• Telecommunications
  • Systems & infrastructure
  • Development & tooling

Monitoring a satellite’s emissions with Grafana

A dashboard that shows when a satellite channel transmits, rebuilds its spectrum in one click and calculates its emission rate over any period

Summary

My department at Airbus Defence and Space monitors the radio emissions of a satellite’s communication channels. Its monitoring application receives the measurements in real time, but it lacks three features the team needs: displaying the spectrum of a chosen moment, setting an emission threshold, and calculating an emission rate. I designed and built a Grafana dashboard that adds them. It reads the measurements from the InfluxDB database, plots the carrier power over a chosen period, rebuilds the frequency spectrum of any moment in one click, and calculates the emission rate against a threshold that follows the channel: its average power outside transmissions, plus a margin chosen from a drop-down menu. It stays responsive over long periods, and it comes with technical documentation and a user guide for the teams.

Written as a technical report, with the structure taught in ENSEEIHT’s written-communication course: a summary, an introduction, four numbered parts, a conclusion and the sources.

Confidential project: the equipment, the software and the frequencies are not named, and every value, screen and example on this page is fictitious.

1. Introduction

My department at Airbus Defence and Space works on a satellite communications project. The satellite carries several communication channels, and the team needs to know, for each one, when it transmits and what its signal looks like.

The measurements of every channel arrive in real time in a monitoring application, which stores them in an InfluxDB database [2]. This application is very complete, but it lacks some features the team needs for its analyses. My manager asked me to add them with a Grafana dashboard [1].

The question

How can the team see, for any channel and any period, when it transmits, what its spectrum looks like at a given moment, and what share of the time it transmits?

Objectives

  1. Plot the carrier power of a channel over a period the user chooses.
  2. Rebuild the frequency spectrum of the moment the user clicks on.
  3. Set the threshold above which the channel counts as transmitting: its average power outside transmissions, plus a margin the user chooses.
  4. Calculate the emission rate automatically over the chosen period.

Part 2 describes what the monitoring application lacks. Part 3 explains the method and the technical choices, panel by panel. Part 4 presents the results, and part 5 analyses them and proposes next steps.

2. What the monitoring application lacks

The monitoring application receives every measurement and stores it in InfluxDB. It shows the state of each channel, but three features are missing for the team’s analyses:

  • A spectrum on demand

    Selecting a moment on a graph of the carrier power does not display the spectrum of that moment.

  • An adjustable threshold

    Nothing lets the team decide from which power level a channel counts as transmitting.

  • An emission rate

    Nothing calculates the share of time a channel spends transmitting over a period.

The dashboard reads the database that the application already fills. The acquisition chain stays untouched: the dashboard writes nothing.

3. Four values, three panels: method and choices

The work follows the path of the data: from the acquisition chain to the database, then from the database to each panel.

3.1 Where the data comes from

The signal received from the satellite reaches a digitiser, which converts it from analogue to digital. The monitoring application collects the measurements of each channel and writes them to InfluxDB 1.8, with the Flux language enabled (Figure 1).

The dashboard reaches this database through two data sources: the InfluxDB source, queried in Flux, and the Infinity source [3], which sends HTTP requests and processes their JSON answers.

Figure 1 – From the satellite to the dashboard Existing chain · generic names Satellite several channels Digitiser analogue → digital Monitoring app every channel InfluxDB 1.8 time series Grafana my part Power over time Flux Spectrum Infinity · InfluxQL · UQL Emission rate Flux · threshold read-only Two data sources InfluxDB source, in Flux power over time, emission rate Infinity source, over HTTP InfluxQL query, then a UQL script: the spectrum A click on a point stores its timestamp clicked_ts Figure 1 – From the satellite to the dashboard Existing chain · generic names Satellite several channels Digitiser analogue → digital Monitoring app every channel InfluxDB 1.8 time series Grafana my part · read-only Power over time Flux Spectrum Infinity · InfluxQL · UQL Emission rate Flux · threshold A click on a point · clicked_ts
Figure 1 – The signal received from the satellite is digitised, then the monitoring application stores the measurements of each channel in InfluxDB. The dashboard only reads that database: through the InfluxDB data source, queried in Flux, for the power over time and the emission rate; through the Infinity data source, which sends an InfluxQL query over HTTP, for the spectrum. The equipment and software of the chain are given generic names.

