RF Fade Margin Explained: Why a Weakening Link Still Answers Polls Right Up Until It Doesn't
Fade margin is the cushion between the signal your radio receives and the weakest signal it can still decode. A link can lose most of that cushion and still answer every poll, which is exactly why it surprises people when it finally drops.
What fade margin is
Every radio receiver has a threshold: the weakest signal it can turn into clean data at an acceptable error rate. This is the receiver sensitivity, and manufacturers publish it on the spec sheet. Your link has a received signal level (RSL), the power that actually arrives at the antenna port after the transmitter, antennas, cable, and the path between the two sites have done their work.
Fade margin is the difference between the two, in decibels:
As an illustration only: if a link receives at -72 dBm and the radio's threshold for the chosen data rate is -95 dBm, the fade margin is 23 dB. That 23 dB is how much extra loss the path can suffer before the receiver can no longer hold the link.
Why a link with shrinking margin still answers polls
Digital radios have a cliff, not a slope. Error correction and retries are built to hide a degrading signal from the application. While the received signal is above threshold, the radio decodes cleanly, and your master station sees a good poll response. At 23 dB of margin, at 12 dB, and at 6 dB, the poll looks identical from the SCADA screen: it answers.
Two things make this worse:
- Retries mask loss. Many polling schemes retry a missed response automatically. A link dropping an occasional packet can still report a healthy "last good poll" time, because the retry succeeded.
- Margin is invisible to the application. A poll tells you the link worked. It does not tell you how close to the edge it worked. Uptime and margin are different questions, and a standard poll answers only the first.
So a link can slide from 23 dB of margin to 5 dB over weeks and show no symptom. Then a heavy rain event, a foliage season, a degraded connector, or a bad night of atmospheric ducting takes the last few dB, and the link falls off the cliff in a single event.
What eats fade margin
Fade margin is not a fixed property of a link. It is consumed by things that change:
- Multipath and atmospheric fading. Reflections and changing air density cause the signal to dip and recover. The longer and flatter the path, the more this matters.
- Rain, particularly at higher frequencies. Rain attenuation grows with frequency; it is negligible at 900 MHz and low UHF, minor through the low microwave bands, and becomes a serious link-budget factor above roughly 10 GHz (11, 18, 23, 38 GHz).
- Obstruction growth. Trees leafing out, a new structure, or a stockpile in the path eats into clearance (see our radio telemetry article).
- Hardware drift and damage. Water in a connector, a failing coax run, a misaligned antenna after a wind event, aging power amplifiers, or a drifting receiver front end all subtract from the margin without announcing themselves.
- Interference. A rising noise floor raises the effective threshold, which shrinks margin even though your received signal looks unchanged.
How much margin is enough
There is no single number, but there is a useful rule of thumb. In flat-fading conditions, every additional 10 dB of fade margin improves link reliability by roughly an order of magnitude in outage probability. Many engineered links target somewhere in the 20 to 30 dB range or higher for critical paths, with more for long, over-water, or high-frequency paths. Treat these as general industry rules of thumb, and validate against your own path design and reliability target.
How to read it
- Get the right two numbers. You need RSL (often a radio status field, sometimes called RSSI, though definitions vary by vendor, so confirm what your equipment reports) and the receiver threshold for your current modulation or data rate.
- Watch the trend, not the snapshot. A single reading at 18 dB tells you little. A reading that was 26 dB last quarter, 22 dB last month, and 18 dB today tells you a path is changing. The slope is the signal.
- Compare to the commissioning baseline. Record RSL at install. Drift of several dB from that baseline, with no config change, deserves a look.
- Separate the two ends. Margin can differ in each direction. A one-way problem often points to a transmitter, antenna, or interference at one end.
- Expect adaptive modulation to complicate things. Radios that step down data rate as the signal weakens will keep the link up while quietly moving the threshold. Read margin against the threshold for the current rate, and note when rate changes occur.
- Mind the vendor differences. Different manufacturers report signal quality in different units, scales, and update intervals. Comparing across vendors needs normalization, or you will compare numbers that do not mean the same thing.
See fade-margin trends on your own links.
Aevus Comms (RF Link Intelligence) is built around this exact problem: reading fade margin as a trend across radios from different vendors in one view, so a shrinking runway is visible while the link is still answering polls.
Related Articles
Bands, degradation modes, and redundancy patterns for SCADA radio.
A modeled scenario: 8 radios degrade for 2 months unnoticed.
A modeled 14-day timeline of a radio drifting toward failure.
Ready to see how this applies to your operation? Start a pilot conversation, no commitment, no field changes.