Try it in one minute
Install, start it, and pick a card on the welcome screen. Every part of the app has a demo with a known
answer – nothing to prepare, nothing leaves your PC.
| Card on the welcome screen | What you will see |
|---|
| A drone flight | The flight on the map and in 3D over real terrain, played back in time with its attitude |
| An ArduPilot flight log | A real .bin log with its automatic health check: vibration, GPS, battery, EKF, motors |
| The same mission flown twice · Three car runs | Every shared signal ranked by how different it is |
| A simulation against its flight | The model beside the real flight, aligned in time, judged within a tolerance |
| Find out what went wrong | A known fault – find which signal changed first, and what led what |
| A sweep: which settings hold | 810 runs of a controller sweep – the 12 settings that hold in every wind and payload, the best one |
| One gain set, three drones | One set of gains that holds for three different drones – and why the others fail |
| Smart sweep demo | The next runs chosen for you, round after round, until it converges – about 64× fewer runs than the full grid |
| A fleet of flights | Ten flights of two copters – the one whose vibration grows, flight after flight |
What it does for you
1 · Find what went wrong
Put two cursors around the moment it went wrong and ask What changed? – every signal that left its
usual range, in time order, the likeliest cause first, and which one led which. Then name the event
(abs(Roll - DesRoll) > 5), find every time it happened, and line them up.
Tutorial: what went wrong, in ten minutes
2 · Judge a run against a golden one
A golden run with a tolerance per signal judges the next recording – pass or fail, the worst moment, the
verdict ready for the report. The same check runs in CI.
signalviewer-cli compare golden.csv runs/*.csv --tol-pct 2 --junit results.xml # exit 1 when one fails
signalviewer-cli check runs/*.csv --checks checks.json # limits, per flight mode
3 · See the flight – on the map, in 3D, as a story
Truth, each navigation solution, each GPS and EKF lane – on an offline map or in 3D over real terrain,
compared in position, velocity and attitude. ▶ Story plays the flight as a short film: the camera
flies to each moment that matters.
4 · Choose parameters that hold everywhere
A results table with a row per run: which parameter sets hold in every scenario, the best of them, and
how the runs spread over each scenario. With several systems, one set for all of them. Then let the
app run the loop: it proposes the next runs, starts your program (or hands them to your farm's
launcher), reads the results and goes on until the best stops changing.
One parameter set for many systems ·
The smart-sweep demo ·
Your farm as the runner
5 · Watch the whole fleet
Point it at the folders of logs. Every flight is checked the same way; ask which flights had errors? or
is the vibration growing over the weeks? – and open the one that needs a look.
Tutorial: what the Fleet is for
6 · Send the answer, not a screenshot
One HTML file that opens with the process in brief – the data, what was checked, what it shows and the
conclusion – then plots, map and 3D on one time bar with Play. Any browser, no install, no network. PDF and
Word too.
Fast on big data – without cutting corners
- A 1.2 GB ArduPilot log opens in seconds; every sample stays available at full resolution, and the display
keeps every spike.
- A sweep of a million runs opens with its answer in about 2 s; switching between parameter sets and
every run does not work the answer out again.
- Every computation is checked against an independent reference – 29 validation cases,
run them yourself with
signalviewer-cli validate.
Get started
- Windows: the installer or the portable
.zip from the latest release – no administrator rights.
- Linux / macOS: the
-portable.7z, or Python 3.9+: pip install signalviewer, then signalviewer.
- From source:
git clone, pip install -r requirements.txt, python run.py.
Nothing needs the internet – not the map, not the 3D terrain. No account, no telemetry.
In your scripts
import signalviewer as sv
rec = sv.load("flight.bin")
sv.scene_html(rec, path="flight-3d.html") # the flight in 3D, one offline page
python -m signalviewer.study loop drones.svstudy.json # the sweep loop, from a terminal or a scheduler
Command line and Python · Scripting API · Everything it does · CHANGELOG
Where it fits
| You mostly… | Also used | What Signal Viewer adds |
|---|
| stream ROS topics live | PlotJuggler, Foxglove | recorded runs compared, pass / fail, live reports |
| work in MATLAB / Simulink | Simulation Data Inspector | no licence; flight logs and CAN next to the .mat; sweeps of runs |
| review one flight online | UAV Log Viewer, PX4 Flight Review | offline, many flights side by side, a fleet, your limits in CI |
בעברית בקצרה
Signal Viewer – מוצאים מה השתבש, מוכיחים מה תקין, ושולחים את התשובה. הכול על המחשב שלכם, בלי אינטרנט.
- לכל חלק באפליקציה יש הדגמה במסך הפתיחה – טיסה, לוג ArduPilot, השוואה לסימולציה, חקירת תקלה, סריקת פרמטרים, הלולאה החכמה וצי טיסות.
- מה השתבש: שני סמנים סביב הרגע – והאפליקציה מראה מה השתנה ראשון, מה הסיבה הסבירה ומי הוביל את מי.
- סריקות: אילו הגדרות עומדות בכל תרחיש, סט אחד לכמה מערכות, ולולאה שמריצה את הקוד שלכם עד התכנסות.
- דוח: דף HTML חי שנפתח ב"התהליך בקצרה" – הנתונים, מה נבדק, מה מוצג והמסקנה.
התקנה: מ־השחרור האחרון, או pip install signalviewer. מדריך: docs/GUIDE.he.md.
License
MIT – free to use, change and ship, also commercially. Third-party parts keep their own licences: uPlot (MIT), three.js (MIT), the report fonts (SIL OFL), CesiumJS (Apache-2.0), map data from OpenStreetMap (ODbL), Natural Earth and the AWS Terrain Tiles (details).
