Insight into the radar trends in 2024: Will the long-term integration lead to separation? Talk about NXP's new generation of RFCOMS radar SAF86xx

NXP's software-defined automotive platform will add a new family member in 2024: RFCOMS radar SAF86xx. This radar product was unveiled at the CES exhibition in Las Vegas not long ago. According to the explanations of Zhai Xiaoshu, marketing director of NXP's Greater China Automotive Electronics Division, and Yang Chang, ADAS product marketing manager of NXP's Greater China Automotive Electronics Division, we will learn about the new radar sensor network and forward-looking distributed aperture radar technology in 2024, as well as the emergence and characteristics of the new product SAF86xx.

Compared with the previous edge computing architecture, that is, the radar RFCMOS completes the complete radar signal processing, NXP's newly proposed distributed radar sensor network does not require complete signal processing in the radar RFCMOS. Shouldn't it save more CPU work if the complete signal processing is done? Why do we have to separate the "live"? The main reason is that after the complete processing, some details will be lost, and the signal given to the CPU will be incomplete. When the CPU fuses the signal based on the lost details, the overall judgment will be incomplete.



Under this "divided activity" architecture, the front-end perception and back-end processing of the radar are physically deconstructed. The front-end radar sensor will become simpler, further improving the RF performance. The back-end data processing part is more convenient for centralized management and tends to be concentrated on the middle ADAS domain controller or an independent radar domain controller, which can achieve more powerful functions than each operating independently. The software can better achieve unified management, as well as later algorithm and function upgrades. For example, a car may usually need 2 long-range radars and 4 corner radars. Then, it can be connected to the central processor through 6 Gigabit Ethernet cables for centralized processing, which is very convenient to manage. The input of the same basic radar signal can achieve more intelligent and more accurate judgment. Therefore, letting the underlying radar signal details be transmitted more instead of excessive processing, and finally concentrating the signal processing on a powerful central processor is a solution to improve sensor capabilities while reducing costs.





Let's take an example of a relatively good technology that has been integrated under this distributed architecture: distributed aperture radar, which is equivalent to a larger aperture antenna of multiple radar sensors at different locations, improving resolution, clarifying detection targets, and achieving better radar detection effects. With this technical foundation, it is possible to further improve the overall resolution of the vehicle radar system, achieve a longer detection distance, and achieve better radar detection results. The comparison between the standard radar and the cooperative radar shown in the figure below clearly shows the improvement in radar detection performance.

The reporter observed that in order to realize the distributed new architecture radar network, the new generation radar chip SAF86xx came into being. As can be seen from the block diagram, the chip contains 4 transmitters, 4 receivers, ADC conversion, phase rotator, low phase noise VCO, BBE32EP DSP, Arm® Cortex®–M7 core and SRAM, and retains a simple MPU, which can output pre-processed radar data, and then transmit it through Gigabit Ethernet for subsequent processors to use.


Figure: SAF86xx block diagram

The figure below shows the evolution of NXP from the second generation RFCMOS radar to the current third generation RFCMOS radar. In November 2022, NXP launched the second generation RFCMOS radar chip TEF82xx series. This chip can achieve 300 meters of ranging and an angular resolution of less than 1 degree. It is mainly used for dual cascade and quad cascade, and can be used with the industry's first 16nm MCU S32R29. This chipset solution has been used and mass-produced by component manufacturers on the market.

This year's third-generation RFCMOS radar SAF86xx series has twice the RF link performance improvement compared to the previous generation. On the one hand, the previous generation used a 40nm RFCMOS design process, and this generation was upgraded to 28nm. It is worth noting that there are very great technical difficulties in implementing advanced processes in high-frequency RF links. The transmitting antenna and receiving antenna have four transmitting links and four receiving links, which can realize more virtual receiving links, equivalent to 4×4 up to 16 virtual channel transmission.


In addition to the performance of the chip device itself, the lower right corner of the above figure also shows that it supports the design of 3D waveguide antennas. Compared with traditional patch antennas, 3D waveguide antennas have a 9dB RF improvement, which can not only improve the perception of objects, but also reduce the size of the radar. It can be seen that this packaged transmitter directly introduces a 3D waveguide antenna through a lower-cost FR4 PCB board to realize a 3D transmitting array and receiving array. Traditional patch antennas are all planar, while 3D antennas can bring better transmission and reception performance while saving PCB area and cost.

The figure above shows the performance difference between the traditional PCB patch antenna and the 3D waveguide antenna. The yellow part is the detection probability of the radar. Based on the detection probability diagram of the PCB antenna, the maximum detection distance for a 5dBsm object (such as a motorcycle) is 200m. For a motorcycle that is more than 200m away, the radar target will appear to be constantly flashing and jumping, and it is uncertain whether the radar target really exists. The 3D waveguide antenna can easily sense a 5dBsm object (such as a motorcycle) at a distance of more than 300 meters, and can know the target more accurately, and give warnings or slow down in time. 3D waveguide antenna technology can significantly improve the confidence of the radar and ensure driving safety in a more comprehensive way.

In addition, Yang Chang, ADAS product marketing manager of NXP Greater China Automotive Electronics Division, explained in detail the other characteristics and parameters of this chip: power output, noise figure, phase noise, ADC sampling rate, chirp bandwidth, and 7-bit phase rotator. Theoretically, the two parameters related to the maximum detection distance are power output and noise figure. The power output of this chip can reach 15 dBm, and the noise figure is 10.5 dB, achieving a maximum detection distance of 300 meters and above. The phase noise is -96 dBc/Hz, which can better distinguish objects from some objects with relatively small reflection values, such as some two-wheeled vehicles, or two cars that are relatively close. The ADC sampling rate reaches 80 MS/s orthogonal sampling, and 40 MS/s can be used in a single chip. The chirp bandwidth is as high as 5GHz. The chirp bandwidth is related to the distance resolution. At present, many radars on the market use a bandwidth of less than 1GHz, but devices with large bandwidth still have their necessity. It is expected that in the future, large bandwidth will be very necessary to be applied in situations such as parking and low speed. The last one is the 7-bit phase shifter (phase rotator). Now many RF chips have begun to add 6 or 7-bit phase shifters. It cooperates with the RF transmission link to realize more complex MIMO waveform design, allowing millimeter-wave radar to take into account both long distance and high resolution, which is difficult to achieve on traditional radars. However, through the combination of phase shifter technology and transmission link, long distance and high resolution are achieved. Of course, the consumption it brings has higher requirements for back-end computing power.

Multiple new SAF86xx, combined with NXP's S32R series MPU, can build a complete radar network. The advantages of distributed radar sensor networks make it easier to realize some new functions, expansions and upgrades, such as 360° sensor fusion, collaborative sensing, artificial intelligence processing, and the realization of hosted software features. There are different winds and

clouds in the separation and combination. It is estimated that the winds and clouds of the distributed architecture of radar sensors have already begun in 2024, and there is also hope in the future.