another small cheap 24GHz doppler radar, but more feature-full

[buta]

Thank you alex_udanis the photos.

It is an interesting implementation with discrete components and distributed elements (layout and pcb substrate).
The frequency and Tx power stability is one of the challenges to ensure the radar meets the ISM band regulatory requirements.
The reflected signal is directly converted to the base band by (maybe) diode mixers so the receiving path offers a conversion loss, but it may be ok for a short range radar with big targets (human and car).

[buta]

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

where v is the radial speed https://www.radartutorial.eu/11.coherent/Radial%20Speed.en.html

[JBeale]

It does not use much silicon but it exceeds its claimed specs for range, and for applications like my driveway motion sensor I would say it has the best price-performance ratio of the low-cost modules I have tried.

In my use, I have not connected the frequency adjust pin. A few comments I have read online say it has no effect, but I don't know what tests they did. I have no test equipment for the 24 GHz band, so I have no idea what the frequency stability is; all I can say is that the basic doppler function works. How much testing was done, and in fact if it has any agency approvals at all, is unknown to me. All I can say for sure is the one module I have uses 155 mW (31 mA at 5 V) so it is not putting out more RF energy than that, and the frequency is not wildly off, given the doppler equation assuming 24.15 GHz gives me roughly the expected answer.

[Marco]

In my use, I have not connected the frequency adjust pin. A few comments I have read online say it has no effect, but I don't know what tests they did.

You could just try to reflect off a nice strong constant reflector at a couple meters and put 5V DC with 4Vpp sawtooth on the Vt input, varying the distance should vary the frequency of the I/Q outputs according to FMCW math you can find online.

[JBeale]

I tried to calibrate my detector for speed by riding a bicycle at a constant rate. That turns out to be difficult to do. I was listening to a metronome app on my phone to help time my left foot at a steady 90 beats per minute, but I'm pedaling with both feet so I was pushing the bike forward 180 times per minute, or 3 times per second. You can see that rate in a small velocity wiggle on the signal that happens 15 times in 5 seconds, so it checks out. However there was also a slower wander of the average speed, probably related to the road not being absolutely flat so it's hard to ride at an absolutely steady pace.  To get a cleaner signal I attached a small corner-cube radar retroreflector to the bike handlebars, so the signal is dominated by that reflection, and not the various motions of my body and legs.  A GPS tracklog says the bike averages 7.9 to 8 mph in 2nd gear with this pace, and various readings of the data gives me a calibration factor from 68.9 to 69.5 Hz/mph, but with the variable speed this may not be any better than 10% accuracy.  From the doppler formula at nominal Ft output, this module should be 72.05 Hz/mph.
I also tried the experiment with a car, but try as I might that doesn't go at a perfectly constant rate either, when the road is not level.

[I wanted a rude username]

Wonder what the maximum detectable normal speed is. Hypothetically this would be limited by ADC sample rate, but these sensors probably aren't designed to work at more than "human" speeds and might have other tradeoffs built in.

[JBeale]

If you're referring to the YH-24G01 model I've been working with most recently, I don't know the module's schematic but you can see from the photos it doesn't have a lot of discrete parts. I'd guess it may have 1-2 poles of simple R/C lowpass filtering for "normal" speed ranges. Apart from that, I'd expect such circuits to potentially be able to detect high speeds, even hypersonic. The underlying doppler shift principle applies to any physically possible speed*. But by itself you just get IF outputs, so designing an appropriate signal chain for any given frequency is up to you.

If you meant the ones I posted about at first, as the original thread title mentions, the cheap ones I have seen are intended to detect typical human motion and I presume they filter out higher speeds, which would be considered noise in that application. I believe Infineon offers other parts designed to measure up to 100 km/hr.

* However, a reflected doppler signal goes to f=0 and disappears completely if an object is receding faster than half of light speed.

