Flex PCB design and assembly

Our engineers start with the product. They look at how the circuit sits in the enclosure, where it folds, and what moves. For a wearable, the flex often runs around a battery or along a strap. For a smart ring, it may wrap inside the shell. For a robot hand, it may cross a joint that moves all day. For a data glove used to train robots, it may run along each finger and flex with every grip. Each of these is a different bend, and each needs a different stack-up. Our engineers model the fold in the enclosure CAD, so the flex outline, the bend line and the parts around it are fixed together.

Next comes the stack-up. Layer count, copper type, grain direction and trace position relative to the neutral bend axis decide how long a moving circuit lasts. The DFM review checks bend radius, trace routing near the bend, coverlay openings, stiffener areas and the parts placed near the fold. Each finding is sorted by when it must be fixed: before the demo build, before EVT, or before production.

Then the circuit is built in the Shenzhen factory, to IPC-6013 Class 2 or Class 3, and assembled on its SMT lines. Boards get automated optical inspection, and firmware test results are kept per unit, by serial number, when the order includes functional test. The demo build tells you whether the fold works in your hand. EVT and DVT then prove it over many cycles before the design is closed, and the DFM review is repeated before DVT so a late change to the bend is caught while it is still cheap to fix.

A flex circuit is right when the interconnect is the problem. Two rigid boards joined across a hinge, a connector stack that fails under vibration, or a cable that must fold into an odd shape are all good cases. Flex removes parts instead of adding a technology. Fewer connectors is usually the real saving.

It is the wrong answer when the board is only small and dense; that is an HDI problem. It is also wrong when the assembly needs the stiffness of a rigid laminate to hold connectors and heavy parts without a stiffener under each of them. See how a smart ring team handled flex in a small shell in our smart ring case study.

Whether the bend is static or dynamic, and the bend radius. A circuit folded once in assembly can be built far more freely than one that flexes many times. For a moving circuit, copper grain direction, layer count and trace position near the neutral bend axis decide its life. Declaring the wrong case costs money or causes field failures.

After that, the design must show where stiffeners go, so connectors and part areas have something rigid behind them. It must show the coverlay openings, which take the place of solder mask. It must also say whether any part of the circuit should be rigid-flex instead. Each item is confirmed against your drawing in engineering review. Our flex and rigid-flex design guide for wearables walks through each one.

Up to 8 layers for flex and up to 40 layers for rigid-flex through engineering review. Single- and double-layer flex is quoted through review too, because the advanced route's layer field starts at 4. A dynamic circuit stays at the low end on purpose, because each added layer stiffens the bend.

The advanced route offers eleven layer counts from 4 to 30, and polyimide is a laminate family that always routes to review. The layer count is a mechanical decision as much as an electrical one, and our DFM services page explains how that finding reaches your demo build. When a design needs more layers than a bend allows, the usual answer is to split it: keep the dense routing on a rigid section and run only the signals that must cross the fold through a thin flex link.