Planning a radio link that works
Path loss, link budget and margin — how to tell before you deploy whether a wireless link will hold, and which lever to pull when it will not.
The question a link budget answers
"Will this reach?" is not a question you should discover the answer to by driving to a site with a ladder. It is arithmetic, and the arithmetic is addition — which is the entire reason radio work is conducted in decibels.
A link budget starts with the power you transmit, adds every gain along the way, subtracts every loss, and reports what arrives at the far end. Compare that to the receiver's sensitivity and the difference is your margin. If the margin is comfortable the link will hold; if it is thin it will work on a still day and fail in the rain.
Prx = Ptx + Gtx − Lcable − Lpath + Grx
Powers are in dBm, gains and losses in dB. Everything is a single column of figures you can add up by hand — which is worth doing at least once, because it makes obvious which terms actually matter.
Path loss, and why frequency appears in it
The dominant term is almost always the path. Free-space path loss is:
FSPL(dB) = 20·log₁₀(d) + 20·log₁₀(f) + 32.44
with distance in kilometres and frequency in megahertz. The constant is purely a unit conversion — use miles and it becomes 36.6, use metres and hertz and it vanishes. If your answer is out by tens of decibels, check which variant you used before looking for anything subtler.
People are often surprised that frequency appears at all, since empty space does not absorb more at 2.4 GHz than at 868 MHz. It does not — the frequency term is about the antenna. For a given gain, a higher-frequency antenna has a smaller effective aperture, so it catches less of the passing wave. The formula folds that into the path term by convention. The practical upshot is real regardless of the bookkeeping: identical hardware reaches noticeably further at 868 MHz than at 2.4 GHz.
One figure worth carrying: doubling the distance costs 6 dB, at any frequency. Four times the range costs 12 dB. That makes it easy to check whether an ambition is plausible before doing any real work.
Free space is a best case you will never see
The FSPL figure describes a clear line of sight through nothing at all. Reality subtracts more:
- Obstruction. An interior wall costs a few dB; a reinforced concrete floor can cost twenty or more. Foliage is worse than people expect, and worse still when wet.
- Ground reflection. A signal arriving by two paths can arrive out of phase and partially cancel. Beyond a certain distance the loss stops following the square law and starts falling off far faster.
- Fresnel zone intrusion. Line of sight is not enough — the radio needs clearance around that line, an ellipsoid whose radius grows toward the midpoint. An obstacle that does not block the optical path can still cost you several dB by clipping this zone.
- Weather. Rain is negligible below about 6 GHz and becomes significant above 10 GHz.
How much margin
Margin is not padding for a calculation you do not trust; it is budget for everything the calculation cannot see.
- Under 6 dB — works on the bench, fails in the field.
- 10 dB — a reasonable floor for a fixed outdoor link with genuine line of sight.
- 20 dB or more — anything mobile, indoors, through vegetation, or that must not fail.
The reason thin margins disappoint is that the losses are not static. A van parks in the path. The antenna rotates a few degrees on its mount. Leaves come out in spring. Each is a few dB, and they arrive together.
Sensitivity depends on data rate — and that is the biggest lever
Receiver sensitivity is quoted per data rate, and the spread is enormous. A module might be −95 dBm at its fastest setting and −137 dBm at its slowest. That 42 dB is worth more than any antenna change and more than any legal increase in transmit power.
This is the mechanism behind long-range modulations like LoRa: they trade throughput for sensitivity by spreading each symbol over more time. Slowing down buys range, directly and dramatically.
It is not free. Longer airtime means more energy per message — which collides with battery life — and it consumes more of your duty-cycle allowance. But when a link is 10 dB short, dropping the data rate is usually the cheapest fix available.
Which lever to pull
When the budget does not close, the options are not equally good:
- Transmit power — doubling it buys only 3 dB, costs battery life, and is legally capped. Usually the worst option.
- Antennas — often the best. Gain helps at both ends, so upgrading both can buy 6–12 dB. A directional antenna buys more still, at the cost of needing to be aimed.
- Height — frequently free and frequently decisive. Getting above the obstruction changes the problem rather than fighting it.
- Data rate — the largest single lever, as above.
- Cable — coax loss at 2.4 GHz is brutal, and a run of thin cable can throw away everything a better antenna gained. Mount the radio at the antenna where you can.
Check the rules before the physics
A link that closes beautifully may still be unlawful. ISM band limits differ by region in ways that constrain the design rather than just the paperwork: EU 868 MHz imposes a duty cycle, commonly 1%, so a device cannot transmit whenever it likes. US 915 MHz instead limits dwell time per channel and expects frequency hopping.
Those rules interact directly with the data-rate decision above — slowing down to buy sensitivity uses more airtime, and airtime is exactly what the duty cycle limits. Check theband table before committing to a reporting schedule, not after.