PRD-032 — Launch (the ascent act of the Mission Arc)

Status · Draft — v2.x epic kickoff Sources · PRD-003 (Mission Arc) · 02_Project_Concept.md §six-screens · ADR-030 · ADR-058 Audiences · curious learner, STEM student (PA §audiences) Promises · real physics, educational at every level, fail honestly (PA §promises) Principles · physics first, prototype is ground truth (PA §principles) Why this is a PRD · Getting off Earth is the single most physics-dense, most cinematically codified event in all of spaceflight — and today /fly skips it entirely, opening the sim already in cruise. The ascent act turns launch from a zero-duration dot into the app's best standalone learning surface: the one place where thrust, gravity, drag, and staging are all legible at once. It is the missing first act of the Mission Arc, and it must feel scientifically true and visually gorgeous in equal measure.

T+00:00. A person is looking down the flank of a Falcon 9 from an onboard camera as the nine Merlins light and the hold-downs release. The HUD reads TWR 1.24, ALT 0.0 km, q ▁. The tower slides past. T+00:12 the vehicle pitches — not straight up, but over, leaning downrange, and a small caption fades in: orbit is sideways, not up. T+01:12 the numbers say Max-Q and the throttle dips. T+02:35 the first stage lets go and falls away against the curve of the Earth. T+08:44 the engine cuts, the vehicle is weightless in a parking orbit, and the camera pulls back until the Earth shrinks to a point and the scene warps into the heliocentric arc the user already knows. One continuous scrub, from the launchpad to Mars.

The problem

Every mission in Orrery "launches" at met_days: 0 — but launch is a point event with no duration, no pad, no atmosphere, no stages. The sim opens with the spacecraft already coasting between planets. The public understands two things about spaceflight: launches and landings. Orrery renders the invisible cruise beautifully and skips the one moment everybody has actually watched.

Worse, the physics of ascent is the most teachable in the whole subject and it's absent. Why do rockets have stages? Why don't they go straight up? What is Max-Q and why does the announcer say "throttle up"? Why is reaching orbit about speed, not altitude? These are answerable in ninety seconds of watching a real ascent with the right numbers on screen — and nowhere in the app answers them.

North star — a live launch broadcast

The bar is not "an educational diagram that moves." The bar is a live launch on television — the immersion, the pacing, the camera language, the tension of a real broadcast (SpaceX webcast / NASA TV / an Everyday Astronaut stream). The user should feel present at a launch, not shown a simulation of one. Every direction call — camera choice, plume, sound, HUD typography, the beat timing — is measured against that bar, the same way /fly measures itself against Voyager/Cassini hero photography. This is a UX-max feature: broadcast-grade or it hasn't landed.

Because the bar is this high, visual direction is locked before build. Slice 0 is a set of concept mockups (art-directed frames for the key beats — pad hold, tower-clear, onboard-looking-down, staging, MECO-over-Earth) presented for a direction lock, per the "visual anchor before UX commit" discipline. No ascent code ships until the look is signed off.

The experience

The user opens /fly?mission=curiosity and, instead of starting in cruise, starts on the pad. The clock reads negative seconds. Ignition, hold-down release, tower-clear — played as a directed multi-camera sequence (the same shot-cutting engine that already cuts flyby montages): a low wide pad shot, a tracking shot following the arc, an onboard shot looking down the body at the Earth falling away.

A telemetry HUD runs the whole time: altitude, velocity (surface- and orbit-relative), downrange distance, dynamic pressure q, thrust-to-weight, remaining propellant, and a running tally of the Δv spent to gravity loss, drag loss, and steering loss — the fuel you burn and never get back. When the Science Lens is on, the live force vectors draw on the vehicle: thrust up the body axis, weight toward Earth's centre, drag opposing velocity — the tug-of-war made visible.

Discrete beats announce themselves as they happen, each with real numbers: liftoff, pitch-over / gravity turn, Max-Q, booster separation / MECO, fairing jettison, second-stage ignition, SECO → orbit, and finally the injection burn that raises the orbit onto the escape/transfer trajectory. That last beat is the seam: the ascent scene zooms out, Earth shrinks to a point, and the view warps into the existing heliocentric transfer arc — which is exactly where /fly begins today.

The user can scrub, pause, and change speed on a single continuous timeline: seconds-resolution through the eight-minute ascent, then days-resolution through the months-long cruise, in one drag from pad to destination.

The launch site as a place

Before ignition, the pad is somewhere the user can be. Rather than a bare geometry, the launch site is an explorable, zoomable, panoramic ground scene — the same treatment /moon and /mars landing sites already get from the surface-scene renderer (RFC-017 / ADR-072): a real place with the vehicle on the pad, the tower, the flame trench, the horizon. The T-minus hold is the moment to establish that place before the physics takes over. This reuse is deliberate — the panorama pipeline exists; pointing it at LC-39A / SLC-40 / Baikonur / Sriharikota / Kourou / Wenchang gives each flagship a recognizable home. (Scope note in RFC-034 §7 — a full per-site panorama is a candidate slice, not a v1 gate; a stylised pad ships first.)

