Antenna Design for IoT Devices: PCB Trace Antennas vs Chip Antennas at 2.4 GHz

At 2.4 GHz on a compact IoT PCB, a well-executed PCB trace antenna (IFA or meander) performs within 1–2 dBi of a $0.15 chip antenna when the fab holds impedance tolerance and you respect keep-out. A chip antenna buys forgiveness on layout mistakes; a trace antenna buys BOM cost at the price of EM simulation time and fab sensitivity. We shipped both on the same product line (environmental sensor, nRF52840, BLE) and compared in an SV chamber and in field RSSI.

Candidates

Antenna	Part / type	BOM	Matching network
Chip	Johanson 2450AT42E010BE	$0.15 @ 10k	2-element PI, 0402
PCB IFA	Custom 12×3 mm inverted-F	$0	3-element PI, tuned per board

Stack-up: 4-layer, 1.6 mm, FR-4 Dk≈4.4, L1 RF feed, solid L2 ground, no components in 5 mm keep-out per Johanson app note (chip) and identical keep-out enforced for trace variant.

Simulation vs reality

Sim: Keysight PathWave ADS with vendor substrate model for chip; trace antenna drawn from Nordic reference then adjusted in ADS Momentum
Chamber: Satimo near-field, 2400–2483 MHz
On-air: RSSI at 10 m LOS outdoor, 100 packets/s, nRF52840 +0 dBm

Results (peak gain, mid-band):

Variant	Sim peak gain	Chamber peak	10 m RSSI median
Chip + correct keep-out	2.0 dBi	1.8 dBi	-68 dBm
Chip, VIOLATED keep-out (cap too close)	—	0.9 dBi	-74 dBm
PCB trace, 5 mil trace width	1.6 dBi	1.4 dBi	-70 dBm
PCB trace, fab ran 4.5 mil (out of tol)	—	0.7 dBi	-78 dBm

1–2 dB in chamber equals noticeable link margin at BLE sensitivity threshold.

Tradeoffs that mattered

Chip antenna

Pros: vendor characterization, less in-house EM expertise, smaller board area if vertical mount. Cons: placement rules sacred—one 0402 cap in keep-out cost 5 dB; height 2 mm affects slim enclosure.

PCB trace

Pros: no antenna line item, coplanar with ground easy on 4-layer. Cons: fab tolerance on trace width and dielectric constant shifts resonant frequency; every PCB spin may need matching tweak. Budget network analyzer time (NanoVNA V2 adequate for go/no-go, not for sign-off).

We added PI matching (Johanson 2450BM15A0018 matching chip on trace design for repeatability).

Layout rules we enforce in ERC

No copper pour under antenna void except ground reference per design
Feed via count minimized—one transition L1→L2 cost 0.3 dB in our A/B
Crystal 32 MHz keep away 8 mm from RF feed (clock harmonics showed in EMI bench testing)

Protocol choice (BLE vs Thread) affects acceptable margin—see MQTT vs CoAP on CC2652 for gateway-side sensitivity to dropouts.

Certification note

Modular approval path favors certified chip modules (FCC ID on module). Custom trace on nRF52840 SoC requires intentional radiator filing on final product—legal/compliance cost, not RF physics. Chip on module (nRF52840-QIAA module variant) vs chip antenna on custom PCB is often a business decision more than 1 dBi.

What I'd do next

Matching fixture in production test—reflectometer spot check VSWR < 2:1 at 2440 MHz, reject bad assemblies.
Dual-source fab qualification—trace antenna designs validated per fab Dk measurement.
Enclosure detuning study—plastic 0.8 mm above antenna shifted trace design 3 MHz; chip less sensitive in our sample of n=6.

Enclosure and human body detuning

Wearable prototype: hand proximity detuned PCB trace antenna −2 dB; chip antenna on ground plane extension less affected. Product marketing wanted smallest PCB—RF lost argument until field drop-rate data returned.

Anechoic vs office RSSI

Chamber gain within 0.5 dB repeatability; office RSSI varied 4 dB day-to-day (Wi-Fi congestion). Certify in chamber; validate in target environment with over-the-air packet loss metric, not RSSI alone.

Conducted vs radiated

Chamber results optimistic vs USB cable acting as radiator—FCC pre-scan showed USB 3.0 hub near DUT raised conducted emissions 3 dB on harmonic. Chip antenna benefit: smaller loop area, less cable coupling in cramped enclosure.

Second source antenna qualification

Johanson lead time spike forced Linx ANT-2.4-CHP alternative—required re-match PI network, one spin, $4k chamber time. Chip antenna modular approval easier to swap with pre-certified module than re-spinning custom trace.

Pre-certification sample size

FCC ID on module variant used 5 samples; custom trace design required 10 for variance on cheap fab. Budget chamber time in NPI schedule—trace antenna is not zero NRE.

Pick chip antenna when team lacks EM bench and enclosure is tight. Pick trace when BOM at 100k+ and you can own matching every spin. Neither works if keep-out is a suggestion.