Xilinx has launched the radiation resistant KINTEX ultrascale programmable chip, which can be used for satellites and other space hardware.
These types of chips, which can withstand harsh radiation in space, may be the brains of orbital space applications such as satellite payloads, and have 10 times better digital signal processing performance than previous versions, salings said. This means that the chip can get data from the sensor and process it effectively.
Xilinx designs field programmable gate array (FPGA) or chips, which can be programmed or reprogrammed after they are put into hardware. The kinetex ultrascale FPGA chip is the first to use a 20 nanometer manufacturing process (nanometer is one billionth of a meter), replacing the older 65 nanometer process chip previously used.
Xqrku060 chip also brings high performance machine learning (ML) into space for the first time. Rich product portfolio of ML development tools, support industry standard frameworks including tensorflow and pytorch, and accelerate real-time airborne processing in space for neural network reasoning through a complete “processing and analysis” solution.
The chip has intensive, energy-efficient computing power, scalable precision and large capacity of on-chip memory. It provides 5.7 Tera operations per second (top) for deep learning optimized peak int8 performance. That’s 25 times more than the previous generation. However, the amount of DSP computation for processing intensive algorithms and analysis is increased by more than 10 times
This allows it to do more processing and significantly reduce size, weight and power consumption, which are all key factors when designing chips for use in harsh space environments where physical space is extremely important. Minal sawant, Xilinx’s systems architect and space product manager, said in a statement that the technology was “ideal for high bandwidth payloads, space exploration and research missions.”.
Xilinx said xqrku060 is the only on orbit reconfigurable chip in the industry. This means that it can be placed in a piece of hardware space and reprogrammed to perform other operations.
In orbit reconfiguration, as well as real-time airborne processing and ml acceleration, enables satellites to update in real time, provide video on demand, and perform calculations “in real time” to handle complex algorithms. According to the company, machine learning can solve various problems such as scientific analysis, object detection and image classification (such as cloud detection).