playing with cheap HB100 doppler radar module for car speed

[JBeale]

For only $7 I figured the HB100 doppler radar module on ebay was worth it to play around with.  I took the IF signal direct from the board through a DC-block capacitor into the mic input of my camera, with no preamp or other circuits. I did not expect there to be much range, but cars (esp. SUV and truck sized cars) seem to be a big enough target to get over 100 feet of range usually.  Putting the recorded audio into a FFT shows a very clear signal peak, and on this 10.525 GHz module you divide the peak frequency by 31.37 to get speed in mph, so a signal at 953 Hz means the car is going about 30 mph. It gets more tricky when the car gets close and you see it more from the side so you're illuminating the wheel spokes, which complicate things with more modulation, not to mention the detector is to the side of the car's path, so the main return signal droops down in pitch with the cosine of the angle from the detector.

If your audio is turned up (and your speakers have enough low-frequency response) you can hear the signal from the legs of the jogger at 0:46, which have a low swishing heartbeat sort of rhythm.  That is a pretty small target, so I didn't expect the radar to see it.  This is with the bare board, I assume it could be made more directional with a horn or reflector antenna.

  https://youtu.be/r2NwtRPYWK4

[JBeale]

The spectrogram of the audio above shows some things more clearly, you can see the jogger's legs around 0:45 for example, and I didn't realize I tracked the third vehicle (from video, actually a dark pickup truck) all the way to stopping at the intersection, which is 500 feet away!  The rear of the pickup must be a strong reflector.  Spectrogram by "spek" http://spek.cc

[edavid]

Very nice!  BTW, if you want more HB100s, they are only $2.30 on AliExpress.

[JBeale]

They are amazingly cheap. What I'd really like is a microwave horn to get better directionality and range. I guess it should be easy enough to make out of sheet metal, or even aluminum foil. 

[JBeale]

Some improvements, in case of interest; this extended detection of a single pedestrian walking at 2.2 mph out to 80 feet.
I used a Teensy 3.2 for the backend and I also posted the code here:
https://forum.pjrc.com/threads/45621-T3-2-with-FFT-improves-performance-of-cheap-doppler-radar?p=150107&viewfull=1#post150107

UPDATE: original schematic had a wrong value, 1.5 k / 100k would give you only x67 gain. I actually started there, but then decided I could use much more gain, and my final circuit used 150 ohm / 100k for 673x gain.  Also, I just now added a printable PDF with the actual size pattern for the horn antenna I made, with the HB100 circuit board sitting inside flat on the bottom of the truncated pyramidal horn. You may or may not want the extra directivity, but it's helpful if you want greater range, and without it you end up with substantial pickup to the sides and back of the board where you may not want it.

Update 3: Add photo of cleaner horn made out of 0.011" copper-clad FR4 and copper tape.  Better sealed in back, so it has better front/back ratio; about the same gain in front though.

Update 4: As seen on Hackaday. Some of the comments there point to interesting ideas. http://hackaday.com/2017/08/10/making-a-cheap-radar-unit-awesome/

[JBeale]

In case of interest, here's what the 24 GHz module CDM324 looks like under the hood (took it apart after I accidently applied +5V to the signal output, which killed it).  I note the PCB is actually marked IDM136.  https://goo.gl/photos/ZZeaj7dLyfe7ahLo7

It is apparently a clone of a device for which there is an actual datasheet:
http://www.innosent.de/fileadmin/media/dokumente/DATASHEETS_2016/Datenblatt_IPM-165_V8.5.pdf

[JBeale]

I wanted to check if it was possible to redirect both the output and input patch antennas of the HB100 into the .9 x .4" aperture of a WR-90 waveguide flange, so I could use a surplus horn antenna with it.  This may be as silly as it looks, but yes it does work to some extent (signals go out, and reflections are detected.)  This lensed conical horn from an old X-band radar speed gun has a beamwidth of 18 degrees (FWHM) and very low sidelobes, so the pattern is quite a bit tighter than either the raw board, or my crude pyramidal horns.

EDIT: I was later able to get about the same directivity with the HB100 just sitting at the base of a somewhat larger, flared exponential horn made from aluminum sheetmetal, without using a dielectric lens like the old speed gun had.  With no need to funnel the signal through the WR90 flange, the feedpoint match and hence signal strength and total range was improved.

[Oldgreybeard]

I know this is an old thread, but I just wanted to post to thank @JBeale for coming up with this idea, and add my slight variations to it.

