Site Survey User Guide

01Overview

Site Survey answers one question for an installer standing in a building: is this installation finished, and if not, which sensors need attention?

Checking mesh connectivity without it means visiting each sensor in turn — open its dashboard, view its map, refresh the mesh map, read the numbers, go back, repeat. On a twenty-sensor site that is twenty round trips, and nothing compares one sensor against another.

Site Survey puts every whitelisted sensor in a single list, colour-coded green, yellow or red against thresholds you control. One button tests all of them. When the whole list is green, the installation is done. Red rows tell you which sensors need the antenna reoriented, the sensor moved, or a repeater added.

02Requirements

What the app needs before it will do anything useful.

What the gateway must provide

Two platform services do the radio work. Both arrive with Atrium 2.6.0:

• Mesh testing — runs the route trace and link test behind Test All Sensors and the per-row retest button.
• Radio scanning — runs the noise-floor sweep behind Scan Noise Floor.

Where a gateway does not offer one of these, the app hides or disables that feature and everything else keeps working. You do not need to uninstall anything.

03Installing

Site Survey installs like any other Atrium app.

1. 1
   Open the Apps page

   Log in to the gateway dashboard and select Apps in the top navigation.
2. 2
   Install Site Survey

   Find Site Survey in the App Center and install it. If you were sent a package file instead, use the upload control on the same page.
3. 3
   Wait for the gateway to reload

   Installing an app restarts the gateway’s services. This takes a few seconds and interrupts nothing permanently; the page reconnects on its own.
4. 4
   Open the app

   Site Survey now appears as a card on the Apps page. Select it to open.

04First-Time Setup

The shortest path from a fresh install to a graded site.

1. Open the app

Select Site Survey on the Apps page. The header names the gateway you are surveying, and the line beneath it reports when the last full survey ran, how many sensors are on this gateway, and the link-test parameters currently set.

Before any test has run, every sensor shows Not tested. That is expected; the app does not test anything until you ask it to.

05The Sensor List

The main screen. Every whitelisted sensor, one row each.

Filtering and sorting

Two pills sit above the list. All shows everything. Needs attention narrows to the rows that are not green — poor, marginal and untested together.

The Sort menu offers three orders:

Order	Behaviour
Worst signal first (default)	Poor, then marginal, then good, then untested; weakest signal first within each grade.
Most hops first	Longest route first. Useful for finding sensors leaning on a repeater.
Name	Alphabetical.

Untested sensors sort last rather than worst. A sensor nobody has tested is not evidence of a bad installation, and burying the genuinely red rows beneath it would defeat the list.

Expanding a row

Select any row to open its detail. You get one card per leg of the route, plus a card summarising the test itself.

Each leg card names its two ends and reports the average signal for that leg, the minimum and maximum seen, how many test packets arrived out of how many were sent, and the retries for that leg. A two-hop route produces two leg cards, so you can see which half of the path is the weak one.

The test card reports the status, how many hops were measured, when the route was last mapped, the packet count and size the test used, and the margin over the noise floor if a scan has been run. Where a test failed or finished only partly, the gateway’s own explanation appears here in full.

06Testing Sensors

What happens when you press the button, and why it takes as long as it does.

What the Status column means

Status	What is happening
Not tested	No test has ever been requested for this sensor.
Waiting for check-in	Queued. Nothing is on the radio yet. The gateway is waiting for the sensor to wake up, which can take up to an hour. The line beneath reports how long it has been queued.
Tracing route	The sensor woke, acknowledged the command that holds it awake, and the gateway is discovering the path to it. It has about 30 seconds.
Link testing	The path is known and the gateway is measuring each hop. The line beneath counts hops measured out of hops on the route.
Complete	Every hop on the route was measured. The row’s numbers are current.
Partial	The sensor went back to sleep before its own hop could be measured, but the other hops were. The data shown is real; retest to catch the rest.
Failed	The test did not produce a measurement. The reason appears beneath, and in full when you expand the row.

07RF Environment

What the installation is competing with.

A signal strength on its own does not say whether a link will work. A -80 dBm link against a -105 dBm noise floor has 25 dB of headroom and is healthy. The same -80 dBm against a -85 dBm floor will drop packets no matter how the antenna is turned. Those two look identical in the RSSI column, and they have completely different fixes.

Select Scan Noise Floor to measure it. The scan takes a fraction of a second. Once it has run, the button reads Rescan.

