About azmth
azmth is a free, real-time satellite tracker. It shows every active satellite orbiting Earth on an interactive 3D globe, predicts when any of them will pass over your location, detects freshly launched Starlink trains, and explains the orbital mechanics behind all of it in plain language. There is no signup and no advertising; the project is built by one developer and supported by donations. It has been featured by PCWorld and in the Morning Brew newsletter, and cited by Scientific American and Il Post.
The name comes from azimuth: the compass direction you look toward to find a satellite crossing your sky.
Data sources
- CelesTrak general perturbations (GP) orbital data, maintained by Dr. T.S. Kelso. azmth refreshes its catalog every 2 hours, matching CelesTrak's update cadence.
- CelesTrak SATCAT for operator and country attribution.
- NASA Visible Earth for the Blue Marble Earth textures (public domain).
- NASA 3D Resources for most of the 3D models shown in the True Scale view (public domain): the ISS, Hubble, GOES, TDRS, Terra, Aqua, Landsat, Sentinel-6, Suomi NPP, Chandra, Fermi, Swift, the Solar Dynamics Observatory, Hinode, SORCE, Cluster II, GRACE, OCO-2, Aura, ICESat-2, Jason, CYGNSS, RADARSAT-1, the SSL-1300 communications bus, and the generic 3U CubeSat.
- Our own models, built in Blender from published dimensions and photographs where no agency model exists and we could not license a suitable one: Meteosat Second Generation, Amazon's Kuiper (now Amazon Leo), and the Soyuz and Progress ferries in Earth orbit; ESA's Mars Express and the UAE's Hope in orbit at Mars; and the Sojourner and Zhurong rovers on the ground there. The four Mars ones are labelled approximate wherever the app credits them, and where a model stands in for a satellite rather than depicting it, the True Scale view labels the shape as approximate too.
- Community 3D models for the Tiangong station (by w29572227) and GPS satellites (by ABC Almaty), both via Sketchfab under the Creative Commons Attribution (CC-BY) licence.
- CelesTrak SupGP for SpaceX-published ephemerides of just-launched Starlink batches, tracked ahead of official cataloging and labeled provisional.
- astronomy-engine for the positions of the Moon and bright planets in the sighting identifier (MIT licensed, accurate to about one arcminute).
- GeoNames for the city database used in manual location entry.
Methodology
Satellite positions are computed in your browser using the SGP4/SDP4 orbital propagation models via the satellite.js library, from each object's latest general perturbations elements. The catalog is baked into the site and refreshed every 2 hours by an automated pipeline, so the app makes no live calls to CelesTrak when you use it. Orbit classification (LEO, MEO, GEO, HEO) uses standard altitude bands derived from each satellite's elements; characteristics like sun-synchronous or Molniya orbits are derived from inclination, period, and eccentricity.
Accuracy
Pass predictions are typically accurate to within about 30 seconds for satellites with fresh orbital elements. Accuracy degrades for objects with stale elements or rapidly decaying orbits. Whether a satellite is sunlit and whether your sky is dark are geometry, worked out from the Sun's position in your own sky. How bright it will look is a prediction: from published measurements where they exist, and otherwise from an estimate based on its size, labelled as one. How each pass label is decided is set out under How we judge a pass, below. Satellite dimensions shown in the True Scale view are documented values for well-known satellites and clearly labelled class-level estimates for the rest.
How we judge a pass
Every pass azmth lists carries one label saying how you could see it if the sky is clear: with your eyes, with binoculars, with a telescope, or not this time and why. The i beside each label gives the reason in one line. We judge every moment of the pass, not only its highest point, because a satellite can be sunlit for the first minutes of a pass and then fade into Earth's shadow. A label describes a prediction, never a promise: clouds, haze, the Moon and your own horizon are not part of it.
The checks, in order
- Is it sunlit while your sky is dark? Satellites only shine by reflecting sunlight, and most only stand out once the Sun is at least 6° below your horizon, the end of civil twilight as the US Naval Observatory defines it and the same cut-off Heavens-Above uses for its predictions. If a satellite is sunlit only while your sky is still too bright, the label is “Sky too bright”. Something as bright as the space station can sometimes be seen a little earlier, and its note says so. If Earth's shadow covers it whenever your sky is dark enough, the label is “In Earth's shadow”.
- Is it high enough? A pass that climbs past 20° is judged on the part above that height, the same floor Heavens-Above sets for the passes it lists. Lower down a satellite shines through more air, and it is often behind buildings or trees. A pass that never gets that high, but is sunlit in a dark sky down there, reads “Low in the sky”, unless it would be too faint anyway.
