Rockets drop geostationary satellites into a long, egg-shaped transfer orbit because it is cheaper; the satellite's own engine then rounds off the orbit at about 36,000 km.
A geostationary satellite circles about 36,000 km above the equator, once a day. Getting there in one go would need a much bigger rocket. Instead, the rocket does part of the job and the satellite does the rest.
The rocket's upper stage fires near Earth and puts the satellite into a geosynchronous transfer orbit (GTO). This is a stretched ellipse whose low point is a few hundred kilometres up and whose high point is near geostationary height.
The satellite then coasts out to the high point. There it fires its own engine — a burn — to speed up and round off the orbit. The same firing can also remove the orbit's tilt, so it ends up flat over the equator.
Chandrayaan-3 used a similar split for a different goal. LVM3-M4 placed it in a parking orbit of about 170 × 36,500 km, and the spacecraft's own propulsion module did the rest of the journey.
ISRO rates its rockets by what they can put into GTO:
LVM3 can carry 8,000 kg to a 600 km low orbit — twice its GTO figure. And GTO is only part of the way: the satellite must still carry its own propellant to finish the climb.
The satellite's engine is as important as the rocket. In January 2025, GSLV-F15 placed the navigation satellite NVS-02 in a transfer orbit of 170 × 37,785 km, inclined at 20.8°. The rocket did its job. But the command signal did not reach a valve on the satellite's oxidiser line, so its engine could not fire. ISRO's review found that a connector had most likely lost contact. NVS-02 could not be raised to its circular orbit and stayed in its elliptical transfer orbit.
Each NavIC satellite you see in its final orbit began in a transfer orbit and climbed the rest of the way with its own engine. Track NavIC satellites
This is a Hohmann-style transfer. Use GM = 398,600 km³/s², a low orbit at 180 km (r1 = 6,558 km) and geosynchronous radius r2 = 42,164 km.
Circular speeds. v_low = √(GM/r1) = √(398,600 / 6,558) ≈ 7.80 km/s v_GEO = √(GM/r2) = √(398,600 / 42,164) ≈ 3.07 km/s
Transfer ellipse. a = (6,558 + 42,164) / 2 = 24,361 km. Perigee speed: √(GM × (2/r1 − 1/a)) ≈ √(398,600 × 2.639 × 10⁻⁴) ≈ 10.26 km/s Apogee speed: √(GM × (2/r2 − 1/a)) ≈ √(398,600 × 6.39 × 10⁻⁶) ≈ 1.60 km/s
The two burns. Burn 1 (rocket, at perigee): 10.26 − 7.80 ≈ 2.46 km/s. Burn 2 (satellite, at apogee, no tilt): 3.07 − 1.60 ≈ 1.48 km/s.
Adding the plane change. A transfer orbit from Sriharikota is tilted, for example 20.8° for NVS-02. Combining the tilt with the apogee burn costs
Δv = √(v1² + v2² − 2·v1·v2·cos i) = √(1.60² + 3.07² − 2 × 1.60 × 3.07 × cos 20.8°) ≈ 1.68 km/s
Doing the plane change at apogee, where the satellite is slowest, is cheapest. That is another reason the last step is left to the satellite.
Splitting the job lets one rocket serve many satellites: the launcher only needs to reach GTO, and each satellite carries the propellant for its own final 1.5–1.7 km/s.
The apogee is at geostationary height, and the satellite is moving slowest there, so speeding it up to circular speed (and removing any tilt) costs the least propellant.
The rocket placed it correctly in a transfer orbit, but the drive signal did not reach the oxidiser-line valve of its engine, so orbit raising could not be done and it stayed in the elliptical orbit.
Reaching GTO needs much more speed than a low circular orbit, so more of the rocket's capacity goes into velocity and less is left for payload.