An antenna in package puts the radiating element inside the module, not on the motherboard. The package becomes the RF front end: the antenna, the feed, the matching network and often the transceiver die all sit in one overmolded or cavity package that is reflowed onto the board like any other component. For the board house this changes the question from "can you route 28 GHz" to "can you place and reflow a module in which every 0.1 mm of warpage matters".
1. Why AiP exists
At 28 GHz and above, the loss between a package and a board-level antenna is large enough to matter. The transition from the die to the package substrate, then to the PCB, then to the antenna, adds loss and reflection at every interface. Moving the antenna into the package removes one interface entirely and makes the module a characterized part: the manufacturer tests it as a unit, and the board is a carrier.
That is the trade. The board design gets simpler and the assembly gets harder, because the module now has tight requirements on coplanarity, thermal path, paste volume and placement accuracy.
2. What the module needs from the board
| Requirement | Typical value | What drives it |
|---|---|---|
| Pad coplanarity | Within 0.05 mm across the land pattern | Module warpage plus board warpage; both stack |
| Placement accuracy | Plus/minus 0.05 mm | Pad pitch and the number of peripheral pads |
| Paste volume | Controlled per pad, uniform across the array | Ground pad area and the risk of tilting |
| Thermal path | Via array under the die area, filled and capped | Power density at the die, not the module average |
| Board flatness | Bow and twist within 0.5 percent after reflow | Module stiffness and the solder joint height |
The coplanarity number is the one that gets missed. A module that is flat when it arrives can be bowed by the reflow profile, and a board that is flat before assembly can warp when the copper on one side is heavier than the other. The two effects add.
3. Stencil and paste
AiP modules usually have a large central ground pad plus a ring of signal pads at fine pitch. That combination forces a stencil decision: a thick stencil for the ground pad and a thin one for the fine pitch. The common solution is a stepped stencil, and the step position has to be designed around the module outline.
Paste on the ground pad is usually partitioned into a grid of small apertures rather than one large opening. This controls the paste volume and gives gas a path out during reflow, which reduces voiding under the die. The void target for an RF ground pad is tighter than for a mechanical one, because a void under the die is both a thermal and an electrical discontinuity.
4. Reflow
The module's own thermal mass matters. A package with a metal lid and a large die heats more slowly than the surrounding board, so the profile has to be measured on the module, not on a test coupon beside it. A thermocouple on the board next to the module reads several degrees lower than the module itself in the soak zone.
Warpage during reflow follows the temperature, not the room. The peak zone is where the module and the board are both softest, and that is where a marginal coplanarity becomes an open joint at one corner. Slower ramp and a longer soak reduce the gradient; too slow and the flux is exhausted before the joints form.
Nitrogen is common for these assemblies. It reduces oxidation on the fine pads, and on a module with a large ground pad it improves wetting spread, which is exactly what you do not want too much of.
5. Underfill and shielding
Some AiP modules require underfill for thermal cycling, especially when the module is large relative to the board and the CTE mismatch is significant. Underfill adds a cure step and makes rework impossible, so it is a decision made with the module vendor, not by the assembly house alone.
Shielding is usually inside the module, but the board may still need a fence or a can around it. A shield can soldered after reflow introduces heat next to a module that has already been reflowed once. The thermal budget has to be checked, and a clip-on or screw-down shield is often the safer choice.
6. Inspection and test
X-ray shows the void percentage under the ground pad and the alignment of the peripheral joints. It does not show a joint that is connected but cracked. Cross-section on first article does, and it is the only way to see the joint shape at the corner where warpage is worst.
RF test is the real acceptance criterion, and it is not a board-house test. It requires a test fixture, a calibrated setup and a known-good reference. On AiP assemblies we run the electrical and functional test the customer specifies, using their fixture or a fixture built to their drawing, and we report the pass record against the fixture. The RF measurement itself is done on a calibrated RF bench, either at the customer's facility or with a partner laboratory; it is not something a PCB assembly line performs.
7. Where AiP assembly goes wrong
Four patterns, all seen in production. A module placed with the correct XY but a 0.02 mm tilt, which produces a good X-ray and a marginal RF result at one edge of the band. A ground pad paste volume that is right on average and wrong at the corners, which produces voids where the die is hottest. A reflow profile measured beside the module rather than on it. And a board whose copper balance warps after the first reflow, so the second side fails.
None of these are visible at incoming inspection. All of them are controlled by the process window agreed before the first build.
Send the module outline, the land pattern and the vendor's process recommendations and we will confirm the stencil, the profile and the inspection plan before the pilot run.
Related reading
- Via in Pad with Epoxy Fill for RF Boards
- GaN SSPA PCBA Manufacturing: Thermal, Assembly and Test Boundaries
- RF PCB DFM Checklist Before You Release the Gerber
- RF Laminate Dk Drift with Temperature and Humidity
- PTFE PCB Process Control
Need a DFM check or a stack-up review on your files? Write to pcba@superb-tech.com or WhatsApp +86 133 9608 1443.