A full size, 63 ft (19 m) tall, Atmospheric Test Vehicle (ATV) was built under contract by Scaled Composites for use in hover test flights. The $2.8 million ATV was not intended as an all-up test article, since it had no rocket engine and no heat shielding. The ATV was rolled out of its Mojave hangar on March 1, 1999, bearing an FAA registry of N990RR.
The rotor head was salvaged from a crashed Sikorsky S-58, at a price of $50,000 — as opposed to as much as $1 million for a new head. Each rotor was powered by a 350-lbf (1,560 N) hydrogen peroxide jet, as intended for the orbital vehicle. The rotor assemblage was tested in a rock quarry before installation on the ATV.
The ATV flew three successful test flights in 1999. The pilot for these three flights was Marti Sarigul-Klijn and the copilot was Brian Binnie (who later gained fame as pilot of Scaled Composites' SpaceShipOne on its second X-Prize flight).
The ATV made its first flight on July 28. This flight consisted of three vertical hops totaling 4 min 40 sec in duration and reaching a maximum altitude of 8 ft (2.4 m). The pilots found the flying extremely challenging for a number of reasons. Visibility in the cockpit was so restricted that the pilots nicknamed it the "bat cave". The view of the ground was entirely obstructed, so the pilots had to rely on a sonar altimeter to judge ground proximity. The entire craft had a low rotational inertia, and torque from the spinning rotor blades made the body spin, unless counteracted by yaw thrust in the opposite direction.
The second flight, on September 16, was a continuous hover flight lasting 2 min 30 sec, reaching a maximum altitude of 20 ft (6.1 m). The sustained flight was made possible by the installation of more powerful rotor tip thrusters and an autothrottle.
The third and last flight was made on October 12. The ATV flew down the flightline at Mojave Airport, covering 4,300 ft (1,310 m) in its flight and rising to a maximum altitude of 75 ft (23 m). The speed was as high as 53 mph (85 km/h). This test revealed some instability in translational flight.
A fourth test was planned to simulate a full autorotative descent. The ATV would climb to an altitude 10,000 ft (3,050 m) under its own power, before throttling back and returning for a soft landing. At this point, given that further funding was then unlikely, safety considerations prevented the test being attempted.