The science it teaches

The ascent act is a curriculum. Each flight phase carries one big idea, surfaced in the HUD and as a Science Lens overlay + ?name= ScienceChip cross-link:

Beat	Big idea	What the user sees
Ignition / liftoff	Thrust-to-weight (TWR) — you only rise when T/W > 1; it climbs as propellant burns off	TWR gauge ticking up at constant thrust
Pitch-over	The gravity turn — orbit is going sideways fast, not up high; it's falling and missing the ground	the vehicle leaning downrange; velocity vector rotating toward horizontal
Ascent	Gravity / drag / steering losses — Δv you spend and never recover	three running counters in the HUD
Max-Q	Dynamic pressure q = ½ρv² — peaks then falls as the air thins; the throttle-down	q bar rising then cresting; throttle needle dipping
Staging	The rocket equation + why we stage — dead tank mass wrecks the mass ratio, so you throw it away	Tsiolkovsky readout; mass ratio jumping at sep
Fairing jettison	Out of the air — drag gone, so dump the shroud's dead mass	petals falling away; drag counter flatlining
SECO → orbit	Orbit = horizontal velocity — ~7.8 km/s sideways, not kilometres up	orbit-relative velocity crossing circular-orbit speed
Injection burn	The energy to leave — raising apoapsis to escape/transfer; the handoff to cruise	the parking orbit stretching into the transfer arc

This is the payoff of "physics first": the numbers on screen are integrated from the vehicle's real mass, thrust, and Isp, not decorative.

Vehicle coverage

Ascent shape is a property of the rocket, not the payload — a Delta II flies the same profile regardless of what it's lifting. Missions already carry their launcher in fleet_refs (role: "launcher"), so the hook exists.

• Flagship tier — fully modeled. ~6–8 hero vehicles across agencies, so the marquee launches are exact: Falcon 9, Saturn V / SLS, Atlas V, Ariane 5/6, Long March 5, PSLV, Soyuz. Per Orrery's global-programs principle this is deliberately not NASA/SpaceX-only — CNSA, ISRO, Roscosmos, JAXA, and ESA vehicles are first-class.
• Generic fallback — everything else. Every other launcher in the fleet gets a parameterized generic 2-stage model derived from its published stage Δv / mass fractions, honestly captioned as a representative profile. No mission is left without an ascent act.

Why now

/fly just finished its throne-of-glory cinematic sweep (multi-camera montage, arrival compositions, iconic-shot composer). The cinematic machinery a beautiful launch needs — shot-cutting, scene handoff, a validated physics harness, the Science Lens, the time-control scrubber — already exists. This epic is largely about pointing that machinery at the first eight minutes of flight, plus one new coordinate frame and one new data library. The gap between "we have the tools" and "we have the feature" has never been smaller.

Success looks like

• A user watches the vehicle pitch over and reads the caption and says "oh — orbit is sideways, that's why they don't just go straight up."
• A student pauses at staging, reads the mass-ratio jump, and can explain why a single-stage rocket can't reach orbit.
• A user scrubs in one continuous drag from the launchpad to the Mars flyby and feels the whole journey as one arc.
• The integrated numbers land on the real flown milestones (MECO altitude/velocity, staging MET) closely enough that a spaceflight fan nods instead of wincing.

Planning the take-off, not just the transfer (/plan)

The same physics that animates the ascent also plans it. /plan — the mission configurator — today plans only the interplanetary transfer (porkchop, Δv budget). It should also let the user plan the take-off: pick a launch vehicle and see its Δv-to-orbit, payload-to-orbit margin, TWR, and ascent losses fold into the mission's total Δv budget — the moment "can this rocket actually lift this payload to this orbit?" becomes answerable in the planner. This is a headless reuse of the ascent engine (no rendering), delivered as the epic's final slice; the physics core is built shared from the start so it costs a consumer, not a rewrite (RFC-034 L-I / S10).

What this sets up next (the sibling epic)

Launch is the ascent bookend of a mission's life. The natural next epic is the other end and the middle: descent & landing (Moon and Mars EDL — the "seven minutes of terror," powered descent, touchdown, feeding into the existing surface scenes) and orbital rendezvous (Earth/Moon orbit insertions, separations, and dockings/joinings — ISS/Tiangong/Apollo CSM-LM, Artemis Gateway). Together with this epic they'd complete the full arc: pad → orbit → cruise → arrival → descent → surface, every phase a continuous, physics-backed, broadcast-grade scene. This PRD deliberately owns only ascent; the ascent engine, the multi-scale clock, and the scene-handoff pattern it establishes are the reusable spine those future epics inherit. Flagged here so the architecture in RFC-034 is built to extend downward, not just to ship ascent.

Non-goals (v1)

• Not a game. No fly-it-yourself throttle/pitch control in v1 — this is a cinematic learning piece you observe and scrub. An interactive mode is a deferred follow-on epic on the same physics core (see RFC-034 §deferred).
• No CFD / 6-DOF fidelity. Planar integration with real vehicle parameters, not a full aerodynamic simulation. Honest, not exhaustive.
• No landing/recovery act. Booster return-to-launch-site is out of scope for v1 (a natural later slice).
• No per-second-accurate telemetry replay. The profile is physically integrated and data-anchored to published milestones, not a reconstruction of one specific flight's black-box data.

Technical architecture — coordinate frame, the multi-scale clock, the integrated ascent engine, the vehicle-profile schema, and the scene handoff — is specified in RFC-034.