Traffic light and diagnostics

The traffic light tells you which link to look at; the diagnosis tells you why. This guide explains how each one judges.

The traffic light

Each station's color is the verdict of the full diagnosis — the worst of its seven axes rules (see below). Its main axis, signal, is judged in two modes, and the difference matters:

Without a reference — the absolute dBm value is judged:

State	Signal
🟢 Good	≥ −55 dBm
🟡 Attention	−55 to −67 dBm
🔴 Critical	< −67 dBm

With a reference — what is judged is how far it drifted from the reference you set (Δ dB):

State	Drift
🟢 On reference	up to −3 dB
🟡 Dropped vs ref	−3 to −8 dB
🔴 Check link	more than −8 dB

An Absent station is not a critical one: it lost its session and keeps its last reading, with the reason in the tooltip.

References: compare against your own network

Even without a reference, the app does not judge blindly: it uses the signal expected at each link's distance (computed by the radio itself) and, only as a last resort, the absolute threshold. The reference adds what no automatism can know: how that link was when the installer left it right — its alignment, its cable, its obstacles. It is your criterion, and it wins over the other two.

When a link is the way it should be, click "Mark OK": that moment's signal becomes that antenna's healthy point, and from then on the signal axis judges the drift against it (Δ dB) instead of the absolute value.

• The reference is per antenna and survives reboots and IP changes.
• With a reference, the cell shows ref −58 dBm and an ×. To update it (after a realignment, for example), remove it with the × and click "Mark OK" again — it is two gestures: while a reference is set, the button is not shown.

Mark while the link is stable

"Mark OK" stores the reading at the instant of the click, without averaging. Marking during a fade pins a bad point as the reference.

Why signal alone is not enough

A −54 dBm signal at 1.5 km can be excellent — and a link with a "good" signal can be suffering noise, interference or sunken capacity. That is why the diagnosis does not look at one number: it combines seven axes and the worst one rules. The result is the row's color plus a list of causes in plain words, shown when hovering the signal cell.

The seven axes

1. Signal in context. With a reference, the drift against it; otherwise, against the signal expected at that distance (computed by the radio itself); as a last resort, the absolute threshold.
2. Channel quality. The margin between signal and noise floor (SNR) or, on devices that report it, CINR — which also sees interference from other transmitters on the channel. This is where causes like "Channel interference: signal looks fine, the channel doesn't" come from.
3. Link quality. The airMAX indicator, averaged so bursts don't raise alarms: below 90 % asks for attention, below 70 % is critical. It includes the air latency measured by the radio: sustained highs or recurring spikes degrade the verdict ("congested channel, signal is fine").
4. Real vs. expected capacity. If the link delivers much less than its modulation promises (under 65 %), something is holding it back: saturation or degraded modulation.
5. Ethernet port. No cable or half duplex means the problem is in the wired stretch — not in the air. Speed is only mentioned when it limits what the radio could deliver.
6. Chain imbalance. The antenna's two polarities should receive almost the same; a difference of 5 dB or more betrays a rotated antenna, crossed polarization or a damaged chain — with the total signal looking normal. No other axis sees this.
7. Radio health. Saturated CPU or exhausted memory: a choking radio drops packets without touching the signal.

How to read a cause

Each cause names the fact and its context: "Signal 12 dB below expected at its distance", "High noise: SNR 14 dB", "Downlink well below expected: 23 of 80 Mbps". The cause tells you where to look — the cable, the channel, the alignment — before climbing to the roof.

The AP is judged differently

An AP does not have "one link" to judge: it has fifteen. Its State column combines channel load, radio health and its sector's latency — the detail is in Access points.

About the thresholds

The numbers in this guide are the current version's, calibrated against real networks. Some — latency's in particular — are still being field-calibrated and may be adjusted between versions; the tooltip always shows the measured values, so you can judge with your own criteria alongside the app's.

The app itself tells how they were set: Help → "Cómo se calibra" sums up the method behind each limit, with real cases — and the invitation to report yours to support@materis.io, which is where the adjustments come from.