3.2 The values used

Each measurement holds many fields. The dashboard needs only four:

ValueWhat it isUsed for
cp Carrier power, in dBm Power over time, emission rate
powerdBm The power values across the channel, in dBm, stored as one list Spectrum
startFreqHz The frequency of the first value of that list, in Hz Spectrum
deltaFreqHz The frequency step between two values of the list, in Hz Spectrum

The queries read only these values: the other fields never leave the database.

3.3 Panel by panel

Carrier power over time

Need
See when a channel transmits.
How

A Flux query reads the cp values of the selected channel over the chosen period. When the period is long, specific conditions in the query reduce the calculation load before the points reach Grafana. Their details stay confidential.

Result
Transmissions stand out above the noise, and the user can zoom on any part of the period (Figure 2).
Fictitious screenshot: a time graph of carrier power over 24 hours, with blocks of points above a dashed threshold line and a tooltip on the selected point. Open the screenshot
Figure 2 – The carrier power of a channel over 24 hours. The dashed line is the threshold, the average outside transmissions plus deltaDB; the selected point is in emission. Reconstruction with fictitious data: the real screens are confidential.

The spectrum of a chosen moment

Need
Check what the signal looks like when a point stands out.
How

A click on a point of the time graph stores its timestamp in a dashboard variable, clicked_ts. The Infinity source then sends an InfluxQL query over HTTP: it fetches the last measurement of the channel at or before that moment (Listing 1). The answer is a JSON array. A UQL script [4] splits the list of powers into numbers and gives each value its frequency with equation (1), where i is the position of the value in the list.

SELECT powerdBm, startFreqHz, deltaFreqHz
FROM channel_measurements
WHERE carrier = '${channel}'
  AND time <= ${clicked_ts}ms
ORDER BY time DESC
LIMIT 1
Listing 1 – The spectrum query, in InfluxQL (names made generic)

frequency(i) = startFreqHz + i × deltaFreqHz (1)

Example with fictitious values: with startFreqHz = 5,000 Hz and deltaFreqHz = 50 Hz, the first three powers sit at 5,000, 5,050 and 5,100 Hz.

Result
Each power value is paired with its frequency, and the spectrum appears under the time graph (Figure 3).
Fictitious screenshot: a frequency spectrum, power against frequency, with a carrier rising about 20 dB above a flat noise floor. Open the screenshot
Figure 3 – The spectrum rebuilt for the selected moment: a carrier rises above the noise floor. Reconstruction with fictitious data: the real screens are confidential.

The emission rate

Need
Know what share of the time a channel transmits.
How

The threshold follows the channel. The dashboard calculates the average carrier power outside transmissions over the period: the noise level of the channel. The user adds a margin in dB, chosen from a drop-down menu at the top of the dashboard (the deltaDB variable). Equation (2) gives the threshold. A Flux query then compares every cp value of the period with this threshold and counts the points above it: the points “in emission”. Equation (3) gives the rate.

threshold (dBm) = average power outside transmissions (dBm) + deltaDB (dB) (2)

emission rate (%) = (points in emission ÷ all points) × 100 (3)

Example with the fictitious values of Figure 4: with an average of −68.4 dBm and deltaDB = 6 dB, every point above −62.4 dBm counts as in emission.

Result
The threshold and the rate update as soon as the period, the channel or the margin changes.

3.4 Choices and reasons

ChoiceReason
Build on the existing database The dashboard only reads: the acquisition chain and the application stay untouched.
Flux for the time graph and the rate The comparison and the count run in the database: only the result travels to Grafana.
Lighter queries on long periods The calculation load stays reasonable, and the dashboard stays usable over days of data.
Infinity and UQL for the spectrum The powers are stored as one list. UQL turns a simple InfluxQL query into a full spectrum, inside Grafana, with no extra service.
A threshold tied to the noise level Each channel gets a threshold that matches its own noise level; the deltaDB margin only sets how strict it is, in one click.

4. Results: the three missing features are delivered

Figure 4 shows the finished dashboard. At the very top, the channel and the deltaDB margin, chosen from drop-down menus. Below them, the emission rate and the average carrier power outside transmissions, on which the threshold is based. In the middle, the carrier power over the chosen period. At the bottom, the spectrum of the selected moment.

Fictitious screenshot of the whole dashboard: the emission rate and the average power outside transmissions at the top, the time graph in the middle and the spectrum at the bottom. Open the screenshot
Figure 4 – The finished dashboard: the two figures at the top, the carrier power over time, and the spectrum of the selected moment. Reconstruction with fictitious data: the real screens are confidential.
  • 3 of 3

    missing features delivered

    A spectrum on demand, an adjustable threshold, an automatic emission rate.

