The dream in the automotive industry is fully autonomous driving, even in poor weather conditions. This requires synchronization of multiple sensor systems, as well as real-time processing of data from these systems. Current radar systems operating at ~24 GHz (K-band) are prevalent in new vehicles with ADAS systems, and there will be an inevitable shift to state-of-the-art 76-81 GHz radar systems to support autonomous vehicles.
These systems already present EMC challenges that must be addressed at the board level, intra and inter-vehicle levels, and signal processing level. Given the EMC and EMI challenges that arise within a single radar module, between radar modules, and between multiple vehicles, how can systems designers ensure multiple nearby vehicles can be sufficiently resolved by an autonomous vehicle? Here, we’ll take a look at the current EMC challenges that arise in vehicles with FMCW radar systems and some future directions systems and board designers can take to overcome these challenges.
Moving to the 76-81 GHz range provides a number of benefits compared to K-band radar. These include broader available bandwidth, lower attenuation, more focused beams, and longer range. As is well known, moving to higher frequencies increases on-board crosstalk and radiated EMI. As newer vehicles will use multiple module for long-range and short-range radar, as well as wireless systems at other frequencies for V2X communication, there are some particular design challenges that systems designers should consider to ensure proper target detection. Particular EMC problems become more apparent as multiple radar-equipped vehicles occupy roadways.
This is primarily an board-level challenge that requires proper stackup design, layout, routing, and grounding techniques. In newer commercially-available 24 GHz and ~76 GHz radar modules, you’ll find that the digital and RF sections are sometimes separated into different boards within a single module. This requires proper grounding, routing, and layout to prevent EMI between analog and digital portions of the module.
Separating digital and analog sections into different boards is a natural choice as it does not significantly increase the footprint while it effectively allows you to have two stacked signal layers (analog and digital) on different boards. Some current radar modules use coplanar waveguide routing on the surface layer to the Tx and Rx series-fed patch antenna (SFPA) arrays to provide sufficient high frequency isolation from other sections of the board.
This is a major EMC challenge in automotive radar design, which arises when multiple vehicles with radar operating in a similar frequency range interfere with each other. Consider a situation with two radar-equipped vehicles (Car A and Car B) operating in the same frequency band. The radar signal from Car B can be received by Car A, which masks detection of any true targets that would otherwise be detectable by Car A. Car B’s radar will also cause multiple false targets to be seen by Car A. Car A will then have the same effect on Car B. The logic here can be extended up to any number of vehicles. A recent extensive study by the NHTSA  looked at the effects of radar congestion in multiple scenarios, and the results provide significant insight into the effects of interference between radar modules on different vehicles. These results are also extremely useful for newer ADAS designers as they provide some important examples of target tracking in different situations.
Automotive radar EMC carries several test and measurement challenges.
This particular EMC challenge is difficult to solve at the PCB level, but it has been addressed at the signal processing level. This is the same EMC challenge that occurs with K-band radar emitters used by DOT to monitor real-time traffic flow in many major cities. Civilian radar detectors overcome this interference problem with traffic sensor rejection (TSR) filtering. With multiple radar-equipped vehicles, this same type of interference problem between radar modules can cause false alarms if not addressed.
The inelegant TSR filtering solution used in radar detectors was to delay the alert time of the K-band radar sensor by more than the radar burst duration; the sensor would only alert the driver if a longer radar burst was detected. This works fine when you know the duration of transmitter’s radar (approximately 1.5 s for real-time traffic sensors), but it becomes ineffective when you have multiple vehicles transmitting radar pulses at unpredictable times. The current solution involves placing an anti-aliasing filter on the Rx side of the signal chain before the signal is sampled with an ADC.
Once false targets from other interfering radar systems are suppressed, the true targets need to be sufficiently resolved. This is primarily a pulse design and signal processing challenge. In terms of pulse design, a designer needs to try and maximize the detection resolution while balancing maximum unambiguous range and velocity. Determining heading is a simple matter of beamforming with an antenna array and just requires tracking the angle at which the beam was emitted, thus it will not be discussed in more depth here.
A radar frame emits multiple FMCW pulses, typically with linear chirp. The beat frequency between the transmitted and received pulses tells you the position of the downrange target, and changes in the beat frequency over time will tell you the velocity. Note that, without velocity detection, the motion of a vehicle will distort the range estimation due to Doppler shift, so proper range estimation requires a corresponding velocity estimation.
It is possible, both with short and long-range radar, that multiple vehicles interact with the emitted beam and produce an echo within the duration of a single chirp in the radar frame. When multiple targets are present, multiple echoes from each target need to be distinguished with some signal processing steps. While position could be distinguished from time-of-flight measurements, followed by calculating the time difference between multiple echoes to determine relative velocity, a more elegant (and accurate) solution involves applying orthogonal fast Fourier transforms to each received chirped echo in a frame. This gives you a 2D grid of range vs. Doppler shift measurements for each vehicle, which can then be converted into a range vs. velocity grid. Although some processing still needs to be done to eliminate the interference for multiple vehicles.
Finally, the signal can be cleaned up with a constant false alarm rate (CFAR) algorithm , and peaks in a velocity vs. range vs. intensity graph can be detected using a standard peak detection algorithm. An example showing the results provided by CFAR for distinguishing two vehicles is shown in Figure 1.
Fig. 1: Resolving the range and velocity of true targets with a CFAR algorithm .
Note that, without CFAR processing and peak detection, it may appear that there are actually four vehicles, as can be seen from the four intersections in the left graph. This important signal processing algorithm produces the traces seen on the right graph, telling us the range and velocity.
