BLE Antenna Tuning in a Plastic Enclosure — From Smith Chart to Compliance

Almost every BLE programme we audit starts with the same surprise: the dev-board antenna pattern looked clean, the production housing arrives, and the pattern collapses. Range drops 30–60%, peak gain falls 4–8 dBi, and the FCC chamber report comes back with a spurious 4 dB out of mask. Then the firmware team gets blamed because "nothing else changed."

What actually changed is the loaded environment of the antenna. A Nordic nRF52 or TI CC2640 reference board has the antenna sitting on the edge of bare FR-4, surrounded by 30 mm of air on every side and a calibrated ground plane underneath. Your production housing has:

• Plastic with ε_r between 2.8 and 4.5 within 2 mm of the antenna — pulling the resonant frequency 80–200 MHz lower (a F-antenna at 2.45 GHz becomes a 2.30 GHz antenna).
• A battery within 5–10 mm — a large metal mass that detunes the antenna and modifies the radiation pattern, often creating a null in the direction you most need.
• A display or sensor cluster that wasn't on the dev board — every conductive surface is a parasitic resonator.
• An LCD backlight ground that connects through the chassis with a different impedance to the BLE ground reference at 2.4 GHz than at DC.

"We had a customer ship 8,000 fitness trackers before realising the leather strap option had a metalised buckle that detuned the antenna by 140 MHz. The plastic-strap variant passed compliance; the leather-strap variant didn't. Both came out of the same factory in the same week." — Pioneer Horizon antenna engineering team

The fix isn't to over-engineer the antenna. The fix is to design and tune the antenna inside the housing it actually ships in, with the components it actually sits next to. That requires three things that the dev board can't give you: a parameterised EM model, a matching network with placeholder components, and a chamber-correlation loop.

Four families dominate BLE-class designs. Picking the wrong one for the housing geometry costs more in firmware workarounds than the right one costs in BOM.

PCB trace antenna (F or inverted-F)

The default reference design from every silicon vendor. Cheap (zero BOM cost beyond the matching network), reasonably efficient (-2 to -3 dBi peak), but extremely sensitive to ground-plane geometry and adjacent dielectric loading. Use these when you control the ground plane and there is at least 8 mm of clearance from the antenna to any conductive surface.

Ceramic chip antenna

1.6 × 0.8 mm SMD parts that come pre-tuned for a specific ground plane size. Examples: Johanson 2450AT43, Murata LDA21. Peak efficiency 50–60% when ground plane matches the datasheet (typically 30 × 40 mm minimum). When the ground plane is smaller or the housing intrudes, you tune through an external L-C network on the feed line.

PIFA (planar inverted-F)

Larger metal antenna often realised as a stamped plate or laser-direct-structured (LDS) feature on the housing itself. Peak gain 0 to +2 dBi, much less sensitive to nearby dielectrics because the antenna volume is larger and the radiation is more 3D. Best choice for wearables and devices where the housing constrains a trace antenna. Higher BOM cost, requires mechanical integration.

LDS / molded-interconnect-device antennas

Laser-direct-structured copper directly on the housing inner wall. Highest design flexibility, highest tooling cost. We reserve LDS for products shipping more than 50,000 units/year where the antenna shape needs to wrap around an awkward enclosure curve.

The selection matrix we use

• Plastic case < 5 mm to antenna, controlled ground plane ≥ 30 mm: ceramic chip antenna.
• Plastic case > 8 mm to antenna, ground plane ≥ 40 mm, low cost critical: PCB F-antenna.
• Wearable, metal/battery within 3 mm of antenna, range critical: PIFA or LDS.
• Mixed material housing (plastic + metal trim), variable battery position: PIFA with tuneable matching.

Tuning a BLE antenna means moving its input impedance to 50 Ω at 2.44 GHz (centre of the BLE 2.4 GHz ISM band) while keeping VSWR < 2:1 across 2.40–2.48 GHz. With dev-board geometry that's usually a one-component fix. With production housing it's a deliberate optimisation across three components.

The pi-network you start with

Series-L, shunt-C, series-L is our default topology for BLE-2.4. The first series-L moves clockwise around the Smith chart (resistive + inductive); the shunt-C moves clockwise on the constant-conductance circle (capacitive); the second series-L fine-tunes the final landing. Three components, three knobs, very controllable.

Measurement setup

• VNA calibrated to the antenna feed point — not the chip pin. Use a 50 Ω coax pigtail soldered at the feed pad, length compensated in calibration.
• Antenna in the actual housing, with battery installed, sensor cluster connected, and the housing closed. Tuning an open housing then closing it for the chamber is the most common mistake we see.
• Sweep 2.30–2.50 GHz, 401 points. Capture S₁₁ and group delay simultaneously.

