POSA_LEAKSMS/raspberry_pi4_pms7003/README.md
cw2007 ec78be0bda feat(pms7003): 누수 노드에 미세먼지(PM2.5/PM10) 통합
센서1·3(누수 노드)에 PMS7003 를 온습도와 동일 sensor_id 로 통합하고 metric_type='pms7003'
로 구분해 수집한다. LED 는 스위칭 없이 상시 점등 방식으로 확정.

- raspberry_pi4_pms7003/: PMS7003 노드 풀세트(드라이버·센서·install·systemd·vendor·배선·연결테스트·프레임 단위테스트)
- 서버: evaluate_metric_thresholds() 에 pm25/pm10 선택 인자 추가(온습도 전용 호출 후방호환),
  sensor_data.php 수집 경로에 pms7003 metric 반영, config.php/예시에 PM 임계 추가
- SQL: migration_pms7003.sql · schema_sht30.sql 정합, tools/migration_leak_add_{pms7003,sht30} 마이그레이터
- tests/pm_threshold_test.php + run_tests.sh 러너 등록
- .gitignore: 노드별 config_local.py 커밋 차단
2026-07-19 10:46:33 +09:00

165 lines
7.5 KiB
Markdown

# Particulate-Matter (PMS7003) Sensor Module — Raspberry Pi 4 Model B only
A module that reads a **Plantower PMS7003** laser particulate-matter sensor over the Raspberry Pi 4
Model B **UART** and periodically reports **PM1.0 / PM2.5 / PM10** (µg/m³, atmospheric concentration)
to the cafe24 server (`php/api/sensor_data.php`). The server stores the measurements and performs
**high-PM2.5 / high-PM10 threshold evaluation, SMS, and dashboard display**.
> This is a **Pi-only** metric — there is **no STM32 firmware** for the PMS7003. The module is a
> faithful mirror of the SHT30 module (`../raspberry_pi4_sht30/`): same measurement loop,
> `startup→periodic` events, raw-body HMAC signing, and fail-closed secret handling. PM is a **new
> metric** on the **existing metric pipeline** — distinguished only by `metric_type="pms7003"`.
>
> The PMS7003 **co-locates on an existing leak node** and reports under the **same `sensor_id`** as
> that node's leak sensor (Pi-C = 3, the other node = 1), exactly like the SHT30 co-location.
---
## 1. Files
| File | Role |
|---|---|
| `pms7003_sensor.py` | Main — periodic measurement → body build/sign → POST (with retry), startup→periodic |
| `pms7003_driver.py` | PMS7003 UART 32-byte frame read + checksum verification + PM decode (pyserial, simulation fallback) |
| `wire.py` | Body JSON + raw-body HMAC signature (`fixed2`/`_esc`/`sign_raw` copied verbatim from the SHT30 module) |
| `config.py` | Non-secret default settings (sensor id/interval/range/UART). Loads secrets from `config_local` |
| `config_local.example.py` | Operational/secret-value template → copy to `config_local.py` for use (not committed) |
| `requirements.txt` | Python dependencies (`requests`, `pyserial`, pinned) |
| `install.sh` | Pi 4 install (air-gapped offline + online fallback, enables UART + disables the serial console) |
| `vendor.sh` | Pre-download dependencies for air-gapped use (run on the build machine) |
| `tests/test_pms_wire_and_frame.py` | Pure unit tests: body byte-string, known HMAC, frame decode + checksum rejection (no hardware) |
| `systemd/pms7003-sensor.service` | Service unit for auto-start on boot |
---
## 2. Hardware (Pi 4 Model B baseline)
- **Raspberry Pi 4 Model B** (BCM2711) + Raspberry Pi OS **Bookworm 64-bit**
- **PMS7003** laser PM sensor + its breakout/adapter cable
- Power: **USB-C 5V/3A**, network: Pi 4 **Gigabit Ethernet** (wired) or WiFi
### 2.1 Pinout (primary UART)
| PMS7003 signal | Pi 4 pin | Note |
|---|---|---|
| VCC | **5V** (physical 2) | PMS7003 needs **5V** power (fan + laser) |
| GND | GND (physical 6) | |
| TXD (sensor → Pi) | GPIO15 / **RXD** (physical 10) | 3.3V logic — safe direct to the Pi |
| RXD (Pi → sensor) | GPIO14 / TXD (physical 8) | optional (only needed for passive-mode commands) |
| SET / RESET | — | optional; leave floating (internal pull-up = enabled/run) |
> The PMS7003 **TXD idles at 3.3V logic**, so it connects **directly** to the Pi RXD — no level
> shifter needed. We only read the sensor's TXD (active streaming mode), so wiring the Pi TXD→sensor
> RXD is optional.
---
## 3. Server compatibility (must verify)
This module reports to **the current repo's `php/` server**. The server requires a **raw-body HMAC signature**.
```
X-Signature = lowercase_hex( HMAC-SHA256( key=API_KEY, msg=request body bytes ) )
```
- `API_KEY` in `config_local.py` must be **exactly identical** to `API_KEY` in the server's `config.local.php`.
