-
another small cheap 24GHz doppler radar, but more feature-full
Posted by JBeale on 05 Nov, 2022 04:59 -
A while back I played around with the cheap CDM324 24-GHz doppler radar, sold for door openers and the like. It comes on a PCB that is 25 mm square. It had only one active device, the oscillator and just one low-level analog output: the mixer signal.
I see within the past few years there are a bunch of similar-sized 24GHz modules, I think all using the same tiny square 16-lead microwave chip, that do more and sometimes even cost less. I got this one: HLK-LD1155H-24G which apparently has an I+Q demodulated output so it could in theory detect direction of motion, although it doesn't. On the back it has (I assume) a dual opamp and a microcontroller that does "human presence detection" and spits out serial data showing either nothing, "presence" or "motion" and a signal strength. It sees me walking out to about 6 meters away, at least if I'm directly in-line with the patch antennas looking out of the plane of the PCB, and moving directly toward or away from it.
I can see what look like I and Q output signals on two pins of the controller chip, but they are not brought out to any board connector pins and it's a fiddly job to solder to the chip. Does anyone know of a similar module that makes these outputs more easily accessible?
-
Is the radar IC Infineon BGT24LTR11?
-
That's a good question, and I don't know the answer. It has the right number of pins (16) and the same shape. I see many small 24GHz sensor modules for sale now on ebay and aliexpress which use a similar-looking chip, maybe the same one. They usually spec 24 GHz around +6 dBm, run on 5V at about 80 mA. Some advertise FMCW modulation, and not just motion detection but also range measurement out to 5 meters or so. The "DFRobot SEN0395", "HLK-LD303 24G" and Seeed MR24D11C10 sensors use a similar looking chip. The page on the Seeed device does specify "Infineon mmWave FMCW industrial radar" so maybe that's it.
https://www.seeedstudio.com/24GHz-mmWave-Radar-Sensor-Human-Static-Presence-Module-p-5267.html -
This one uses BGT24LTR11 but it is expensive, US$65.90.
https://www.seeedstudio.com/Grove-Doppler-Radar-BGT24LTR11-p-4572.html
It has IF IQ pins.
https://wiki.seeedstudio.com/Grove-Doppler-Radar/
-
Thanks, that one looks good although it costs more. I see that page says "This product came into life with the joint efforts between Seeed Studio, Infineon Technologies, and Honmax Technology." It certainly has a lot more technical info available than the cheaper items.
I also found a ~$20 part labelled "AM182" with I+Q outputs and claiming +12 dBm output, but it may be just a variant of the older-style CDM324, with a simpler and perhaps lower-performing RF section than what the Infineon part gives you.
-
Well it is pretty sensitive to small motions if you're close. Here are the I/Q outputs I tapped into on the LD1155H board I showed in my first post, with it about 35 cm away from my face. I'm trying to hold still but you can still see motion signals from involuntary movement, I guess that's what is used for "presence detection". Every time I blink, it's definitely visible on the signal. It goes offscale at the end where I move my hand to stop the scope trace.
-
Here's my heartbeat measured from a distance of 1 meter, after the addition of a small 3D-printed exponential horn antenna. At this point I'm sitting very still in a chair and trying not to breathe, because a deep breath causes off-scale movements at this gain.
(I had just run up the stairs and back, it's not normally quite that high.)
-
HLK-LD1155H-24G may use a BGT24LTR11 clone.
If FFT the IF / baseband signal, you can estimate the velocity (CW radar) or velocity and distance (FMCW radar) from the frequency domain.
It is possible to capture the IF / baseband signal with a sound card and post process on a PC. -
I came across another one that looks promising, and does provide pins with IFI and IFQ signals. It also claims +15 dBm EIRP which is higher than the others I've seen, and a detection range of 15 m for people and 30 m for cars. The one I have now is more like 5 meters at most, for people. "The YH-24G01 module is a 24GHz microwave radar module used to detect moving vehicles/people." https://www.aliexpress.us/item/2255801030958625.html
I see this is also listed on a page with some notes on several different 24GHz modules: https://revspace.nl/FMCWRadar
-
I can confirm the YH-24G01 sensor does basically work. I would compare it to the CDM324 as far the circuit design and the output signal levels, except the antenna pattern is definitely more narrow, and of course it has the two separate I & Q outputs. I measured a current draw of 31 mA at 5 V. It does not use the newer ICs like the Infineon part that some other devices do (those also draw more current). The photo shows me waving my hand first towards it and then away, and as the doppler frequency goes to zero you see the phase shift from +90 to -90. The yellow scope trace is what the pinout labels the IFI output, purple is IFQ.
-
Here is a spectrogram output from the YH-24G01 I/Q doppler module, where I can see the difference between walking away from the sensor, and walking toward it. With the log (dB) intensity, the not-quite-90-degree phase shift of the I/Q outputs confuses the picture with the reversed image, but subtracting off a fraction of the reversed +f/-f spectrum allows to tune that out and get a more clear indication of motion towards and motion away (lower plot). It does see me pretty well walking out to 11 m at least, without using any extra horn or other antenna structure so that's a better range than some of these cheap modules, no doubt in part because the antenna pattern is more narrow than most. I'm using a +20 dB audio preamp with AD828 opamp, right into a "line in" audio input of a laptop.