When these signal processing steps are taken together, we now have a system that is robust against interference from other vehicles. The remaining consideration for systems designers requires selecting the appropriate components that are robust against some common signal integrity problems in RF systems.
5G designers are already familiar with passive intermodulation, and the same effect can occur in amplifier stages used in RF signal chains for frequency modulated signals. The third-order intermodulation products are the most important as they lie closest to the FMCW bandwidth, although other odd-order products will be present and can lead to interference in wideband signal chains. If you aren’t using an SoC with an integrated amplifier, or you are designing your own amplifier as part of a signal chain, you’ll need to suppress intermodulation products created on the Tx side from interfering with the Rx side.
As higher frequency radar systems use relatively wideband amplifiers and ADCs, intermodulation products can propagate into the Rx side as crosstalk and reduce the usable range in a radar module system. This is particularly problematic with power amplifiers on the Tx side, which typically run near saturation. This is another problem that should be addressed at the board level and circuit level. A higher order filter on both sides can be used for intermodulation product suppression, while eliminating interference requires sufficient isolation between the Tx and Rx sides of the RF signal chain.
Although we already looked at the signal processing and layout challenges in automotive radar systems and modules, we still haven’t considered EMI from the rest of the vehicle. Strong electromagnetic fields in electric cars leads to EMI that increases the noise figure in an FMCW radar module, which can reduce the maximum available range based on the required SNR value.
Greater innovation in automotive radar that provides higher spatiotemporal resolution with a large number of Tx/Rx modules in new vehicles requires suppressing EMI or avoiding it altogether. 5D radar imaging is likely to become the state-the-art method for gathering 3D images of nearby vehicles with high spatial resolution, as well as Doppler and time tracking. This technique is already well-known among the medical community for gathering spatially resolved hyperspectral images. These images then need to be gathered within some solid angle (i.e., angular resolution) and then reconstructed extremely quickly.
Fig. 2: Example results from 5D radar imaging. Note that these images are actually dynamic; they change over time as a target is tracked .
At the radar module and ECU levels, all signals required for MIMO radar must be phase and frequency synchronized so that a coherent beam can be constructed for angular resolution with a phased array antenna. SoCs and embedded signal processing within radar modules aid in synchronization for beamforming, but synchronization among multiple radar modules is already critical for preventing interference between modules on a single vehicle. In the process of locating and tracking vehicles in the nearby environment, a significant amount of data gets generated and must be sent to an ECU. Sending these data throughout a vehicle with coaxial or twisted pair cabling incurs losses and makes them susceptible to high frequency EMI from other systems. One can expect that data sent throughout a vehicle will reach gigabit levels as more ADAS modules are placed in new vehicles.
Much like the telecommunications industry has done, this motivates a move to fiber for vehicular radar systems and for moving data around a vehicle. Using fiber with radar systems is a natural choice (i.e., radio over fiber) to eliminate susceptibility to EMI within a vehicle. Note that fiber does not have to be limited for signal routing in a radar system; fiber can be used for in-vehicle networking as it provides higher bandwidth and data rates while eliminating interference problems that will continue to arise in new vehicles. Radar transceiver module proofs-of-concept that are built with COTS components are already being reported . The schematic below shows a module built on the AWR1243 radar transceiver module from Texas Instruments.
Fig. 3: Proof-of-concept for a microwave photonic radar module for use in the 76-81 GHz band. The total link gain for the electrical signal from the master output to the RC input is zero, including electrical-optical and optical-electrical conversion, and all pre and post amplifiers. LDC: laser diode current controller, LD: laser diode, TEC: temperature controller, µC: microcontroller, PD: photo diode, EDFA: Erbium-doped fiber amplifier, RC: radar chip .
In addition to the AWR1243 transceiver IC, this proof-of-concept module uses COTS components and can be easily integrated into a small radar module. The use of photonic circuits eliminates the intermodulation problems and many mixed signal interference problems noted above. If there is a shift to microwave photonic systems at the radar module level, expect this to coincide with greater use of fiber for in-vehicle networking.
This particular EMI/EMC challenge is currently being solved primarily with signal processing, appropriate component selection, and layout to produce accurate angle-resolved range and velocity estimates for nearby vehicles. The move to fiber and further advancements in microwave photonics-based radar systems represents a significant opportunity for innovation in this space. Obviously, any new mmWave system must pass EMC checks before it hits the market, and generalized in-lab test procedures are being developed for evaluating high frequency automotive radar systems . The next advance in automotive radar my come in the form of partially coherent radar systems . Prepare to see more innovation in the design, processing, and testing arenas as these systems proliferate new vehicles.
If you’re looking for a knowledgable firm that offers cutting-edge technology research services and thought leadership marketing services for electronics companies, contact NWES for a consultation.
 U.S. Department of Transportation, National Highway Traffic Safety Administration, Radar Congestion Study. September 2018. [Link]
 Dr. Rivas-Torres Wilfredo, Automotive Radar Fast Chirp System Analysis. Keysight Webinar, January 2019.
 University of South Alabama, BioImaging and BioSystems Center, “5D Imaging.” Retrieved September 2019. [Link]
 Preussler, Stefan, et al., Photonically synchronized large aperture radar for autonomous driving. Optics Express 27.2 (2019): pp. 1199-1207. [Link]
 Lei, Jianmei, et al., Automotive mmWave Radar EMC Test Developments and Challenges. 2019 13th European Conference on Antennas and Propagation (EuCAP), pp. 1-6. IEEE, 2019. [Link]
 Komissarov, Rony, Vitali Kozlov, Dmitry Filonov, and Pavel Ginzburg, Partially coherent radar unties range resolution from bandwidth limitations. Nature communications 10, no. 1 (2019): p. 1423. [Link]