Unmanned Systems Technology 022 | XOcean XO-450 l Radar systems l Space vehicles insight l Small Robot l BMPower FCPS l Prismatic HALE UAV l InterDrone 2018 show report l UpVision l Navigation systems

39 Radar systems | Focus about a third of the weight of the system, but moving to heat pipes and specialist substrates can reduce the weight. There are two variants, one with an integral fab and a lighter one at 730 g that needs fan cooling, giving designers the option to add their own types of cooling. The second generation, with customised components such as more efficient switched mode power supplies with the exact voltage rails for the array, GPU and FPGA, can produce a system weighing 500 g (1.1 lb). The metamaterials can be made from standard laminated RF printed circuit board (PCB) technology but using customised structures within the boards. That allows for high-volume production on existing manufacturing processes. While the phased array technology is initially being used in airborne designs, variants are being developed for driverless cars and trucks. The automotive designs have a more tightly specified requirement, allowing the radar to be optimised for high-definition radar imaging. RF CMOS Compared to existing 24 GHz radar used in airborne systems, the 76- 81 GHz millimetre-wave (mmWave) radar used for cars provides a wider bandwidth, which enables better range and velocity resolution. Previously, most mmWave chips on the market integrated only the radar transmitters and receivers, while developers had to add additional components such as the ADC and digital signal processor (DSP). This increased the cost and size of the radar system. By contrast, the latest single-chip mmWave sensor has two transmitters, four receive antennas and four-channel high-speed ADCs. It also comes with CAN bus, CAN-flexible data rate and UART data output interfaces. An ARM Cortex-R4F real-time microcontroller core and a DSP handle the data processing for collision sensing applications. The integrated DSP provides enough computational power to run signal- processing algorithms that calculate range, azimuth and velocity information for several objects, while the ARM core is responsible for configuring the chip and formatting the output data. At the system level, the sensor enables smaller UAVs to detect and monitor their surroundings. The processed information can then be sent to a flight controller to calculate and execute avoidance commands. The additional processing power in the radar sensor helps reduce the data rate and simplifies the interfacing comms between the sensors and the flight controller. That allows design engineers to decouple the sense-and-avoid tasks and distribute them across two processing units in the UAV system. Radar imaging Radar is essentially a 4D sensor, so it requires high resolution in all dimensions, including time. While a 4K resolution camera will produce an image of around 20 Mbytes (160 Mbits), a high-resolution radar image can be more than 100 Gbits. That creates a significant signal processing challenge at the sensor level and for the system architecture. Moving to high-definition imaging requires significant processing power in the sensor node, as it is not practical to move all the raw data to a central processing unit. Existing devices with four receive channels and three transmitters cannot provide such high resolution, so one system designer has developed an array of 48 receivers and 48 transmitters to deliver 2500 virtual channels. This compares to the 192 virtual channels possible using super-resolution techniques with the smaller arrays. This level of complexity requires new signal processing designs, with a time- multiplexed MIMO (multiple in, multiple out) antenna array on the PCB of the radar system that does not make use of metamaterials. The system architecture for the high- resolution radar imaging system has three chips: one for the transmitters, one for the receivers and one for the baseband, which can handle a total of 72 receive channels and 96 transmitters. This allows the radar’s front end to be scaled with additional receiver and transmitter chips with a single baseband processor. These are all built on Unmanned Systems Technology | October/November 2018 Integrating 77 GHz radar transceivers and processing into a single CMOS chip is bringing down the cost of developing radar systems for collision detection (Courtesy of Texas Instruments)