Globe view: smoothed trails spiral instead of orbit on longer time windows

[peterlewis]

Symptom

In the Globe view with Smooth Trails on, satellite ground tracks appear to spiral rather than trace repeating orbital arcs. The effect becomes obvious once the time-window filter is long enough to include multiple passes — reportedly visible at ~2h windows and worse at 24h / 7d.

With smoothing off, the same data renders as the expected orbit-like arcs (with visible 1° NMEA quantization, but no spiral).

Suspected area

Smoothing is applied per-pass in `SatPass.groundTrackPoints` via `smoothedAzEl`, called from `SkyGlobeView.swift:350`:

```swift
let samples = pass.groundTrackPoints(observerLat: lat, observerLon: lon,
maxPoints: .max,
smoothingWindow: smoothTrails ? 0 : 1)
```

`smoothedAzEl` is a two-pass centered moving average on (az, el), with az unwrapped across the 0°/360° seam. Output is re-wrapped via `truncatingRemainder(dividingBy: 360)`.

Candidate causes to investigate:

1. Edge bias from the centered window — near the start/end of an observation array the window is truncated (`max(0, i-half)…min(n-1, i+half)`), which pulls endpoints toward the interior. For partial passes this could systematically bend the start/end of each ground track.
2. Az unwrap / re-wrap interaction with the projection — if the unwrapped azimuth accumulates across a 360° cycle (e.g. for a near-zenith pass) and the average lands in a non-canonical range, the re-wrap + projection may land the sub-sat point on a shifted longitude.
3. Adaptive window size scaling with pass length — `adaptiveSmoothingWindow` widens to 11 taps on long passes. A wider window amplifies any endpoint bias, which could explain why the spiral grows with longer time-window filters (older = longer-recorded passes).
4. Decimation after smoothing — unlikely given `maxPoints: .max` is used on the globe, but worth ruling out.

Repro

1. Run PCC with GPS fix acquired, recording enabled.
2. Leave it running for 2h+ to accumulate multiple passes.
3. Globe view → time window ≥ 2h → Smooth Trails on.
4. Observe ground tracks spiralling rather than repeating.
5. Toggle Smooth Trails off → spiral disappears (trails become jagged but orbit-shaped).

Fix direction (TBD)

Probably one of:

• Use a symmetric reflection / repeat-edge strategy at the endpoints of `smoothedAzEl` so truncated windows don't bias the output.
• Or: don't smooth the first/last few samples at all and let the raw values bracket each pass.

Related

• The Map view shares the same `groundTrackPoints` code path and may show the same artefact — worth confirming.
• The polar view doesn't use smoothing at all (`SkyView.swift:524`) so it's unaffected, but its toolbar toggle is still wired to the same global setting — minor UX inconsistency to address separately.