Flex and Rigid Flex PCB Design

In short: A flex or rigid-flex board survives only if copper is kept at the neutral axis of a generous bend radius, traces cross the bend perpendicularly in smooth arcs, stiffeners and coverlay stop short of the bend, and the stackup is built so the flex layers are never forced to carry plated through-holes inside the bend zone.

Why flex is different from rigid boards

A rigid FR-4 board is a stiff plate: you route it, and it stays flat for life. A flexible printed circuit (FPC) is a thin polyimide film, typically 12.5 to 50 µm, with rolled-annealed (RA) or electrodeposited (ED) copper of 12 to 35 µm. It is a mechanical part as much as an electrical one. Every time it bends, the outer copper is stretched and the inner copper is compressed, and copper fatigues exactly like a paper clip bent back and forth. The design job is therefore to keep strain in the copper below the fatigue limit for the number of bend cycles the product needs.

Rigid-flex combines both worlds: rigid sections carry connectors and BGAs, while flex tails replace cables and connectors. That removes solder joints and connector contact resistance, which is why you find it in cameras, hearing aids, foldable phones and avionics. If you are new to layer construction, read our guide to PCB stackup design first, because rigid-flex stackups are an extension of the same ideas.

Bend radius: the number that decides reliability

Surface strain in a bent copper foil is approximately ε = t / (2r), where t is the total thickness of the bending section and r the inside bend radius. Halve the radius and you double the strain. RA copper tolerates roughly 0.3 percent strain for millions of dynamic cycles, while ED copper, with its columnar grain structure perpendicular to the bend, cracks far sooner.

The commonly used guidance, in line with IPC-2223, is expressed as a multiple of the flex thickness:

Example: a 0.1 mm single-layer flex that must fold once during assembly needs an inside radius of about 0.6 mm. The same strip used as a laptop-hinge cable cycling 200,000 times needs roughly 10 mm. This is why a multilayer flex is bent only once; for dynamic use you split the signals across several thin single-layer or double-layer films.

Key fact: for a dynamic flex, specify RA copper, not ED copper, and keep the copper thickness at 18 µm or less. Both choices lower strain and improve cycle life by an order of magnitude.

Trace orientation and routing inside the bend

Rules that interviewers love because they are easy to forget:

• Cross the bend at 90 degrees. Traces parallel to the bend axis are loaded along their length and see no relief; perpendicular traces share strain across the whole width.
• Use curved routing. Never use 90-degree or even 45-degree corners in the bend zone. Stress concentrates at a corner and starts a crack. Arcs or teardrops are preferred.
• Stagger traces on double-sided flex. Do not stack top and bottom copper directly over each other, which creates an I-beam effect that stiffens the section and raises strain. Offset them so the film can bend between them.
• Keep the width constant. Avoid necking down or widening inside the bend; transitions belong in the flat zone, and pads need generous teardrops.
• Prefer hatched ground. A solid copper plane is stiff. A cross-hatched plane (about 50 percent coverage) keeps flexibility, though it changes the return-path impedance, so model it in a field solver.
• No vias in the bend. Plated holes are rigid anchors and crack their barrels in a bend.

Impedance control on flex is its own subject: the dielectric is thin (25 µm polyimide, Dk about 3.4), so a 50-ohm microstrip is only about 90 to 100 µm wide with an adjacent plane, and hatching the reference plane moves that number again. Cadence Allegro, Siemens Xpedition and Altium all provide rigid-flex stackup managers where you define zones with different layer counts and bend areas, then run 3D clearance checks. The general workflow still follows the PCB design process steps, with a mechanical fold simulation added before release.

Stiffeners, coverlay and the transition zone

Coverlay is a polyimide film with acrylic or epoxy adhesive, laminated over the copper and opened by laser or pre-punched windows over pads. It is the flex equivalent of solder mask, but it is far tougher. Photoimageable (PI) flex solder mask is cheaper but cracks in dynamic bends, so use coverlay wherever the board moves.

Stiffeners add rigidity under connectors and component areas. FR-4 (0.2 to 1.6 mm) is common, polyimide stiffeners (125 to 250 µm) suit ZIF connector tails that need an exact 0.3 mm total thickness, and aluminium or stainless steel serves for heat spreading and flatness. Rules of thumb:

• Keep the stiffener edge at least 0.5 to 1 mm clear of the bend start, otherwise it becomes a hinge point where all strain concentrates.
• Taper or step the stiffener edge, or extend the coverlay over it, to spread the stress.
• Where a rigid section meets flex, add a bikini-cut coverlay or an adhesive fillet that covers the copper exit so peel forces do not lift it.

Key fact: the number one field failure of rigid-flex is a crack at the rigid-to-flex transition, not in the middle of the bend. Add a teardrop to every trace there and keep the flex layers continuous from one rigid section to the next.

Manufacturing process

The flow starts like a normal board and then diverges:

1. Flex core imaging. Copper-clad polyimide (adhesiveless is preferred for reliability) is laminated with dry film, exposed and etched. Thin handling needs carriers and panel frames.
2. Coverlay lamination. Pre-routed coverlay is aligned with pins and pressed at about 180 °C for roughly one hour, which also cures the adhesive. This step moves registration, so design pads with enough annular ring.
3. Rigid layer bonding. No-flow or low-flow prepreg is cut with clearance windows over the flex zone. Ordinary prepreg would squeeze out and glue the bend area solid.
4. Drilling and plasma desmear. Polyimide and acrylic smear differently from FR-4, so plasma etchback cleans the barrels before copper plating.
5. Plating, outer imaging and finish. ENIG or ENEPIG finish is standard because HASL thermal shock distorts thin film.
6. Depth routing and profile. The rigid skin is controlled-depth milled or laser-cut to reveal the flex arm, then the outline is cut with a tight tolerance, typically ±0.1 mm.
7. Bake before assembly. Polyimide absorbs moisture; bake at 120 °C for two to four hours before reflow to avoid delamination.

Because material, drill and lamination choices change panel utilisation and yield, involve your fabricator early. Our article on PCB design for manufacturing explains the checklist, and the multilayer PCB design guide covers the plated-hole rules you inherit in the rigid zones.

Cost, density and when not to use flex

Rigid-flex costs three to ten times a comparable rigid board, driven by low panel utilisation, hand-lamination steps and tighter inspection. It pays back when you eliminate connectors and harnesses, reduce assembly labour, or fit into a 3D enclosure. For dense fine-pitch breakouts, combine it with HDI PCB design microvias in the rigid sections. If a board never needs to move after installation, a rigid board plus a simple FFC cable is often cheaper.

If you want guided practice with layout tools and stackup decisions, our CourseTron courses include hands-on PCB modules, and the ChipXpert training programmes add placement-linked VLSI and board-level tracks.

Ready to design real boards? Explore the PCB design courses at CourseTron and build a rigid-flex project from stackup to Gerber.

FAQ

What is the minimum bend radius for a flex PCB?

As a guide, use 6 times thickness for a single-layer static bend, 10 times for double-layer, and 20 times for multilayer. Dynamic applications need about 100 times thickness or more.

Can I place vias in the bend area?

No. Plated through-holes are rigid and crack under flexing. Place vias and components in flat or stiffened areas only.

Why use coverlay instead of solder mask on flex?

Coverlay is polyimide film and tolerates repeated bending, while liquid or photoimageable mask is brittle and cracks in dynamic bends.

Why use no-flow prepreg in rigid-flex?

It keeps resin from squeezing into the flexible area during lamination, which would otherwise make the bend zone stiff.