The major breakthrough came when I realized that I could start with a high end power supply outputting 14 volts and terminate each parallel line with a device known as a UBEC. If you want to power 70 Pis in a constrained space, you need to devise a customized power distribution system with adequate heat venting.ĭue to the thin effect, voltage drops over distance, so the distance a Pi was from the power would impact the voltage and therefore the speed. However, you need to take pains to make sure that each Pi is getting the same amount of voltage or else they will run at different speeds. For example, since Pis do not have hardware clocks, you have to rely on NTP. By the end, I was a legitimate domain expert in several niches within niches of Pi dark arts. I started a company that constructed booths which used 70 accurately sync'd Pis with custom PiCams (fitted with lenses) to take pictures of human subjects, for turning them into avatars like this:Īt the beginning of the project, I'd barely powered up the Pis I had collecting dust in my drawer. In theory this could all live in a single Pi, but for various legacy reasons I ended up with a dedicated Pi for data processing, and a dedicated Pi for data gathering. The long-term database is running a custom Rails app I wrote to report on the data and show graphs, compare weather over years, etc.
Then, I have another Pi that, every minute, reads those values and stores them in a long-term database (I have about 5 years of weather data at this point). Why 2 temperature sensors? Turns out the BME280 runs about 2☏ hot, and the 18B20 is right on, so I end up using the 18B20 as my primary temp sensor.Įvery minute a cronjob runs and the Pi takes readings from the sensors and stores them locally. The Pi is indoors, and the wires run through the wall. The sensors are outdoors in a small actively-ventilated enclosure I made. It's also connected to a Dallas 18B20 temperature sensor. I have a Pi connected to a BME280 breakout board that I got from Adafruit ( ). Primary motivation to do the project was so that wife can have pice of mind knowing plants are taken care of while we are away on longer trips. (though I just bought my when visiting Fry's) Non-contact Liquid Level Sensor -XKC-Y25-NPN. Key parts I ended up using for the project:Ģ. saves all metrics to sqllite db for future reference summarizes watering event and sends me notification with link to the recording and other helpful stats (temp, humidity, moisture level etc.) uploads recorded video automatically to youtube channel (if pi has internet connection) activates camera to record watering (+2s buffer to have all on the video) via usb cam connected to the pi (leveraging opencv lib)
Watering duration is determined by current soil moisture level Python script activates relay to run the pump for the right duration.
Soil moisture level is below certain threshold (measured by 2 soil moisture capacity sensors embedded in the plant)ģ. There is water in tank / source (checked by non-contact liquid level sensor)Ģ. Triggers for watering plant are defined as follows:ġ. Water pump is connected to a clear vinyl tube for water distribution to the plant.Īvailability of water level at source is checked by a non-contact liquid level sensor. I'm using a cheap submersible water pump controlled via relay by a python script running on raspberry pi.
0 kommentar(er)
