▶ Spin it in 3D · Pitch decks · The numbers · Build it
The moment that kills every single-antenna link — played by the live demo: head-down → NO SIGNAL → the dipole takes over. ▶ Drag the jumper yourself.
Today, the ground sees only a dot in the sky. Spectators, the waiting area, your own team — they follow the jump with the naked eye, and the footage only arrives after landing. The moment itself stays invisible.
SkyDive·Live puts the jump on the screen as it happens. A helmet-mounted transmitter the size of an action cam sends a digital HDZero picture from ~4 km up to a receiver at the landing zone — straight onto the big TV in the waiting area. Its own radio link, no internet, ~14 ms latency. Not a recording. The present tense.
Camera (MIPI) → 1 W VTX → U.FL → antenna(s) → ~4 km of air → ground antenna array → diversity RX → HDMI → monitor / public-viewing TV
Eleven housing generations and hundreds of scripted CAD checks, distilled into two purpose-built designs — a deliberately simple foundation, and a leap that attacks the one moment that breaks every single-antenna link.
The complete printed system, and the proof the concept holds together — GoPro form factor, tool-free battery swap, real off-the-shelf RF parts.
A body is a shadow. Belly-down (face down), the antenna points cleanly at the ground. Go head-down (falling head-first) and the jumper's own body slides between transmitter and ground — a single antenna tears off, right at the most spectacular moment. So Gen 2 carries two: a patch flush in the side end-cap and a dipole up top, and an RF switch that picks the better one in real time. Both sit flush in the shell — screwed in, no stuck-on bump, no snag risk.
🛰️ Feel it yourself — the interactive dual-antenna demo (the GIF at the top is this demo, playing itself): rotate the jumper head-down, watch the single antenna drop to "NO SIGNAL", then switch on the second and watch the link hold.
Every block has its place — justified thermally and by RF. Colour = component identity, the same key used throughout the pitch deck:
| part | what it does | real off-the-shelf part | |
|---|---|---|---|
| 🟠 Camera | the eye | HD wide-angle skydive POV; lens flush through the front wall — nothing protrudes to snag | HDZero Micro V3 |
| 🟦 VTX | the radio heart | turns the picture into a 1 W signal, ~14 ms, sized for 4 km (margin is heading-dependent — see the numbers) | HDZero Freestyle V2 |
| 🟩 Antenna | the link | patch flush in the side end-cap + Gen 2's up-facing dipole — the receiver always rides the stronger one | TBS 5G8 RHCP patch |
| 🟦 Battery | the energy | 3S LiPo in a protected, tool-free swap tray; externally charged | 3S LiPo + BMS |
The numbers that matter → ENGINEERING.md
This is not a hobby gamble. Every critical path is calculated, and honestly split into calculated vs to-be-measured.
| value | ||
|---|---|---|
| 📡 Transmit power | +30 dBm (1 W) — PMSE short-term assignment for the event, 25 mW SRD for all tests | planned path |
| 📡 Free-space loss @ 4 km | 119.8 dB (5.8 GHz, Friis) | ✅ derived |
| 📡 Link margin @ 4 km | ≈ +8 dB belly (favourable heading) · ≈ +2 dB worst heading via the dipole · head-down rides the threshold | calculated, v5 geometry |
| 📡 The honest twist | the side-mounted patch makes belly margin heading-dependent — re-run 2026-06, assumptions & sensitivity in ENGINEERING.md | self-corrected |
| 🌡 Thermals in freefall | ΔT ≈ 6 K — ram-air convection carries the VTX heat | calculated |
| 🔋 Runtime | ~40 min theoretical / ~32 min practical | dimensioned |
| ⚖️ Sender mass | ~200–250 g | dimensioned |
The hard case (head-down — falling head-first) is identified and attacked at the source: dual antenna at the sender plus diversity at the ground — the architecture of the only proven precedent (Corliss/Vislink 2016). The recalculated margins are thinner than the first pass and say fluctuating link, not dead link; the turntable pattern measurement of the assembled sender decides it. No overclaiming — when our own numbers got worse, we published the worse numbers.
Build it yourself → BUILD.md
Two builds, one idea — pick yours: Gen 1 · MK2 (simplest & most robust — 3 PETG parts, the natural first build) or Gen 2 · v5 (dual-antenna, flush — 7 ASA parts, holds the link in any orientation). Each gets its own chronological step-by-step plan — print → prep → assemble (in order) → wire & power → fit tests — in BUILD.md.
