Failure modes of automotive radar sensor

[johnnygoode1]

Hi everyone,

I got my hands on a Bosch mid-range radar (MRR) sensor that I believe belonged to a Tesla model S a while back. Now I'm working on a school project where we try to get the sensor to fail (i.e., output wrong or unusable data). Those of us involved in the project have defined a goal of understanding more about the mechanisms behind how radars fail in general but also what aspects of Bosch's specific sensor design might contribute to a failure.

So far, some preliminary research has revealed the following common failure types for automotive radars:

- External RF noise
- Ghost targets (e.g., multiple reflections from the same object resulting in targets with integer multiples of speed/distance)
- Obscurants on the sensor (e.g., snow, mud)

That's all we can come up with at the moment but I'm sure someone out there knows more about this stuff.

Thanks!

[CopperCone]

I assume you are looking at what can happen on the PCB they have to or analyzing the semiconductor structure they used for areas of current density etc to determine what  partial failure modes can possibly happen due to 'normal' electrical circuit failure/degredation?

If its one chip behind a NDA you probably won't get far unless your gonna reverse engineer it and have good expertise.

It sounds crazy but you can get a fab that does a run for a small company that just sent them some kinda simulation data that possibly does not have the expertise in designing chips that you are used to buying from people that have been in the bushiness for a long time with their own fabs. Anything designed like that is going to have back and forth between two companies meaning that it can get chowdered more easily. This reasoning is pretty paranoid though, since you have two sets of eyes on it.

Also there is the whole issue of quality (good design being implemented haphazardly or with process problems)

Before someone starts a shit storm over this,
https://www.darpa.mil/program/trusted-integrated-circuits

This kind of thing is an issue be it malicious or poor design. That one focuses on malicious though.

Or are you only interested in what can happen to a theoretical electrically functional module ? i.e. how the sensor data can mislead

[CopperCone]

well if a gain stage degrades in gain for instance it can cause erroneous readings that might be detectable with various software algorithms, i.e. degraded operation but 'safe enough' or at least to cause a warning

something like a filter that tracks various process variable spans in a box car or something like that, maybe the shape is correct but the magnitude is wrong or vise versa, then it can signal a flag

[TheUnnamedNewbie]

Look at the output bands it is operating on. See if you can detect the signal and use it to retransmit a false signal. After all, the radar is looking for a frequency shift (doppler) and a time shift (distance). Frequency shift can be easily done with a mixer, and time delay... Well, do you happen to have a lot of coax laying around?

Alternatively you can just see what happens by spewing wideband noise at it (ofcourse, do it in a shielded environment, even though the range at these frequencies is abysmal, you don't want to get into trouble). Even if they use narrow band filters, if you hit it with enough power to push the LNA into nonlinear operation, I suspect the performance will take a hit.

A lot of what you can do does depend on what equipment you have access to. What frequencies is this operating at? Mixers, oscillators, etc do not come cheap.

[johnnygoode1]

blueskull is correct in that we're generally looking for black-box failure modes. That is, how might the radar output wrong information given that it is doing what it's designed to do.

That said, any new information would be interesting and sometimes the line between a black-box and internal failure mode may not be as clear. For example, sometimes a sensor can be saturated by a strong input signal -- the internal components may be driven into their non-linear regions but there may be still a useful output (which just may not be as accurate as before). So the internal components haven't really failed and it's still sort of doing what it's supposed to do but at the same time the radar isn't really being "tricked" -- the sensor just cannot handle the strength of the input signal.

So... all ways that the sensor can fail are probably worth discussing. Thanks for your posts!