10 GHz CW Radar


This is a modified burglar alarm, based on a X-band Gunnplexer type design. It's interesting, because it has two detector diodes 1/4 wave apart, so you effectively get I/Q demodulation, which allows you to distinguish between approaching and receding doppler. That price was for a pair.

With I/Q detectors, you can just feed one to the X axis and the other to the Y axis on an oscilloscope. It's pretty nifty to watch the doppler trace out a circle as a target moves towards or away from the horn. For walking people, the doppler frequency is a few tens of Hz (66 Hz = 1 m/sec), and the spot traces out a circle going clockwise for one direction, and counterclockwise for the other. This is a fairly good illustration of positive and negative frequency in the complex domain. I've written up a page with more information on how the I/Q works in this gunnplexer.

These motion detectors don't have the varactor in the cavity like the M/ACom Gunnplexers do, however, you can vary the output frequency over a several MHz range by changing the bias on the Gunn diode. There's a web site from TBD with some stuff that he did using another Gunnplexer, also without varactor, where he fed 10 Mbps digital data through it. Conceivably, with a pair of these, you could rig up some sort of AFC to change the bias to track the frequency changes. These kinds of devices are notorious for poor temperature stability (hundreds of kHz/degreeC). In all, this probably isn't the ideal device for a narrow band comm application.

First off, I did some tests to see what sort of voltage vs frequency characteristic the oscillator had. I just coupled the output into a spectrum analyzer while varying the voltage on the Gunn. I wouldn't trust those kHz digits too much, because the oscillator was just sitting on the bench, and the frequency varies signficantly from "load pulling" (i.e. as stuff in the vicinity changes the amount of power reflected back, the frequency changes).

Gunn diode Voltage Measured output frequency (GHz)
7.0 10.520
7.3 10.522 7
7.5 10.524 06
7.7 10.525 18
8.0 10.526 74

I then started modifying, ripping off the existing circuit boards and replacing them with a simple IF output DC block and isolation and a fixed voltage regulator. Here are some thumbnail photos of the modified system. Click to bring up a page with bigger, higher resolution images with comments.

Overall Insides view Detail of interface side view of interface


Circuit details

Here's a simple interface to a PC sound card. At first, I thought I might need to use some opamp hi-Z buffers, but, a bit of experimenting showed that a 1k resistor directly across the detector didn't seem to make much difference, so my worries about the sound card loading down the mixer were unfounded. The final circuit is a 1k resistor across the detector diode, a 2.2 uF series DC block, and a 47k load resistor.
The original motion detector had two fairly large PC boards stuffed with hundreds of parts. I dumped all that and used a 7805 and a couple of resistors to make a voltage regulator to drive the oscillator. The voltage turns out about 8.3V. I was trying for 7.5V, but I forgot to include the bias current through the regulator, so it came out a bit high.


Frequency Modulation

Next, I plan to modulate the voltage with a ramp or sawtooth to generate a frequency ramp. With this, I can do actual ranging, bu looking at the difference frequency created from the delay between outgoing and incoming signals. 300 nSec is 100 meters, one way, so 1 microsecond is around 160 meters, two ways. If I want 160 meters to be around 5-10 kHz, my ramp needs to be, then 5 kHz/microsecond, or, 5 MHz/millisecond. This might be a bit tricky, since I only have about 10 MHz of adjustment range. A delta F of 500 Hz would give me a 20 millisecond ramp period (=50 Hz)...

Perhaps another approach is a stepped frequency radar. You transmit on a series of frequencies, measure the phase and amplitude of the return, then inverse Fourier transform to get the time domain response. Range ambiguity is a problem, of course.

Better antennas

One could always try to use this to feed a dish, but a bit of research showed that feeding a dish with a rectangular horn isn't all that great of an idea. For one, you don't know where the phase center of the horn is, the horizontal and vertical beamwidths aren't usually the same, and a horn doesn't provide a good illumination of a reflector. Not to say that it can't work, but, all the same, probably not optimal.

An interesting idea I just ran across is making a metal lens for the radar. It looks like one can get a system gain of around 30 dB by illuminating a suitable lens, made from layers of styrofoam and aluminum foil, with a horn feed.( http://www.w1ghz.org/antbook/chap3.pdf )

More info

radar10g2 - Bigger pictures of the insides

radar10g3 - An explanation of how using 2 detector diodes can tell the difference between forward and reverse

radar10g4 - Some actual data from the radar

radio/radar10g.htm - 1 May 2003 - Jim Lux
revised 30 Nov 2006, Jim Lux, to add the links at the bottom, especially the page with real data
(radio main page) ( Jim's home page)