- Is it bright enough for what your sky shows then? We compare its predicted brightness, as a magnitude (the lower the number, the brighter), with the faintest thing your eyes can pick out in that sky:
- “Easy to see”: a published measurement of this very object, or of new Starlink trains, puts it at magnitude 3.0 or brighter, and within your eyes' reach. The 3.0 leaves room for a suburban sky and for how much harder a moving light is to follow than a star (Bortle's dark-sky scale; Bob King, Sky and Telescope).
- “Could be visible”: within your eyes' reach on our figures, from any source. The fainter ones need a dark spot away from streetlights.
- “Try binoculars”: too faint for your eyes, within reach of binoculars (see below).
- “Needs a telescope”: slow, distant satellites only (see below).
- “Too faint to see”: none of these, on our figures.
- “Brightness unknown”: sunlit in a dark sky, but we have no size to estimate its brightness from, or it is lit from behind, where our size-based estimate is too unreliable to use.
A satellite that binoculars would reach in a fully dark sky, but that passes while twilight is still too bright for something that faint, is marked “Sky too bright” as well.
Your eyes' reach as the sky darkens
Just after sunset only the brightest stars show, and fainter ones come out as the Sun sinks further. For each depth of the Sun below the horizon, we take the faintest star you can see straight overhead from a figure in Weaver's 1947 study (Publications of the Astronomical Society of the Pacific). We read the values off the printed curve ourselves, so they carry some reading error, and we cap them at magnitude 6.0, the dark-sky figure Mallama and Cole (2025) use as their reference. By the time the Sun is 12° down, the end of nautical twilight, Royal Museums Greenwich counts the sky as dark for most purposes.
| Sun below your horizon | Faintest star overhead |
|---|---|
| 6° (end of civil twilight) | 1.3 |
| 7° | 2.3 |
| 8° | 3.1 |
| 9° | 3.5 |
| 10° | 3.8 |
| 12° (end of nautical twilight) | 4.1 |
| 14° | 5.1 |
| 16° | 5.6 |
| 18° | 5.8 |
| 20° or more | 6.0 |
These are values for straight overhead, in a clear sky. The sky towards the sunset is brighter, and town light brightens it everywhere, so from most places the faintest thing you can see is brighter than these figures. Bortle's dark-sky scale sets out by how much, from a truly dark site to a city centre.
Where the brightness figures come from
- Named objects. The ISS and China's Tiangong station use their brightness as measured in the MMT archive of Kazan Federal University (Karpov and others, 2016). Hubble uses the figure Heavens-Above publishes, an observers' figure rather than a peer-reviewed one. Where two published figures differ, we use the fainter. Each is carried to the pass by its distance and by how much of the satellite's sunlit side faces you.
- Fleet averages. Starlink, OneWeb, Qianfan and Guowang use the averages Mallama and Cole (2025) measured, Amazon Leo those of Mallama and others (2026), and navigation satellites the middle of the MMT archive's measurements, about magnitude 10 at their distance. These are averages over a whole fleet, carried to the pass by distance alone, and single satellites vary. Fleets also change over the years, so we check the table again every year.
- New Starlink trains. A batch counts as new only while it is at most 7 days from first being tracked and still below 357 km, the height below which Mallama and others (2024) measured the new Starlinks of the time at their brightest. Height alone is not enough, because the FCC has since authorised working Starlink orbits below that height too.
- Everything else gets an estimate from its size: a plain, evenly reflecting ball with an assumed reflectivity of 0.2, the default the Stellarium planetarium uses, and a size that is a guess for its kind of satellite unless we hold its published dimensions. An estimate is shown in square brackets, and never earns “Easy to see”. Where we hold no size at all, the label is “Brightness unknown”.
When we suggest binoculars or a telescope
Binoculars are suggested when a satellite is too faint for your eyes by no more than 2 magnitudes, and never past magnitude 8.0. That gap is the one the observers' FAQ at satobs.org gives between the naked eye and binoculars on satellites under a dark sky; using the same gap in twilight is our own assumption. From a town, binoculars reach less (Mallama, 2022), which is why the label says “try” and its note says they “may” show it.
A telescope is suggested only for satellites that crawl across the sky, slower than 0.03° a second, such as navigation and geostationary satellites. Only once the Sun is 12° down, and only with a place to point it: a time, a height and a direction. A 100 mm telescope reaches about magnitude 12.7 under a dark sky at high magnification (RASC Observer's Handbook), and Bob King of Sky and Telescope finds many geostationary satellites in telescopes that size. We never suggest a telescope for a fast satellite in a low orbit: it crosses a telescope's small field almost at once, and following one takes a computer-driven mount (satobs.org).