What the panel reports

The floor is reported as the median channel rather than an average. The radio hops across the band, so a link competes with the typical channel; an average would be dragged around by one loud interferer, and the lowest reading would flatter a genuinely busy site. The peak is reported separately because a single hot channel is a different problem — and often a fixable one — from a raised floor.

Margin over noise

Once a scan exists, each sensor’s expanded row gains a Margin over noise figure: its weakest leg’s signal minus the noise floor, in dB. The panel also names the sensors with the least margin, so you do not have to do the subtraction across every row.

Margin deliberately does not change a row’s colour. A sensor’s grade would otherwise shift the moment somebody ran an unrelated measurement, and a link that has not changed should not appear to have.

08How Results Are Graded

Where the colours come from.

Four measurements are graded for every sensor. Each is compared against its Good threshold, then its Ok threshold:

The row takes the worst of the four. One measurement in the red band makes the row red, however good the other three are. A sensor with a strong signal and a flat battery is still a sensor that will stop working.

RSSI comes from the weakest leg rather than an average of the route, because a strong gateway-to-repeater hop must not mask a marginal repeater-to-sensor one. The chain is only as good as its weakest link.

A sensor with no link measurements at all is untested rather than graded. The app never infers a grade from telemetry signal strength — that is the strength of the last data packet, which says nothing about the tested quality of the link.

09Survey Settings

The gear button at the top right. Link-test parameters, and the thresholds that colour every metric.

Link test

These two control how thorough each hop measurement is.

This setting is not cosmetic. Each test packet the sensor misses is retried by the radio, so the cost of a large test grows sharply on a weak link — and a slow test can outlast the short window a sleeping sensor stays awake for, which is what produces a Partial result. On a site where sensors keep coming back partial, lowering the packet count is often the fix.

Thresholds

Six pairs. Values meeting Good are green, values meeting Ok are yellow, anything else is red.

Each pair must be internally consistent — Good has to be at least as demanding as Ok. Where it is not, the dialog explains the problem under that card and Save is unavailable until it is fixed. Packet count and size must be at least 1.

Reset to defaults restores every value in the table above. Cancel, the close control, clicking outside the dialog, and the Escape key all discard changes.

10Exporting the Report

A commissioning record you can hand over.

Select Export Report to download the current survey as a CSV file, named for the gateway and the date. It opens in any spreadsheet.

One row per sensor, carrying the name, address and type; the worst-leg signal, success rate and retries; the hop count and the full route; when the sensor was last heard; the battery level; the margin over noise; the test status; how many hops were measured; and any error message. Comment lines at the end record the gateway name, the export time, the link-test parameters, and the noise floor and peak if a scan has been run.

11Automations and Data

How Site Survey relates to the rest of the gateway.

Automations

Site Survey does not appear on the Automations page. It publishes no events for other apps to react to, and offers no actions that can be triggered by one. This is deliberate: every action it performs puts traffic on the shared radio, and a noise scan stops the gateway hearing sensors entirely. Neither belongs on a timer.

A survey is something an installer runs deliberately, reads, and acts on.

What it reads

Everything the app shows comes from data the gateway already holds:

• The whitelisted sensor list, with each sensor’s name, type, battery level and last-seen time.
• The mesh routes and per-hop link measurements the gateway stores when a test runs.
• The gateway’s own name, for the header and the exported report.

It writes nothing to your telemetry history, and it never changes a sensor’s configuration. Its own storage holds only your thresholds, a record of each full survey, and the noise scans.

Other pages

The gateway’s own Mesh Map page draws the same underlying route and link data as a diagram. Site Survey grades it and compares sensors; the Mesh Map visualises the topology. Running a test in Site Survey updates what the Mesh Map shows.

12Troubleshooting

Symptoms seen in the field, and what they usually mean.

13Reference

Values, ranges and units in one place.

Units and ranges

Test status values

Not tested, Waiting for check-in, Tracing route, Link testing, Complete, Partial, Failed. Described in What the Status column means.

Grades

Good, Marginal, Poor, Untested. A row takes the worst grade of its four measured values. These grades are defined by your own thresholds, not by any published standard.

Route chips

GW is the gateway’s own radio. A Repeater chip is a mains-powered sensor relaying traffic onward. The final chip is the sensor’s own address, shortened. A route with no repeater chip is a direct link.