RF, wireless and connectivity
Getting the radio to work on a crowded board — LoRa, Wi-Fi 6, cellular, GNSS and RTK.
RF is where embedded projects quietly go wrong. The digital section works, the firmware works, and the range is a third of the datasheet figure because a ground plane was cut under a feed line.
The parts that decide whether it works
A 50 Ω feed is only 50 Ω if the stack-up supports it and the fabricator holds it. Keep-outs have to be respected on every layer, not just the top. Ground stitching around the feed has to be dense enough at the operating frequency. The antenna needs the ground plane its datasheet assumed. None of this is exotic — it is just unforgiving, and it cannot be fixed in firmware.
Link classes, and what each one costs you
| Class | What it buys | What it costs |
|---|---|---|
| Sub-GHz (LoRa, 868 MHz) | Kilometres of range on a battery that lasts years | Almost no bandwidth; duty-cycle limits shape the protocol |
| Wi-Fi 6 / BLE | Bandwidth and a device everyone already owns | Range, power, and a crowded 2.4 GHz neighborhood |
| 5G / LTE | Coverage without owning infrastructure | Peak transmit current that the power tree has to absorb |
| GNSS with RTK | Centimetre positioning for autonomy | Antenna siting, and a reference the other radios must not pollute |
| AIS | Marine traffic awareness | One more mast-mounted radiator competing for ground |
| RF-PLC hybrid | A fallback path where either link alone fails | Two front ends on one board that dislike each other |
The selection conversation is short. The design conversation — feed geometry, keep-outs, ground stitching, co-existence, band planning — is the whole job.
Regulatory reality, early
Which bands you may use, at what power, with what duty cycle, depends on where the product ships. That constrains the radio choice, which constrains the antenna, which constrains the board. Decided at architecture stage it costs nothing. Decided after layout it costs a respin.
Common questions
Straight answers
Which radios have you actually integrated?
Sub-GHz LoRa and LoRaWAN at 868 MHz; Wi-Fi 6 with Bluetooth LE; 5G and LTE cellular modules; GNSS with RTK for autonomy-grade positioning; AIS reception for marine work; and RF-PLC hybrids for smart grid. At the MCU end, integrated Wi-Fi and BLE on ESP32-class parts. The specific module matters far less than the feed, the ground and the co-existence plan around it.
Can you design a custom antenna?
No — and you should be suspicious of a small design house that says yes. We integrate proven antennas and modules, control the feed and the ground, and route the board so the antenna performs to its datasheet. Custom antenna design and chamber-verified tuning belongs with a specialist, and we will bring one in rather than improvise.
Several radios in one small enclosure — is that a problem?
It is the problem. Co-existence gets planned at architecture stage: band separation, duty-cycle coordination, physical separation, shielding and filtering. Discovering an interference issue at pre-compliance is expensive; discovering it in the block diagram is free.
Do you handle certification?
We design for it and support you through it — band planning, pre-compliance measures, and the design changes that come out of a chamber session. The certification itself runs through an accredited test house.
Evidence
Where this has been applied
Edge-compute architecture for an unmanned surface vessel
Three architectures compared on real bandwidth budgets, so the platform decision was made on evidence rather than TOPS figures.
Flight-controller carrier for a tube-launched UAV
Teardown, clean-room re-implementation, and a rigid-flex carrier that fits inside a launch tube.