- The server blocks replays via `timestamp` freshness (±600 seconds), so the Pi clock must be **NTP-synchronized** (enabled by default on Raspberry Pi OS).
- It reports with the co-located leak node's `sensor_id` (**3** by default, or **1** on the other node) and `metric_type="pms7003"`.
- Threshold alerts (high PM2.5 / PM10), SMS, and recovery evaluation are **all performed by the server**. This module only reports measurements.
- The Pi-side `metric_status` is **physical plausibility only** (`normal` / `out_of_range`); the server never trusts it for alerting.
### 3.1 Wire body (fixed order)
```json
{"device_id":"..","device_location":"..","sensor_id":3,"sensor_name":"..","event_type":"..","timestamp":0,"metric_type":"pms7003","pm1_0":8,"pm2_5":12,"pm10":18,"metric_status":"normal","app_version":".."}
```
`pm1_0` / `pm2_5` / `pm10` are **bare integers** (µg/m³, atmospheric concentration, range 0..1000).
The body is built directly and raw-POSTed so the **signed bytes == sent bytes**.
---
## 4. Installation
### 4.1 Prepare configuration values
```bash
cp config_local.example.py config_local.py
# Open config_local.py and fill in API_URL / API_KEY / DEVICE_ID / DEVICE_LOCATION
# On the id=1 node also set SENSOR_ID = 1 (Pi-C keeps the default 3).
```
If `config_local.py` is missing, execution **stops immediately with a clear error** (fail closed).
**Never commit** `config_local.py` or `API_KEY` (handled by `.gitignore`).
### 4.2 Online install (internet-capable environment)
```bash
chmod +x install.sh
./install.sh
sudo reboot # reboot once — the installer enables UART and disables the serial console
```
### 4.3 Air-gapped offline install
On an internet-connected machine of the **same architecture (aarch64, ideally the same Pi 4 + Bookworm 64-bit)**,
download the dependencies first to build a bundle, then move it to the target Pi 4 and install.
```bash
# (build machine, internet YES) generate wheelhouse/ + debs/
./vendor.sh
# after copying wheelhouse/ and debs/ into this folder on the target Pi 4:
# (target Pi 4, internet NO)
./install.sh # if debs/ · wheelhouse/ exist, it installs offline automatically
```
> `pyserial` is pure Python (arch-independent), so it can be downloaded on an x86 build machine.
> Other wheels/debs such as `requests` must be the **aarch64** builds (running `vendor.sh` on the target arch is recommended).
---
## 5. Run / check
```bash
# UART present? (should symlink to ttyS0 or ttyAMA0)
ls -l /dev/serial0
# pure wire/frame self-check (passes without hardware or config_local.py)
python3 tests/test_pms_wire_and_frame.py
# actual run (simulation mode if pyserial missing / UART unavailable)
source venv/bin/activate
python3 pms7003_sensor.py
```
In environments without `pyserial` installed or with the UART unavailable (dev PC, etc.), it automatically runs in **simulation mode**.
---
## 6. Auto-start on boot (systemd)
```bash
# assumes the folder is placed at /home/pi/raspberry_pi4_pms7003 (otherwise edit the path in the .service)
sudo cp systemd/pms7003-sensor.service /etc/systemd/system/
sudo systemctl daemon-reload
sudo systemctl enable pms7003-sensor
sudo systemctl start pms7003-sensor
sudo systemctl status pms7003-sensor
journalctl -u pms7003-sensor -f
```
`SupplementaryGroups=dialout` is required for `/dev/serial0` access.
---
## 7. Security notes
- Secrets (`API_KEY`), server URL, and device info live only in `config_local.py` (not committed).
Committed files contain only placeholders.
- The report body is protected against tampering by an `HMAC-SHA256(API_KEY, raw_body)` signature (`X-Signature` header)
(immune to length-extension attacks). The legacy body-field sha256 fallback is disabled by default on the server.
- The signed bytes and the transmitted bytes are **always identical** (`wire.py` builds the body directly,
signs those bytes, then raw POSTs). `fixed2` / `_esc` / `sign_raw` are copied verbatim from the SHT30 module
so signing stays byte-for-byte identical to the shared reference / server `verify_signature_raw`.
- If `config_local.py` is exposed, immediately rotate `API_KEY` (change it on the server at the same time).