(This is the 7-part Gen 2 build. Prefer the simpler Gen 1 · MK2 — 3 PETG parts? Its own plan is in BUILD.md.)
Seven printed parts (ASA), each watertight and collision-checked. STEP files for SolidWorks (both builds) + STL + 3MF live in cad/; printable STLs are also on the v1.0 release.
| part | role | print note |
|---|---|---|
| Body | main shell · integral heat-wall · side battery door · flat GoPro mount (M5×0.8) | open-top-up · tree-support under the 2 GoPro fingers |
| Cover | top lid · GORE pressure-vent · 4× M3 into heat-sets | flat, no supports |
| Electronics sled | carries the VTX + camera, drops in above the heat-wall | no supports |
| Antenna module | screws onto the top (4× M3) — holds the λ/2 dipole + RF switch (Gen 2) | minimal |
| Antenna end-cap (−X side) | flush RF window — holds the patch (the aluminium body is its ground-plane) | minimal |
| Battery tray | 3S LiPo · slide-in, push-detent + lanyard (won't open in freefall) | no supports |
| Battery door | side, tool-free — swap a battery between jumps | minimal (hinge) |
Assembly: ① heat-sets into the body → ② patch into the −X side end-cap → ③ load the tray, slide it in → ④ VTX + camera on the sled, drop in and wire through the heat-wall → ⑤ dipole + RF switch into the module → ⑥ cover on (4× M3) → ⑦ close the door.
Print: ASA (never PLA — it softens too low) · +0.8 % isotropic shrink · 0.2 mm layers · perimeters that fully fill the 3.0 mm wall · enclosure + heated bed. Fasteners: M2/M3 brass heat-sets (no self-tappers) + an M5×0.8 GoPro thumbscrew. Full step-by-step in BUILD.md.
A monitor on a tripod catches the signal over two antennas (omni + directional patch) and always shows the stronger one — true diversity, the same approach professional systems use. Daylight-readable with a sun-hood; an external recorder grabs an instant-playback copy; HDMI runs the same picture onto the big public-viewing TV. Two safeguards against the same dropout — the picture gets through.
Everything is parametric and scriptable:
- 🧊 Interactive 3D — spin the model in your browser (assembled ↔ exploded). GLBs ship with the v1.0 release.
- 🛠 Stack —
build123d(parametric CAD in Python) · custom build/verify pipeline · RF link-budget · thermal (convection / flat-plate) · regulatory (PMSE / SRD / AFuV) · DFM for FDM printing. - 📋
BOM.md— full bill of materials (sender, ground station, measurement gear). - 🎞 Pitch decks — tap to open on any device (phone · tablet · Mac · Windows, nothing to install): open the pitches →. The flagship is the playable v5 "never lose the picture" (EN/DE).
Ambitious engineering project in development — not a finished product, not a closed validation.
- ✅ Done: CAD (watertight, 0 collisions, fastening verified), electrical compatibility on paper, every budget calculated, both generations designed and collision-checked.
- 🔜 Next: first print → VTX thermal measurement → S11 (antenna) measurement → test jump.
- 🎯 Target: first official tests, summer 2026.
Nothing here carries a measured badge yet — that is the point of what happens next: first print, thermal measurement, antenna pattern on a turntable, then the first test jump.
- ⭐ Watch / star this repo — releases will carry the first real measurements and, eventually, the first freefall footage from the system itself.
- 🔧 Building one, or flying camera and have opinions? Open an issue — helmet-setup input from real jumpers shapes the next housing revision.
- 🪂 Drop zone, federation or event? Reach out via the GitHub profile.
A solo-built, prototype-stage project shared in full. Renders are from the project's own CAD; calculated values are marked as such and separated from what still has to be measured — and when a recalculation makes a number worse, the worse number gets published (see the 2026-06 link-budget revision in ENGINEERING.md). Transmit power is regulated: the plan of record is a PMSE short-term frequency assignment for event operation and licence-free 25 mW SRD for all development tests. See DISCLAIMER.md.
License: CC-BY-4.0 · Renders & 3D: own CAD (build123d) · Made by @SchoenTom