[buta]

I watched a YT video using flying fox pulley to test the radar but I can't find its URL now.
Alternative the Doppler shift can be checked by tuning forks.
Fd = 160 * v at 24GHz
Fd = 160 Hz when v = 1 m/s (3.6 km/h)

The YH-24G01 does not provide the IF bandwidth, it states the Rx conversion LOSS = 6 dB typ at IF = 500 Hz (approx 3 m/s)

[S57UUU]

Radar Doppler is simply twice the number of wavelengths passed per second.
Example:  13mm wavelength, 1m/s  -> 1000/13*2 = 154Hz
Wavelength = 300/f [MHz]

Marko Cebokli

[JBeale]

Nice idea with the pulley! Amazingly, I have a similar thing all ready to go, it is still set up under the eaves of the house, left over from an animated ghost display I made for Halloween.  I hadn't even thought about that, but now you point it out, it is an obvious thing.  I'm not sure how fast I can make it go, but probably a useful speed- I know the motor is reasonably steady, and I can measure the true speed independently in various ways. Thanks!

[buta]

If you just want to check the speed (or Doppler shift), tuning forks may be alternative.  The pulley system properly can be used to check the radar SNR and the target speed.  Looking forward to watching your implementation.

[JBeale]

Here is one thing a small doppler module can do: show the actual effect on vehicle speed of a traffic-calming speed hump on a 25 mph residential road. The leading edge of the hump is located at about 140 feet (radial distance) from the radar.  This plot showing the trajectory of several different vehicles, records the measured vehicle speed as a function of radial distance (which was calculated by stepwise integration of the speed vs time data, starting from signal acquisition, which with this antenna pattern roughly corresponds to where the vehicle passes abreast of the sensor).

Looking at this data I would say that slower vehicles, which tend to be larger ones,  slow down more dramatically approaching the hump, which I also notice by eye. Medium-speed vehicles have a variety of slightly different behaviors, and a few seem to simply ignore it.  The road goes over the crest of a small hill around 225 ft where most cars pass out of view, so the one trajectory shown here extending beyond 300 ft. must have been a larger van or truck to still have been visible to the radar at that distance. 

The lower displayed speed from 0-50 feet is just an artifact of a cosine-angle term, because the radar is positioned off to the side of the road and records apparent (radial) speed, and not true along-direction-of travel speed, but those become similar beyond about 50 feet as the radial vector and vehicle motion vectors approach alignment.  I am showing cubic-spline-fits to the real data to reduce noise, but the curve fits are pretty good (except for initial scatter, where from looking more sideways you see more wheel and hubcap reflections).

EDIT: Ah, the hazards of fitting real world complexity into simplified curves. Looking at the raw data, it looks like the >300 ft trajectory was not a single track, but in fact a combination of two separate vehicles passing by, without enough space between for the algorithm to separate them out.

[JBeale]

A bit more data now with both raw data points and best-fit, 5-knot cubic spline curve overlaid, so you can see that past 50 feet out, the velocity curve fit is usually pretty good.

[buta]

It is interesting to know drivers' behavior.  Will you put up a sign to show drivers their speed?

I think the amplitude of the Doppler shift can estimate the size of the vehicles, one of the facts affecting the amplitude is the object's radar cross section.

[JBeale]

One thing I didn't expect was that the system can (very roughly) estimate the length of the vehicle, from combining the speed when it is more distant from the sensor, with the duration of the near-0-radial-velocity part of the signal (approaching freq=0), which shows how long it takes for various parts of the vehicle to pass immediately by the side of the radar. This is with the assumption it isn't changing speed much overall. It's not more consistent than about 25% but you can generally tell a smaller car from a larger truck on that basis, and something like a garbage truck or school bus is pretty clear.

[JBeale]

Here's a comparison of two different doppler signals from oncoming traffic. Time increases to the right, speed is zero at the bottom edge,  increasing upward. The signal falling towards zero indicates when the traffic passes right by. One from a normal-size car, and another from one of the larger trucks I've seen on this residential road.  The signal from an oncoming car shows up about 200 feet away, while this particular truck was detectable at very nearly twice that distance. You can see from the width of the vertical part at the right edge of each signal, that the truck takes longer to completely pass by.  Looking at the truck signal, it slowed down and sped up a few times as it approached and much of the velocity curve appears to show changing speed with nearly straight line segments, maybe hinting that this truck has a manual transmission (?)

This image doesn't have a speed scale, but in this case the car shown reached a maximum apparent speed of 23 mph and the truck reached 19 mph.  The apparent measured speed is radial to the doppler sensor. As both vehicles were increasing in speed as they passed by, it's probably safe to assume they reached some higher speed later.

Also visible toward the bottom on the left-hand side under the car is rain, I think the larger lines are individual rain drops that happen to fall right in front of the antenna, which generate discrete blips in the signal.