A useful control system starts by defining its boundary. In this project, the goal was not to build a complete autonomous farm. It was to create a prototype that could regulate one measurable environmental variable—humidity—inside an aeroponic chamber.
That decision shaped the architecture. Nutrient chemistry, pH, EC, crop selection, light, and temperature still mattered to the plant system, but they were treated as operating context rather than as controlled outputs of the humidity loop.
1. Define the controlled variable first
The controlled variable was relative humidity in the aeroponic environment. A reference humidity could be selected, the sensor measured actual humidity, and the controller acted on the difference.
This sounds obvious, but it prevents a common design mistake: choosing hardware first and only later deciding what problem the system is supposed to solve. Once humidity is the objective, the sensor, actuator, controller, and logging requirements can be evaluated against that objective.
2. Build the block diagram around cause and effect
The block diagram forces each element to have a clear role. The ultrasonic mist maker is a process input because it creates the aerosol. The fan is the manipulated actuator because its speed is changed by the controller. The chamber is the plant being controlled. The DHT11 measures the response.
3. Separate mist generation from mist delivery
The fan moves mist produced by the ultrasonic mist maker. That is an important architectural choice.
If the controller directly cycled the mist maker, the process input would be atomizer duty. Instead, this prototype used airflow as the control handle. That can make the moisture source more continuous while allowing the controller to vary transport into the chamber.
However, airflow also changes mixing and evaporation. Increasing fan speed can transport more droplets, but it can also change local evaporation, chamber leakage, and sensor exposure. The process gain is therefore not simply “more fan = proportionally more RH.”
4. The actuator chain includes more than the motor
The physical actuator path was:
Controller output → PWM command → motor driver → DC fan → airflow → mist transport → chamber humidityEvery stage can introduce a nonlinearity or limit. A fan may not rotate below a minimum PWM duty. Motor speed does not equal airflow exactly. Duct resistance changes operating point. Root growth or chamber geometry can alter flow over time.
Thinking in this chain is more useful than treating the fan as an ideal actuator.
5. Sensor placement is part of the design
The humidity sensor should represent the environment being controlled. Mounting it directly beside the mist inlet may make it respond rapidly but may over-read the local plume. Mounting it in a stagnant corner may under-represent the root-zone condition and add long delay.
A practical prototype should test several positions and compare their response during the same mist/fan cycle. The best position is not necessarily the one with the fastest change; it is the one that most consistently represents the control objective.
6. Keep nutrient chemistry outside the wrong loop
Plant nutrition involves nutrient composition, acidity, and concentration, but those are not directly the humidity variable being controlled.
That means pH and EC should not be silently mixed into the humidity controller. They can be monitored independently and could later have their own dosing or reservoir-control logic.
| Variable | Role in this prototype | Possible future control |
|---|---|---|
| Humidity | Primary controlled variable | PI loop through fan airflow |
| Fan speed / command | Manipulated actuator | PWM output, optional RPM feedback |
| Mist generation | Moisture source | Could later be cycled or modulated separately |
| pH | Operating context | Acid/base dosing loop |
| EC | Operating context | Nutrient / dilution management |
| Temperature | Disturbance and RH influence | Heating/cooling/ventilation loop |
7. Design the electrical interfaces safely
The controller should not drive the fan or ultrasonic transducer directly from a logic pin. The fan needs a suitable transistor or motor-driver stage with a freewheel path for inductive current. The mist-maker module typically has its own higher-frequency driver matched to the piezoelectric transducer.
Power wiring and sensor wiring should be separated carefully. PWM motor switching and high-frequency atomizer drive can inject electrical noise into a low-cost digital sensor or controller supply. Local decoupling, sensible grounding, short current loops, and physical separation reduce the chance that electrical interference becomes a false humidity signal.
8. Think about faults before tuning gains
Disable the mist maker before dry running damages the transducer.
Do not allow an invalid humidity reading to command maximum fan or continuous mist indefinitely.
The controller may demand more output while actual airflow remains zero. RPM or current monitoring can reveal the fault.
High local RH or direct droplet deposition can wet the sensor and create misleading feedback.
A safe fallback state should be defined separately from the normal PI control law. Control optimization and equipment protection are different layers.
9. A practical commissioning sequence
- Run the humidity sensor alone and log the chamber without mist.
- Run the fan at fixed duty-cycle steps and observe airflow and sensor response.
- Run the mist maker at a safe constant water level without closed-loop control.
- Combine mist and fan in open loop and map fan command to humidity response.
- Identify delays, saturation, dead zones, and the approximate process time constant.
- Only then tune the PI controller.
- Test several setpoints and disturbances, not one successful operating point.
Define one variable, identify the actuator that truly influences it, measure the response, keep unrelated variables outside that loop, and only add complexity after the first loop is understood.