The rigid-board assumption does not hold on flex
Flex discussions concentrate almost entirely on bending: radius, conductor routing, coverlay choice. All of that matters, but none of it explains a very common class of rework — the board has not been bent yet, assembly mated the connector a few times, and the connector-area pad lifts off with its copper attached; or a ring crack appears at a hole mouth before the bend test is even finished.
The root cause of that class sits somewhere else. The rigid-board assumption about what holds a hole and a pad in place simply does not carry over to flex.
Two things are taken for granted on a rigid board. First, the hole wall has glass-reinforced laminate on both sides. That laminate is comparatively dimensionally stable in plane, so after drilling the hole wall is a channel with a skeleton behind it, and the plated copper sits against that skeleton. Second, the copper under a pad is laminated onto glass weave. The copper-to-substrate interface has woven texture to key into, so a vertical pull is distributed across a patch of surrounding laminate.
Flex removes both at once. The substrate is polyimide film — no glass skeleton, thin, and dimensionally active in every direction with temperature, humidity and stress. The copper-to-film interface is nearly smooth, either adhesive-bonded or directly metallised. Consequently:
- The plated copper in the hole wall is no longer backed by a stable skeleton. It is the load-bearing structure of the channel, and through-thickness movement of the film acts directly on that copper barrel.
- A pad no longer has a sheet of laminate sharing the vertical load. The load concentrates on the small copper-to-film interface under that pad, and the interface is the ceiling.
- The whole board still gets bent, rolled and pushed into an enclosure in service, and those loads travel into wherever the holes and pads happen to be.
So holes and pads on flex are not "the rigid-board approach applied to a thin board". They are a structural problem where anchorage has to be deliberately created. The sections below cover failure modes, structural measures, the separate rules near a bend area, and finally evidence and drawing statements. General structural and manufacturing boundaries are covered separately in flex construction and manufacturing limits and are not repeated here.
The failure modes change in kind, not in degree
Removing the glass weave does not mean "the same problems happen more easily". It introduces several different mechanisms.
Circumferential cracking of hole-wall copper. On a rigid board, hole-wall cracking is usually discussed in terms of aspect ratio, plating uniformity and thermal shock, with the laminate as the reference frame. Flex adds a mechanism: the film's through-thickness expansion differs markedly from copper and nothing constrains it, so thermal cycling repeatedly pushes and pulls the copper barrel axially. Cracks tend to appear where the hole mouth meets the surface foil — the corner where the channel transitions from barrel to planar copper.
Whole-pad lifting. On rigid boards, pad loss often brings a crater of substrate with it. On flex the more common picture is the pad leaving the film along a clean interface, with no substrate damage visible from the back. That tells you the failure is at the copper-to-film bond, not inside either material — in other words, this is not a material strength problem, it is an interface area and load distribution problem.
Separation between barrel and annular ring. On flex, the hole-wall copper and the surface annular ring are two parts of the same plated layer. When the film wrinkles, blisters or delaminates around the hole, those two parts are the first to part company, showing up as intermittent continuity or drifting contact resistance while the external appearance may look unremarkable.
Early failure from assembly loads. Connector zones, mating zones and areas pressed against the enclosure carry real mechanical load during final assembly. A rigid board's own stiffness intercepts those loads. Flex cannot, so the load reaches the interface directly.
The common thread is clear: most hole and pad failures on flex occur at interfaces and transition corners, not in the bulk of the copper or the film. That sets the direction for every measure that follows — not thicker copper or tougher film, but more anchorage area, better load paths, or keeping load off the interface altogether.
Structural measures that put anchorage back
Since anchorage is not inherent in the materials, it has to be built structurally. Each of the common measures addresses a different load path, and each trades away something.
Capturing the pad under coverlay (anchor ring). The coverlay overlaps the pad edge, forming a mechanical hold-down. It does not improve the copper-to-film interface, but it converts part of the vertical pull into shear carried by the coverlay — to lift, the pad must first push that overlap away. The cost is reduced exposed pad area, tighter soldering and registration windows, and a new dimensional chain between coverlay opening and pad. How much hold-down stiffness the overlap can provide depends on the coverlay itself, discussed in coverlay and polyimide material selection.
Anchor tabs. Copper extensions outside the pad run under the coverlay and get captured with it. They route load from the pad centre out to several captured anchor points, effectively enlarging the anchorage area. The cost is keep-out space around the pad, which fine-pitch devices often cannot spare.
Stiffeners. Bonding a stiffener under a connector zone, mating zone or component area restores locally near-rigid behaviour — the load is taken by the stiffener instead of travelling to the hole and pad interfaces. This is the most effective measure against assembly loads. The cost is thickness, weight, bond-position accuracy, and the fact that the stiffener edge becomes a new stress concentration line. Drawing requirements for stiffeners are developed in stiffener drawing review.
