On a rigid-flex board, almost all of the risk sits on a single line: where the rigid section ends and the flex begins. Our Rigid-Flex PCB page states the reason plainly — the practical difference from a conventional multilayer board is the transition between rigid and flexible areas, because that is where two materials with different thermal expansion and different moisture behaviour meet. Our rigid-flex design guide calls the same place a stress concentration zone and the most failure-prone spot on the assembly.
That leads to a conclusion which is routinely missing from inquiries: this line has to be on the drawing. Not as an intention, not as a sentence in the accompanying notes, but as dimensioned geometry tied to a datum. The reason is not formality. It is how manufacturing works.
This article covers only those few millimetres. Whether rigid-flex or a pure flex circuit is the right construction principle is covered in Rigid-Flex vs Flex: construction boundaries. Routing rules inside the bend area are in Conductor routing and bend area rules, and coverlay material selection in Coverlay and polyimide material selection. Those topics are linked, not repeated.
1. Manufacturing has tolerances, not intentions
A fabricator reads geometry. It does not read design goals. If the drawing does not say where the transition line sits, manufacturing cannot work backwards to where you wanted it — for a concrete reason: three mutually independent fabrication steps converge on that line, and each brings its own tolerance.
The values are published on our manufacturing tolerances page, stated per IPC-6012 acceptance standards:
- Board outline: ±0.005" (≈0.13 mm). That is the profile, which includes every routed edge that releases a flex arm from the rigid stack.
- Internal cutouts: ±0.005" non-plated, ±0.010" plated. That covers every window opened in the rigid build-up to expose the flexible area.
- Solder mask registration: ±0.003", depth ±0.005". Cited here not as a coverlay tolerance, but to make one point: every applied layer has its own registration and does not land coincident with the profile.
Overlay those independent values and you get a scatter band several tenths of a millimetre wide. State the basis clearly: that is pure arithmetic stacking of published standard tolerances, and explicitly not a Sunking-committed tolerance for the transition zone. We do not publish a transition-zone tolerance, because it depends on the specific build-up and is confirmed per project.
The point is independent of the exact number: if your design only works when that line sits in one specific place, then that place has to be dimensioned — otherwise tolerance stacking decides it, not your design. Manufacturing will work correctly inside its tolerances and still ship a part that misses your intent. That is not a fabrication defect. That is an incomplete drawing.
2. What the drawing has to fix at this location
Three pieces of information, mutually independent, none derivable from another.
First, the position of the transition line itself. Dimensioned to a functional datum, not to an adjacent edge that carries a tolerance of its own. The document list on our product page explicitly asks for the *definition of rigid and flexible areas* — that is this item, and it is not the same thing as the board outline.
Second, the coverlay anchor. How far does the coverlay reach into the rigid section? Our design guide gives the direction unambiguously: the coverlay opening should extend a short distance into the rigid area to form an "anchor" that prevents delamination. That overlap is functional. Terminating flush on the transition line is the worst option, because then the tolerances of both features land in the same seam and, in an unfavourable combination, the layer ends short of the line — exactly where stress is highest. We deliberately give no millimetre value here: the dimension depends on the build-up and is confirmed in engineering review. What the drawing must state is that there is an overlap and what the dimension references.
Third, the stiffener boundary. Where does the stiffener end, and does it end before the transition zone or on top of it? A stiffener reaching onto the transition moves the bend to its own edge and creates a very small, unspecified bend radius there. How to dimension a stiffener completely — function, side, orientation, closed contour, adhesive boundary — is covered in stiffener drawing review. For the transition zone only one question matters: where is the boundary, and is it dimensioned so that it cannot migrate into the zone under an unfavourable tolerance combination?
The product page document list additionally requires bend radius and bend direction. Both matter on this line: direction determines how the copper must cross the boundary. The guide states it as a rule — run copper traces perpendicular to the bend direction when crossing the boundary, avoid diagonal crossings, add teardrops and rounded corners, avoid right-angle routing. Without a stated bend direction, "perpendicular" is undefined.
3. Where it flips from "difficult" to "not manufacturable"
Four situations where no amount of process optimisation helps, because the problem is in the geometry.
Vias and pads inside the zone. The design guide leaves no room here: keep vias, pads and components away from the bend line, and place vias well outside the flex area. A via in the stress concentration zone is not a manufacturing challenge; it is a designed-in fracture origin. That it can be fabricated does not change that.
Diagonal crossing of the boundary. A trace running at an angle across the transition line concentrates stress, and manufacturing cannot compensate for it. It is a layout property, not a process parameter.
Flex layer count in the wrong place. The product page publishes 1–8 flexible layers, while the design guide recommends minimising layer count in the flex area and preferring single or double layer. Both hold at once: 8 flexible layers are within specification, but they are not a good build-up for a tightly bent location. The published ≥100,000 bend cycles apply explicitly to static installation and limited dynamic bending within the design rules; actual life depends on bend radius, copper thickness and trace routing in the flexible area, coverlay execution, and whether assembly introduces torsion. A thick flex stack at a tight bend does not make use of that specification.
The tolerance chain is not closed. If the transition line is dimensioned to the outline, the stiffener to a hole, and the coverlay to no defined datum at all, there is no continuous datum framework. Every part is fabricated inside its individual tolerances and still does not fit in the assembly. This is the most common case and the easiest to avoid.
4. Why the flex page values do not apply here
One substitution worth clearing up. Our Flexible PCB page publishes 1–8 layers, PI base material, 2mil minimum trace width, bend radius ≥6× board thickness, copper 0.5oz–2oz, and PI/PSA coverlay.
The Rigid-Flex PCB page publishes a different set: 2–20 rigid layers, 1–8 flexible layers, 3mil minimum trace width, 0.3–4.0 mm finished thickness, ≥100,000 bend cycles, base material FR-4+PI.
