The copper on the inside of the bend cracks first, and it is almost always the inner layer of a plated flex section rather than the outer one. Everything else that gets blamed, the lifted coverlay, the squeezed out adhesive, the cracked via barrel, arrives after the copper has already started to neck. Once that order is understood, the bend radius stops being a drawing habit and becomes a number that can be calculated before the panel is ordered.
The arithmetic behind the radius
Bending a flex section strains the copper on the outside of the curve in tension and the copper on the inside in compression. The surface strain is the foil thickness divided by twice the bend radius. A 35 micron copper foil on a 1 mm inner radius therefore sees about 1.7 percent strain at every bend. Rolled annealed copper will take that once; it will not take it a thousand times. For a part that flexes in service, the working limit is closer to 0.2 percent, which is why the usual guide is not one number but a multiple of total thickness:
| Construction in the bend | Radius guide | Typical use |
|---|---|---|
| Single sided flex, static bend | 1x total thickness, 3x preferred | Fold during assembly, then fixed |
| Double sided flex, static | 2x total thickness | Folded interconnect inside a housing |
| Multilayer flex, static | 10x total thickness | Rigid-flex with 4 or more flex layers |
| Any flex, dynamic | 100x total thickness | Hinge, moving head, drawer cable |
The multiplier applies to the total thickness of the flexible section, coverlay included, not to the polyimide alone. A 0.3 mm multilayeneeds a 3 mm radius, and a customer who asks for a 1 mm radius on that stack is asking for a fatigue failure with a predictable date.
What else fails, and in what order
| Order | Failure | What drives it |
|---|---|---|
| 1 | Inner copper crack at the trace edge | Strain above the foil limit, plus etch undercut that thins the edge |
| 2 | Coverlay lift or adhesive squeeze out | Coverlay is stiffer than the flex, adhesive flows under bend pressure |
| 3 | Via barrel crack | Plated barrel in the bend zone, 20 to 25 micron wall in tension |
| 4 | White spots and delamination at the transition | Bend line closer than about 1.5 mm to the rigid to flex transition |
| 5 | Joint crack at a pad in the bend | Any component, stiffener or solder fillet inside the bend zone |
Item four is the one that surprises people, because the board looks fine and the failure appears after thermal cycling. The rigid section cannot bend, so all the movement concentrates in the last millimetre of flex next to it, and the strain there is several times what the nominal radius suggests.
Three mistakes that cause it
A via inside the bend zone. A plated barrel is a ring of copper in tension with nothing behind it, and it opens long before the trace beside it. Vias and stitching rows belong outside the bend, with the ground plane hatched rather than solid through the curve so the flex stays compliant. If a signal has to cross the bend, it crosses as a trace, not as a via pair.
A corner where the drawing should have an arc. A bend formed over a sharp edge concentrates strain at the edge radius, which is often under 0.2 mm regardless of what the note says. The drawing needs three things stated: the bend line, the formed radius, and whether the bend is static or repeated. A note that says "bend as required" leaves the radius to whoever builds the fixture, and the fixture is usually made from what is in the shop.
Foil weight and copper type left to default. Electrodeposited copper at 1 oz in a two layer flex will crack where 0.5 oz rolled annealed copper survives, and the difference is not visible until the bend test. The same applies to the bond ply: a no flow prepreg and an acrylic adhesive give different thickness and different stiffness in the bend, so the radius that worked on the prototype may not match the production stack. Ask for the flex stack to be agreed with the bend radius calculation, not after it.
How SUPERB PCB builds rigid-flex
SUPERB PCB builds rigid-flex to 32 layers in a 5 plus N plus 5 construction, 0.1 to 12 mm total thickness, HDI from 1 to 3 stages, laser drilled microvias from 0.075 mm and mechanical drills from 0.15 mm, impedance held to plus or minus 5 percent with plus or minus 2 percent on critical nets, and warpage held to 0.5 percent or better on rigid-flex panels. Panels run to 622 by 1092 mm with up to seven press cycles, and cavities or steps from 2.5 mm are formed in the rigid area where an OCXO or a connector needs depth. Every panel carries impedance coupons, and the flex stack is confirmed against the bend radius on the drawing before release.
On the assembly side the flex section is supported through reflow rather than left to hang, stiffeners are placed outside the bend, and connectors are kept out of the flex zone so that no solder fillet sits where the part will move. Environmental testing follows the product: thermal shock, vibration and a bend test to the radius and cycle count stated on the drawing.
- Rigid-flex PCB solutions: layer counts, stack options and the bend radius rules applied to a quote.
- PCB capability: the process window behind the numbers above.
- Environmental reliability testing: thermal, vibration and bend testing after assembly.
Frequently asked questions
Is a 1 mm bend radius realistic on a 0.3 mm multilayer rigid-flex?
Not for a static multilayer bend, where 10x thickness points to 3 mm. If the mechanical envelope really only allows 1 mm, the stack has to change: a two layer flex section, 0.5 oz rolled annealed copper and a thinner coverlay bring the required radius down. The radius and the stack have to be chosen together.
Does a stiffener make the bend stronger?
A stiffener makes the rigid area stiffer and the transition sharper. Placed inside or next to the bend it moves the failure forward, so it belongs at the connector and component areas only, with a gap of at least 1.5 mm to the bend line.
Can a flex section be bent again after assembly?
It can be moved, but the copper at that line has already worked. A second forming operation at the same radius shortens the fatigue life and should be written into the test plan rather than discovered in the field. If the unit is assembled and then folded, say so before the stack is agreed.
Next step
Send the Gerber data, the flex stack, the bend line drawing and the mechanical envelope. The reply states the minimum radius for that stack, the copper type and weight for the flexible layers, where stiffeners and vias will sit relative to the bend, and the bend test that will be run on the first articles.