Cantenna radar busted

Making your own microwave frequency circuits is still as expensive as it ever was.


That's the heart of a 2.4Ghz heterodyne receiver. It's $47.





That would be the 2nd most expensive part, at $43.




These parts were ubiquitous in college, but you didn't realize how expensive they were. You can't design microwave circuits without government money or outside Asia. We might have used cheaper 1Ghz parts because Wifi had yet to get big.



It was a bit strange that they used an attenuator followed by an amplifier. The mane change was using a 100uF instead of a 1uF to decouple the IF output. That lowered the minimum frequency it could detect, thus the minimum velocity.



A simple test of the MIT radar in doppler mode revealed just how sensitive to moving objects even 2.4Ghz can be. It had no problem detecting the slightest movement of an object. It was directional enough to detect only movement towards & away from the antenna. It didn't detect movement not directly in front of the antenna.


Unfortunately, flying based on doppler mode was busted. When tasked with detecting its own movement based on the relative motion of the walls, there was no directionality. When looking straight down, there was no isolation of altitude from horizontal movement. There was no isolation of translation movement from turning.


For the horizontal sensing, there's no way it would detect the right motion when pointed at a corner or at clutter. For every case, it was too sensitive to any objects near the antennas. That might explain the lack of directionality.

Knowing exactly why it doesn't work requires a spectrum analyzer. It would be affordable for 2.4Ghz, but not the 60Ghz automotive radar.


An FFT would take too many computations to detect the slowest motion at 2.4Ghz. It would need a simpler way to detect the rate of change in the waveform. Chirping mode would give better results for the slow speeds. Fortunately, automotive radar uses 60Ghz which should increase all the beat frequencies.

Finally, there's no way for the doppler mode to determine if it's moving towards or away from the object. Quadrature output theoretically solves this.


There is hope a proper automotive radar will be more directional, but simple doppler navigation is busted. A very directional beam pointed at the floor might still give the desired velocity reading, as long as it avoided walls.

Most likely, 2 doppler modules need to be on a turret which is always aligned with 2 completely unobstructed walls & sonar still needs to be used for altitude. Any synthetic aperture or chirping would not fit in the available computing power. It's probably time to start investing in an FPGA development board.