Water in a twisting CubeSat

SLOSHSat was a CubeSat designed by my group for a space engineering course at the University of Sydney in 2024. Its payload was a small clear tank of red-dyed water with cameras looking in, to watch how liquid moves in microgravity, which matters if satellites are ever going to be refuelled in orbit. It turned with magnetorquers alone. We took it through critical design review and built prototypes; it never flew.

I designed the custom boards for sensing, localisation, experiment control and driving the torquers, wrote the firmware on FreeRTOS, and printed the structure in carbon-fibre polycarbonate. Sending video down was out of the question, so the plan was to measure the water on board and send the numbers. I wrote the water tracking: it straightens out the wide-angle image, crops to a green square taped round the tank, picks out the red water and finds its middle. One run printed xbar = 1.27 mm, ybar = 1.10 mm.

Drawn in 2D, in the plane the camera looks at. Earth’s field is taken to lie in that plane and to hold still through a turn, so a rod’s torque is its dipole crossed with the field, and the vertical coil, which could only tip the satellite out of the page, stays off. Assuming 0.2 A·m² per rod, a 30 µT field and 0.002 kg·m² of inertia, a quarter turn would take about a minute in orbit; here it takes about two seconds. The water is 230 particles with cohesion and wall adhesion, tuned by eye rather than to water: at real torques it would creep across the tank more than slosh. The camera’s mask is drawn at 22 × 16 pixels so the pixels show. A camera frame is taken as 640 × 480 colour pixels, 921,600 bytes uncompressed; the packet is the centroid and spread as four 16-bit numbers in hundredths of a millimetre.