Flares, and what we leave out
Satellites can flare: a flat panel catches the Sun and flashes far brighter for a moment. We do not predict flares, so “Too faint to see” means too faint on our figures, not impossible to see. Satellites are also sometimes harder to see than predicted, and every label assumes a clear sky. The sighting identifier uses the same figures only to rank what you may have seen, never to rule a satellite out, for the same reason.
azmth Luna
The Moon half of azmth: an interactive 3D Moon at azmth.space/moon with the real terminator and libration, every verified landing and impact site, the lunar orbiters we track live, and a catalog of landmark Moon missions with sourced facts and photos. Its data and credits:
- The mission catalogue: landmark Moon missions and 11 explainers (Understanding the Moon) at azmth.space/moon/missions, every claim carrying its source and the date it was last checked.
- NASA SVS CGI Moon Kit for the Moon colour and terrain maps, built from Lunar Reconnaissance Orbiter camera (LROC) and laser altimeter (LOLA) data (public domain).
- NASA/JPL Horizons for the lunar orbiter ephemerides, refreshed twice daily. A probe whose ephemeris lapses shows as no data, never a guessed position; Queqiao-2, which has no published ephemeris at all, is drawn only as an approximate orbit from the one public orbit determination (an amateur radio Doppler fit, September 2024).
- astronomy-engine for the Sun and Earth geometry that drives the terminator, phase, and libration (MIT licensed).
- Mission photo credits appear with each image across the mission catalog; landing coordinates come from the sources cited on each mission page.
- Orbiter 3D models: LRO and the ARTEMIS probes are official NASA 3D Resources models (public domain). Danuri and the Chandrayaan-2 orbiter we built ourselves in Blender from their agencies' published dimensions and imagery; both are labelled approximate in the app.
- Sounds (the replay ambient and UI chimes) are our own.
azmth Mars
The Mars half of azmth: an interactive 3D Mars at azmth.space/mars with the real day and night line and the axial tilt that gives Mars its seasons, every place human hardware has reached, the orbiters and both moons positioned live, and one clock you can speed up or run back to 1960 to watch the whole Mars programme arrive. Its data and credits:
- The mission catalogue: 110 Mars missions and 13 explainers at azmth.space/mars/missions, every claim carrying its source and the date it was last checked. Where it mattered the research went to the Russian, Chinese and Japanese sources themselves rather than to English summaries of them.
- USGS Astrogeology and NASA MOLA for the surface: the Viking colour mosaic for what Mars looks like, Mars Orbiter Laser Altimeter data for the terrain (both public domain).
- NASA/JPL Horizons for the orbiter and moon ephemerides, refreshed twice daily. A probe whose ephemeris lapses shows as no data, never a guessed position; Tianwen-1, which has no published ephemeris at all, is drawn as an approximate orbit and labelled as one.
- PDS Small Bodies Node for the measured shape models of Phobos and Deimos (P. C. Thomas), so both moons are drawn as the irregular bodies they are instead of as spheres.
- astronomy-engine for the Sun and Mars geometry that drives the terminator, the sub-solar point, and the season (MIT licensed).
- Spacecraft and rover 3D models: the Mars Reconnaissance Orbiter, 2001 Mars Odyssey, the ExoMars Trace Gas Orbiter, Perseverance and Ingenuity are official NASA 3D Resources models (public domain). Curiosity and the twin Mars Exploration Rover, which drove as Spirit and Opportunity, come from NASA's Eyes on the Solar System model set, because NASA 3D Resources publishes no glTF of either. Mars Express, Hope, Sojourner and Zhurong we built ourselves in Blender from published dimensions and photographs, because no agency publishes a model of them we could use; all four are approximate and are labelled so wherever the app credits them. Every rover in the site view stands at true size, its mesh measured against the published figures. Phobos and Deimos need no model: their shapes come straight from the PDS measurements above.
- Rover traverses and site imagery: the path each rover drove and every flight Ingenuity made, from NASA's MMGIS layers, the Analyst's Notebook at Washington University in St. Louis, Pathfinder telemetry in the Planetary Data System and the published Tianwen-1 NaTeCam positions (CNSA/GRAS, CC BY 4.0), drawn over HiRISE and CTX pictures of the ground itself (NASA/JPL/University of Arizona and NASA/JPL-Caltech/MSSS).
- Mission photo credits: every image is self-hosted, and its caption carries the credit as a link to the page the picture came from.
- Free data: the places, the catalogue, and the live orbiter and moon positions are published as CSV and JSON at azmth.space/mars/data under CC BY 4.0. The vocabulary a Mars page assumes (the sol, solar longitude, the areoid) is defined at azmth.space/mars/glossary.
Contact
Feedback and corrections: hello@azmth.space. If azmth is useful to you, you can support it at buymeacoffee.com/langguesser.