Automotive Radar And The Doppler Effect

With more and more cars driving themselves, there is an increasing demand for precise environment aware sensors. From collision avoidance to smooth driving, environmental awareness is a must have for any self-driving cars. Enter automotive radar: cool, precise and relatively cheap. Thanks to a donated automotive radar module, [Shahriar] gifts us with a “tutorial, experiment and teardown.”

Before digging into the PCB, [Shahriar] explains the theory. With just enough math for the mathmagically inclined and not too much for the math adverse, [Shahriar] goes into the details of how automotive radar is different from normal stationary radar.

Only after a brief overview of the Doppler effect, [Shahriar] digs into the PCB which reveals three die-on-PCB ASICs responsible for generating and receiving 77GHz FMCW signals coupled to a 2D array of antennas. Moreover, [Shahriar] points out the several microwave components such as “rat-race couplers” and “branchline couplers.” Additionally, [Shahriar] shows off his cool PCB rulers from SV1AFN Design Lab that he uses as a reference for these microwave components. Finally, a physical embodiment of the Doppler effect radar is demonstrated with a pair of Vivaldi horn antennas and a copper sheet.

We really like how [Shahriar] structures his video: theory, followed by a teardown and then a physical experiment to drive his lesson home. If he didn’t already have a job, we’d say he might want to consider teaching. If the video after the break isn’t enough radar for the day, we’ve got you covered.

47 thoughts on “Automotive Radar And The Doppler Effect

    1. True story: We used to play “Radar Love” by Golden Earring in the lab while doing engineering and testing on 77-80 GHz radar. And during other exotic projects I can’t talk about ;-)

      Also, ZZTop’s “Antenna Head”.

      1. Limited range means limited sensing, right ?
        What happens when you have a high density of vehicles traveling at or near speed limit on a +4 way lane ?
        Decreasing sensing range to keep up with density, yet increasing the collision risks ?

    1. Duty cycle, these radar are not allowed to operate 100% of the time, but rather a few percent of the time. You should also select your time slot randomly to avoid being in sync with antoher radar (or monitor actively when others are transmitting)

      1. Its right in the title of the video… “FMCW”
        It’s constant wave radar, there is no pulsing. Range is limited because of the high frequency and presumably, power output.

        1. Continuous Wave radar does not mean that you have to transmit power 100% of the time. Of course you can, but in this case, the risk of jamming other radar is very high.

          The ramp used in an FMCW radar is actually quite short. You can transmit a few of these ramps, shutdown the radar, wait for a few milliseconds, then transmit an other ramp and repeat.

          Regarding the range, you are right, the main limiting factor is the power.

    2. I think the tight pencil beam antenna pattern they create with their array helps too. Anything comming from behind the car is attenuated. Now for the cars comming the opposite direction….. I’ve bet they got some advanced DSP going on in there to descriminate between chirps.

    3. In better radar they use spread spectrum so they avoid being on any frequency long enough to be properly tracked and shot down. It also puts less power into any frequency which reduces the signature while keeping the power high overall. Since the ‘window’ for the receiver is open for the transmitted frequency for only a short time the chances of detecting another radar signal is small. Since the range and speed and signal power are known from previous returns, one can screen out anything that is far outside what was happening before.

    4. I’m an automotive hardware engineer, specifically for the forward looking radar. I can talk about this topic with reasonable authority.

      I cannot speak for this hardware (I am very aware of this specific hardware), but automotive radar products use chirp modulation – a type of fingerprint, such that the reflected signals can be distinguished from other units. It is absolutely not due to radiation pattern or power limitations (limited range). On top of this the duty cycle is very low.

    1. This technology is just making it easier for more advance terror attacks.

      P.S. The Manchester and London terror attacks were genuinely relgion-nutjob based (as opposed to the Boston incident)

      Ariana Grande – At least the theo-trio had “Documentries” calling her demonic for a while now, so no supprise that such an event was targeted, heck even some non-religious people were calling her (alongside many other funny-handshake-club highly promoted singer-celebs) demonic among other things, i.e. bad influence on the youth of today…

      Just a bunch of maniacs on a Hate-n-Kill rampage spree using (And thus dirtying, read: Mocking) their religion as their excuse.

