Avionics system for experimental water-powered rockets. Real-time telemetry · Triple-redundant apogee detection · Wi-Fi ground station · 7-state flight FSM · Onboard LittleFS black-box logging
Atl-1 is a flight computer for water-powered PET bottle rockets. It is the hardware validation platform for ongoing research into replicable low-cost EMI characterization methodology for student avionics — quantifying the effect of conducted and radiated noise sources on I²C sensors using a factorial 2×2 experimental design with A/A validation.
- Real-time telemetry over Wi-Fi WebSocket at 10 Hz
- Triple-redundant apogee detection (velocity zero-crossing + altitude drop + safety timeout)
- Hardware IIR pressure filtering on BMP280 to suppress aerodynamic noise
- Passive gravity parachute recovery on the flight-1 build — servo-based deployment logic is present in firmware, but the servo was omitted to save mass (see Hardware)
- Onboard black-box logging to flash via LittleFS — survives WiFi loss
Status: First instrumented flight completed — June 1, 2026, Veracruz, Mexico.
Date: June 1, 2026 — Veracruz, México
Launch site: Open field — pressure estimated, no gauge available
| Metric | Value |
|---|---|
| Apogee (barometric) | ~14 m |
| Peak velocity | ~4.5 m/s |
| Peak G-Force (liftoff) | ~1.6 G |
| Flight duration | ~30 s |
| Parachute deployment | ❌ Not triggered — see note |
| LittleFS data recovery | ✅ CSV recovered via Serial DUMP post-landing |
Note on parachute:
LIFTOFF_Gis set to 2.5G. Actual peak was 1.6G — rocket was under-pressurized (no gauge available). FSM remained in STANDBY throughout the flight. Fix for Atl-2: proper pressurization with calibrated gauge.
Raw flight data: flights/atl1_flight_20260601.csv
Atl-1 also serves as the flight demonstrator for an ongoing study on EMI behavior in compact low-cost avionics. The experimental design, dataset, and results are in preparation and will be released together with the publication.
| Component | Part | Notes |
|---|---|---|
| Microcontroller | ESP32 DevKit (standard) | Wi-Fi AP, WebSocket server |
| IMU | GY-521 module — chip: MPU-6500 (WHO_AM_I = 0x70) | ±16G / ±2000°/s — see library patch below |
| Barometer | BMP280 (0x76) | Hardware IIR FILTER_X16 enabled |
| Power Boost | MT3608 DC-DC step-up | LiPo 3.7V → 5V |
| Battery Charger | HW-373 (TP4056 + DW01) | USB-C charging with simultaneous load support |
| Battery | LiPo 1S — 3.7V / 1200 mAh | Minimum flight voltage: 3.0V |
| Actuator | (none on flight-1) — SG90 on GPIO 13 is firmware-defined only | Flight-1 flew passive gravity recovery; physical servo omitted to save mass — planned for Atl-2 |
| Switch | SPDT (GPIO 10) | Master power |
| Buzzer | Piezoelectric active (GPIO 3) | State audio feedback |
I²C bus: SDA → GPIO 21, SCL → GPIO 22. Pull-ups: 5.1 kΩ to 3V3.
Power path:
USB-C ──► HW-373 ──► LiPo 3.7V ──► MT3608 ──► 5V ──► ESP32 DevKit (5V pin)
└──► DevKit onboard LDO ──► 3V3 ──► Sensors
As-built decoupling: one electrolytic cap at the MT3608 VOUT and one ceramic cap at the sensor supply. Atl-1 has no LC filter — the LC filter belongs to the bench experiment's noisy config, not this flight build.
WiFi is disabled on LANDING to reduce idle current from ~180mA to ~20mA.
Avionics Bay design history (v1 → v4): docs/docs_avionics_bay_evolution.md
gForce ≥ 2.5G (4 consecutive)
┌──────────┐ ────────────────────────────► ┌──────────┐
│ STANDBY │ │ ASCENT │
└──────────┘ └──────────┘
│
┌────────────────────┼────────────────────┐
│ │ │
vel ≤ 0 alt drops 1m timeout
(5 samples) (4 samples) 12,000 ms
└────────────────────┼────────────────────┘
▼
┌──────────┐
│ APOGEE │ servo → 90°
└──────────┘
│
┌──────────┐
│ DESCENT │
└──────────┘
│ alt < 0.75 m
┌──────────┐
│ LANDING │ WiFi OFF · log closed
└──────────┘
| State | ID | Entry condition | Action |
|---|---|---|---|
STANDBY |
0 | Boot | Transmit, log, wait for liftoff |
ARMED |
1 | (future — remote arm) | — |
IGNITION |
2 | (future) | — |
ASCENT |
3 | gForce ≥ 2.5G × 4 | Track maxAltitude, evaluate apogee |
APOGEE |
4 | Triple-redundant trigger | servo.write(90) |
DESCENT |
5 | Post-apogee | Monitor alt, log |
LANDING |
6 | alt < 0.75 m | Close log · disable WiFi |
Recovery, as-built: the FSM executes
servo.write(90)at apogee, but the flight-1 build has no servo installed — recovery is passive gravity. Servo-actuated deployment is the original design and the planned path for Atl-2.