-
photo of doppler module and audio-freq preamp
-
I would think walking away would show a downward frequency shift and walking toward an upward? This is strange? If there is an I.F. frequency involved I could see it but your module doesn't have an I.F., only the doppler beat frequency. Can you explain why this seems mirrored?
-
Swapped I and Q reverse the spectrum. On average, you have a 50% chance of getting it right :-)
-
I would trust the accepted laws of physics including doppler shift, sooner than I would trust whether I ended up swapping the left and right channels when soldering my audio cable. :-)
You are of course right; I was too excited about getting a result that clearly shows a difference, to think about the sign error. -
Chinese auction sites are not known for understating the performance of goods for sale. The description of this one read "The detection range is 15 meters for human and 30 meters for vehicle". I added on a small aluminum foil horn with 100 x 45 mm opening, and was both surprised and pleased to see I was able to detect myself jogging 39 m away and back. Also visible is a car that passed by going the other way, as I was coming back.
-
Happy NY,
Interested to know how do you generate the FFT plots, the sampling frequency and FFT size.
I guess:
top: FFT of complex baseband signal (I - jQ), its vertical range is -2000 to 2000
bottom: FFT of (I - Q), its vertical range is 0 to 400
The horn may increase the antenna gain and reduce the field of view.
Have you removed the shield case and check the internal circuit?
-
Yes, it's just two ways of looking at the I/Q spectrum, and the bottom plot uses the I-Q difference.
The code I'm using to make these plots plot is "IQ-plot1.py" at https://github.com/jbeale1/doppler
Yes, I think the passive gain of any antenna always comes at the expense of the width of its pattern. My crude measurement suggests the patch antenna array on this board, bare with no horn, has a full angle of 20 degrees vertically, when oriented as shown (the AliExpress item page had the gain plot H and V labels reversed.) My horn makes that only slightly more narrow, but reduces the sidelobes.
I extended the horn a few more cm on top and bottom, and tried it out by the side of the road. It looks like I am seeing cars out to a range between 60 and 120 m, depending on the size and shape of the vehicle.
I only have one of these boards and I didn't want to damage it. I do have some excellent photos another person took, I'll ask if I can make them public. It's only got a few parts, but it has that strange curvilinear layout that microstrip circuits use to create filters and resonators on the PCB.
-
Not sure about HLK-LD1155H，but the HLK-LD2410 module from the same company uses S3KM111L from ICLEGEND MICRO，not BGT24LTR11 from Infineon
https://www.iclegend.com/en/product/list/S3KM111L/
-
Do these modules work well for speed measurement of passing traffic?
-
Yes. I haven't tested the accuracy or stability over time. You have to do some coding to calculate the speed, of course.
The principles described in this Infineon note apply here also:
https://www.infineon.com/dgdl/Infineon-Radar%20FAQ-PI-v02_00-EN.pdf?fileId=5546d46266f85d6301671c76d2a00614
Assuming your sensor is directly in-line with the path of travel, the equation is: v = (c / 2 * Ftx) * Fd
v = speed of car, m/s
c = speed of light, m/s
Ftx = radar Tx frequency, Hz
Fd = received doppler frequency, Hz
For this module, Ftx = 24.15 GHz (nominal, no idea how accurate) so it simplifies to v = 0.00621 * Fd (m/s)
For a typical car on my road I get Fd = 1500 Hz => 9.3 m/s = 21 mph
There are also larger and more expensive ($100+) modules designed for the purpose that give you a speed readout directly.
However the spectrogram plot gives you not just one number, but a plot of the car's speed over time with high resolution. For example, here is the spectrogram from a car passing by my house on my road this morning. The car signal (lighter yellow part of the plot) becomes faintly visible around t=25 seconds at a frequency around 1300 Hz, so that's 8 m/s or 18 mph. The signal fades out where the road drops down behind a hill, it comes back into sight around t=35 seconds near 700 Hz so that's about half the speed. The car slows down to a near-stop at 42 seconds, looks like 200 Hz, 1.2 m/s which is less than 3 mph, then speeds up somewhat until it passes my sensor around t=52 seconds. At that point of course the relative speed drops to zero just from the cosine-angle term. You see a bunch of different frequencies as a car passes by, I think from the rotating wheels and hubcaps as the radar illuminates them more from the side. There was a light drizzle of rain at this time, which appears as the short noise spikes looking like grass near zero frequency. The bright horizontal line at f=0 is an artifact because my input signal has a non-zero DC offset, I should remove that with a highpass filter.
This was at 6:35 am before sunup, most cars go around 20-25 mph here but maybe it was a newspaper delivery, or a heavy truck being extra cautious over a speed hump in the road.
-
Here are some photos of the inside of the radar module.
-
So entirely discrete then RF wise?
-
It certainly looks that way to me; basically the same idea as the CDM324 which has been around for years, but with more antenna patches for a tighter pattern, and the I/Q outputs separate. So it doesn't have the fancy features that the newer chips have, but in practice I have found it has a significantly longer range in the same situation (seeing a moving person or a car) than at least a few of the cheap newer modules. Like the CDM324, the output signals are very low, so in most situations it benefits from an external preamp.
-
Also tunable frequency, how precise is the center frequency though? PCB features with a soldered pHEMT and PIN diode for tuning (presumably) aren't exactly high precision compared to an ASIC's internals. Not when 1 wavelength is a couple mm.
Maybe they calibrate the module at the factory?