4-, 6-, and 8-Layer Rigid-Flex PCB Stackups

A common 4-layer rigid-flex build uses a 2-layer flex core (polyimide with copper on both sides) running continuously through the design, with rigid FR-4 layers laminated onto the flex core in the rigid regions to bring the total to 4 layers where rigidity is needed. This is the simplest way to achieve a genuine bend zone with enough routing on the rigid sections for moderate component density.

At this layer count, the main design decisions are bend-zone placement, stiffener strategy at connector and component areas, and whether the two flex-layer signals need any impedance control. There is limited room to selectively terminate layers, since there are only two flex layers to begin with.

Moving to 6 layers typically adds either a second flex layer pair (a 4-layer flex core) or additional rigid-only layers laminated onto a 2-layer flex core, and this is where the stackup decision starts to matter more: does the added routing capacity need to exist in the flex zone, or only in the rigid regions? Carrying unnecessary layers through the flex zone adds thickness and stiffness exactly where the design needs to bend, which works against bend-radius and reliability goals.

The common pattern at 6 layers is to keep the flex zone at 2 layers (just the signals that genuinely need to cross the bend) and add the extra rigid layers only in the rigid sections, terminating them at the rigid-to-flex boundary with proper copper clearance and acute-angle avoidance at the transition.

At 8 layers, rigid-side routing density is usually high enough that most of the layer count exists to serve the rigid sections, not the flex zone. The central stackup question becomes which specific nets must cross the bend (power, a handful of control signals, maybe one differential pair) versus which can be routed entirely within the rigid regions and connected through vias that never enter the flex zone.

This selective approach—sometimes described as book-building or selective layer termination—keeps the flex zone as thin as the minimum required layer count, which is good for bend radius and long-term reliability, rather than letting flex-zone thickness grow in lockstep with total layer count. It takes more upfront stackup planning than simply extending every layer through the flex region, but it is the difference between an 8-layer rigid-flex board that bends reliably and one that does not bend well at all.

If a controlled-impedance signal needs to cross from a rigid section into the flex zone, the stackup review has to account for the fact that dielectric thickness, copper weight, and reference-plane structure typically all change at that transition—a trace that was 50 Ω on the rigid side will not automatically stay 50 Ω once it enters the flex core unless the geometry is re-solved for the flex-zone dielectric. Flag any impedance-controlled nets that cross a rigid-flex boundary explicitly during stackup review rather than assuming continuity.