Each boot creates a new file: /flight_1.csv, /flight_2.csv, ... (auto-increment, never overwrites).
Capacity: ~60 bytes/row × 10 Hz × 60s ≈ 36 KB/flight. ~1.5 MB available → ~40 flights before FORMAT.
| Command | Action |
|---|---|
LIST |
List all files and sizes |
DUMP |
Dump all CSV files to Serial |
DUMP:flight_1.csv |
Dump specific file |
FORMAT |
Erase all log files |
BMP280 configured with hardware IIR FILTER_X16 at ~28 Hz. Boot sequence:
- 100 reads discarded — IIR filter needs ~70 samples to converge from cold start
- Reads 101–200 averaged →
basePressure - 50 samples averaged →
groundAltitudeoffset (zero reference)
rawAlt = bmp.readAltitude(basePressure) - groundAltitude;
filteredAlt = filteredAlt * 0.8f + rawAlt * 0.2f;Calibration must be done outdoors at the launch site, stationary and uncapped. Indoor calibration causes negative altitude readings.
Adafruit MPU6050 — requires WHO_AM_I patch (see below)
Adafruit BMP280
Adafruit Sensor (dependency)
ESP32Servo
WebSockets by Markus Sattler
LittleFS (ESP32 built-in)
The GY-521 module on this build contains an MPU-6500 (WHO_AM_I = 0x70), not an MPU-6050 (0x68). Apply this one-line fix:
File: Arduino/libraries/Adafruit_MPU6050/Adafruit_MPU6050.cpp — line ~93
// BEFORE:
if (chip_id.read() != MPU6050_DEVICE_ID) {
// AFTER:
if (chip_id.read() != MPU6050_DEVICE_ID && chip_id.read() != 0x70) {- Board: ESP32 Dev Module (classic ESP32 DevKit — confirm against your Arduino IDE selection)
- Partition Scheme:
Default 4MB with spiffs (1.2MB App/OTA, 1.5MB SPIFFS)← required for LittleFS - Baud: 115200
- Power on outdoors at launch site, stationary and uncapped
- Wait for
=== READY FOR FLIGHT ===(~2.6s calibration) - Ground offset printed — confirm within ±1.0 m
-
LISTvia Serial → confirm new/flight_N.csvcreated - Ground station connected — altitude reads near 0 m
- Only one browser tab open (multiple clients = duplicate CSV rows)
- Parachute packed for passive gravity deployment (no servo on flight-1 build)
- LiPo at 4.2V (HW-373 blue LED solid)
- Cap and pressurize after READY confirmation
Atl-I_Flight-Computer/
├── firmware/
│ └── HydroRocket.ino
├── flights/
│ ├── atl1_flight_20260601.csv
│ └── graphics.py
├── hardware/
│ ├── Schematic_HidroRocket_v0_2026-04-20.pdf
│ ├── Schematic_HidroRocket_v0_2026-04-26.pdf
│ └── Schematic_HidroRocket_v0_2026-04-26-(actual).pdf
├── mechanical/
│ ├── AvionicsBay_v4.1.0.1.stl
│ └── archive/
├── docs/
│ └── docs_avionics_bay_evolution.md
├── media/Prototype Gallery/
└── README.md
| Issue | Status |
|---|---|
| LittleFS black-box logging | ✅ Implemented |
| BMP280 IIR warm-up before calibration | ✅ Fixed |
| Ground altitude zero offset | ✅ Fixed |
| MPU-6500 WHO_AM_I library rejection | ✅ Patched |
| WiFi power drain post-landing | ✅ WiFi off on LANDING |
| Item | Priority |
|---|---|
| Proper pressurization with gauge (Atl-1 under-pressurized → 1.6G, threshold 2.5G) | ✅ Gauge acquired |
| Verify CG/CP with electronics in OpenRocket (tumble at apogee) | 🔴 High |
| Replace BMP280 → MS5611 (±0.1m vs ±8m) | 🟡 Medium |
| EMI characterization — factorial 2×2 bench experiment (4 cells × 10 reps, A/A validation) | 📄 Paper |
Jesús Alberto Perea García
Mechatronics Engineering — IEST Anáhuac, Tamaulipas
Founder — Vértice Labs Research Program
github.com/DonJechu
Atl-1 is the flight validation platform for experimental research on EMI characterization methodology in compact low-cost avionics. Paper in preparation for IEEE regional conference.