By way of background, our drive joins a single track lane on the inside of a corner, and visibility up the lane is zero, so we have no idea  if any cars, cycles, etc are coming down the hill towards where we want to pull out.  Nothing ever goes that fast, apart from the occasional cyclist, so we can usually get away with just edging out slowly, and hoping any traffic coming sees us first.  I've been looking at a way of sensing cars and bikes coming down the lane, and giving a reliable traffic light indicator by our gate, showing when it's safe to pull out.  Doppler radar seems ideal, but the beam width of the antennas on the standard HB100 modules is both too wide and there are side lobes that detect motion from the sides and behind the module.

As a twist on the horn shown here, I decided to 3D print something similar, sized to fit into an off-the-shelf IP66 box.  I found that using self-adhesive copper foil, as used for screening inside guitars, was ideal for lining a horn antenna, as long as the 3D printed surface is sanded smooth and the copper foil burnished down well to give a smooth surface.

The basic horn worked OK, and certainly extended the sensing range, and got rid of the unwanted detections from behind and to the side, but the beam was very skewed to one side, I found.  This isn't that surprising, perhaps, because the transmit and receive patch antennas are close to the walls either side, so my guess is that there are some pretty odd reflections happening inside the horn.  I'm no microwave expert, by any stretch, but I know that a patch antenna can feed a waveguide OP, so what I thought was needed was a short waveguide section, matched to the polarisation and frequency of each patch, to match each antenna to separate transmit and receive horns.  I also thought that, as well as increasing  the directivity a bit, this might reduce the slight remaining susceptibility to movement at the sides. 

It turned out that I could just about fit a 19mm x 9.5mm rectangular waveguide section to each patch antenna.  A waveguide of these dimensions has a cut-off at about 7.89 GHz, a lower working frequency of 9.86 GHz and an upper working frequency of 14.91 GHz, so just about OK for the 10.525 GHz from the module, and best of all, this size placed each patch antenna just about smack bang in the centre of the waveguide section, leaving just enough room for a small, tapered, divider to hopefully match each transmit or receive patch to it's horn a bit more effectively.

Initially I just added a central wedge divider, between the transmit and receive antennas.  I found that the dielectric of the PLA that I used upset the microstrip transmission lines to each antenna, so had to cut away a section around the feed, just for 1mm, to avoid this.  The central divider, on it's own, very significantly increased the range, and got rid of the offset directivity.   Adding the tapered dividers between each of the pair of transmit and receive antennas improved things still further, perhaps because there was now a better match to the horns.

I cheated when it came to the signal amplifier, and bought HB100 modules already fitted to an amplifier board, so the output is at TTL level, and easier to process.  The board I used is the DFRobot SEN0192 module.  This comes with two LEDs and a sensitivity pot mounted on the antenna side of the board, which didn't allow it to fit flat to the 3D print, but it was easy to de-solder both and fit them to the other side of the board, leaving a flat face to screw down to the back of the antenna.  These DFRobot modules aren't that much more expensive than the bare HB100 modules (I paid around $9 each for them).  having the amplifier, filter and comparator on the one module makes for easy interfacing to any microcontroller, too.

The next stage is to knock up some simple code to only detect moving vehicles, and ignore the occasional spurious trigger from movement of trees, etc.  The plan is to just use a simple pulse counting approach, to detect targets within a range of speeds, say from 5mph up to maybe 30mph.  Nothing will be going faster than this down this narrow lane, I'm sure, so this should be good enough to filter out unwanted returns that are only there for a quarter of a second or so.  Tested range with me just doing a walk test seems to be around 50m, so plenty, I hope.  I'm sure that the range will be better for cars, but 50m is enough, really, as it will only be able to see around 70m up the lane, anyway, because of a bend.

The finished unit will be solar powered, and to save power I'm going to have it off most of the time, only turning on when another sensor detects a car coming down our drive.  I'm hoping I can use a very low drain PIR sensor to do this, so that the unit can run OK with a small battery and solar panel.

Below are some photos of the prototype horn antenna, together with the modified HB100 module.  Thanks again for the inspiration you gave me from this thread, @JBeale:

[JBeale]

Wow! Thanks for reporting on  your experiments, that looks great!  Seems like the microwaves always want to see gradual transitions in waveguide or horn geometry, to reduce reflections. 

When it rains, I found it also works as a (short-range) weather radar, the raindrops reflect enough energy to show up.  The signal looks different enough from a car (more broad-spectrum noise than the single frequency of a big target with fixed velocity)  that the software could probably be written to not trigger on rain, but your working car detection range would still go down whenever it rains, I think.

[mimimomonz]

Reviving this old thread again, thanks for your great detailed description @Oldgreybeard. I have the same scenario with a blind corner on our driveway exit, and thought about a radar triggered traffic light - and you have done all the design work for me :-)

Would be interested to hear an update on how this project has progressed since last year

Cheers, Wayne