Measuring SNR on a LoRa Mesh

The number appears on every packet: +9 dB, +4 dB, –1 dB. When you first start watching Live Packets, it’s background noise — just another column in the stream. But SNR is the one number that tells you, for each link, whether your network can be depended on.

How to measure SNR on a LoRa mesh is not a complicated question, but it has layers. Here’s what the number means, where to read it, and what to do with it.

What SNR measures

Signal-to-Noise Ratio is the difference, in decibels, between the signal your observer’s receiver hears and the background noise floor at that location. A +10 dB reading means the signal arrived 10 dB stronger than the noise. At +2 dB, it’s barely above it.

LoRa uses spread-spectrum encoding that allows it to decode signals below the noise floor — this is the engineering that makes MeshCore links work at distances that look improbable on paper. But “the radio can decode it” and “this path is reliable for community infrastructure” are different standards. As SNR approaches zero, decode reliability becomes conditional: the link works eight times out of ten, then six, then varies unpredictably with conditions.

The practical target for a link you want to depend on: +5 to +7 dB or above in typical conditions. Below that is working territory, not reliable territory.

Where Waev surfaces SNR

Three surfaces, each with a different purpose.

Live Packets shows the raw observation stream as it arrives. Each row is one packet: timestamp, source node, hop count, and SNR at the enrolled observer. This is where you measure SNR per link in real time — scroll back a few minutes to see the variance on a given link over a short window.

Network Stats aggregates the data into a path-quality view. Instead of individual packets, you see a row per node with its typical hop count, average recent SNR, and a status flag (healthy / watchlist / degraded). This is the faster way to find which links in your network are underperforming — one scan of the table reveals the candidates.

Outpost is the per-node detail: SNR history over time for a single node, including the heard-by log showing which observers recorded it and when. When a node’s behavior is inconsistent, Outpost is where you diagnose whether the pattern is tied to time of day, weather correlation, or a specific relay going offline.

Reading the measurements

A single SNR reading is a snapshot. What you want is a pattern. Here’s how to read it:

Consistent high SNR (>+7 dB) with stable hop count: solid link. You don’t need to touch this one.

Consistent low SNR (+3 to +5 dB): marginal link. It works under normal conditions. A weather event, a nearby interference source, or a small shift in antenna angle can push it into failure territory. If this is a critical path, improve the margin; if it has a redundant path, monitor it.

Hop count higher than expected: the mesh rerouted around something. If a node was reaching your observer in 2 hops last month and now shows 4, a link in the shorter path degraded or disappeared. The hop count didn’t grow because you changed something; it grew because something changed.

Inconsistent SNR on the same link: usually means intermittent obstruction (vegetation in wind, weather fronts), an antenna that shifted, or a power issue on the relay. Use Outpost’s time-series view to see if the variation has a pattern.

Turning the measurement into action

The measurement is only useful if it leads somewhere. Here’s the decision chain:

A link that consistently reads below +5 dB and carries important traffic needs improvement. The options: raise the antenna on either end, reduce the path length by adding a closer relay, or check for nearby interference with stats-radio from the node’s serial CLI.

A link that developed higher hop count than expected needs investigation. Look at the previous shorter path: which relay is no longer heard? Check that relay’s power, placement, and recent Live Packets history.

A link at negative SNR that was positive six months ago is the one to prioritize. That’s structural degradation — something in the environment changed — and it doesn’t self-correct. The change in Reading the Signal gives the full SNR interpretation framework; this series entry is specifically about the measurement method and where the numbers live in Waev.

For acting on what you find — whether the answer is a new relay, a better antenna, or a smarter placement — repeater placement starts from the same signal data.