PCB Warpage: Causes, Consequences and Controls

A board that is flat when it leaves the fabricator is not necessarily flat when it reaches the assembly line. Warpage develops across the whole life of a panel — during pressing, while solder mask and legend are cured, through the thermal shock of the surface finish, and again in the reflow oven — and it arrives at the placement machine as a shape the equipment has to cope with rather than a dimension the design specifies.

Warpage, Bow and Twist

The terms are used loosely, and it is worth separating them. A board can bend in a single direction like a shallow arc, which is bow, or it can be distorted diagonally so that corners do not lie in the same plane, which is twist. Both are measured against a reference plane, and both are expressed as a maximum deviation over the board’s own diagonal or length.

The measurement itself is not complicated: the panel is placed on a flat surface or in a fixture and the deviation is measured, with the result normalised to the board size so that the tolerance means the same thing on a small board and a large one. What makes it important is that the acceptance limit is a fabrication specification, and a board that exceeds it is a reject even when every electrical measurement passes.

Why the Material Moves

A laminate is a composite, and the constituents do not expand at the same rate. The property that describes this is the coefficient of thermal expansion, and it differs not only between copper and laminate but within the laminate itself between its plane and its thickness.

Copper expands by roughly seventeen parts per million per degree. The resin and glass laminate behaves differently, and its behaviour changes with temperature: below its glass transition temperature it expands through the thickness at a rate in the tens of parts per million, and above that temperature the rate rises several times over as the resin softens.

That difference is the origin of most warpage. A board with copper on one region and almost none on another will expand unevenly as it heats, because the two regions are made of different proportions of the same two materials. Nothing in the process can remove that difference; it can only be balanced by making the distribution more even.

Where the Stress Is Applied

Four process steps contribute most of the residual stress.

Pressing is the first. Resin cures at different rates in different parts of the panel depending on how the heat reaches them and how much copper they contain, and the layers are bonded while the whole stack is constrained. When the pressure is released and the panel cools, whatever asymmetry was built into the stack appears as a bend.

Mask and legend curing is the second, and it is easy to overlook. Baking temperatures are set to cure the ink, and if they approach or exceed the laminate’s transition temperature the resin is no longer rigid. In that state the panel is supported only by its own stiffness and by whatever fixture holds it, and gravity or airflow across an oven can deform it permanently.

Surface finish is the third. Processes that immerse the panel in molten solder and then cool it rapidly impose a thermal shock in which the two faces of the board are at different temperatures and the copper distribution on each face differs. The resulting differential strain is retained in the finished board.

Assembly adds the fourth. In the reflow oven, the panel is hot enough for the same effects to appear again, and a large board with heavy components is being supported by its own edges while gravity acts on the unsupported middle.

Design Choices That Cause It

Three layout decisions change how much the board will move.

Copper balance is the first. Large areas of copper on one side of a layer and almost none on the other create a difference in thermal expansion and in heat absorption across the panel. The same applies between layers: a design in which a power plane occupies one layer heavily and the opposite layer is mostly bare laminate is unbalanced. Copper distributed as evenly as the circuit allows, with areas of unused copper filled and connected, is the general remedy.

The stack-up itself is the second. Layers that are symmetric about the centre line, with matching thicknesses and matching prepreg types, are less prone to warping than a stack whose construction is heavier on one side than the other.

Mechanical features are the third. A V-cut intended to make the board easy to separate also interrupts the material that gives the board its stiffness, and the resulting line of reduced section concentrates stress. Where the shape of the board is irregular or the panel is large, the mechanical design has to account for the fact that the stiffest parts are not necessarily where the loads are.

At the extreme, this is a matter of layer count and thickness together. Very thick, high layer count boards are rigid but are pressed with high forces and accumulate large internal stresses; controlling their flatness is described in the note on warpage control for large thick boards.

Storage and Handling

Warpage can also be imposed mechanically after the board has been made. Thin panels stacked flat sag under their own weight and take a set. Panels stored in a rack that grips them too tightly are held under a bending load. Panels that are handled by an edge while long and thin bend far enough to exceed their elastic range.

The practice for thin boards is to store them vertically in a rack that supports them along their length and does not clamp them, to keep them in their packaging until they are needed, and to avoid stacking them under load.

What Warpage Costs

The cost appears at several stages.

In assembly, a non-flat board does not sit at a known height under the placement head, and the result is placement offset at best and contact with the board surface at worst. In the oven, a board that is already bowed tends to sag further, which changes the temperature the joints see and can leave joints at the centre of the panel under-heated.

In the finished product, components mounted on a warped board carry stress at their solder joints from the moment they are assembled, and that stress adds to whatever the product imposes in service. Cracks that appear later in a joint or under a large package often trace back to a board that was never flat.

In the factory, a board that does not meet the flatness specification has to be scrapped or reworked, and both of those are more expensive than the measures that would have kept it flat.

Keeping a Board Flat

The controls follow directly from the causes, and they are spread across three parties.

In design: distribute copper evenly on each layer and between layers, use a symmetric stack-up, avoid unnecessary mechanical features along lines of high load, and consider the panel layout with the board’s rigidity in mind rather than only its material usage.

In fabrication: control the press cycle so that cure is even across the panel, keep the temperatures of subsequent baking operations below the point at which the laminate loses rigidity, and support the panel through processes that involve large thermal excursions.

In handling: store thin boards vertically and unclamped, and avoid stacking them under load.

The review that catches the design-side contributors is the same one that catches other manufacturability issues. Our layout quality checklist covers the checks to run before the files are released, and the assembly consequences of a board that was not flat are covered in the guide to SMT PCB assembly.

FAQ

Is warpage a fabrication defect or a design problem? Usually both. The design determines the imbalance the process has to cope with, and the process determines how much of that imbalance becomes permanent deformation.

Does a thicker board warp less? A thicker board is stiffer, so it deflects less under the same stress. It does not have less internal stress, and thick high layer count boards are pressed with high forces that can leave more of it.

Can a warped board be flattened? Not reliably. Applying heat and pressure to flatten a finished board may bring the shape within tolerance temporarily and will not remove the underlying stress, so the deformation tends to return during assembly.