White spots inside the laminate are measling: the resin has pulled away from a glass yarn bundle and left a small lens of air that scatters light. It is not a stain on the surface, it cannot be wiped off, and it is not delamination, which is a separation between a copper layer and the resin or between two plies. Measling matters in two situations only. When the lens sits between two conductors on a high voltage gap, and when it grows after thermal stress, because at that point it is the first stage of a delamination that will not stay cosmetic.
IPC-A-600 treats measling as a workmanship condition with limits on size and count per unit area, tied to whether it crosses or reduces conductor spacing. A few small spots in a clear area between traces pass in every class. The same spot inside a 0.5 mm gap on a 300 V circuit does not, because an air lens in the dielectric is a partial discharge site and a path for moisture, and the gap the customer calculated is no longer the gap the board has.
How the air lens forms
FR-4 is glass fabric plus resin, and the resin has to fill the space between yarn bundles at every crossover during lamination. Three things stop it. Prepreg that has taken up moisture, stored above 23 C and 50 percent relative humidity, releases steam at the 180 to 200 C press temperature and the vapour occupies the space the resin needed. Resin content that is low for the weave, often on a cheap high Tg product, leaves the crossovers starved before the press closes. And a press cycle that applies full pressure before the resin reaches flow temperature pushes resin to the panel edge, so the centre of a large panel measles while the border looks perfect.
The other half of the story is thermal. Z axis expansion is 50 to 70 ppm per C below Tg and 200 to 300 ppm per C above it, so a panel that goes through reflow at 245 C moves its thickness by around 40 um on a 1.6 mm stack. An air lens between yarns is a stress raiser with no strength at all, and repeated passes grow it. That is why the spots are usually found after assembly rather than at incoming inspection: the first reflow is what makes a marginal laminate visible.
Why incoming inspection misses it
A measled area reads as a faint cloudy patch under the surface layers, seen through translucent laminate under a strong lamp, and it is easy to confuse with a handling mark from the AOI lamp or a shadow from the weave itself. It becomes obvious after the first reflow, when the lens has expanded and the resin around it has moved. The practical test is a microsection at 200x plus a thermal stress screen: solder float at 288 C for 10 seconds, or the T260 and T288 times by TMA, where a laminate that measled in the press usually shows a short time to delamination at temperature.
Three reactions that make it worse
Baking the assembled board to dry it out. Measling is a void, not moisture. A bake above Tg spends thermal life that the panel cannot get back, and it does not close the lens. The bake that belongs in the process is the J-STD-033 floor life bake before the first reflow, at 125 C, driven by the moisture sensitivity level, not a bake after the symptom appears.
Scrapping or accepting without a section. A white mark can be a lens between yarns, a fine crack in the resin at a bundle boundary (crazing), a gap between plies (delamination), or a mark on the surface. The disposition is different for each, and the only way to tell them apart is a polished section. Sorting a lot on a visual call leaves the real cause in the press.
Calling it cosmetic on a high voltage net. Clearance and creepage were calculated on solid dielectric. A 0.2 mm white lens inside a 0.5 mm gap is a reduction in the dielectric and a place where tracking starts after a few hundred hours of damp heat. The cost of a re-laminate is trivial next to a field return on a 400 V board, which is why the acceptance call belongs to the customer's QA against their own spacing.
What keeps it out of the panel
Prepreg stored cold, brought to room temperature sealed, and used inside floor life. Vacuum bag lamination. A press cycle that applies kiss pressure until the resin reaches flow, then full pressure, with a ramp held around 2 to 3 C per minute so volatiles leave the stack before the resin gels. Resin content and volatile content checked per lot rather than per supplier. And a section from the panel centre as well as the edge, because the centre is where starvation shows first on a large panel.
On the SUPERB side the laminate is A grade from Shengyi, Kingboard and Jin'an Guoji with Tg 130, 150 and 170 stocked, builds run up to seven press cycles, thickness is held to plus or minus 0.10 mm below 1.0 mm and 10 percent above, warp to 0.75 percent, and boards ship with a section report so the laminate condition can be read rather than argued. Assembly lines bake to J-STD-033 at 125 C before the first reflow, and the quality system covers IPC-6012 and IPC-6018 along with ISO 9001 and ISO 13485 for medical work. Laminate windows are on FR-4 process parameters, layer count and press limits on multilayer PCB, and the reliability side on high reliability PCB.
Who makes the acceptance call
The honest boundary: acceptance of a measled area is not ours to grant. We can show the section, the T260 and T288 data and the panel position, and say whether the pattern is press related or handling related. Whether a given spot is acceptable on a given net is a decision that has to be made against the customer's own spacing and voltage.