AZB-12 AN-07 · wit.notazizelse.xyz

← AN-07 WIT stratospheric payload

Technical document · AN-07

Overview and decisions

wit.notazizelse.xyz · assets/README.md on GitHub

Reference material for the electronics, PCB, firmware and mission design. Generated 2026-09-22. Nothing here touches the KiCad schematic or board files.

Contents

assets/
├── README.md                       this file
├── bom/
│   ├── BOM_full.csv                every part, 22 columns, 74 lines
│   ├── BOM_pcb_jlcpcb.csv          machine-assembled subset (56 lines, $89/board)
│   ├── BOM_handsolder.csv          you fit these yourself (15 lines, $125)
│   ├── BOM_offboard_from_full.csv  off-board items extracted from the full list
│   ├── BOM_external_system.csv     battery, antennas, structure, ground station, recovery
│   ├── BOM_tools_test_equipment.csv  what you own, what to buy, in priority order
│   └── BOM_notes.md                WHY each part, and why not the alternatives
├── power/
│   └── power_budget.csv            per-load average and peak current, with margins
├── mechanical/
│   └── mechanical_constraints.md   100×100×300 envelope → board size, height and mass budgets
├── pcb/
│   └── stackup_and_layout.md       4-layer stackup, floorplan, grounding, RF routing, DRC, fab outputs
├── learning/
│   └── pcb_learning_path.md        2-layer -> 4-layer: KiCad vs Altium, ~9h of verified videos, practice plan
├── firmware/
│   ├── firmware_architecture.md    scheduler, tasks, memory experiment state machine, fault handling
│   └── telemetry_packet_spec.md    91-byte binary packet, the single source of truth
├── testing/
│   └── test_plan.md                Stage 0 (before power) through flight-day checklist
├── missions/
│   └── mission_definitions.md      7 missions: hypothesis, graph, falsification, limitations
├── procurement/
│   └── suppliers.md                local Tashkent / distributors / AliExpress / JLCPCB, and order sequence
├── datasheets/                     (empty — fill as you download)
├── reference_designs/              (empty)
└── app_notes/                      (empty)

Confirmed decisions

Item Decision
Envelope 3U, 100 × 100 × 300 mm
Board 4-layer, JLC04161H-7628 stackup, ENIG, ~90 × 90 mm
MCU STM32F407VET6, LQFP-100
Radios Both — MicoAir LR900-A (5 V, external) and Ebyte E22-900T22S (SMD, 850–930 MHz), separate UARTs
Battery 4 × Energizer L91 AA — −40 °C rating confirmed from packaging, available locally
Geiger CAJOE module + SBM-20 — confirmed orderable from AliExpress
Camera None on this PCB
Assembly JLCPCB PCBA + a deliberate hand-solder list
EDA KiCad 9 — stay here for this board; learn Altium after it flies
Power Single buck-boost → 5 V, then LC + LDO → 3.3 V (one hot loop, quiet rail for M7)

Costs

USD
PCB components, per board ~168
Bare PCB + assembly, per board at qty 5 ~43
Assembled board, each ~211
5 assembled boards 1055
External flight hardware, ground station, recovery ~600
Test equipment still to buy (priority items) ~245
Spares, consumables, shipping and customs ~400
Project total ~2300

Component list is at v2 — see bom/COMPONENT_CHANGES_v2.md for the eight availability changes.

Open items blocking progress

  1. Does 100 × 100 mm mean the structural box or the finished insulated exterior? This sets the board width at either 90 mm or 70 mm. See mechanical/mechanical_constraints.md §1.
  2. Structure material, wall thickness, tube or box.
  3. Launch date.
  4. Whether a competition imposes requirements.
  5. Frequency council decision (in progress).
  6. Balloon payload mass/aviation clearance — same authority, separate question.

Order these today

Nothing below depends on the PCB being finished.

  1. STM32F407VET6 dev board × 2 (~$24, AliExpress) — the largest schedule lever in the project
  2. Geiger module + SBM-20 (AliExpress) — longest, least predictable lead time
  3. u-blox SAM-M10Q × 2 (Mouser/DigiKey) — longest distributor risk
  4. MicoAir LR900-A × 2 — stock already uncertain
  5. Bench PSU + ST-LINK × 2 + multimeter + logic analyser — needed before the board arrives

See procurement/suppliers.md for the full sequence and supplier links.

The three things most likely to ruin the flight

  1. Not setting the GNSS airborne dynamic model at every boot. Every measurement above 18 km is lost, silently. firmware/firmware_architecture.md.
  2. Using a BMP-series barometer. Specified only to 300 hPa ≈ 9 km. bom/BOM_notes.md.
  3. Transmitting on both radios simultaneously. Desense or front-end damage. Needs a firmware interlock, not a convention.