Moving the hole instead of reinforcing it. Often the most effective action is relocation: away from the bend area, away from stiffener edges, away from profile cut-outs. On flex, hole position is a structural parameter, not merely an output of routing.
Shaping the copper around the hole. The corner at the hole mouth is a crack initiation site. Making that transition gentler — a larger annulus, no abrupt copper-thickness step at the mouth, conductors entering the pad as a taper rather than a right-angle stub — changes the degree of stress concentration. Rules for conductors entering bend areas and pads are covered in conductor routing rules in bend areas.
Avoiding the problem by construction. If a connection area genuinely needs rigid-board behaviour, rigid-flex may be a better answer than "flex plus stiffener". The boundary between the two is compared in rigid-flex versus flex construction boundaries; choosing rigid-flex then introduces its own set of transition-zone constraints, see rigid-flex transition zone manufacturing.
These are not a checklist to stack up. Coverlay overlap costs area, anchor tabs cost routing space, stiffeners cost thickness, relocation costs layout freedom. Which combination is right depends on what load that location actually sees — soldering heat, assembly mating, or service bending. The three call for different measures.
Holes near a bend area follow separate rules
Putting a hole inside a bend area, or close to its edge, is a distinct problem that an anchor ring does not address.
The load is of a different nature. The failures above are mostly vertical pull or thermal cycling. Bend-area load is in-plane tension and compression that reverses with every flex. A hole there is a stiffness discontinuity: the copper around it is thicker and harder than its surroundings, so deformation across the area is interrupted at that hard point and strain concentrates at the hole edge.
Worse, the hole is also a geometric notch. Strain at a notch edge exceeds the nominal strain, and once a crack starts it propagates along the bending direction. There is no rigid-board counterpart to this, because rigid boards are not bent repeatedly.
The practical rules are therefore layered:
- As a default position, no holes inside the bend area. This is a position, not a suggestion.
- Holes near the bend area keep their distance, and the references are not only the bend line but also coverlay edges, stiffener edges and profile cut-outs — all of them stiffness discontinuities a hole should not coincide with.
- Prefer hole-free connections through transitions. If a net can cross the bend area on one layer, do not place a hole on each side just to change layer.
- When a layer change is unavoidable, put it in a stiffness-stable region — typically under a stiffener or well away from any edge.
Bend radius and copper-foil type change how much strain the bend area can tolerate, and therefore how much margin these distance decisions have; that is treated in bend radius and copper foil selection.
What evidence shows the anchorage holds
Anchorage is an interface property and cannot be seen. Incoming inspection cannot certify it by appearance, but that is not a licence to sign off on "visual OK" alone. The usable evidence falls into categories that answer different questions.
Structural conformance evidence. Coverlay opening position relative to the pad, whether anchor tabs are genuinely captured, stiffener bond position and edge condition. This shows whether the design intent was implemented. It says nothing about interface strength, but if this layer fails, nothing after it matters.
Interface integrity evidence. Wrinkling, blistering, voids, delamination or unbonded areas around holes, pad edges and coverlay overlaps. These are direct signs that the interface is already compromised, and they are usually detectable in process.
Destructive verification evidence. Cross sections showing hole-wall copper continuity, hole-mouth corner geometry and copper-to-film contact; plus pull-type testing of interface load capacity. This is the closest to the real property, but it is destructive, so it can only be sampled — and the sampling location must be agreed in advance, because a section taken in a flat area and one taken in a hole-dense area will support different conclusions.
Process stability evidence. Plating uniformity, hole-wall quality, lamination parameter control. Anchorage is not an attribute you can inspect in at the end; it depends largely on a stable process. Hole-wall inspection methods are set out in drilled hole wall quality inspection, and plating release logic in plating quality inspection and release.
Load representativeness evidence. The easiest to overlook: the load applied during verification must be of the same kind as the service load. If the real risk is assembly mating, thermal cycling alone is not sufficient evidence. If the real risk is long-term flexing, a single pull test is not sufficient either.
Organise the judgement this way: use structural conformance and interface integrity at high inspection coverage because they catch definite nonconformance; use destructive verification and process data to support release confidence; then confirm that the verification load matches the service load in kind. How inspection stages are organised on this site is described under quality and inspection.
What the drawing and the order must state
A high proportion of flex hole and pad failures trace back to the statement stage: the drawing gave hole size and pad dimensions but no anchorage intent. The fabricator cannot see which pad will be mated and which hole sits near a bend, so it applies generic rules. To let the fabricator work backwards to a process route, these items need to be explicit:
- Which areas carry which loads. Mark bend areas (with a qualitative note on direction and order of magnitude of cycles — no numbers needed), mating zones, pressed zones, component zones. This is the input to every anchorage decision.
