MIT's coffee can radar is of a Frequency Modulated Continuous Wave (FMCW) design. In this design, frequency is generated that is swept over a range. The signal is split, with part sent to the transmitter antenna. A receiver antenna picks up reflections, and its signal is mixed with the other part of the original generated signal. This heterodyning generates a much lower frequency which can be analyzed to extract both relative speed and ranging data. This stands in contrast to Doppler radar designs which can only measure relative speed. The MIT radar operates around 2.4GHz, as parts for this range are available and cheap due to WiFi, Bluetooth, and other common protocols operating here.
Although conceptually simple, the hard part for designing this radar on a low-cost 2-sided PCB is the layout. I am never done much RF design, and some of it can seem like the dark arts. But I have done some research and have some idea of the challenges. First, the FR4 used in typical PCBs is quite lossy at microwave frequencies. Worse, its relative permissivity is highly variable across samples, so tuning the layout for proper impedance matching can be difficult. I am planning to design for proper impedance matching and since high performance is less a concern than learning for this project, it should be OK.
Based on the RF components and heterodyned signal filter design from the MIT radar project, I designed the first version. Instead of sending the signal to a computer's sound card for analysis, I sent the signal to an analog input of a Feather board. The signal is filtered with a 15kHz cutoff, which is reasonable to analyze on a fairly low-power MCU. I took care to place all of the RF components under a shield, I kept all the RF traces smooth, with significant ground stitching.
I have not purchased parts yet as I have other projects ongoing, but hopefully soon I can build and test this.