Sunking PCB

Flex PCB Bend Radius: How Board Thickness, Copper Weight and Layer Count Lock Your Routing

Design Guide2026-09-29Sunking Circuits Technical Team
Flex PCB Bend Radius: How Board Thickness, Copper Weight and Layer Count Lock Your Routing

Start with the sentence that trips most designs: "bend radius ≥6× board thickness" is not an absolute dimension, it is a ratio.

That means when you mark R1.5 on the mechanical drawing, you have not specified a feasible bend. You have specified a board thickness ceiling. The reverse is equally true: when you add two copper layers to the stack-up, or move copper from 0.5oz to 2oz, you are quietly raising the minimum allowable bend radius without changing a single character of the R value on the drawing.

The SKPCB flex PCB capability page publishes six items:

SpecificationValue
Layers1-8 layers
Base MaterialPI (Polyimide)
Min. Trace Width2mil
Bend Radius≥6× board thickness
Copper Thickness0.5oz-2oz
CoverlayPI/PSA

Of those six, only bend radius is a relative quantity. The other five are absolute windows. So the entire coupling in this specification hangs off one intermediate variable: thickness. This article answers only that: how these numbers convert into each other, and which one to move when you need a tighter radius.

It does not cover bend-area routing rules (see conductor routing rules in the bend area) or the stack-up ladder (see flex construction and manufacturing limits).

1. Read "≥6× board thickness" as an inequality

In a form engineering can use:

```

R_min ≥ 6 × t_total

```

t_total is the total thickness of the flex section, not the rigid-area board thickness and not the base film alone. The flex section thickness is a stack of contributions:

```

t_total = PI base + copper layers (× layer count) + adhesive system + coverlay (PI/PSA)

```

Two directions follow, and they carry entirely different engineering meaning:

Forward (stack-up known, solve for radius): once the stack-up is fixed, t_total is fixed and R_min falls out directly. Cheapest path, provided the mechanical design leaves you enough bend room.

Reverse (radius given, solve for stack-up): the enclosure has already locked the R value, therefore

```

t_total ≤ R_given / 6

```

This is the one line worth remembering. The R value a mechanical engineer hands you is not a routing constraint. It is a stack-up constraint. R1.5 means the flex section total thickness must not exceed 0.25mm. R3.0 means it must not exceed 0.5mm. Once that thickness budget is set, layer count and copper weight can only be allocated inside it.

2. Layer count and copper weight spend the same thickness budget

This is where most rework originates: layer count and copper weight are decided by different people for different reasons, but they draw down the same budget.

  • Layer count (1-8): driven by routing density and shielding needs, usually set by the layout engineer.
  • Copper weight (0.5oz-2oz): driven by current carrying and voltage drop, usually set by the power or electrical engineer.

Each side stays inside its own published window and is fully compliant. But the two choices stack, t_total rises, R_min rises with it, and the result collides with the R value the enclosure gave you. Nobody exceeded a limit, and the result is still unmanufacturable.

That is why the correct response to "I need a tighter bend" is not to go back and change routing. It is to return to the thickness budget and ask three questions:

  1. Can layer count come down? Going from 4 layers to 2 removes two copper layers plus the adhesive and dielectric between them, usually the single largest line item in the budget. The cost is routing density, and the 2mil minimum trace width window gives you room to trade finer lines for one fewer layer. Check that first.
  2. Can copper be thick only where it is needed? Current demand usually lands on a handful of conductors. If an entire layer's copper weight is being dragged up by a few power traces, confirm whether widening can replace thickening — the same cross-section achieved by widening does not enter the thickness budget; achieved by thickening, it does.
  3. Can the bend section carry fewer layers? Bending only happens locally. If the mechanical design allows, route fewer layers through the bend and keep the thick stack-up in the non-bending area. That points toward rigid-flex construction, covered below.

3. What fails first when you exceed the window

A boundary statement first: the following is public engineering background, not an SKPCB numerical commitment. The only bend-related figures SKPCB publishes are two: flex at ≥6× board thickness, and rigid-flex at ≥100,000 bend cycles. This article does not invent a life figure for flex.

When a laminate bends, there is a neutral axis where strain is approximately zero. The farther material sits from that axis, the more tension or compression it carries. Two directional conclusions follow:

  • The greater the total thickness, the farther the outermost conductor sits from the neutral axis, so strain at a given radius is higher. That is the physical reason R_min scales with t_total.
  • Copper reaches its limit before PI does. Polyimide is far more ductile than copper foil, so the first visible consequence of over-bending usually appears in the copper conductors: the outer, tension-side copper micro-cracks first, then propagates to an open circuit under repeated flexing.

Note carefully: this failure is often not immediate. Electrical test at assembly may pass cleanly, with failure showing up only after repeated in-service bending. That makes "bend radius out of spec" a category of problem that outgoing inspection is poorly placed to catch. It is not a manufacturing defect, it is a design input that exceeded a limit, so it has to be resolved at review, not left for quality inspection to find.

Copper foil type itself (rolled-annealed versus electrodeposited, whose grain structures give different flex endurance) is genuinely a variable in engineering practice. But the SKPCB flex page publishes only the copper thickness window 0.5oz-2oz; it does not publish foil type options. If your application is dynamic flexing, foil type must go into the review documents as a question, not an assumption.

