A plant-watering reminder has a simple job: help someone decide whether to water a pot. For a small team building one, the first challenge is getting from a sensor reading to advice a person can use. A number on a screen is a start. It still needs a meaning that holds for the pot, the soil and the probe position in front of you.
You can see that gap in real projects. In an Arduino forum discussion started in August 2023, a builder mapped dry and wet readings to a percentage, then found that dry-soil values could come out negative. The question of what a reading means comes up before a useful reminder can be defined.
On our bench we connected a Mysentech MSE capacitive soil probe to a Raspberry Pi Pico (RP2040) on a GraftPort carrier and pushed the probe into a potted plant. This note follows that setup to one question: what does the prototype tell us already, and what would it still need before telling someone to water a plant?
Decide what the reminder will mean
For an early product, "check this plant" may be a better promise to test than "this plant needs water now". The first leaves the watering decision with the person. The second requires more evidence about the plant, the growing medium and how water reaches the roots.
Existing software makes a threshold interface easy to build. Home Assistant's Plant Monitor integration accepts moisture in percent and flags readings outside configured limits. Its example limits are software settings, though; they do not calibrate a new probe or establish the right watering point for your plant.
Projects also have to decide what happens when a measurement stops being useful. The author of a ZoneFlow irrigation integration update describes ignoring stale dry readings and asking the user to check a probe that keeps reporting wet. Those are that project's features, not something our MSE test demonstrated. For a reminder, they suggest a practical requirement: show when a reading was taken, and keep "check the sensor" separate from "water the plant".
Understand the measurement before choosing a threshold
The MSE reports temperature and capacitance to the controller. Capacitance is an electrical property, reported here in picofarads (pF). Water, soil and air have very different dielectric properties, so the water around a capacitive probe changes its reading, which makes it useful for following changes in soil conditions. That principle does not give this pot a ready-made moisture conversion.
METER's guidance on soil-specific calibration explains that soils differ electrically and recommends calibrating its own sensors for the soil in use. That supports checking your growing medium. It does not supply an MSE calibration equation, and METER's accuracy figures do not transfer to this module.
For a first experiment, keep the raw pF value visible. It lets you compare observations without hiding them behind a percentage whose meaning has not been established.
Connect the hardware you intend to observe
The probe talks Modbus over UART, the serial link that carries commands and responses. A small adapter joins the probe cable to the carrier's UART0 connector; it makes the connectors fit, but the signal labels still decide which wire goes where.
Our example used GPIO16 as the controller's transmit pin and GPIO17 as receive, at 9600 baud with Modbus address 1. The supplier's interface table marks the probe's R pin to connect to the host's TX, and T to the host's RX, where the host is the RP2040. Follow those explicit signal directions rather than inferring them from cable colors, and check the module's and controller's supply requirements before powering your own assembly.
The probe carries printed minimum and maximum insertion-depth marks. For a reminder prototype, record the placement and keep it fixed while observing the pot: moving the probe adds another change to explain. The supplier advises against forcing the probe into hard soil or leaving it immersed in water for long periods.
Install the driver, then read what it returns
We used version 1.0.0 of our MSE driver package, installed onto the board with mpremote mip. The package contains the sensor module, an example main program and a Modbus library it depends on, so copying only the main script is not enough. MicroPython's mpremote documentation covers the host-side tool; use the documentation that matches your firmware release, since our setup ran MicroPython v1.23.0.
The driver handles the Modbus messages, so the application asks for named values, temperature and capacitance, instead of decoding bytes. Our example program also printed configuration and version fields, and it wrote the temperature-compensation B coefficient to 0.5 before reading. Treat that write as part of the experiment's configuration, not as a universal calibration value. If you only want to read, leave it out deliberately and record that you did.
What the pot experiment actually tells us
The console reported 26.5 °C and 31.07 pF in two consecutive blocks. It also showed a capacitance center of 15.0 pF, a range of 15.49 pF, a filter count of 5 and an averaging window of 6. Those fields document the configuration behind the reading; they are not dry and wet reference observations.
The milestone is that the connected probe returns readable data in a real pot. Two identical readings and a set of configuration fields do not give a dry-to-wet series, so 31.07 pF cannot yet tell the user whether to water. It certainly should not be relabeled as 31.07% moisture.
Before choosing a threshold, observe the same pot through watering and drying, keep the probe position fixed, and record the raw reading alongside an independent assessment of the soil. We have not run that series yet; it is the next experiment.
Turn those observations into a reminder people can try
Agree on the decision you want the user to make. If it is "check the pot", record which observed conditions should produce that prompt and how the prompt clears after watering. Keep timestamps so an old value cannot appear to be a fresh observation. If repeated observations do not separate the useful conditions, investigate placement and measurement before polishing the interface.
That keeps the build small enough to learn from. The probe setup can be the measurement starting point; a light, display or phone notification is a later integration step. A pump adds another task, controlling water delivery, which this experiment has not tested.