Coffee is one of the most popular beverages for people. The separation of ripe coffee beans from immature ones is one of the important features in coffee processing. When this process is done manually, it will usually consume a lot of time and effort. The purpose of this research is to develop a prototype microcontroller-based device to detect and separate ripe and unripe coffee automatically. The automatic separation system is designed to classify coffee according to its color using TCS3200 sensor and ESP32-based system. The components of this system include a conveyor to move the coffee to the sensor, a DC motor that moves two pieces of coffee used as movers, and a servo as a tool to move the coffee into the appropriate container. This tool is also equipped with a loadcell to calculate the weight of the coffee fruit that has been separated and monitor the weight of the tool with a notification when the coffee fruit separated by the container has reached its capacity limit. Based on the results of research and testing, the detection accuracy of the tool is 77,14%, with a light intensity of 83-96 and 74,28% for a light intensity of 121-127 from 35 samples tested consisting of 3 main color categories, namely red, green, and yellow. The determination of the RGB value is very influential on the light intensity, For load cells that receive loads can function properly, the average error value obtained by the load cell for raw is 4.08% and the cooked one reaches 2.68%. and the system in this program runs well, which can monitor and there is a notification in the telegram when the load reaches 250g. Some factors that affect inaccuracies are fruit that is not in the right position to the sensor, unevenly ripe fruit and peeled fruit skin.
Keywords: coffee, coffee bean separation, microcontroller, color sensor,, ESP32, load cell, Telegram.