opinions on reasonably cheap mmWave radar module?

[JBeale]

I just noticed a fairly cheap "mmWave Radar - Human Presence Detection" module being offered by DFROBOT SKU:SEN0395. I won't directly link to it or state the price so this doesn't look like an advertisement but it should be easy to search for.  It claims to be able to detect a person even if they are not moving (from micro-motions like breathing) and claims a range of 9 meters.  You can configure the detection range with up to four zones with 15 cm resolution, but once configured it does not report actual range, only overall "present" or "not present" with 1 Hz update rate.

I know about the mm-wave modules used for automotive applications but they start north of $500 so this is the first board like this I've seen at its price.  I'm puzzled by the fact that it apparently knows the range to target, a rare ability for cheap boards, but it won't directly tell you.  The sub-$10 24GHz doppler motion detectors are CW only, so they can't do ranging. This board claims FM-CW mode as well as CW.

Curious what people think is the likelihood this product is for real. Has anyone tried this out?  The company does have its own forum but doesn't seem to have a lot of traffic.

[Marco]

Interesting, here's a page from someone who's started to look into it I found on google (not that specific one, but there's a lot of FMCW claiming modules out there).

The SRK1101A seems to be the IC on that module working in the RF domain. I assume they are using with techniques similar to the ones from this paper.

[JBeale]

Thank you very much for those links, so it looks like there are many FMCW modules recently, but it's unclear how well they perform. Interesting...

[Marco]

Ultimately a cheap IC with a 24 GHz VCO and I/Q mixers isn't that far out of the realm of possibility and with those the methods from the paper could be implemented with a cheap microcontrollers with DAC&ADCs.

I judge it plausible

[JBeale]

Indeed, it does seem plausible. I'll look forward to someone reporting more definite experimental results.

Here is another product with an IC doing the 24 GHz Tx/Rx: http://www.iflabel.com/product/28.html 
https://www.seeedstudio.com/Millimeter-wave-Doppler-radar-SYH24A-p-4392.html

They have some PDFs including something they call a datasheet (sketchy and light on details), and they are all dated Feb 2021 so pretty recent. They also talk about "micro-motion" and person-sensing even if they are asleep, via breathing movement, but they don't mention ranging or FM-CW.

There is also this, https://www.seeedstudio.com/Grove-Doppler-Radar-BGT24LTR11-p-4572.html
"based on the BGT24LTR11 Silicon Germanium MMIC which is a 24GHz radar transceiver. It is driven by an XMC1302 MCU based on Arm® Cortex®-M0"

That Infineon part has a real data sheet. It does generate I/Q outputs from 24 GHz xmit/rcv, and it does have a VCO although they only suggest it for temperature stabilization, not FM-CW mode. https://www.infineon.com/dgdl/Infineon-BGT24LTR11-PB-v01_00-EN.pdf?fileId=5546d4625696ed7601569d063799153e
Apparently that chip has been around since 2018. https://www.eenewseurope.com/content/bgt24ltr11-mmic-smallest-24ghz-radar-market
bgt24ltr11-mmic-smallest-24ghz-radar

[JBeale]

I  also came across this Acconeer XM112, it is 60 GHz and "pulsed coherent", not FM-CW.  States 0.5 - 10 meter range on the brochure, but only mentions 2 m range in the data sheet. They claim to measure distance in mm, and distance changes of microns. A little above my budget just to play with something that I have no actual use for, but the technology seems interesting.
https://www.robotshop.com/en/high-performance-radar-module-xm112.html
https://www.robotshop.com/media/files/content/a/acc/pdf/high-performance-radar-module-xm112-datasheet.pdf

[JBeale]

I got one of these DFRobot SEN0395 "Human Presence Detection" devices.  I've only played with it an hour so far, but with the example Arduino code it more or less works. So far it is mostly, but not always correct about whether I'm standing or sitting nearby.  I would say it may well be usefully accurate. 

As far as output goes, it's one of the lower bandwidth information sources: effectively only one bit per second. Every second, the Arduino demo code sends a single ASCII character out; a "0" or a "1" and that's all there is.  There's an onboard LED that also blinks once per second while running (but not when initialized, takes several seconds). 

