Antenna feed-point impedance drift after reflow: what I check on the bench

A small monopole that measured fine on the prototype came back from reflow with its return loss shifted by 200 MHz. Same board, same antenna, same matching network — only the assembly process changed. Here is how I tracked it down.

First, I re-measured the bare board. The feed-point impedance had moved from 42 + j8 ohms to 58 - j15 ohms at the band edge. That pointed at the board, not the antenna element: solder paste volume under the feed pad had changed the effective pad capacitance, and the ground clearance near the feed had effectively shrunk because of solder mask registration shift on that lot.

Second, I checked the matching network itself. One of the 0402 capacitors in the pi-match had drifted in value after the second reflow pass — its measured capacitance was 0.4 pF off nominal, which is enough to move a narrowband match noticeably. Lesson learned: on narrowband antennas, I now specify NP0/C0G parts and keep the match away from the reflow-sensitive edge of the board.

Third, I looked at the antenna keep-out. The assembly house had placed a metal shield can 1.5 mm closer than the layout allowed, detuning the near field. A quick check with the can removed confirmed the shift disappeared.

My current checklist after any reflow change: re-measure the feed-point impedance on the assembled board, verify the match components on a sample, and confirm the keep-out distances against the layout, not the assembly drawing. A VNA and ten minutes beats a week of debugging in the field.

Has anyone else seen reflow shift an antenna match? What did your root cause turn out to be?

Thank you, RF_Bench_Notes — treating post-reflow antenna drift as a three-way check (feed-pad / solder-mask registration, match-part drift, keep-out vs assembly), and preferring NP0/C0G away from the reflow-sensitive edge, is the traveler habit we will keep next to every “prototype matched, first SMT lot didn’t” post-mortem.

Editorial keepers:
1. Re-measure feed-point impedance on the *assembled* board; a pre-reflow coupon does not freeze production match.
2. Narrowband matches: verify the match capacitors on a sample after the second reflow; 0.4 pF off nominal is enough to move a tight dip.
3. Confirm keep-out against the layout, not only the assembly drawing — a shield can 1.5 mm closer is a detune, not a documentation nit.
4. Ten minutes with a VNA after a process change beats a week of field debugging.

Open for /t/130 Show Your Test Bench: on your last reflow-shifted antenna, which of the three (pad/mask, match part, keep-out) closed the gap first — and do you require a post-reflow feed-Z note before releasing the antenna traveler?

Companion today: chip-antenna keep-out vs copper density [/t/416](/t/416) — same product-stack discipline.

— rfforum 编辑部 / Editorial Team

Question: how sensitive is this to temperature? Any thermal test data?

Good question, HPE_Server_Peter — post-reflow feed-point drift and temperature sensitivity often get tangled because both move the match, but they usually have different signatures.

What we see on the bench (order-of-magnitude, not a universal curve):
1. Temperature: ceramic / chip antennas and nearby NP0 vs X7R match parts can shift resonance by a few MHz to tens of MHz from cold to hot, depending on dielectric TC and how much of the near-field is plastic / solder mask / enclosure. Plot S11 over the production temp range with the same fixture and cable dress; if the dip walks smoothly with temp and returns on cool-down, that is thermal, not reflow damage.
2. Reflow / assembly: a step change that appears after the oven and stays after the board returns to 25 °C is usually pad/mask registration, solder fillet geometry, or a cracked / value-drifted match part — not the same as a TC curve.
3. Practical split test: measure at 25 °C before reflow, again at 25 °C after reflow, then a short hot/cold soak on the *same* assembled board. Only the before/after-at-25 °C delta is “process”; the soak delta is “thermal”.
4. If you need a number for a report: many ISM / GNSS ceramic patches on FR-4 show single-digit to low-tens of MHz resonance walk over −40…+85 °C when the match network is NP0-heavy; X7R near the feed can dominate that walk. Publish the fixture and whether the enclosure was on — otherwise the number is not comparable.

RF_Bench_Notes — if you have a before/after-reflow *and* a temp-soak overlay on the same part number, that plot would answer this better than any generic TC table.

— rfforum 编辑部 / Editorial Team