      Those were primative attacks, whereas those carrying out the attacks are vulnerable people themselves targeted due to their easily swayed thoughts. Thus if they were educated beyond picking up a market item (Guns on black-market) and they had enough brain-power to pull off such modern possible attacks, they’d have enough brain power to reason with them selves and stop them selves before they do something stupid.

      Could be used for terrorism, however the run-of-the-mill terrorists are typically too simple-minded to achieve such things.

    2. Because the radar is the only thing that a driverless car will rely on for driving?

      Jeeezus, people. I said it already in another HaD post and i wil say it again: You alone are not smarter than the thousands of engineers working on these things. There is no way that you’re the first one that thought of that one major unavoidable flaw in the system.

      Also quit trying to see the devil in each and every thing that you do not understand completely.

      1. Can you so explain me why there was no way to open the cockpit door on the germanwing flight, 1000 engineers wont replace sometimes some “idiot questions”. (I was sure there was a way to open those doors from ground control, engineers didnt think about crazy pilot…) (i was also sure there was anticollision system on this tower crane that crashed 30 meters away from me and killed 7 ppl 2 month ago :

        1. Involving failure of human in this discussion is considered moving the goalpost. I dare to say there is no way of circumventing human failure to an acceptable extent. This is why there are pentesting companies out there who will not accept a job that focuses on human error, e.g. social engineering. It’d be a waste of travel time for them, because it’s just too easy.

          The issue with the germanwings flight was, as you said, that no one has thought about a crazy pilot. The door lock was implemented to keep people out of the cockpit. That was about after 9/11, everyone just thought about people that would want to enter the cockpit to cause harm.
          Also, ground control is completely out of play here, it’s not like with NASA missions where one GC controls one vessel at one given time. Even airport towers have zero direct control over the planes, they just tell the pilots where to go. Any data link able to control the airplane would be an attack vector. Attack vectors are bad, mkay.
          Which means the only possibility to allow access to the cockpit in the case of crazy pilot would be an emergency protocol that can be executed by the flight attendant. And this is where we get back to 9/11.
          TL;DR: No way to win here.

          And a quick search shows no results for crane anti-collision systems that are designed for stationary rotating cranes. Just some for overhead cranes, but that’s a fairly easy thing to implement, because those run on the same rail. But on rotating cranes that’d be very hard to near impossible.

      2. But what if those “thousands of engineers” are working in an environment that punishes “thinking outside the box”?
        You know, like under the old socialist governments where questioning the process could get your head lopped off…

        1. I think we’re long past that.

          We’ve long moved towards capitalist thinking. Failures in critical systems might lead to deaths. Deaths are always bad for the capital. “Failure of this part might lead to death? We better make sure that the vehicle remains safe when this part fails.”

  1. The scary part is people with his analysis capability with more nefarious intentions have not yet tested most of these ADAS systems in the car. I’m sure with proper effort, one could simulate a bad reflection to the car and create all sorts of havoc for the driver.

    I’d much rather see R&D investment into passive stereo-scopic 3D vision systems. It seems like as processing power increases and thus the ability to turn z-depth buffers from stereoscopic vision into point clouds and ultimately real-time 3D models of your surrounding would be a better and safer approach.

    1. > I’d much rather see R&D investment into passive stereo-scopic 3D vision systems.

      There is. Loads of it. And there are cars with that out there. They also still have the radar because you can’t have enough sensors if you want to really be sure about something.

      Source: I happen to work in automotive. At exactly that one business units of ours that happens to already be selling stereoscopic 3d vision systems, among other things. We do sell the radar systems, aswell.

  2. These things are in a lot of cars. Higher end Cadillacs and GMCs for example. They aren’t just for self driving cars, but also for automatic braking and blind spot notification.

    Radar can be rather dumb, like police speed enforcement radar, which transmits a CW signal at a set frequency and then looks for a shifted return frequency (Doppler shift) to determine your speed. the receiver is always on, but only looking for shifted frequencies, so they don’t jam themselves – in fact, CW radar is very difficult to jam effectively without knowing what you’re doing. Things like phase and such come into play, plus needing a high power transmitter that can operate at just the right frequency. X-band would be fairly easy, but cops rarely use that band anymore. K-band is high enough it would be hard to transmit a power enough signal at the right frequency without spending A LOT of money. Ka band is wide and you never know exactly the frequency you might encounter and it is even higher than K – again, a lot of money… and all of it WILL send to jail along with a hefty “FCC weight” fine (in the states, probably worse elsewhere in the world)

    Then there are pulsed radars like air traffic and ship radars that transmit pulses, either at a fixed rate for simpler ones or variable rates for special modes of operation. They fire a pulse (or pulse stream), start a timer, open up the receiver and wait for a return, the return pulse stops the timer, do the math and get a range to the target, where the beam is pointed during a pulse/return is the general direction and altitude of the target. There are a whole slew of jamming and anti-jamming techniques at play.

