Measuring Your LoRa Link Margin

Part 3 of the measure your mesh series. Earlier installments covered SNR and hop count and why packets disappear. This one covers link margin: the number that tells you how much headroom you actually have.

What link margin actually is

A link budget accounts for every gain and loss between your transmitter and receiver:

Link budget (dB) = Tx power + Tx antenna gain − cable losses − path loss + Rx antenna gain − receiver sensitivity

The result tells you how much stronger the received signal is than the minimum the receiver can decode. That difference is the link margin.

A link margin of +20 dB means the signal arrives 20 dB stronger than required. You have 20 dB of headroom before the link fails. A margin of 0 dB means you’re exactly at the threshold — any degradation in path conditions will drop packets.

The diagram shows a typical outdoor deployment. Path loss is the dominant term — it dwarfs every other component combined. This is why antenna placement matters so much: a 3 dB improvement in either antenna position (height, clear line of sight, fewer obstructions) translates to a 3 dB improvement in link margin. Conversely, a connector in bad condition consuming an extra 2 dB matters.

Reading margin from your SNR data

You don’t always have the full link budget available. But if you know your SNR and your spreading factor, you can estimate the margin above the demodulation floor:

Margin above floor (dB) = Measured SNR − Minimum decodable SNR for this SF

From Semtech’s specifications, the minimum decodable SNR values are:

SF	Min decodable SNR	So at SNR = +5 dB
SF7	−7.5 dB	12.5 dB margin
SF8	−10 dB	15 dB margin
SF9	−12.5 dB	17.5 dB margin
SF10	−15 dB	20 dB margin
SF12	−20 dB	25 dB margin

A link running at +5 dB SNR on SF9 has 17.5 dB of margin above the absolute demodulation floor. Whether that 17.5 dB is “enough” depends on what you’re protecting against.

How much margin is enough?

This depends on what you’re protecting against. Three reference points:

10 dB (absolute floor for static deployments). A link with 10 dB above the demodulation floor should work in dry, clear conditions. It will be inconsistent during heavy rain, fog, or when seasonal foliage thickens. If this is an indoor-only link in a stable RF environment, 10 dB may be workable. For outdoor mesh — no.

15 dB (working minimum for outdoor static links). This is the practical minimum for a link you intend to rely on. It accommodates normal weather variation, moderate seasonal change, and typical multipath fading. Consistent with recommendations from LoRa deployment literature — 15 dB is frequently cited as the lower bound for outdoor static deployments.

20 dB (target for critical paths). For a relay that multiple downstream nodes depend on — a single point of failure in your topology — 20 dB gives you meaningful resilience. This is the level where a link survives the kinds of degradation events (a weather front, a new obstruction, equipment aging) that would drop a 15 dB link.

These are rules of thumb, not specifications. Your actual failure threshold depends on your RF environment, spreading factor, and local interference floor.

Where margin disappears without warning

Most link failures aren’t sudden. They’re the result of margin that was always thin getting eroded over time:

• Seasonal foliage. Sub-GHz signals are attenuated by leaves and branches. A link that cleared 18 dB margin in winter may be at 12 dB in midsummer with the same equipment.
• Weather. Precipitation adds absorption. A heavy rain event can cost several decibels on marginal paths.
• New reflectors. Multipath fading from nearby objects — a parked truck, a new outbuilding, a fence — can shift the interference pattern and either improve or degrade links you weren’t planning to touch.
• Connector degradation. A connector that’s been outdoors for two years in freeze-thaw cycles may have significantly higher insertion loss than when it was installed.

This is why SNR history matters. A link showing +8 dB SNR consistently over six months is very different from one that has been trending from +14 dB down to +8 dB over the same period. The trend is the signal; the snapshot is just a reading.

Waev’s Outpost view shows per-node SNR history over days and weeks. The baseline work described in the grid-down post is about building enough history to see these trends before they become failures.

A practical measurement workflow

If you want to assess a specific path in your network:

1. Pull the last 30 days of SNR readings from Waev’s Outpost or Network Stats for the path you’re evaluating.
2. Note the median (typical operating condition) and the minimum (worst observed). The minimum matters more — your link needs to survive its worst days, not its average ones.
3. Translate to margin above the demodulation floor using the table above.
4. Compare against the thresholds: if your minimum SNR is in the “marginal” zone (+0 to +7 dB), you have less margin than you might think.

The path that has been consistently at +9 dB on SF9 has approximately 21.5 dB of margin above the floor. Healthy. The path that’s at +1 dB on SF10 has 16 dB — technically within the working minimum, but a single bad weather event can push it below threshold.

Questions about a specific link you’re trying to evaluate? Tell us what you’re seeing.