RF range testing that predicts real-world coverage
How to plan RF range testing that predicts real coverage: link budgets, conducted checks, field methods and the mistakes that skew numbers.

Every radio datasheet quotes a range, and almost none of them describe your product. RF range testing is how you find out what your device actually achieves once it sits in a plastic enclosure, on a wall, next to a person, running from a half-flat battery. This post is for founders and engineers who need a coverage number they can put in a spec sheet without crossing their fingers.
We see the same pattern often: a development kit reaches across the office car park, the first enclosed prototype struggles to reach the next room, and nobody can say which of a dozen changes caused it. A structured test plan turns that guesswork into a few decibels you can attribute and fix.
Why datasheet range rarely survives the enclosure
Range is the output of a link budget, not a property of a chip. The budget is transmit power plus antenna gains on both ends, minus path loss and every other loss along the way, compared against the receiver’s sensitivity. A datasheet figure assumes ideal antennas, line of sight and no obstacles. Your product has none of those.
The losses that eat the margin are mundane:
- Antenna detuning. Plastic, ground plane size and nearby components shift the resonant frequency. A few dB of loss is common; a badly detuned chip antenna can lose considerably more.
- Enclosure and mounting. Glass-filled plastics, metallic paints and, above all, a metal surface behind the device change the pattern and efficiency.
- The human body. At 2.4 GHz, a wearable or handheld device can lose 10 dB or more in some directions simply because a person is in the way.
- Supply voltage. Some radios and power amplifiers lose output power as the battery sags, so range at end of life is worse than on a fresh cell.
None of these show up on a development kit sitting on a desk, which is exactly why kit-based range estimates are optimistic.
Turning a link budget into a range estimate
Before testing anything, write the budget down so you know what you are trying to verify. Free-space path loss in dB is 20·log10(d in km) + 20·log10(f in MHz) + 32.44. At 2.44 GHz that gives roughly 80 dB at 100 m, and every doubling of distance adds about 6 dB in free space. Indoors, path loss grows faster: 9–12 dB per doubling is a reasonable working assumption once walls and furniture are involved.
As an example, an nRF52840 transmitting at 0 dBm into a receiver with a sensitivity around −95 dBm on the BLE 1M PHY has about 95 dB to spend. LoRa radios such as the SX1276 reach sensitivities in the region of −137 dBm at SF12 and 125 kHz bandwidth, which is where their long range comes from, at the cost of airtime and battery. Whatever the radio, subtract antenna and enclosure losses, then reserve a fade margin of 10–20 dB for multipath and changing conditions. What is left is the path loss you can tolerate, and that maps to a distance.
If the paper estimate already looks thin, testing will not rescue it. Go back to antenna choice and placement first; our notes on PCB antenna placement for small connected devices cover the usual layout mistakes.
Conducted tests before radiated ones
We start on the bench with a cable, not in a field. Connect the radio’s RF output through a calibrated step attenuator to a reference receiver, and reduce the signal until the link starts failing. This isolates the radio, matching network and firmware from the antenna and environment.
A conducted test answers three questions quickly:
- Is the transmit power what the firmware claims? Measure it with a spectrum analyser or power meter, at both nominal and minimum battery voltage.
- Does the receiver reach its datasheet sensitivity, or is the board’s own noise (switching regulators, clocks, displays) degrading it? Run the test with every subsystem active, not just the radio.
- Where does packet error rate climb? Sensitivity is usually specified at a defined error rate, so measure errors, not just RSSI.
Many boards need a test pad or a U.FL footprint for this. Adding one on the first prototype revision costs very little and saves days later.
A repeatable RF range testing method in the field
Once the conducted numbers are sound, move to radiated tests with the final enclosure, final battery and production firmware. Firmware should have a test mode that sends sequence-numbered packets at a fixed interval and logs RSSI, SNR where the radio reports it, retries and missed sequence numbers on both ends. Without sequence numbers you cannot calculate packet loss honestly.
Line of sight
Find an open area and walk the devices apart in fixed steps, holding height constant at roughly the installed height. Ground reflections matter: a device at 0.5 m behaves very differently from the same device at 2 m, particularly for sub-GHz links. Record where packet loss passes your threshold, typically 1–10% depending on how the protocol handles retries.
Orientation and body effects
Rotate the device under test through eight positions, 45° apart, at a fixed distance and log the average RSSI for each. A deep null in one direction is common with small antennas and is worth knowing about before a customer mounts the product that way. For wearables and handhelds, repeat with the device worn or held.
Representative indoor runs
Pick two or three buildings that resemble real installations: a timber-framed house, a concrete apartment block, an office with metal partitions. Use the same routes each time so later firmware or hardware revisions can be compared directly.
| Test | What it tells you | When to run it |
|---|---|---|
| Conducted, step attenuator | Radio and matching health, real sensitivity | Every board revision |
| Line of sight walk | Antenna and enclosure efficiency | First enclosed prototype |
| Orientation sweep | Pattern nulls, mounting guidance | After antenna or enclosure changes |
| Indoor route | Realistic coverage claim | Before writing the spec sheet |
Mistakes that make range numbers meaningless
The most common error is using a phone as the reference receiver for BLE products without saying so. Phones vary widely in antenna performance and in how they report RSSI, so a figure measured with one handset can differ noticeably with another. If the product pairs with phones, test with several, including an older budget model.
Other traps we look for:
- Testing with a development kit on the other end when the product will talk to a gateway with a different antenna.
- Comparing RSSI values across different chips, which are not calibrated to the same reference.
- Single runs on a single day. Repeat each point and note conditions; vegetation, rain and people moving all change results.
- Forgetting regulatory limits. EU 868 MHz LoRa is generally limited to 25 mW ERP with duty-cycle restrictions, so turning the power up is not always an option.
Range work also overlaps with emissions work. A noisy regulator that hurts sensitivity often shows up in pre-compliance scans too, so it pays to run both on the same revision.
What we would do before quoting a range
A coverage figure ends up in sales material and support conversations, so it should be defensible. Before committing to one, we would make sure that:
- A written link budget exists, with fade margin, and the measured results are compared against it.
- Conducted transmit power and sensitivity have been checked at minimum battery voltage with all subsystems running.
- Radiated tests used the final enclosure, battery, mounting and firmware, with packet loss measured from sequence numbers.
- Orientation nulls are known and reflected in installation guidance.
- The published figure is the indoor result with margin, not the best line-of-sight run.
If the numbers fall short, the fix is usually antenna tuning, placement or enclosure material rather than more transmit power. That is part of the hardware and testing work we do, and it is far cheaper to settle at prototype stage than after tooling.