Where are the first Starlink V3 satellites right now?
On 28 September 2026 Starship reached orbit for the first time and released 26 Starlink V3 satellites. Until official orbital data is published, this page tracks them using an orbit reconstructed frame by frame from SpaceX's launch webcast.
Upcoming ground track
How the position is measured
No orbital elements for the V3 satellites were public when this page was built. Instead, the orbit is reconstructed from three independent sources: the published mission timeline, the on-screen telemetry (speed and altitude) of the SpaceX webcast, and the small globe in the bottom-right of the webcast, which marks the ship's position throughout the orbital coast. The satellites were released from the ship between T+34 and T+65 minutes, so for the first hours the ship's orbit is their orbit.
01Mission timeline and webcast sync
Liftoff was at 12:48:59 UTC from Pad 2 at Starbase. In the webcast recording, the mission clock reads T+00:00:00 at video time 0:35:51, confirmed at two other points (T+00:25:23 at 1:01:14 and T+00:50:00 at 1:25:51). The orbital insertion burn ends at video 1:01:11 (T+25:20) and the deorbit burn begins at 2:48:17 (T+2:12:26). Everything between those is free, unpowered orbital flight.
02On-screen telemetry
The webcast overlay shows speed (km/h, Earth-relative) and altitude. Frames were sampled every 10 minutes from the 4K (3840×2160) recording. During the coast, altitude oscillates between 262 km (near T+66) and 277 km (T+36), consistent with the reported 262 × 277 km orbit; speed moves inversely, 26,367–26,436 km/h. After the deorbit burn both change sharply.
Telemetry readings (table)
03Reading the webcast globe by pixel registration
The globe widget is only about 270 pixels across even in 4K, so reading it by eye gives errors of 3–15°. Instead, every frame was measured automatically:
- Rim fit. The globe's bright upper limb was traced along 71 rays and fitted with a circle: centre (3529.6, 1974.7) px, radius 134.5 px, residual 0.5 px rms. The position marker sits at a fixed screen point 5 px below the centre.
- Camera model. Each pixel inside the disc is projected onto a sphere through a perspective camera with unknown sub-camera latitude, longitude, roll and distance. Camera distance and roll were calibrated on two clean frames: the widget is effectively orthographic, north-up (roll 0°).
- Coastline matching. A 1° Natural Earth land mask is compared with the frame's brightness, row by row (removing the widget's top-to-bottom shading), and the view orientation maximising the correlation is found by grid search plus local refinement.
- Marker to coordinates. The fitted camera converts the marker pixel into a latitude and longitude. Varying the camera parameters over their plausible range moves the result by only about 1°.
This was repeated every 3 minutes from T+28 to T+2:10, giving 35 positions spanning 1.1 orbits.
04Orbit fit
A circular orbit with J2 (Earth-oblateness) secular drift of the node and argument of latitude was fitted to the 35 positions by least squares (Nelder–Mead) over inclination, right ascension of the ascending node, phase and mean altitude. Ground positions use Greenwich sidereal time.
Measured positions (table)
05From the ship's orbit to the satellites
The satellites were all released from the same ship, so they start at essentially the same point. Two effects pull them away from the ship's fitted orbit over the first hours:
- Separation velocity spreads them into a train. At 0.1–1 m/s, the train is roughly 5–50 km long after five hours (along-track drift ≈ 3·Δv·t).
- Atmospheric drag at 270 km makes them lose altitude and fall behind. With plausible density and solar-array orientation, that is 4–126 km behind after five hours; the middle case is about 30 km.
Both are smaller than the fit uncertainty, so the page shows one marker with a ±1° (≈110 km) uncertainty ring, offset slightly behind for the middle-case drag.
| Density (kg/m³) | Edge-on 0.01 m²/kg | Typical 0.03 | Face-on 0.06 |
|---|---|---|---|
| 3×10⁻¹¹ | 4 km | 13 km | 25 km |
| 7×10⁻¹¹ | 10 km | 29 km | 59 km |
| 1.5×10⁻¹⁰ | 21 km | 63 km | 126 km |
Along-track lag ≈ ¾·n·(ȧ/a)·t², with ȧ = √(μa)·ρ·B. B is the ballistic coefficient CdA/m. Values are assumptions, not measurements.
V2 Minis start raising within days, pause near 335–350 km, and reach their operational shells in roughly 3–6 weeks. A V3 only needs to climb about 80 km, from 270 to about 350 km, so arrival in 2–4 weeks is a reasonable guess, not a published figure. Once raising starts, this page's model becomes steadily less accurate.
V2 Mini climb data (table)
06Limits of this estimate
- It models the ship, not each satellite. The 26 satellites are shown as one marker; their individual spacing is unknown.
- Accuracy decays with time. About ±1° along-track now; once the satellites begin orbit raising (expected within days), the real positions will drift ahead of this track.
- Inclination disagreement. The globe fit gives 30.5°, press reports say 32°. The fit uses the webcast value.
- Official data supersedes all of this. When SpaceX publishes ephemerides (mirrored by CelesTrak) or the US Space Force catalogs the objects, this page should switch to them.