I’ve been tinkering with some of the raw telemetry data from Venera-15, an radar imaging satellite that the Soviet Union put into orbit around Venus in 1983.
Here are some examples of the stages of processing. The radar system records 2540 complex (in-phase, quadrature) samples of a reflected radio signal. It just looks like random noise if you visualize it.
Correlation of that signal with the pseudo-random code sequence converts the continuous signal into a series of pulses, and that reveals the beginnings of an image. The curved stripes are due to phase shifting as the spacecraft moves away from the surface in its orbit.So, the next step is to correct for that, and that gives us a very noise image of a strip of the Venusian surface. There’s a lot of speckle noise, because the radar beam is coherent illumination, like laser light.
The 2540 samples are 20 looks at the surface, and we can do image stacking to get a better image (the fourth image). However, if we apply Fourier transform, we can separate the signal into 20 Doppler frequency shifts, which represent narrow strips of azimuth. These can be shifted into alignment and stacked, to produce a higher resolution image as seen on the right.