Tuning loop

1. Mark the un-tuned impedance on the Smith chart at 2.44 GHz. Most designs land in the lower-right quadrant — capacitive and below 50 Ω — after housing loading.
2. Solve algebraically for series-L1, shunt-C, series-L2 values that land at the centre. We use a Smith chart calculator (the free Iowa Hills tool works well) and round to E12 values that are actually on our reel.
3. Populate those three components. Re-measure. The first iteration usually lands within 5 Ω of the centre.
4. Trim. Swap one component at a time, never two — otherwise you can't tell which one moved the dot.

Document the final values in the BOM as antenna-tuning components, not generic passives. We've seen production substitutions of "a 1.5 nH for a 1.8 nH because the line ran out" detune the antenna by 50 MHz and push devices out of FCC mask.

Tolerance budget

0402 inductors at 1–10 nH carry ±5% tolerance typical. Across a three-component network that compounds to roughly ±30 MHz at the resonance peak. We always tune with the centre of the tolerance window and verify on the worst-case corner units from the first build.

The fastest BLE programmes we run pair pre-build EM simulation with chamber measurement on the first physical prototype. Skip either and you're either guessing or iterating expensively.

What we model in EM-sim

The model includes the antenna, the full PCB stack-up, every component within 8 mm of the antenna feed (including its body permittivity), the battery as a copper block, the housing as a parameterised dielectric, and any metal trim or screws within 15 mm. We use HFSS for parametric sweeps and CST for time-domain when we need to characterise impulse response (rare on BLE).

Outputs at this stage:

• S₁₁ sweep 2.30–2.50 GHz at the chip pin — used to specify the matching network starting point.
• 3D radiation pattern at 2.44 GHz — used to confirm the antenna polar pattern in the housing has no deep null in the device's expected orientation.
• Total radiated power (TRP) estimate against an isotropic reference.
• Specific Absorption Rate (SAR) estimate for wearable products against a tissue-equivalent phantom.

Chamber measurement on the first build

We bring three prototypes to the anechoic chamber for the first measurement run — one nominal, one with worst-case component tolerances, one with worst-case housing assembly tolerance (mis-aligned snap-fit, off-centre battery). Each gets the same scan:

• 3D radiation pattern with 5° resolution in azimuth, 10° in elevation.
• TRP integration across the BLE band.
• Conducted emissions at the antenna feed (with a known cable loss) per FCC 15.247.

Correlation criterion

Simulated and measured TRP agree within 1.5 dB; simulated and measured pattern peak direction agrees within 15°. If they don't, the model is missing a coupling structure — usually a flex cable or a previously ignored screw — and we update the model before any production tooling commits.

We run this loop in roughly 8 engineering days from EM-sim kick-off to chamber correlation. Customers who try to skip it typically lose 6–10 weeks at compliance.

An antenna that works in the chamber on three hand-built prototypes is not yet an antenna that works on 50,000 mass-produced units. Five disciplines we wire into the production process before any BLE design ships.

1. End-of-line antenna test

Every unit, no exceptions. A simple coupled-loop fixture on the SMT line measures S₁₁ at 2.44 GHz with the housing open. Limits are ±2 dB of nominal. Units outside the window are kicked out for matching-network rework. Setup cost: roughly ₹180,000 for the fixture + software; recurring cost: 2.5 seconds per board on the line. Yield improvement on a representative wearable programme: 96.2% → 99.4%.

2. Housing-closed verification on a sample

Once per shift, pull three random units and test them in a small anechoic test cell with the housing closed. Verify peak TRP within 1 dB of the chamber-validated baseline. This catches housing-assembly drift (snap-fit tolerance, screw torque variations) that the bench S₁₁ test can miss.

3. FCC/IC/CE strategy

• BLE 2.4 GHz devices fall under FCC 15.247 (intentional radiators) and EN 300 328 for CE.
• If you use a pre-certified BLE module (Nordic, u-blox, Murata), the radio is already FCC-listed and you only need to declare the integration — but only if the antenna is the one tested with the module. Change the antenna and you lose modular approval.
• Custom antennas require full intentional-radiator testing. Plan 8–12 weeks and roughly $25,000 in chamber + lab fees.

4. Long-tail failures we've seen

• Adhesive that out-gassed and crystallised on the antenna feed pad over six months — drove range loss in 2% of the field fleet. Switched to a non-out-gassing adhesive grade.
• Hand cream and sunscreen residue on the antenna of a fitness tracker — caused detuning in some user populations. Mitigated with an additional 0.5 mm of housing thickness above the antenna.
• Battery swelling at end-of-life moving 0.4 mm closer to a PIFA antenna — detuned by 60 MHz and dropped throughput by half. Increased the clearance spec in the next generation.

5. The hand-off to the customer

Every BLE programme we ship comes with a tuning report, a 3D radiation pattern, and a matching-network change procedure for future housing revisions. If you're starting a wearable or consumer BLE build, see our companion piece on RF shielding and thermal trade-offs for the heavier-power analogue, or share your mechanical CAD and BLE silicon choice with our team and we'll come back with an antenna-and-housing co-design plan inside five working days.