes are adequate where the flex is soldered once and never touched again. The designer should also specify whether the flex region may be held under tension during assembly, because a flex run that is pulled tight behaves very differently from one that is left with a service loop.

Transition Zones and Bend Radius

The transition from rigid to flexible is the hardest part of the design. Conductor widths should be reduced gradually rather than stepped, plated through holes must be kept well away from the window edge, and the coverlay should extend into the rigid region so that the copper is supported right up to the point where the rigid material stops. A common rule is to keep any hole at least three times the dielectric thickness away from the flex opening, and to stagger the ends of the rigid layers so that stress is not concentrated on one line.

Bend radius is a specification, not a suggestion. A widely used starting point for a one time bend is ten times the total flex thickness, and for a dynamic application that flexes continuously the figure rises to one hundred times or more. The bend should be placed away from plated holes and pads, and it should be formed across the width of the flex rather than along a diagonal. Adding a stiffener, a bonded stiffener or a bead of potting compound at the flex tail is a standard way to release strain.

Connectors, Stiffeners and Assembly

Stiffeners serve two purposes. They provide mechanical support for a connector or for a component placed on the flex, and they define the region that must not bend. Polyimide stiffeners suit thin tails, FR-4 stiffeners are cheaper where thickness allows, and stainless steel stiffeners are used when the tail needs rigidity plus a defined ground reference. The stiffener must be shown on the fabrication drawing with its material, thickness and bond adhesive, because it is added after lamination.

Assembly strategy should be settled early. A rigid-flex board that is designed to be folded into a three dimensional shape needs a fixture and a defined fold sequence, and the fold should not be reworked more than once. If the design instead uses a zero insertion force connector at the tail, the tail needs a stiffener, a defined insertion depth and a strain relief feature. Deciding this at layout time avoids a late change to the stack-up, which is expensive on a rigid-flex part.

Documentation and Verification

The fabrication drawing for a rigid-flex board carries more information than a rigid equivalent. It should show the flex outline, the rigid to flex windows, the stiffener locations with tolerances, in the same spirit as the manufacturable design guidelines, the bend line with its radius and direction, the coverlay openings and the surface finish for each region. A cross section view of the stack at both the rigid and the flexible area removes most of the ambiguity that otherwise leads to questions during quotation.

Verification should include a bend test on a representative coupon, an impedance check on any controlled impedance flex net and a visual inspection of the transition zone under magnification after assembly. Thermal cycling is worth running where the flex carries power, since the polyimide core and the rigid epoxy expand at different rates and the transition zone is where that mismatch shows up first. Reviewing these results against the assumptions made at design time closes the loop.

FAQ

Is rigid-flex worth the cost? It is usually justified when the interconnect is dense, when connector height is critical or when the assembly must survive vibration and repeated movement. For a simple low count connection, a flex cable with stiffeners remains cheaper.

How many flex cycles can a rigid-flex board survive? That depends on the bend radius, the copper type and the coverlay. A static bend may survive only a few hundred cycles, while a properly designed dynamic flex with rolled annealed copper and a generous radius can be specified for tens of thousands.

Can plated through holes be placed in the flex region? They can, but they are a stress riser and should be avoided or kept far from the bend. Where a via is unavoidable in a flex area, it is usually filled and capped, and the bend line is moved away from it.