Skip to main content

How to predict satellite pass data volume

The data a pass delivers is the rate the link can hold, times how long it holds it, discounted by the chance the pass works out. Predict each piece per pass: when the satellite is above the antenna's mask, the SNR at each step, the rate that SNR supports, and whether the pass delivers.

The chain, one step at a time​

orbit + mask link budget + telemetry weather rate chance
──────────── ──► ─────────────────────── ──► ─────── ──► ────────── ──► ───────
AOS, LOS, SNR through the pass rain, gas, capacity success
peak elevation cloud loss │
▼
expected GB = capacity × success
StepQuestionWhere it comes from
PassWhen is the satellite above the antenna's elevation mask, and how high does it rise?SGP4 from the satellite's element set and the antenna's min_elevation_deg. See Passes.
SNRWhat SNR will the link see at each step?The link budget from your RF values, with a model trained on link telemetry. See Link SNR.
WeatherHow much will rain, gas, cloud and scintillation take?Weather attenuation
CapacityHow much data can that SNR carry over the pass?The workflow's capacity step
Expected GBHow much will the pass deliver?capacity multiplied by success

Do it with the pass-yield workflow​

The pass-yield template runs this chain for every pass in a window and sorts the best passes first.

  1. Describe your fleet in a network config, or load the demo fleet.
  2. Read the template with GET /predictions/workflows?templates=true and save its graph under your own slug with PUT /predictions/workflows/{slug}.
  3. Test-run it over the next day. A test run computes everything and delivers nothing:
curl --fail-with-body -sS -X POST -H "x-api-key: $CONSTELLATION_API_KEY" \
-H "content-type: application/json" \
"https://api.constellation.space/predictions" \
-d '{"workflow": "my-pass-yield", "from": "2026-11-10T00:00:00Z", "to": "2026-11-11T00:00:00Z", "deliver": false}'
  1. Read the run with GET /predictions?run_id=<run_id> until its status is done.

In Platform, Passes does the same without any code: Price the next 48 hours runs this workflow as a test run and shows every pass it priced. The API detail is in Workflows.

Reading the answer​

Each pass comes back as an item whose value is the expected GB as p10, p50 and p90. p10 is the low case: nine passes in ten deliver at least that much. p50 is the middle.

The item's explain is its ledger: the value each step gave the pass at its peak elevation, with the unit and the source, such as physics:clear-sky-budget or a model release. A source starting unavailable: means that step could not price the pass. See Items.

What makes the forecast better​

  • Accurate RF values. The link budget needs each satellite's frequency_ghz, bandwidth_mhz and eirp_dbw, and each antenna's g_t_db_k. Nominal datasheet values can be several dB off a real link.
  • Telemetry on every pass. Link SNR learns each antenna from the telemetry you send. On the data evaluated so far, quick adjustments settle within 10 to 30 passes per antenna and the model settles after a few hundred. See Telemetry.
  • Real elevation masks. Set min_elevation_deg on each antenna. Defaults rarely match a real horizon.

Common questions​

How do I estimate the data volume of one pass by hand?

Find AOS, LOS and the elevation profile from the satellite TLE with SGP4, compute the link budget at each step, subtract the expected rain and gas loss, map each SNR to the highest MODCOD that closes, and sum the data rate times the step length. The pass-yield workflow does those steps for every pass and keeps each value in a ledger.

What does p10 mean for a pass forecast?

p10 is the low case: nine passes in ten deliver at least that many gigabytes. Plan commitments on p10 and expectations on p50.

Why does the forecast differ from my datasheet link budget?

Datasheet RF values can be several dB off a real link. Link SNR learns each antenna from the telemetry you send, so its forecast moves toward what the link actually measures.

Next: price the next 48 hours of the demo fleet pass by pass, with no code.