12The Precision Desk

Who this is for: anyone who wants to go past "design a clean orbit" and work with the actual objects in space today — real two-line element sets from the catalog, orbit determination, conjunction screening, and the standard file formats the professionals use. Plain language first, real terms alongside.

Everything in this chapter lives on the Flight Deck (/deck), in the rail tabs on the left. It is the part of Delta V that earns the comparison to STK and GMAT — and, in keeping with the Honesty Ledger, it tells you exactly how far to trust each number.

1. Track a satellite — fly a real object from today's catalog#

Open Track a satellite. Pick a preset (ISS, Hubble, a GPS bird…) or paste a TLE (a two-line element set — the little text block that describes where a satellite is). The engine flies it with the real NORAD propagator:

• SGP4 for near-Earth satellites (the ISS, Starlink, weather sats).
• SDP4 for deep-space ones — orbits that take 225 minutes or longer to go around, like GPS, Galileo, GLONASS (the navigation satellites, ~12 hours) and anything in geostationary orbit (~24 hours).

Plain-language version: SGP4/SDP4 is the math the US Space Force actually publishes the catalog for. It's not our approximation — we ported the reference algorithm and checked it against the reference implementation (python-sgp4) to sub-micrometre agreement across LEO, MEO and GEO. When you fly a GPS satellite here, it's where GPS actually is.

You get the ground track on a world map, the live facts (altitude, speed, inclination, period), and two honesty readouts:

• Quick-look drift — how far a simple two-body model would wander from the real propagator after a day (hundreds of km for the ISS — that's why SGP4 exists). For deep-space objects this reads "n/a" because the simple model doesn't apply there.
• TEME → J2000 frame — SGP4/SDP4 works in a frame called TEME; the rest of the engine uses J2000. The Earth's pole slowly wobbles, so the two frames drift apart (tens of km over decades). We show you the size of that gap, and the CCSDS export can convert between them.

There's also a Ground-station passes (SGP4) card: pick a station, a window, and an elevation mask, and it tells you exactly when that station can see this real satellite — rise time, duration, peak elevation, slant range — flown with the real propagator.

2. Coverage — what does a whole constellation see?#

The Coverage tab paints a global heatmap of who-can-see-whom and how often ("revisit gap"). Three modes:

1. This satellite — the duty cycle of one orbit.
2. Walker — a synthetic constellation you design (the inc : total / planes / phasing notation real systems publish).
3. Real (SGP4/SDP4) — load an actual constellation from the live catalog (Iridium, Globalstar, Starlink, OneWeb, and the GPS/Galileo/GLONASS nav systems) and fly every member with the real propagator.

Try Real → GPS: 32 real satellites, flown with SDP4, give you the classic 100% global, zero-gap navigation coverage — on a flat map and painted onto the 3-D globe. Iridium (near-polar) covers the poles; Globalstar (mid-latitude) shows a real polar gap. This is live data, honestly computed.

3. Determine orbit — orbit determination from observations#

Real operations rarely start with a perfect state vector; they start with noisy measurements and solve for the orbit. The Determine orbit tab does exactly that — batch least-squares orbit determination (the workhorse of real OD), with a live residual plot showing the fit collapse from a bad guess down to the measurement-noise floor.

Three sources:

• Simulate — pick a truth orbit + a sensor-noise level; watch the solver recover it (e.g. LEO 13 km guess → 85 m fit in 3 iterations).
• SGP4 truth — fit a two-body model and a J2 model to a real SGP4 ephemeris and overlay both residual curves. The gap between them is exactly what J2 (Earth's oblateness) explains; the remainder is drag. This is the fidelity ladder, drawn on real data.
• Upload OEM — import a CCSDS ephemeris file (yours, or one exported from the Track desk, or from STK/GMAT) and fit it.

It also reports the solution covariance — how well the orbit is known, and how that uncertainty grows over time (the in-track "cigar" stretches).

4. Conjunctions — close-approach screening + collision probability#

The Conjunctions tab is conjunction screening: will two objects come close, when (the TCA, time of closest approach), how close (miss distance), and how risky (the collision probability, Pc)? Three modes:

• Pairwise — screen two objects; see the separation-vs-time dip at the TCA.
• Catalog — one primary vs a generated catalog → a risk-ranked worklist, with an apogee/perigee pre-filter (the standard first sieve).
• Live (SGP4) — screen a real constellation (flown SGP4/SDP4) against one of its own members — real intra-constellation screening on live catalog data.

The risk model turns a miss distance into a Pc using a Foster B-plane integral over the hard-body disk, with a covariance from a tracking-quality model or from a real orbit determination on the two objects. A key, honest lesson it teaches: how risky a fixed miss is depends mostly on how long you've tracked the objects (which sets the velocity knowledge, which sets how the uncertainty grows) — not just the sensor.

Honesty (Constitution Art. XI): this is screening, a quick-look filter — not operational collision avoidance. The UI says so. Use it to learn and to triage, not to fly a real maneuver decision.

You can export a CDM (Conjunction Data Message) for any approach.

5. CCSDS — talking to the rest of the world#

Throughout the Precision Desk you'll see CCSDS export/import:

• OEM (Orbit Ephemeris Message) — a whole trajectory.
• OPM (Orbit Parameter Message) — a single state.
• CDM (Conjunction Data Message) — a close-approach record.

These are the international standard interchange formats. Export an SGP4 ephemeris as an OEM, and it opens in STK, GMAT, or Orekit. Import an OEM into the Determine desk and fit it. Delta V doesn't try to replace those tools — it speaks their language, so it fits into a real workflow.

6. How much to trust all this#

• SGP4/SDP4: validated to sub-micrometre vs the reference implementation.
• Orbit determination: batch least-squares, the real method; residuals collapse to the noise floor.
• Frame conversion (TEME→J2000): validated to ~1 metre vs Vallado's worked example; the nutation model matches the IERS reference (ERFA) to a fraction of a micro-arcsecond.
• Conjunction Pc: the standard 2-D short-encounter probability — labeled as screening, never as operational CA.

This is genuinely professional-grade astrodynamics, running in a browser tab, with every claim carrying its evidence. The one thing it is not is a certified operational tool — and it will always tell you that.

Next: the Flight Deck reference for every other tab, or Fidelity & honesty for the full story on how the accuracy ladder works.