LEAFLAB
A corn leaf can look perfectly healthy while the plant is already experiencing water stress. I built a handheld optical sensor that uses two wavelengths of infrared light to investigate what a farmer's eyes might miss, with a price small farming households could realistically afford.





This is my handheld leaf-water sensor, built around a 3D-printed enclosure and a small optical measurement chamber. A leaf goes into the clamp, infrared light travels through its tissue, and the electronics turn the transmitted signals into a water-content estimate.

Through Agrove, roughly 40 units were produced for use by farmers in the cooperative network. Taking the instrument from the workbench into soybean and corn fields introduced a new set of challenges, especially sunlight, leaf positioning, and differences among individual plants.
HIGHLIGHTS
UNDER THE PLASTIC SHELL Here are the optical decisions, circuit components, and abandoned ideas that shaped the device through successive versions.



The sensor uses dual-wavelength near-infrared transmission to estimate leaf water content through differences in optical absorption. Its 1450 nm channel targets a relatively strong absorption band associated with water, while the 1300 nm channel provides a weaker-absorption reference for compensating for some structural and optical variation.
An InGaAs PIN photodiode detects transmitted light, with an OPA657 transimpedance amplifier and ADS1115 16-bit analog-to-digital converter supporting low-noise signal acquisition. An ESP32 microcontroller controls wavelength switching, baseline correction, signal processing, and data communication.
By combining the two measurements through normalized absorbance, the model reduces sensitivity to a shared optical-path factor under the assumptions of the simplified absorption model. Reference water-content measurements are obtained by weighing sampled leaves before and after oven drying at 105°C, providing a basis for calibration and evaluating the relationship between the optical signal and actual water content.
My early design tried to measure leaf thickness using an additional infrared distance sensor. Until I realized that ordinary distance modules resolve distances on the scale of millimeters, while leaves can be only a few tenths of a millimeter thick.
The extra sensor produced unstable readings, and differences in leaf reflectance made the problem worse. I eventually abandoned direct infrared thickness measurement and redesigned the chamber around controlled clamping pressure, dual-wavelength normalization, and empty-chamber baseline correction.
Sunlight posed another obstacle. Near-infrared radiation from the surrounding environment could interfere with the detector, so I developed an enclosed optical chamber and a measurement sequence that accounts for background signals. These revisions also reduced the number of components the farmer would have to maintain.