Any recommendations for reasonably priced FMCW radars?

[mightyfunkster]

I'm doing a university project revolving around using sensors to aid the visually impaired. I'm yet to narrow down complete specifics but essentially I want to be able to detect velocity and range of objects. If possible as well, I would like to be able to differentiate between different objects (although it doesn't need to be detailed classification, maybe just three labels: car, human, something else).

I've been looking at various available mmWave/FMCW radars out there but I'm having difficulty choosing something suitable. There seem to be two categories for mmWave radars as far as I can tell. The expensive ones like Texas Instruments and Aconeer that are £100+ but these are definitely FMCW and allow you to read the raw data from the radar. The others seem to be around £20 but they often don't give much information about how they actually work.

For instance, there's this £20 one from DFRobot: https://wiki.dfrobot.com/mmWave_Radar_Human_Presence_Detection_SKU_SEN0395 but as far as I can tell you have no access to the raw data and it seems to be pre-configured to detect only humans and outputs a '0' or a '1' if presence has or hasn't been detected with no access to the raw data which isn't particularly suitable.

There are a couple of others like this Grove - Doppler Radar https://wiki.seeedstudio.com/Grove-Doppler-Radar/ that says at the start that this radar can perform distance measurement but I don't actually see an evidence of distance measurement, only velocity and angle.

There is also this radar https://www.seeedstudio.com/Millimeter-wave-Doppler-radar-SYH24A-p-4392.html which says it's mmWave but the only data you can get is speed of movement and the status of personnel which suggests you can't do much with it except human presence detection and it doesn't actually mention ranging.

I was wondering if either anyone had more information on some of these cheaper radars and whether they would work for object detection and ranging or if not is there a method of using just a CW radar and some combination to allow distance measurement?

[geggi1]

Go looking for a HB100 Microwave radar module.
This is a 3cm (10Ghz) doppler unit that is used for all kind of hobby project.
You can get them in packs of for less than 30$.

[mightyfunkster]

Hi, thank you for this! I've just taken a look online and it is very good value, however as far as I can tell the only operating mode is CW which presumably means it wouldn't be able to distance calculation?

[TheUnnamedNewbie]

Go looking for a HB100 Microwave radar module.
This is a 3cm (10Ghz) doppler unit that is used for all kind of hobby project.
You can get them in packs of for less than 30$.

As far as I'm aware, that just gives you a 'is there movement'. Looking at what the OP wants to do I think this is hardly enough?

As for the OP: Is there any option that you can use an expensive radar and just argue 'In a real product, in bulk, the cost would go down significantly'?

[thundertronics.com]

Another option would be STM32 development board (F4/H7/etc. with ADC inputs, DAC output and enough ram) + some sensor with FM modulation input and IQ outputs (https://www.rfbeam.ch/product?id=9, innosent.de, etc.). Note that some sensors with FM input can only perform FSK but not FMCW. Then you connect IQ outputs to ADC, and FM input to DAC. You've already mentioned TI's evaluation boards, check Infineon's BGT24 series evaluation modules too.
On a budget (need good soldering experience): take https://wiki.dfrobot.com/mmWave_Radar_Human_Presence_Detection_SKU_SEN0395 , de-solder main MCU, find out which pins are for IQ, VCO, etc., put wires to some 32bit ARM development board.
Cheapest possible option: take HB100 module, de-solder 100uf capacitor from power, by varying input voltage you can shift it's frequency by some amount, maybe few MHz max. It would be enough to perform FSK modulation. It uses dielectric resonator oscillator for 10GHz, but by altering Vds we can shift S-parameters a little (S21 phase in parallel feedback loop), which will result in frequency shift. Voltage may by applied through dac+opamp. Performance would be not great though, because amplitude will vary, and it is limited to FSK, and there is no prescaler output to know current frequency for precise distance calculation. For a university project I would recommend buying some good evaluation board (TI or Infineon). Check if you can record data and perform some processing in Matlab or other software before buying.

These products are already reasonably priced, let's not forget that ro4350 substrate is pretty expensive, and they usually combine with FR4 into 4 layer to arrange MCU placement, which makes it even more expensive. Manufacturing cost of single tiny PCB prototype easily will cost $150+. Mass produced cheap sensors use cheap components, there is no much functionality.