Static and dynamic bending must be separated in the documents

  • Static bend: folded once at assembly, then held. The concern is surviving a single bend without cracking.
  • Dynamic bend: repeatedly flexed in service (hinges, sliders, robot joints). The concern is cycle life.

These two are not the same order of requirement on radius, but the specification table carries only one value: ≥6× board thickness. That published figure is the lower bound of the manufacturing window, not a recommendation for dynamic applications. Designing a dynamic application at the 6× bound spends all of your margin. Bend type is a field you must declare — SKPCB cannot read it out of Gerber data.

4. When to switch to rigid-flex construction

If the thickness budget will not close no matter how you allocate it — 6 layers of routing, 2oz for current, and R1.5 — the problem usually is not in the parameters. It is in the construction choice.

Rigid-flex publishes: rigid layers 2-20, flex layers 1-8, minimum trace width 3mil, board thickness 0.3mm-4.0mm, bend cycles ≥100,000, base material FR-4+PI. Its value here is exactly that it splits the thickness budget in two: dense routing and heavy copper stay in the rigid area, and the flex area carries only the layers that must cross the bend.

Two cautions:

  1. Minimum trace width moves from 2mil to 3mil. The rigid-flex routing window is looser than pure flex, so switching construction is not a pure gain.
  2. ≥100,000 cycles is the figure on the rigid-flex page and cannot be transferred to pure flex. They are different product constructions, and SKPCB does not publish a corresponding cycle life for pure flex. For the detailed construction comparison see flex versus rigid-flex construction boundaries.

5. Fields you must declare before review

Both the flex page and the rigid-flex page carry the same note: materials, stack-up, board and copper thickness, via construction, finish and acceptance criteria are interdependent, and fabrication data must be submitted for a project-specific manufacturability review. For bending specifically, the following cannot be inferred from Gerber and must be stated explicitly:

FieldWhy it is required
Minimum bend radius RThe only input to t_total ≤ R/6
Bend type: static / dynamicDecides whether 6× is a bound or needs margin added
Expected bend cycles for dynamic useNo published life figure for pure flex; requires project review
Bend direction and bend section locationDecides which conductors sit on the tension side
Maximum allowable total thickness of flex sectionA hard constraint if assembly clearance is limited
Whether copper weight is a full-layer or local current needDecides whether widening can replace thickening
Whether coverlay may be adjusted in the bend areaCoverlay enters the thickness budget

Anything not declared is judged against the default windows in the tolerance table and manufacturing capabilities. For the material system see the materials overview. Once parameters are complete, submit through the quote entry.

6. One worked example

Enclosure gives R2.0, dynamic flexing.

  1. Reverse the thickness budget: t_total ≤ 2.0 / 6 = 0.333mm.
  2. A dynamic application is not designed at the bound; after margin, plan to t_total ≤ 0.25mm.
  3. Allocate inside 0.25mm: PI base + copper layers + adhesive + coverlay. More layers and heavier copper both tighten it.
  4. If routing needs 6 layers and current needs 2oz, the budget will not hold. At that point you do not change routing — you return to the three questions in section 2, or switch construction per section 4.
  5. Put all seven fields from the table above into the review package, together with where the R value came from (enclosure clearance or hinge radius).

The order is: enclosure R value → thickness budget → layer and copper allocation → routing. Do it backwards and you discover the bend will not close after routing is finished, which means reworking the whole stack-up.

FAQ

Q1: The spec says bend radius ≥6× board thickness. Which thickness is that?

The total thickness of the flex section, including PI base, copper layers, adhesive system and coverlay. Not the base film alone, and not the rigid-area board thickness. For a rigid-flex construction it is the flex section that is measured.

Q2: I changed the radius from R3 to R1.5. How should routing change?

Do not start with routing. Halving R also halves the allowable total thickness of the flex section, from ≤0.5mm to ≤0.25mm, which is a stack-up level change. What moves is layer count, copper weight, or how many layers pass through the bend. Routing is the last step.

Q3: Does raising copper from 0.5oz to 2oz change the bend radius?

Yes. Copper counts toward total thickness, total thickness rises, and R_min = 6 × t_total rises with it. The R value on the drawing did not change, but the feasible minimum radius grew and may no longer be satisfied. If the copper increase was purely for current carrying, first confirm whether widening the conductor can replace thickening it — widening does not enter the thickness budget.

Q4: Rigid-flex states ≥100,000 bend cycles. What is the number for pure flex?

≥100,000 cycles is published on the rigid-flex page and belongs to that construction. The SKPCB flex page publishes no corresponding cycle life, so this article gives no figure. For dynamic applications, declare the expected bend cycles in the review documents for project-level confirmation.

Q5: Can I design both static and dynamic bends to ≥6× board thickness?

6× is the lower bound of the manufacturing window, not a recommendation for dynamic use. A static bend sitting near the bound carries relatively controllable risk. A dynamic bend designed at the bound has spent all its margin, and the failure typically does not appear at outgoing electrical test but after repeated in-service flexing. Bend type must be declared as a field.

Prototyping a flex or rigid-flex board?

Bend radius, coverlay and stiffener placement decide reliability. Tell us the bend cycles and assembly method and we will propose a construction with pricing.