When placed facing up on the desk in my downstairs office, and then walking around in the room directly above, I get occasional "present" hits, so it is either seeing through the (wood) floor, which is plausible, or tracking the motion of the floor itself from my weight; also plausible.
When I am not in the room with the sensor or the one directly above, it reports consistently 0, but when I enter the room it does report 1 so that's pretty good.

Maybe it does do FMCW mode internally, but with their current firmware there is no method of actually determining a range. I asked at the DFRobot forum and got a reply simply restating the spec, which does not include any provision to determine a range except for indirectly by the initial configuration. Changing config takes four separate commands, each of which must be separated by one second, so to manually poll for range would take at least 5 seconds per reading. However there seems to be a ~10 second startup time so it might really be 15 seconds after any configuration change before valid data is available.

At any rate, so far in very initial testing it does seem to do the basic "person detect" function that they claim.

Note: the two unlabeled through-holes on the left-hand edge of the board in the photo, may be an internal power rail. I powered the board with +4.6 V, and measured the voltage across those two points as +3.29 V. The board is spec'd to run from a supply ranging from +3.3 to +5 V.

[tszaboo]

I have some Acconeer eval boards at work, jut designed a board that should use the IC for a proof of concept board.
Do you have any specific questions about it?

[JBeale]

I have some Acconeer eval boards at work, jut designed a board that should use the IC for a proof of concept board.
Do you have any specific questions about it?

Thanks for the reply. Mostly I'm curious if the claim about detecting small sub-mm position changes turns out be achievable in practice, for typical objects, for example a person at 2 m range. Also what the tradeoff looks like for detectable target size vs motion vs range, but exploring that would probably take a lot of time.

[tszaboo]

I have some Acconeer eval boards at work, jut designed a board that should use the IC for a proof of concept board.
Do you have any specific questions about it?

Thanks for the reply. Mostly I'm curious if the claim about detecting small sub-mm position changes turns out be achievable in practice, for typical objects, for example a person at 2 m range. Also what the tradeoff looks like for detectable target size vs motion vs range, but exploring that would probably take a lot of time.

It will do 1-2mm for sure, I tested that.
For sub mm changes, I think the measurement range needs to be narrowed down. So it doesnt measure between 100mm and 2m, but say 300mm and 310mm. But I haven't tried to do this. It has also several measurement methods, maybe its under another one? And in some modes, it doesn't return 1 value, but an entire few kb buffer with reflected signal strength.

[dkim5nu]

I was looking into this chip for application of measuring speed of nearby objects, something that works similarly to a chronograph, except velocities that exceed 4000 ft/s or more than 1000 m/s. Do you think these chips can pick up things moving at these velocities upto the range that they specify it works?

[GadgetGoof]

Could someone recommend a radar for human following robot? I want to install it on Arduino-controlled RC car, so it follows me without using remote control. I thought about putting several radar modules together and comparing signal strength, or even showing a map with obstacles. And yes, I know it's more complex than that.

[AVGresponding]

Could someone recommend a radar for human following robot? I want to install it on Arduino-controlled RC car, so it follows me without using remote control. I thought about putting several radar modules together and comparing signal strength, or even showing a map with obstacles. And yes, I know it's more complex than that.

Have a look at this, it may be something like what you're after: https://youtu.be/TpnBQ9bW4K8?si=ItCTamYOPOTrSQdW&t=490

[GadgetGoof]

Thanks, Bluetooth based positioning looks promising. Haven't seen such devices before..

[MathWizard]

How would detecting a human at a few meters with ultrasonics, or a thermal camera, compare to this mm RADAR ? I know the wavelengths are way different, but somehow they can get a lot of resolution with ultrasonics, like an ultrasound. There's tons of data processing that goes into it.

I could use radar in front of my fridge.


I was expecting the radar units in links above to cost a lot more. Some day I'll have to get one. So far I haven't tried my ultrasonic units with Arduino, too busy with other projects.