Radar Help Needed

Soundcard & libfftw/gnuplot/little C program? A card capable of 96Khz sample rates is not expensive and this will give you almost DC to 40 odd Khz at any frequency resolution you like, with 90+db dynamic range. You can also trivially write code to calibrate out small delays in the system (take one FFT (of a 0 range target), then subtract it from all the real plots. Hell you have 2 channels of input available and know roughly what the carrier frequency is, so you know how much IF delay is 90 degrees, therefore you can fairly easily generate a Q channel as well as an I channel.....

Why not get a commercial Gunn/Pin diode module they are readily available and not expensive. This thing contains a gunn osc in the rear part, a slot for coupling and a pin diode across the waveguide. Fairly crude but probably good enougth. I have seen them on the surplus market rated at up to a 100mW CW. This then solves all your RF problems, except the horn....

You may not have enougth rf at the pin diode with the circulator approach as they like quite a bit of IF drive. Your delayed signal is at constant frequency right? So why not remove it in your post processing? Or if the IF is leading on the leakage just increase the length of the IF feed line?

Coil of low loss coax (or length of waveguide) fed from attenuator and with one or more impeadence discontinuities to create reflections? The coax is convinient as it is inherently lossy at these frequecies and you can just short it at the apropriate distance.

And I would appreciate if you can teach me how to create a

Seems like IF to one input of the mixer, audio to the other, output to the recever should work? Actually this generates a DSB signal around the IF, but as I doubt your rig is smart enougth to tell the difference....

Regards, Dan.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Thank you very much for the help, Dan! However, do you know any companies that sell this gunn/pin module? I actually tried to search "gunnplexer" (which combines gunn diode, varactor diode, and mixer diode), but I have not found many of them yet! Stanley

Hi, Stanley. I happened to stumble onto your post. I've been away from FMCW radar almost since I posted the message that you referred to but I'll try to provide something of an answer for your questions.

You might want to locate a radar handbook and refer to it for radar cross-section information. A "big" copper sheet isn't really necessary at the short ranges you're dealing with. Small objects can have relatively large RCS. Check references such as Skolnik.

I'm sorry but I'm not an RF engineer so I can't comment on the subtler aspects of mixers. The FMCW systems I'm familiar with used two antennas.

This is extremely difficult to explain without drawings but I'll try. I'm used to radars used as altimeters so I'll describe the spectra from that perspective.

Suppose an aircraft is flying above a perfectly calm/smooth surface of water. Then the only return the receiver will see will be from directly below the aircraft, the nadir. Any radiation away from the nadir will glance off the surface of the water, as visible light would from a mirror, and be directed away from the aircraft's receive antenna. So imagine a single ray of energy leaving the transmit antenna, bouncing off the water and returning to the receive antenna. In that case, the receive spectrum will look like a single line spectrum. Of course, the farther away the target (the water) is, the larger will be the time difference between the transmitted waveform and the received waveform, and the higher will be the difference frequency. But the detected spectrum will be a single line.

Suppose now that the aircraft is over rough terrain. In this case, even radiation away from the nadir can strike an object on the ground and be reflected back to the receive antenna. But that radiation will have to travel farther than the radiation to the nadir. Since the signal angled away from the nadir will travel farther, two things will happen to it: 1) the range as indicated by the difference frequency in the receiver will be higher than the signal from the nadir; 2) the signal will be weaker than the signal from the nadir. Also, the signals angled farther from the nadir will strike the earth (the target) with a more glancing blow and will reflect more poorly. The result of this is a received spectrum something like I've drawn here (you must use fixed point font to make this clear):

| |* | * | * | * | * | * ________| * * * * * * * * ^ ^ | | Return from Returns at increasing the nadir. angles from nadir. Each Closest to the is higher in frequency aircraft and, and lower in amplitude. therefore, strongest.

So the spectrum of the difference signal at the receiver will appear to have a sharp rising edge and a slowly decreasing tail. The smoother the reflecting surface, the smaller the width of the tail. For your metal plate, the ideal spectrum will look like the leading edge without any tail. But you're probably in a lab with tables and chairs and cabinets (and you!) so other things will be reflecting back to the receiver. The result can be a bumpy tail and a leading edge which might not be so vertical. And, of course, the smaller the RCS, or the bigger the range, the lower the amplitude of the received signal.

Don't use the antennas in this case. Connect a suitable attenuator to the transmitter port, connect the output of the attenuator to a delay line and, finally, connect the output of the delay line to the receiver port.

Be sure the attenuator drops the signal level low enough so that you don't damage the delay line but not so low that the signal is too weak to detect it at the receiver. Some delay lines produce a lot of attenuation.

Connect an attenuator to the antenna port, connect the output of the attenuator to a delay line, maybe add some length of transmission line, and short the other end of the transmission line to cause the signal to reflect. Again, watch how much total attenuation, including the delay line, that you use. Remember that the signal has to pass through everything twice. You can get away without a delay line if you use a reasonably long length of cable. Remember that the speed of the signal in the cable is less than the speed in free space so that a given length of cable will yield a longer range than its length might indicate.

Keep in mind, also, that one weakness of FMCW radars is that they have trouble at short ranges. This is because the leakage between antennas, which are close together, tends to cover up the true return signal. This might be a serious problem for you when your target is only 3-5 feet away.

Some FMCW systems use a high-pass filter after the receive mixer to attenuate close-in returns, like leakage, and not attenuate returns from more distant targets.

It's been a while for me but I'll try to get it right. Connect an attenuator to the transmit port, connect a double-balanced mixer to the output of the attenuator, and connect the output of the mixer to the receiver port. Drive an audio generator into the other mixer port. (I can't recall which port is used for which connection. You'll have to figure that out yourself.) Make sure that you use an appropriate signal level from the audio generator. The audio signal (since it is modulating the transmit signal) will look to the radar like a delayed return; the higher the frequency of the audio generator, the larger the apparent range to the target.

Your reference to gating functions suggests that you mean weighting function, something that's used in digital processing to deal with the sidelobes on sampled spectra.

My reference to the gating function was to only allow the transmitter to transmit during the controlled part of the sweep. Sometimes sweep circuits take time to settle, for example. So a clean way to do things is, for example, to let the up sweep begin, then gate the transmitter on. Then, just before the up sweep ends, gate the transmitter off. If you have a well controlled sweep generator you might not have to worry about this.

One point of interest is Doppler. If you only sweep up, say, the a Doppler shift due to target velocity will introduce an error into the range measurement since it adds an unwanted frequency component. The error would be in the opposite direction if you only swept down. So by sweeping both up and down and averaging the range measurements of the two sweeps, the Doppler error is cancelled. This shouldn't be an issue for your stationary targets in a lab, however.

I hope this has been of at least some help to you. Please do not contact me by e-mail. These discussions are best conducted in this forum.

Good luck,

Bob

Continental microwave IIRC have 10 & 24Ghz units, but are fairly pricey. RF parts way be worth a look (not sure). 'Gunnplexer' is IIRC a trade name for the unit I am thinbking of. Also try local ham radio meetings, someone normally has some for sale.

IIRC

Regards, Dan.

Thank you for the tips that you provided, Bob! I think I can make more progress on my fmcw radar project!! Regards, Stanley

Hi! Bob, If you somehow see this message again, do you know anything about using phase detector (comparator) with in-phase and quadrature channels to determine either an approaching or receding target? In addition to that, could I use them to build an anti-jamming system? I actually see lots of radar system with "simplified" diagram. They really don't give readers many insights! Thank you! Regards, Stanley Chien