Industry’s First Space-Grade 20 nm FPGA Can Be Reprogrammed Mid-Orbit

May 20, 2020 by Gary Elinoff

No need to come back to earth for reconfiguration. Xilinx's new Kintex UltraScale FPGA can be reprogrammed mid-orbit.

Xilinx has announced what it terms the industry's first space-grade 20 nm FPGA intended for space applications—satellites, in particular. 


Xilinx Kintex UltraScale FPGA

The Kintex UltraScale XQRKU060 FPGA. Image (modified) used courtesy of Xilinx


Some of the unique features of this FPGA is its ability to be reconfigured mid-orbit and its capacity for machine learning in space.


The Hurdle of Prepping FPGAs for Space

Designing FPGAs for space is no easy task. According to David S. Lee with Sandia National Laboratories, ionizing radiation is the greatest threat to FPGAs operating in space. The effects of this radiation can include single-event latchup (SEL), single-event upset (SEU), single-event transient (SET), and total ionizing dose (TID).

Lee argues that in order to ensure long-time reliability of an SRAM-based FPGA in space, designers must prevent SEUs in the configuration memory. This can be accomplished by "scrubbing" (or identifying errors in the configuration and addressing them), error-correcting codes (ECC), and triple modular redundancy.


External scrubber of Xilinx's XQRKU060

External scrubber of Xilinx's XQRKU060. Image used courtesy of Xilinx (page 10)

These concerns about radiation are directly addressed in Xilinx's new space-bound 20 nm FPGA.


Key Specifications of the 20 nm FPGA

The radiation-tolerant Kintex UltraScale XQRKU060 FPGA brings machine learning (ML) to space with a portfolio of ML development tools that support frameworks such as TensorFlow and PyTorch. These frameworks may enable neural network inference acceleration for on-board processing in real-time.

The XQRKU060 includes 2,760 UltraScale DSP slices and provides up to 1.6 TeraMACs of signal processing compute. The unit’s 32 high-speed SerDes transceivers can run up to 12.5 Gbps each to deliver 400 Gbps of aggregate bandwidth. The device also delivers 5.7 tera operations per second.


RT Kintex UltraScale architecture.

RT Kintex UltraScale architecture. Image (modified) used courtesy of Xilinx


One of Xilinx's space systems architects Minal Sawant explains, “With our extensive history in developing leading-edge, radiation-tolerant technology and deploying this in reliable space-grade solutions, Xilinx continues its lead with the launch of the world’s most advanced process node for space.”


Re-Configurable in Orbit

The XQRKU060 is based on SRAM memory, which allows it to be reprogrammed not only during hardware development in the lab but also in-orbit after launch. Other space-grade, anti-fuse FPGAs can be configured only once, making prototyping and rendering verification costly and difficult. 

While other space-grade FPGAs are essentially frozen once deployed, the ability to re-configure the XQRKU060 any number of times in orbit offers operators great flexibility. New and improved communications standards, for example, can be uploaded at any time for improved system performance.

In the case of technology demonstrator satellites, multiple experiments can be run from a single payload. On-board re-programming may serve to reduce the size of the payload hardware. Launch times can also be moved up since the latest improved FPGA firmware can be deployed as soon as it’s ready.

Re-programming can be achieved while the XQRKU060 is active and in use.


Partial re-configuration

Partial re-configuration. Image used courtesy of Xilinx (page 11)


This is achieved by partial reconfiguration, which allows for the modification of a specific region within the FPGA without affecting applications running elsewhere within the device. 


Beamforming Technology

In a previous article on the Navy's interest in optical beamforming, we discussed how low-cost LEO satellites may be increasingly replacing multi-million dollar geosynchronous spacecraft. These space vehicles must have (at their core) the ability to achieve beamforming—the ability to modify the direction of their radio signal to adapt to changing conditions.

The XQRKU060’s on-board processing capacity allows satellite operators to effect beamforming. Further, the devices' reconfigurability allows it to adapt to changing modulation and carrier specifications.


Improvements over Previous Generations

The 20 nm XQRKU060 builds on Xilinx's long tradition of space-targeted, radiation-hardened FPGAs, including the 65 nm Virtex-5Q and the 90 nm Virtex-4QV.


Three generations of Xilinx Space grade FPGAs

Three generations of Xilinx Space grade FPGAs. Image used courtesy of Xilinx (page 9)


Compared to the previous generation Virtex-5QV, the XQRKU060 is said to feature a five-fold increase in logic cells. The older FPGA offers 18 x 3.125, or 56.25 of aggregate bandwidth while the new 20 nm device includes 400 Gbps of aggregate bandwidth.


Support Tools for the New FPGA

Xilinx offers a rich set of design aids to speed up the development process. 

The Vivado Design Suite provides a simplified development environment for the XQRKU060. The system serves to eliminate routing congestion, which allows for greater than 90 percent of the device to be used without any performance degradation. 

The ADA-SDEV-KIT2 is a development kit for the XQRKU060. It includes LPC FMC and HPC FMC+ interfaces and space-grade power and temperature sensing solutions.



ADA-SDEV-KIT2. Image used courtesy of Xilinx

The Vitis Unified Software Platform features a core development kit to enable users to more easily build applications. It includes a set of hardware-accelerated open-source libraries aimed at Xilinx hardware platforms.


A Game-Changer in Satellite Technology?

Satellite systems are rapidly evolving from huge, solitary, expensive, and inflexible leviathans to small, expendable swarms of highly flexible team players. Despite their low cost and small size, the ability to change parameters often, and in mid-mission, is an essential feature of Space 2.0. The XQRKU060 aims to provide the reprogram-ability, radiation resistance, and computational power to make it possible.