Aim for 50 to 70 micrometres of finished solder under the pad, and get there with area, not with one big opening. On a 0.12 mm stencil that thickness needs roughly 85 percent paste coverage; if the pad is split into a windowpane to let gas out, the same volume needs a 0.15 mm stencil at about 70 percent.
The arithmetic is the whole argument. Solder paste is about half metal by volume and half flux and solvent, so a deposit of 1.0 cubic millimetre leaves about 0.5 cubic millimetres of solder. On a 5 by 5 mm pad, 25 square millimetres, that is 20 micrometres of joint. To reach 60 micrometres the deposit has to be 3.0 cubic millimetres, which at 0.12 mm stencil thickness is 25 square millimetres of aperture, in other words the full pad.
Why the pad cannot simply be printed solid
Flux gas has to leave. When the paste melts, the flux boils off and the volatiles from the pad finish and from any residue underneath expand. Under a solid print the only escape route is the perimeter, and the centre of a 25 square millimetre pad is a long way from the perimeter. The gas collects there, the solder cannot wet through it, and the void sits exactly where the thermal path is supposed to be.
A windowpane fixes that by giving the gas a channel every few millimetres. Four openings on a 5 mm pad, or nine on a 10 mm pad, with webs of 0.20 to 0.30 mm between them, let the volatiles out while the solder is still liquid and before the joint closes. The webs also control the paste height: they act as a dam and stop the deposit from slumping into one pool while the stencil is still on the board.
The cost of the windowpane is area, which is why the stencil thickness has to go up to compensate. Take the same 5 by 5 mm pad at 70 percent coverage: 17.5 square millimetres of aperture. At 0.12 mm that is 1.05 cubic millimetres of solder, about 42 micrometres of joint, on the thin side. At 0.15 mm it is 1.31 cubic millimetres, about 52 micrometres, which is inside the window. The stencil thickness is doing the work that the webs took away.
There is an upper limit as well. Too much paste and the package floats on a molten pool, the perimeter pins at 0.5 mm pitch do not reach their pads, and the failure is an open on the outer row with a perfectly good thermal pad underneath. That failure mode is why the thermal aperture and the pin apertures have to be treated as one set: the pin aperture area ratio at 0.4 mm pitch is already at the 0.66 release limit, so the pins cannot carry extra paste to make up for a starved pad.
Three ways the deposit goes wrong
One aperture over the whole pad. The area ratio is fine, the print looks perfect, and the void ends up in the middle. A single opening also slumps more, so the joint is thick at the edges and thin at the centre, which is the opposite of what the thermal path needs.
Ignoring the vias under the pad. If the thermal vias are open or only partly filled, paste wicks into them during reflow and the pad is starved from below. Our via fill recess limit is 5 micrometres for vias that must stay flat, and a via with a 60 micrometre dimple will take paste out of the joint. Either fill and cap, or expect the void.
Printing onto a part that was not baked. A moisture sensitive package that has been out of its bag takes the paste heat and gives it back as steam at the interface. The profile looks correct, the reflow is clean, and the void is at the die paddle. Baking to J-STD-033 for the package level is part of the paste calculation, not a separate step.
What SUPERB Assembly controls
Paste volume is measured per aperture with 3D SPI before reflow, not judged by eye after, and the thermal pad apertures are on the SPI program as their own group with their own limits. X-ray void reporting follows IPC-7095 with a 5 percent limit on thermal pads, and the report separates the thermal pad from the pin rows so that a good pin result cannot hide a bad pad. Reflow is a twelve zone SAC305 profile under nitrogen below 500 ppm oxygen, recorded per run.
First articles of a QFN build get a cross section through the thermal pad, because X-ray shows the void area and the section shows the joint thickness. A pad that is 90 percent wetted at 25 micrometres passes X-ray and fails thermally, and only the section separates the two. The same discipline runs through turnkey PCBA assembly and the inspection sequence described in reading an X-ray report.
The honest boundary: the 50 to 70 micrometre target assumes a pad of 5 to 10 mm on a side and a package with a flat die paddle. A large exposed pad on a thin package tolerates less, because the package itself bows, and the answer there is a thinner joint and a stiffer stencil support, not more paste. Where the datasheet gives a standoff, the standoff wins over any rule of thumb.
What to send with the BOM
The package outline with the exposed pad dimensions, the pin pitch and pin width, the stencil thickness you expect and whether it is stepped, the thermal via pattern and whether the vias are filled and capped, the moisture level of the package, the void limit you will accept on the thermal pad, and whether the board carries a heatsink on the other side. Those seven items decide the aperture set before the first print, and changing them afterwards means a new stencil.