  • 1 click

    from a point in time to its spectrum

    The timestamp of the point drives the query.

  • 4

    values read from each measurement

    Out of the many fields each measurement holds.

  • 2

    documents for the teams

    Technical documentation and a step-by-step user guide.

Delivered

  • The dashboard, with two drop-down variables: the channel and the deltaDB margin
  • Technical documentation that explains every query and every calculation
  • A procedure that guides the teams, step by step, through using the dashboard
Fictitious screenshot: a time graph of carrier power over 24 hours, with blocks of points above a dashed threshold line and a tooltip on the selected point.
Figure 2 – The carrier power of a channel over 24 hours. The dashed line is the threshold, the average outside transmissions plus deltaDB; the selected point is in emission.

5. Analysis and next steps

What works well

  • It meets the need without changing anything in the acquisition chain: it only reads the database.
  • The spectrum is rebuilt from the measurement itself, so it matches the selected moment exactly.
  • The threshold follows the noise level of each channel, so the same dashboard serves every channel; the user only chooses the margin.

Limits, and what I would do next

  • The margin is chosen by hand

    Too small, it counts noise peaks as emissions; too large, it misses the weakest ones.

    Next step: Suggest the margin from the spread of the noise, for example a multiple of its standard deviation, or learn it from emissions already confirmed.

  • One moment at a time

    The spectrum shows a single moment: comparing two moments means switching between them.

    Next step: Show two spectra side by side, or a waterfall: the spectrum over time, as a colour map.

  • Someone has to look

    The dashboard shows the emissions, but it warns no one.

    Next step: Grafana alerts when a channel transmits outside its expected windows; later, a model trained to spot unusual emissions, where AI helps to watch over a system.

Conclusion

The dashboard answers the question set at the start. For any channel and any period, the team sees when the channel transmits, displays its spectrum at any moment in one click, and reads its emission rate against a threshold that follows the channel’s noise level. It does so without changing the monitoring application, by reading the database the application already fills.

I carried the project from the need to the documentation: the queries in Flux and InfluxQL, the processing in UQL, the dashboard itself, and the documents that let the teams use and maintain it. The next steps would be a margin suggested automatically and alerts on unusual emissions.

The project lasted several months and put many of my skills to the test: organising my work, developing, testing, improving. It taught me to start from what users lack, to justify each technical choice, and to write documentation that others can rely on.

Sources

  1. [1] Grafana Labs, “Grafana documentation”, grafana.com/docs/grafana/latest/ (opens in a new tab)
  2. [2] InfluxData, “InfluxDB 1.8 documentation”, docs.influxdata.com/influxdb/v1/ (opens in a new tab)
  3. [3] Grafana Labs, “Infinity data source plugin”, grafana.com/docs/plugins/yesoreyeram-infinity-datasource/latest/ (opens in a new tab)
  4. [4] Grafana Labs, “UQL”, in the Infinity data source documentation, grafana.com/docs/plugins/yesoreyeram-infinity-datasource/latest/query/uql/ (opens in a new tab)

Terms

dBm
A power in decibels relative to one milliwatt. −60 dBm is a million times weaker than 0 dBm. A difference between two powers in dBm is in dB: the unit of deltaDB.
Carrier power
The power of the signal that carries a channel. It rises clearly above the noise when the channel transmits.
Frequency spectrum
The power of a signal frequency by frequency, at a given moment.
Flux and InfluxQL
Two query languages for InfluxDB: Flux can filter, calculate and aggregate in the database; InfluxQL is close to SQL.
UQL
The query language of Grafana’s Infinity data source, used to reshape a JSON answer before it is displayed.

Skills demonstrated

Each skill links to the skills map on the Projects tab.

Command palette

Welcome
Projects
Career
Mobility
Passions
Civic Engagement
At a glance
About
Journey
Contact
Skills
The job I am aiming for
CV
Cover letter
Experience
Education
Credentials
Languages
Mobility projects
Abroad so far
Secure IoT telemetry
Monitoring a satellite’s emissions with Grafana
Secure multi-site architecture
BâtIA — predicting building failures
5G Standalone network
System and network hardening
Vulnerability assessment and exploitation
Multi-site network design and security
Regenerative repeater for satellite links
Use the light theme
Use the dark theme
Use the system theme
Lire en français
Copy my email address
Download my CV (PDF)
Open my LinkedIn profile
Back to the top