The flex page's 2mil is not a rigid-flex value, and the ≥6× factor is not published on the rigid-flex page. For a rigid-flex build-up, 3mil is the published figure, and bend radius is reviewed per project — which is exactly why the document list asks you to state it rather than carry it over from another product page. Sizing a rigid-flex flex arm with 2mil and a blanket radius factor means sizing a different construction.
The same applies on the material side: FR-4 in the rigid areas, polyimide in the flexible ones. Per the product page, the two differ in thermal expansion and moisture behaviour, which is why adhesive design and lamination parameters in the transition area govern reliability. Final material types, adhesive system and acceptance criteria are confirmed in the project documents. An overview of material classes is on the materials page.
5. The acceptance standard belongs on the drawing, not in a follow-up question
The last item on the product page document list is the *applicable acceptance criteria*. At the transition zone that is not a footnote. Our published IPC acceptance classes are IPC-6012 Class II and Class III; which one applies is information from your drawing or order confirmation, not an assumption manufacturing makes for you. The standard values listed on manufacturing tolerances are stated per IPC-6012, and tighter tolerances are available on request after review with the engineering team — but only if the requirement is stated before fabrication. After lamination, the transition zone is no longer accessible.
The quality framework behind this is ISO 9001:2015 and IATF 16949:2016, both certified by DNV, plus UL recognitions ZPMV2.E326765 and ZPMV8.E326765. How the inspection chain is structured is on the quality assurance page; the corresponding manufacturing limits are summarised under capabilities.
6. Short list before engineering review
A defensible manufacturability review of the transition needs the following. It mirrors the product page document list, scoped to the transition zone:
- Definition of rigid and flexible areas, dimensioned to a functional datum — not merely inferable from the outline.
- Complete layer stack-up, including the split of rigid and flexible layers. The flexible layers generally run through the entire board, so the split determines the lamination build and the finished thickness.
- Bend radius and bend direction, plus the expected number of bend cycles.
- Coverlay and stiffener requirements, stating the overlap into the rigid area and the stiffener boundary.
- Gerber or ODB++ data plus drill and profile data, so that outline, internal cutout and layer registration can be reviewed together.
- Surface finish requirements, see surface finish.
- Quantity, panelisation and the applicable acceptance criteria.
With those inputs the transition zone can be assessed per project before tooling exists. Submit through request a quote; an overview of all board types is under products.
FAQ
How wide does the transition zone have to be?
We do not publish a universal width, because it depends on the stack-up, on the split of rigid and flexible layers, and on the finished thickness — and per the product page that thickness spans 0.3–4.0 mm, a range over which a single number would mean nothing. What you have to fix is not a width copied from a table, but the position of the transition line on your drawing, dimensioned to a functional datum. The width of the zone then follows from your build-up and is confirmed in engineering review against Gerber/ODB++ data, profile data and the complete stack-up.
If the drawing leaves the transition line open, can manufacturing set it?
No — not in the sense of your design intent. Manufacturing works with geometry and tolerances. On that line, board outline (±0.005"), internal cutout (±0.005" non-plated, ±0.010" plated) and the registration of applied layers converge, each per IPC-6012 acceptance standards and each independent of the others. Without a dimension, the position is decided by that tolerance combination rather than by your design. The result can be within every individual tolerance and still fail to work in the assembly.
Should the coverlay terminate flush with the transition line?
No. Our design guide gives the direction unambiguously: the coverlay opening should extend a short distance into the rigid area to form an "anchor" that prevents delamination. A flush termination is the worst variant, because the tolerances of both features land in the same seam and, in an unfavourable combination, the layer ends short of the line — inside the stress concentration zone. We confirm the specific overlap dimension per project; what the drawing must state is that there is an overlap and what the dimension references.
Can the stiffener reach onto the transition zone?
That is precisely the case the drawing should rule out. A stiffener reaching onto the transition moves the bend to its own edge and creates a very small, unspecified radius there. Dimension the stiffener boundary so that it cannot migrate into the zone under an unfavourable tolerance combination. The full dimensioning method for stiffeners — function, side, orientation, closed contour, adhesive boundary — is covered in stiffener drawing review.
Does bend radius ≥6× board thickness apply to rigid-flex as well?
That value is on our flex PCB page and belongs to a pure PI build. It is not published on the rigid-flex PCB page, which states 2–20 rigid layers, 1–8 flexible layers, 3mil minimum trace width, 0.3–4.0 mm finished thickness, ≥100,000 bend cycles and FR-4+PI. So do not carry the factor across — state bend radius and bend direction on the drawing so the flex-area build-up can be reviewed against them per project. The same applies to the 2mil on the flex page: for rigid-flex, 3mil is the published value.
What happens if the drawing does not state an acceptance class?
Then an input is missing that manufacturing cannot decide for you. Our published IPC acceptance classes are IPC-6012 Class II and Class III, and the standard values on the manufacturing tolerances page are stated per IPC-6012. Tighter tolerances are available on request after review with the engineering team, but only before fabrication: after lamination the transition zone is no longer accessible. Carry the class on the drawing or the order confirmation.
Are 8 flexible layers acceptable at a tight bend?
The specification is 1–8 flexible layers, so it is within our manufacturing limits. At the same time the design guide recommends minimising the layer count in the flex area and preferring one or two layers. These do not contradict each other: the specification describes what can be fabricated, the design rule describes what makes sense at a tightly bent location. The published ≥100,000 bend cycles apply to static installation and limited dynamic flexing within the design rules, and depend on bend radius, copper thickness, routing in the flexible area, coverlay execution and whether torsion is introduced during assembly. A thick flex build at a tight bend does not make use of that specification. Submit the complete stack-up together with the bend area definition — entry point at request a quote.
---