Signal Viewer

A fast, offline desktop app for exploring, comparing and reporting on time-series data
CSV · Parquet · MATLAB .mat · ArduPilot .bin · MAVLink .tlog · PX4 .ulg · ROS 2 .mcap · HDF5 · CAN + .dbc

⬇ Download for Windows  ·  pip install signalviewer  ·  Documentation  ·  מדריך בעברית

For flight-test, drone and simulation engineers who have recordings to judge: several runs of the same test, a flight log after a flight, a simulation next to the real thing – often on a closed network. Open the files, see what is wrong, compare, report; then make the same check pass / fail in CI.

• Plot – up to 9 linked plots per tab: Time, X-Y, FFT, spectrogram, map, 3D; two cursors with Δ and statistics.
• Compare runs – every shared signal ranked by how different it is, time-aligned, also per segment (the 2nd AUTO of every flight); the work done on one file is copied to another in one step (its changes, new signals and plots - any of them), linked files keep getting it, and one file's work can be cleared alone.
• Flight logs – ArduPilot, PX4 and MAVLink read directly: flight modes, events, parameters and an automatic health check.
• Live – MAVLink over UDP, TCP or a serial radio, recorded to a .tlog as it arrives.
• Map and 3D – the vehicle, its attitude and its track on an offline map and terrain, in step with the timeline.
• Math – formulas (2*Speed - RPM/95), filters, derivatives, integrals, coordinate frames (NED / ENU / body / ECEF / LLA); the same change on many signals at once (x*8 or sin(x) on each one selected).
• Reports – offline HTML, PDF and Word; one per file from the command line.
• CI – signalviewer-cli check (limits) and compare (against a golden run) with exit codes and JUnit XML.
• Any recording – CAN with a .dbc, raw binary packets from a layout (ICD), or your own format as a short Python reader.

Under the hood: a Python core (FastAPI + NumPy) and a React / TypeScript UI in one native window. Multi-gigabyte recordings open in seconds, every sample stays available at full resolution, and nothing needs the internet – not even the map and the 3D globe.

---

See it in action

Open a file and build a view

Open a file, choose a layout, click signals into plots, then zoom and read values with two cursors.

A flight on the map and in 3D

One template turns latitude / longitude / altitude / attitude signals into a map and a 3D view. Press play and both follow the timeline; Up to the time draws the track as it is flown.

Compare runs

Pick several files: every shared signal is ranked from least to most similar. Click a row to overlay it, align the runs in time automatically, or open the table of all signals.

Flight-log health check

Open an ArduPilot .bin and the Log panel finds vibration, clipping, battery, EKF and motor problems. Click a finding to jump to it; click an event to move the cursor there.

Live from the vehicle

Live ▸ MAVLink: connect to a vehicle, a simulator or a ground station that forwards (UDP, TCP or a serial radio). The link is recorded to a .tlog that opens like any file and grows in live mode; Follow keeps the last seconds in view.

Higher-quality MP4 versions of these clips are in docs/media/.

---

Quick start

git clone --depth 1 https://github.com/DavidAlpert11/SignalViewer.git   # without the history: much smaller
cd SignalViewer
pip install -r requirements.txt
python run.py                      # opens the desktop window (or the browser)

The welcome screen opens each example in one click (a drone flight · the same mission flown twice · three car runs to compare · an ArduPilot flight log), or press Ctrl+O and open the files in examples/:

File	Try
examples/run_012.csv	click Speed, RPM, Throttle, DriveMode into a 2 × 2 layout; press Space and D for two cursors
examples/run_012.csv + run_013.csv + run_014.csv	open all three, then Compare in the side rail
examples/uav_flight.csv	Templates ▾ → Position on the map & 3D, then press P to play
examples/uav_flight.csv + uav_flight_windy.csv	the same mission flown twice, the second in a cross-wind: both tracks on the map, Compare shows where they differ (roll, ground speed, battery)
examples/copter_flight.bin	an ArduPilot quad-copter flight: the start card suggests ArduPilot · Full review and points at the vibration burst it found; the Log button has the events, parameters and the whole health check
a report per file	python cli.py report --template "ArduPilot · Full review" --out reports/ logs/*.bin (HTML, --format pdf or docx; python cli.py templates lists them)
pass / fail in CI	python cli.py check examples/run_01*.csv --checks examples/checks.json --junit results.xml - exit code 1 when a limit is broken, 2 when a file or signal is missing; --template NAME uses the limits saved in a template
against a golden run	python cli.py compare examples/run_012.csv examples/run_014.csv --tol-abs 15 - every shared signal, the most different first; outside the tolerance = exit code 1
your own format	python cli.py readers --example writes ~/.signalviewer/readers/example_kv_reader.py: a function marked @reader(extensions=[".kv"]) that returns the time and the signals; files of that type then open like the built-in ones
raw binary packets	python cli.py readers --example-icd examples_run.tlm writes a layout (example_bench.icd.json: sync, ids, fields with type / scale / unit / bits / names) and a recording; open the .tlm like any file
live data	ArduPilot SITL or MAVProxy on UDP 14550, then Live ▾ → MAVLink → Connect and record

The demo files are synthetic and are regenerated by python tools/make_demo_data.py (the .bin by tools/make_example_bin.py, a real DataFlash log; tools/make_test_bin.py makes test logs of any length).

---

Install

• Windows: SignalViewer-Setup-X.Y.Z.exe from the latest release (installs for you only, no administrator rights), or the portable .zip. Linux / macOS: the -portable.7z for your system.
• Python 3.9+: pip install signalviewer, then signalviewer - every reader and report; it opens in your browser, pip install "signalviewer[desktop]" gives it its own window. Short of disk space? pip install signalviewer-lite is the core (about 350 MB instead of 760) with the rest as extras (signalviewer-lite[ros], [parquet], [all]…); a file that needs one says which. Install one of the two, not both.
• From source: pip install -r requirements.txt and python run.py.

