A warped lot reaches the line and the argument starts. The board supplier says the panels were flat when packed; the assembly house says the first panel would not sit on the printer. Both can be right, because warpage appears or grows after heat. Unless the purchase order states a limit, a method and the point of measurement, there is nothing to measure against.
The limits buyers should quote
| Board type | Warp and twist limit | Why this number |
|---|---|---|
| Boards carrying surface mount parts | 0.75% maximum | The figure in IPC-6012 and the one assembly houses quote |
| Through-hole boards only | 1.5% maximum | Wider allowance because wave solder tolerates more |
| HDI and rigid-flex builds | 0.5% maximum | Thin cores and stacked microvias move more under heat |
The percentage is taken against the diagonal, which is where most specifications fall apart. A 0.75% limit on a 300 mm diagonal allows 2.25 mm of movement, generous. On a 0.6 mm thick board with a 100 mm diagonal it allows 0.75 mm, and the board behaves like a potato chip. The percentage has to travel with the diagonal, the thickness, and a statement of whether it applies to the panel or the single board.
How the number is measured
IPC-TM-650 method 2.4.22 puts the board on a flat surface, supports it at three points, and measures the greatest distance between board and plane. Twist is read with the board resting on two diagonally opposite corners so the other two lift. The reading is divided by the diagonal and expressed as a percentage.
Three conditions change that reading, and a supplier who does not state them is quoting a number nobody can compare:
- When it is measured. After a reflow simulation or a bake, not only on the day of packing. Lamination stress and absorbed moisture both relax with heat, so a board that read 0.4% at goods-in can read 0.8% after the first pass.
- Before or after depaneling. A panel can be flat while the single board inside it is not. V-score depth and tab count decide how the piece behaves once separated.
- Room conditions. A board measured in a dry cold room does not read the same as one measured on a humid floor.
The clean way to close this is to put four things in the drawing: the limit, the method, the condition, and the sample size, for example five pieces per lot with the reading in the outgoing report.
What actually drives warpage
- Copper imbalance. Heavy copper on one side, or a ground plane at layer 2 with nothing comparable near the bottom, pulls the board as it cools. This is decided before the first panel is etched, and it is the strongest single lever.
- Asymmetric stackup. Different prepreg styles above and below the center line, odd layer counts, or cores of unequal thickness. A balanced build with the center line in the middle of the thickness range stays flattest.
- Mixed laminates. A PTFE or hydrocarbon RF material bonded to FR-4 puts two expansion rates in one board, so the hybrid moves at reflow in a way neither material does alone.
- Thick copper and large area. Four and ten ounce outer layers etch differently and carry more stress, and a 500 by 600 mm board multiplies any imbalance by its diagonal.
- Mechanical features and press cycle. Long slots, large cutouts and V-score cut through most of the thickness create local hinges, while cure profile, cooling rate and lamination cycles accumulate stress.
- Moisture. Boards absorb water, swell, then move during reflow. Baking removes it, which is why a reading taken after a bake is the honest one.
What an assembly line can absorb
| Symptom on the line | What it usually means | What reduces it |
|---|---|---|
| Paste smeared along one edge of a panel | Panel not seating on the printer supports | Support pins under the panel, vacuum block, flatness checked on the panel |
| Parts shifted or tombstoned at one end | Board lifting or heating unevenly in reflow | Better support, warpage read after the first pass, profile reviewed |
| Head-in-pillow on a BGA | Board or package moving at reflow temperature | Coplanarity checked on both, profile adjusted, CT scan on the joint |
| Second side fallout, or handler jams on singulated boards | Warp from the first reflow, or warp released by depaneling | Carrier or pallet, thermal balance reviewed, tab count and V-score depth reviewed |
A line can usually work around 0.5 to 0.75% with support tooling, carriers and profile work. Beyond roughly 1%, especially on a thin board, the loss stops looking like a clean reject and shows up as placement errors, rework and scrap on the second side. That is why the limit belongs in the purchase order.
How SUPERB PCB and SUPERB Assembly handle it
SUPERB PCB builds bare boards to IPC-6012 and ships the standard report set with every lot: outgoing inspection record, microsection for layer count and plating thickness, impedance coupons per panel where controlled impedance is specified, and the flatness reading when the drawing asks for one. Board thickness runs 0.6 to 3.2 mm on the standard line and 0.1 to 12 mm on the extended side, multilayer work goes to 68 layers, outer copper tops out at 10 oz, and hybrid RF plus FR-4 stacks have up to seven lamination cycles available.
SUPERB Assembly verifies incoming boards against IPC-6012 before release to the line, so flatness is answered at goods-in rather than after the first panel fails. X-ray counting confirms reel quantities, moisture sensitive parts go through a 125 C bake to J-STD-033, and finished assemblies are inspected to IPC-A-610 Class 2 or Class 3 with AOI, X-ray and ICT or FCT.
- Multilayer PCB: layer counts, stackup balance and the limits that keep a thick build flat.
- High reliability PCB: where flatness, plating and inspection records carry more weight than price.
- Environmental reliability testing: thermal cycling and the tests that expose a marginal build.
- PCB solutions: the board types and finishes available for a quote.
Frequently asked questions
Is 0.75% enough for a thin board?
Often not. On a 0.6 mm board the percentage allows very little absolute movement, so the board can pass on paper while still failing on the printer. Below 1 mm, add an absolute limit in millimetres and state which one governs acceptance.
Does baking fix a warped board?
It fixes the part caused by moisture and relieves some lamination stress. It does not fix copper imbalance or an asymmetric stackup. If a lot straightens after a bake and warps again after reflow, the cause is in the build.
Should the limit apply to the panel or the single board?
Both, with separate numbers. The panel is what the line handles; the single board is what the customer receives. State in the drawing which of the two is measured for acceptance, otherwise two readings can both be true and neither is usable.
Next step
Send the Gerber data, the stackup with copper weights per layer, and the assembly drawing. The reply states the flatness limit, the method and condition used to measure it, the lot sample size, and the reports that ship with the lot. Controlled impedance and mixed laminate builds come back with coupon data.