    Then there are mixtures and variants of all of the above. Like this collision avoidance radar. It uses FMCW, where the signal is CW, but swept in frequency, typically in a ramp-up pattern in a few hundred Hertz. It gives the radar more effective bandwidth and allows the radar to “know” that a target is there even if it appears relatively stationary in relation to the radar. Because this is a “CW” radar, unless the target was moving, you wouldn’t know because Doppler = 0. The sweep fixes that and gives you a “hey, something is there!” part. The FM part also makes it very difficult to jam because the radar is setting its own frequency, and it is very difficult to guess where that frequency is going to be in the next instant. So: “my x-mit frequency is X and the return should look like Y but is actually Z, so something is wrong!”, There are techniques for attempting to jam radar that is “smart” but there are anti-jamming techniques to thwart them.

    If my radar is getting jammed on a highway, I’d simply wake the big hunk of meat at the wheel and let them know “hey, drive this thing, someone is jamming us!” – problem solved.

    1. >”If my radar is getting jammed on a highway, I’d simply wake the big hunk of meat at the wheel ”

      How do “you” know when you’re being jammed? AI isn’t smart enough yet to have self-doubt. It just takes all data in at face value.

      1. “This hunk of data that i just received from doesn’t make any sense. I better light up that check engine light and put the engine into limp mode” is working perfectly so why wouldn’t it be able to detect radar data that doesn’t make any sense and wake the hunk of meat?

  3. At USD$14 for Digikey singles, the BGT24MTR11 is rather nicely priced for a complete radar RF platform in the 24GHz band, with everything there except the power supplies, antennas, and baseband/MCU.

    It’s just a shame that small-scale, cheap board fab in Duroid/Rogers or similar advanced dielectrics is not accessible, because it’s really hard to work with a 24GHz chip without it. That’s a shame, because it’s a really nice looking very low cost, cheap, accessible radar.

    1. Looks like Infineon intended to or does sell a development kit. I don’t see a price. There’s another board, DEMO SENSE2GO2, which has 0 qty at all the Infineon suppliers.

      Apparently they chirp the frequency so they know know how to bin the return frequency. It’s not clear how much Doppler shift affects the distance estimate.

  4. In World War II, airplanes used a “radio altimeter” known as an APN-1 that created a UHF signal with a vacuum tube (a 954 acorn triode) that was frequency modulated in a triangle wave by using a voice-coil actuated member in a microwave cavity. In the 50s they were available on the war surplus market for $5.00. There was no need to correct for Doppler effects, because the range to the ground usually varied very slowly in controlled flight.

    1. They still use radio altimeters I think, although normally they use RADAR altimeters now I understand.
      The RADAR ones work at 4.-2 to 4.4 GHz I see, so not so capturable by SDR TV-sticks without some sort of frequency divider. Plus they only use them near airports. might be interesting though to see if you can capture them, maybe even with some raw detection circuit rather than more fancy stuff that can analyze the signal.
      Same for cars, you could measure the pervasiveness of RADAR using cars on the road and see how the growth is in your region for instance.

      1. As I recall, the APN-1 ran at 450 MHz. I think they were called “radio” altimeters because the term “radar” had not come into wide use. The operating principle, of course, is what we call “radar” today.

    1. I’ve evaluated the related BGT24MTR11. These chips don’t really have a software instruction set. They are basically microwave tranceiver front ends with 16 bit SPI register for configuration. To get these older chips to work right, you need to get a lot of external analog stuff exactly right. In contrast, current generation 77 GHz radar MMICs hare much more fully integrated, with internal DDS chirp generators, control processors, ADC’s, and signal processing. Those chips are so complicated that you basically only interact with them through a highly-developed API library. The older ones require you to provide all that stuff externally. Infineon does offer a reference design. They REALLY want you to source the processor from them instead of someone elses DSP or ARM chip, in which case they’ll provide some reference code.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.