Nothing needs the internet at run time - the packages carry Python, every library, the UI, the map and the 3D terrain. The packages are not code-signed yet: on the first start Windows may say Windows protected your PC (More info → Run anyway) and macOS may refuse to open the app (what to do). Building the installer and the portable packages yourself, and running from source: Install.

Command line and CI

signalviewer-cli export flight.bin --signals "ATT.Roll,ATT.Pitch" --out roll.csv
signalviewer-cli report --template "ArduPilot · Full review" --out reports/ logs/*.bin
signalviewer-cli check  runs/*.csv --checks checks.json --junit results.xml     # exit 1 on a broken limit
signalviewer-cli compare golden.csv new/*.csv --tol-pct 2 --allow-outside 0.5   # against a golden run
signalviewer-cli validate                                                       # the numerical validation, here

import signalviewer as sv
t, roll = sv.load("flight.bin").signal("ATT.Roll")
sv.show(t, {"roll x2": 2 * roll})          # into the open app

More - the checks and tolerances files, your own readers, binary layouts, CAN: Command line, Python and formats.

Where it fits

You mostly…	A good fit	Where Signal Viewer adds
judge test runs, flight logs and simulation output, often offline	Signal Viewer	–
stream ROS topics live	PlotJuggler, Foxglove	comparing recorded runs, pass / fail, reports (it reads their .mcap, .ulg)
work inside MATLAB / Simulink	Simulation Data Inspector	no MATLAB licence, flight logs and CAN next to the .mat (a worked example)
review one ArduPilot / PX4 flight online	UAV Log Viewer, PX4 Flight Review	offline, several flights side by side, your own limits in CI
write the final analysis as a script	Python / MATLAB	a quick look first; sv.load() / sv.show() go both ways

Nothing leaves the PC: no account, no telemetry; English and Hebrew.

Documentation

The documentation site - the user guide (עברית), what works (every feature and the keyboard), command line and formats, install and build, offline map and 3D, performance, known limitations, validation (every computation checked against a reference), gallery, development. The same guide is in the app: Help → User guide (F1). Changes: CHANGELOG. Found a problem or want to help? Help → Report a problem… in the app, or CONTRIBUTING.

בעברית בקצרה

Signal Viewer – אפליקציה לצפייה, השוואה ודוחות על נתוני זמן: ריצות ניסוי, לוגים של טיסה, תוצאות סימולציה, לוגים של CAN והקלטות בינאריות. עובדת כולה בלי אינטרנט, והקבצים לא יוצאים מהמחשב.

התקנה

• Windows: קובץ ההתקנה SignalViewer-Setup-X.Y.Z.exe מ־השחרור האחרון (בלי הרשאות מנהל), או ה־zip הנייד.
• Python: pip install signalviewer ואז signalviewer (נפתח בדפדפן; חלון משלו: pip install "signalviewer[desktop]").
• מקוד המקור: pip install -r requirements.txt ואז python run.py.

מה אפשר לעשות

• לגרור קבצים לחלון: CSV, ‏MATLAB, לוגים של ArduPilot ו־PX4, ‏MAVLink, ‏ROS 2, ‏HDF5, ‏Parquet, ‏CAN עם קובץ ‎.dbc.
• לבנות תצוגה: עד 9 גרפים מקושרים בטאב, שני סמנים עם הפרש וסטטיסטיקה, FFT, ספקטרוגרם, מפה ותלת־ממד.
• להשוות ריצות: כל הסיגנלים המשותפים מדורגים לפי מידת השוני, יישור זמן אוטומטי, אותו מקטע בכל ריצה – גם סימולציה מול הטיסה שלה. העבודה שנעשתה על קובץ אחד מועתקת לקובץ אחר בצעד אחד (שינויים, סיגנלים חדשים, גרפים - כל אחד מהם), קבצים מקושרים ממשיכים לקבל אותה, ואפשר לנקות עבודה של קובץ אחד בלבד.
• לחשב: נוסחאות, מסננים, נגזרות ואינטגרלים – גם על סיגנלים רבים בבת אחת – ופעולות Python משלכם.
• לשתף: דוחות HTML, ‏PDF ו־Word, סשנים ותבניות.
• לבדוק אוטומטית ב־CI: הפקודות signalviewer-cli check ו־compare מחזירות עבר/נכשל ו־JUnit XML.
• לסמוך על המספרים: כל חישוב נבדק מול ערך ייחוס (signalviewer-cli validate), והקוראים נבדקו על לוגים של כלי טיס אמיתיים.

מדריך מלא בעברית: docs/GUIDE.he.md, וגם בתוך האפליקציה: עזרה ← מדריך למשתמש (F1). מה חדש בכל גרסה: CHANGELOG. מצאתם בעיה? עזרה ← דיווח על בעיה… באפליקציה. רוצים לתרום? CONTRIBUTING.

License

Signal Viewer is released under the MIT License: free to use, change and ship, also commercially, as long as the copyright notice stays with it.

It ships with third-party parts under their own licences: uPlot (MIT, backend/app/report_assets/uPlot.LICENSE), the report fonts (SIL Open Font License, backend/app/report_assets/fonts/OFL-LICENSE.txt), CesiumJS (Apache-2.0), and map data from OpenStreetMap (© OpenStreetMap contributors, ODbL), Natural Earth (public domain) and the AWS Terrain Tiles (see Offline map data). Python and JavaScript libraries keep their own licences (requirements.txt, frontend/package.json).