- Whether anchorage measures are specified or delegated. Are you prescribing coverlay overlap and anchor tab form, or authorising the fabricator's standard practice? Either is workable, but one must be chosen, and the scope of delegation stated.
- The dimensional relationship between coverlay opening and pad, plus its priority — when registration and solderable area conflict, which gives way.
- Stiffener position, extent and edge requirements, in particular where stiffener edges fall relative to holes and bend lines.
- Hole classification. Which holes are purely electrical interconnect, which carry mechanical location or fastening, which sit in loaded areas. These three are handled differently; if the drawing does not distinguish them, the fabricator can only treat them as one class.
- Acceptance method and sampling location. Where destructive sections are taken, how many, and which form of evidence governs. Without this, the two sides will argue about the same lot using sections from different places.
- Concession priority. When anchorage requirements, bend performance, thickness limits and solderable area cannot all be satisfied, which is sacrificed first. The fabricator can see the conflict but cannot decide it for you.
Five of those seven are qualitative statements requiring no numbers at all, yet they determine whether the process route can be set correctly. Site capability scope is listed under manufacturing capabilities, the flex product page is flexible PCB, and the full board-type list is under products.
A workable decision order
- Mark the loads first: where it bends, where it mates, where it is pressed, where components sit.
- Move holes away from every stiffness discontinuity line — this step is free, so do it first.
- For holes and pads that must stay in loaded areas, select anchorage by load type (thermal: overlap and transition geometry; assembly: stiffener first; bending: change the layout first).
- Check whether the area, thickness and routing space consumed by those measures still fits. If it does not, return to step 2 and change the layout rather than stacking on more anchorage.
- Fix acceptance: evidence types and sampling locations go into the drawing.
- Write the concession priority.
The value of that order is that "it will not hold" surfaces before design freeze rather than during final assembly. Assembly-stage loads also involve carrier design, see flex assembly carriers; notches and stress concentration from profiling are covered in flex profiling and die cutting.
FAQ
Q: Why do flex pads come away with their copper attached when rigid boards rarely do this?
A: Because the load path differs. Copper under a rigid-board pad is laminated onto glass weave, the interface has woven texture to key into, and a vertical pull spreads across surrounding laminate. On flex the copper-to-polyimide interface is nearly smooth with no skeleton to share load, so everything concentrates on the small interface under that pad, and the interface capacity is the ceiling. That is why the fracture surface is usually clean with no substrate damage visible from the back.
Q: Will thicker copper or a tougher film solve anchorage?
A: Wrong direction. Flex hole and pad failures occur mostly at interfaces and transition corners, not in the bulk of either material. Strengthening the bulk changes neither the interface area nor how load reaches it. The effective directions are increasing anchorage area (coverlay overlap, anchor tabs), taking the load away (stiffener), or keeping load off the interface entirely (relocate the hole, change the layout).
Q: Should I apply both coverlay overlap and anchor tabs to be safe?
A: You can, but both consume area — overlap takes exposed pad area, tabs take keep-out space outside the pad, and fine-pitch devices often cannot spare either. More importantly, both address the same load type (vertical pull). If the real risk is assembly mating, a stiffener beats stacking the two. If the real risk is bending, change the layout first.
Q: Can a hole ever go inside a bend area?
A: The default position is no, and that is a position rather than a conservative suggestion. In a bend area a hole is both a stiffness discontinuity and a geometric notch; strain concentrates at its edge and a crack propagates along the bending direction. When a layer change is unavoidable, move the change point into a stiffness-stable region — under a stiffener or well away from any edge.
Q: Is keeping holes away from the bend line enough?
A: No. The references include coverlay edges, stiffener edges and profile cut-outs as well as the bend line. All of them are stiffness discontinuities, and a hole coinciding with any of them produces the same strain concentration.
Q: Can incoming inspection judge anchorage visually?
A: Not directly — anchorage is an interface property and cannot be seen. But visual inspection is not useless: wrinkling, blistering, voids, delamination and unbonded areas are direct signs that the interface is already compromised, and they catch definite nonconformance. Layer the judgement: structural conformance and interface integrity at high coverage, destructive sections and pull testing as samples to support release confidence, then confirm the verification load matches the service load in kind.
Q: Why must the cross-section sampling location be agreed in advance?
A: Because different locations support different conclusions. A section from a flat area versus one from a hole-dense area or near a stiffener edge can show markedly different hole-wall geometry and interface contact. Without an agreed location, supplier and customer argue about the same lot using sections from different places and never converge.
Q: Does the drawing have to carry numbers?
A: Five of the seven statements listed here are qualitative: load-zone marking, specification or delegation of anchorage measures, hole classification (electrical / mechanical / loaded), acceptance sampling agreement, and concession priority. None require a number, yet they determine whether the process route can be set correctly. Give only hole size and pad dimensions with no anchorage intent, and the fabricator can only apply generic rules.

