10 GHz Doppler Radar Data

This page has some actual data from the 10 GHz Gunnplexer style doppler radar described here. The radar was set up pointing along the rail of a riding arena where kids were going both directions on horses (now there's a nice big target) at various speeds. I captured about a minute of data using DL4YHF's spectrogram program (http://www.qsl.net/dl4yhf/spectra1.html). Here's a screen capture typical of what you see while capturing data.

Since my targets are moving fairly slowly (a few m/s at most) I've set it up to look only at the bottom 500 Hz, and I'm only looking at one channel here (although recording both channels at 11.025 kHz). At this radar's frequency, a doppler frequency of 66 Hz corresponds to a velocity of 1 m/s, so you can see some targets moving by at about that speed. The higher frequency data is probably returns from faster moving things on the horse (like their feet or the rider thrashing about). Some day, maybe I'll stick some corner reflectors on a target and figure it out.

At the time I recorded this (January 2003), I hadn't figured out how to treat Left and Right as I/Q signals for DL4YHF's program (in fact, it might not have had that capability back then). In any case, I now had a stereo .wav file about 2.5 MB long (which I will send you if you send me an email asking for it). I loaded it into Matlab, and used Matlab's capabilities to do the spectrogram, as well as turning left and right into I/Q complex samples. Here's the whole sample epoch:

As you can see, we've got fairly good dynamic range, and the "interesting stuff" is down in the sub 1kHz region. You can also start to see that you can distinguish between things coming towards the antenna and going away have positive and negative frequencies. In this picture, the negative frequencies appear at the top, because that's the default for Matlab's spectrogram function. Next, I decimated the signal down by a factor of 10 and did some rescaling of the axes to make a more conventional plot.

This one has time running horizontally, as well. The solid line in the middle is the DC offset in the sampled data from the sound card. The noise out to a few tens of Hz is probably mostly the close in phase noise of the Gunn oscillator (which are notoriously noisy). However, here you can easily see the targets moving by. The ones with positive frequency are moving towards the antenna, the ones with negative frequency (around 27 seconds) are moving away. The curve is because the target is moving at an angle, but fairly close, to the radar. In fact, from this data, you can tell that the velocity sense is correct.

You can see that the "image rejection" of the system isn't all that hot (maybe 10-20 dB), so every target actually shows up in two places. I haven't fooled with it to see if you could improve it (say by changing the relative gain of I and Q). The matlab code can be downloaded from here.

radio/radar10g4.htm - 30 November 2006 - Jim Lux
(radio home page) (Jim's home page)