Kyocera's New Antenna Tool Picks the Part — But It Won't Fix Your BLE Range Without These Layout Rules

You picked a chip antenna with 70% efficiency on the datasheet. Your prototype measures 15 dB down from where the link budget said it should land. The part didn’t lie. Your ground plane did.

Vendor selection tools, including Kyocera’s antenna selector, are good at one job: narrowing a catalog by frequency, footprint, gain, and efficiency spec. They hand you a shortlist. None of that shortlist survives contact with a bad copper pour, a shield can 2 mm too close, or a matching network shoved off in a corner because it was easier to route that way.

This is a layout checklist for BLE antenna tuning that applies no matter which part you chose or which tool pointed you at it. If your node lives somewhere you can’t easily reach it (a remote sensor, or something as unreachable as a device talking to Hubble’s satellite network), the range you skip designing in up front is range you’ll never get back.

What the selection tool actually does (and where it stops)

Kyocera’s selector, like the ones from Molex, TE, Johanson, and Abracon, filters a parts catalog. You feed it a band (2.4 GHz), a size budget, a target gain, and an efficiency floor. It returns parts that meet those numbers on the eval board the vendor characterized them on.

That’s genuinely useful. Sorting 40 chip antennas by footprint and peak efficiency by hand is tedious, and the tool does it in seconds.

Here’s where it stops. The tool doesn’t know your ground plane geometry. It doesn’t model your enclosure, your battery placement, or your matching network. The efficiency number it shows you was measured in free space on a reference board that looks nothing like your wearable.

Treat the output as a starting BOM line. It tells you which part can work. Whether it does work is decided entirely by the five layout decisions below.

Layout Rule 1: 2.4 GHz antenna placement and the ground plane

Ground clearance is the highest-leverage variable on the board. Get it wrong and no matching network will save you.

The rule flips depending on antenna type, which is where a lot of respins come from.

For a chip antenna, you want copper gone directly underneath and around it. Most ceramic chip antennas are tuned assuming a clearance zone with no ground plane, no fill, no traces on any layer within the keep-out. Flood a pour under one and you’ll detune it hard, sometimes shifting resonance clean out of band.

For a PCB trace antenna (an IFA or monopole variant), the ground plane is the counterpoise. The radiator needs a solid ground reference adjacent to or below the feed, sized per the datasheet. Starve it of ground and efficiency craters.

   Chip Antenna                PCB Trace (IFA) Antenna
 ┌──────────────┐            ┌──────────────────────┐
 │  ▓ antenna ▓ │            │  ═══╗   trace         │
 ├──────────────┤            │     ║                 │
 │  (no ground) │ clearance  │  ═══╝  ← feed         │
 │              │            │                       │
 ├══════════════┤            ├══════════════════════┤
 │ GROUND PLANE │            │      GROUND PLANE     │
 └──────────────┘            └──────────────────────┘
   ↑ keep copper OUT           ↑ ground as counterpoise
     under antenna               below/beside per DS

The trade-off: for counterpoise-style antennas, more ground area up to a point buys you better efficiency and a more stable pattern. Past that point, or on the wrong antenna type, added ground just adds detuning risk.

Treat the datasheet clearance dimension as a floor, not a target. Nordic and TI’s BLE layout guidelines both push more clearance than the antenna vendor’s bare minimum, because they assume real-world loading. Give yourself margin here if the board size allows it.

Layout Rule 2: keep-out zones and PCB antenna matching space

The keep-out is 3D, and people who nail the 2D version still lose dB to the vertical one.

Inside the specified keep-out region: no copper on any layer, no components, no routed traces, no metal shielding, and no battery. The battery one bites hardest in wearables and small sensors, where the cell wants to sit right where the antenna needs air.

The vertical dimension matters just as much. A metal enclosure wall, a shield can lid, or a display bezel sitting close above the antenna will load it and pull the tuning. Check clearance to the nearest metal in the assembled stack, not just the PCB.

The datasheet keep-out gives you the rated number in free space. Real products add plastic, a battery, and a hand. Every one of those loads the antenna. The datasheet keep-out is the minimum that produces the spec. Margin is what keeps you in band once the enclosure goes on.

If your mechanical team hands you a fixed volume, negotiate the antenna corner early.

Layout Rule 3: matching network placement and RF layout

Place the pi-network footprint as close to the antenna feed as routing allows. Every millimeter of feed trace between the radio and the match is transmission line you’re now matching through, and it moves your tuning target.

Lay out a full 3-component pi footprint. Every time. Even if simulation says you’ll ship a single series element, populate the board with a proper pi and set the unused positions to DNP.

  ANT ───┬──[ C1 ]──┬── RF_OUT (to radio)
         │  series  │
       [ C2 ]     [ C3 ]   ← shunt, DNP-optional
         │          │
        ═╧═        ═╧═      via to solid GND
        GND        GND

  Rule: full pi footprint, place < feed-trace-length from antenna feed

Retuning after enclosure integration is the norm, not the exception. When your measured Smith chart shifts once the plastic and battery are in place, a full pi lets you retune with a reel of 0402s and a soldering iron. A single-component layout means a board respin to add the shunt you now need.

The feed trace itself: short, 50-ohm controlled impedance, referenced to a solid ground plane the whole way. Don’t route it over a plane split or a gap in the reference, that turns your clean 50 ohms into a mystery. If your stackup makes 50 ohms awkward, fix the stackup, because everything downstream assumes it.

Ground the shunt components with dedicated vias straight to the reference plane. Don’t share a via or snake a trace to ground. Inductance in that path shows up as a shifted, lossy match, and it’s invisible until you’re on the network analyzer wondering why the shunt isn’t doing what the model said.

The pre-fab checklist

Run this against the board in front of you before you release for fab.

The last row is the one people skip. Tuning on a bare eval board tells you the part works in free space, which you already knew from the datasheet. It tells you nothing about your product. Budget lab time to measure and retune in the final mechanical assembly, with the real battery, real plastic, and ideally a phantom or hand nearby.

Two of these mistakes together (say, a pour under a chip antenna plus a single-component match) can cost you 10 dB or more. That’s a coverage cliff, not a rounding error.

Why datasheet range never equals field range for BLE range optimization

The efficiency number in the datasheet came from an eval board in free space. Your product is not that.

Three things eat into it once you build the real thing. Plastic has a dielectric constant above 1, which pulls resonance down and adds loss. Metal near the antenna, from the battery or PCB, detunes and blocks. And a hand or wrist near a 2.4 GHz antenna can drop efficiency by several dB and reshape the pattern entirely.

Plan your link budget with margin against the datasheet number, not equal to it. How much depends on the product, but assume you’ll keep all of the datasheet efficiency and your prototypes come back short. If you’re feeding a BLE advertising packet into a network where every dB of link margin extends usable range, that margin is the difference between a reliable node and an intermittent one.

A node you can walk over and re-flash is forgiving. A sensor in a wall, a tracker on a shipping container, or a device talking to a satellite gives you no second chance to retune. The range has to be in the board before it ships.

Use the vendor tool to pick the part. Then own the layout, because the board is where your range actually lives.

Hubble Network turns the same BLE advertising packet into satellite-reachable range, so a node that ships with tight link margin can still connect from the field. See how it works →