The residence where I currently live relies on an older water supply system that demands the filling of a rooftop tank to maintain adequate water pressure within the house. Rather than opting for the costly route of replacing the pump and pressurizing the entire system, I chose to enhance it by integrating sensors.
A HomeAssistant instance is responsible for monitoring water levels and activating a pump by sending commands to a Shelly Pro 1PM device (by the way, Shelly devices are exceptionally well-supported in HomeAssistant).
To keep track of tank water levels and potential overflows, we employ an ESP32-based sensor node. Given the various infrastructure challenges in the area, I've implemented an HTTP endpoint on the Shelly device to facilitate communication with the sensor. Integration with HomeAssistant is done by defining MQTT sensors. This setup ensures that the Shelly and the sensor can continue to operate and manage the water supply even in situations where HomeAssistant (or the Raspberry Pi Zero2 it operates on) may become unavailable. A script running on the Shelly checks if HomeAssistant API is reachable and falls back to a direct mode if it isn't.
FireBeetle ESP32 based and has the following sensors:
- Analog pressure sensor image. 15psi, 0.5-4.5V output for full input range. Usual listing at ebay would be:
1/8NPT Stainless Steel Pressure Transducer Sender 0-4.5V Oil Fuel Air stw B2AE - Fluid flow counter sensor image
- BMP280 for environment information
The water tank is a cylinder lying horizontally, diameter - 500mm, length - 1000mm, volume: ~197 liters
Tank physical size is hardcoded in the tank_volume.c
Formula for calculating tank volume given depth of water is known:
V = L(R2arccos((R-D)/D)-(R-D)sqrt(2RD-D2))
Where:
- R - is the radius of the cylinder
- D - is the depth of water in the cylinder (calculated based on water pressure measurements)
- L - length of the cylinder (would be height if it was standing as in math books)
All of the above is explained at https://www.mathopenref.com/cylindervolpartial.html
Firmware is based on Mongoose OS.
A GPIO pin is using the counting periphery. Precise gate timing is achieved using the RMT periphery, and this requires sacrificing a GPIO pin. Count results and reporting is handled in a separate FreeRTOS task.
Additionally for testing the logic of the counter LEDC periphery is used by redirecting it's output via the device multiplexing control to the same counting GPIO pin on which the external sensor is attached. This functionality is optional and can be turned off by changing this line in mos.yml
cflags:
- "-DFREQUENCY_TEST_MODE=1"
Provided that the pressure sensor can output from 0.5V to 4.5V for 0 to 5 psi [0 to 0.34 atm] and the maximum water column can be 0.5m e.g. 0.05atm of static pressure (1 atm for every 10m of water) output of the sensor can not overshoot the maximum 3.3 V value for the ADC input. When overflow occurs dynamic pressure will raise above the maximum but empirically established that voltage will not overshoot.
Not much is provided for hardware filtering thus oversampling and exponential moving averaging is applied to stabilize reported readings.
- MQTT (configurable) topic device-id/status
- Webhook (configurable) - http client on device will hit a web server url (configurable)
- WebSocket
- http://device-address/status - status for device
- http://device-address/raw - raw values from adc pressure readings and counter
Polling data can be done using:
- HTTP endpoint http://device-address/status (configurable)
- Calling
Device.StatusRPC method
JSON containing tank status data readings
{
"timestamp": 1698212727,
"air_temperature": 0.0,
"air_pressure": 0.0,
"air_humidity": 0.0,
"tank_liters": 0.0,
"tank_percentage": 0.0,
"tank_status": "low",
"tank_overflow": false
}
JSON containing raw tank status data readings from pressure and flow sensors
{
"timestamp": 1698183816,
"tank_pressure_adc": 0,
"tank_overflow_count": 0,
"tank_overflow_frequency": 0.0
}
Any significant change in water volume or overflow status will cause a notification to be published on MQTT topic, over WebSocket and POST data to be sent if a HTTP WebHook URL is configured.
Setting device config can be done over http using the mos tool:
mos call config.set '{"config":{"mqtt":{"server":"garagepi4:1883"}}}' --port http://tanksensor2/rpc
mos call config.save --port http://tanksensor2/rpc
Device configuration "app" is available when accessing the built in web server.
Single page app will display current device status as well will allow to configure threshold values.
- make build
- for building the project
- make localflash
- for flashing over a wire
- make flash
- for OTA over http
To change device name pass a DEVICE_ID option to make:
DEVICE_ID=my_trinket make build
According to the docs debug levels are respectively:
- 0 ERROR
- 1 WARNING
- 2 INFO
- 3 DEBUG
- 4 VERBOSE_DEBUG
There are two make targets - debug_info and debug_debug that will change the output in the logs
You can use UDP debug option by setting the option debug.udp_log_addr to an ip:port of a receiver, where you can capture
the debug log using nc
File debug.json
{
"config" : {
"debug": {
"udp_log_addr":"192.168.19.17:9966"
}
}
}
Set over http
curl http://tanksensor2/rpc/config.set -d @debug.json
Capture debug output over IP network
nc -ul 9966