Adapting the Reconfigurable Spacecube Processing System for Multiple Mission Applications
Total Page:16
File Type:pdf, Size:1020Kb
Adapting the Reconfigurable SpaceCube Processing System for Multiple Mission Applications David Petrick, Daniel Espinosa, Robin Ripley, Gary Crum, Alessandro Geist, and Thomas Flatley NASA Goddard Space Flight Greenbelt, MD 20771 david.j . [email protected] Abstract-This paper highlights the methodology and slower computer currently used on ISS, as the main avionics effectiveness of adapting the reconfigurable SpaceCube system for two upcoming ISS experiment campaigns. This paper will to solve complex application requirements for a variety of show how we quickly reconfigured the SpaceCube system to space flight missions. SpaceCube is a reconfigurable, modular, meet the more stringent reliability requirements. compact, multi-processing platform for space flight applications demanding extreme processing power. The SpaceCube system is suitable for most mission applications, TABLE OF CONTENTS particularly those that are computationally and data intensive such as instrument science data processing. We will show how 1. INTRODUCTION ............................................... 1 the SpaceCube hybrid processing architecture is used to meet 2. HYBRID FLIGHT COMPUTING ......................... 2 data processing performance requirements that traditional 3. SPACE CUBE v1.0 DESCRIPTION ...................... flight processors cannot meet. 2 4. MISSION USE CASES........................................ 6 This paper discusses the flexible computational architecture of 5. CONCLUSIONS ............................................... 17 the SpaceCube system and its inherent advantages over other FUTURE WORK ............................................. 17 avionics systems. The SpaceCube v1.0 processing system 6. features two commercial Xilinx Virtex-4 FX60 Field REFERENCES •••••••••••••••••••••••••••••••••••••••••••••••••••• 17 Programmable Gate Arrays (FPGA), each with two embedded BIOGRAPHY ...................................................... 18 PowerPC405 processors. The FPGAs are mounted in an innovative back-to-back method, which reduces the size of the circuit board design while maintaining the added benefit of 1. INTRODUCTION two FPGAs. All SpaceCube v1.0 cards are 4" x 4", yielding a small, yet powerful hybrid computing system. The SpaceCube is a family of Field Programmable Gate Array architecture exploits the Xilinx FPGAs, PowerPCs, and (FPGA) based on-board science data processing systems necessary support peripherals to maximize system flexibility. Adding to the flexibility, the entire system is modular. Each developed at the NASA Goddard Space Flight Center card conforms to a custom mechanical standard that allows (GSFC) [1]. The goal of the SpaceCube program is to stacking multiple cards in the same box. provide lOx to lOOx improvements in on-board computing power while lowering relative power consumption and cost. This paper will detail the use of SpaceCube in multiple space SpaceCube is based on the Xilinx Virtex family of FPGAs, flight applications including the Hubble Space Telescope which include processor, FPGA and digital signal Servicing Mission 4 (HST-SM4), an International Space processing (DSP) resources. These processing elements are Station (ISS) radiation test bed experiment, and the main leveraged to produce a hybrid science data processing avionics subsystem for two separate ISS attached payloads. platform that accelerates the execution of science data Each mission has had varying degrees of data processing complexities, performance requirements, and external processing algorithms by distributing computational interfaces. We will show the methodology used to minimize the functions among the elements. This approach enables the changes required to the physical hardware, FPGA designs, implementation of complex on-board functions that were embedded software interfaces, and testing. previously limited to ground based systems, such as on board product generation, data reduction, calibration, This paper will summarize significant results as they apply to classification, event/feature detection, data mining and real each mission application. In the HST-SM4 application we time autonomous operations. The system is fully utilized the FPGA resources to accelerate portions of the image reconfigurable in flight, including data parameters, software processing algorithms more than 25 times faster than a and FPGA configuration, through either ground standard space processor in order to meet computational speed commanding file transfers or autonomously in response to requirements. For the ISS radiation on-orbit demonstration, the main goal is to show that we can rely on the commercial detected events/features in the instrument data stream. FPGAs and processors in a space environment. We describe our FPGA and processor radiation mitigation strategies that Background have resulted in our eight PowerPCs being available and error The SpaceCube processing system was started at GSFC in free for more than 99.99% of the time over the period of four 2006 with Internal Research and Development (IRAD) years. This positive data and proven reliability of the SpaceCube on ISS resulted in the Department of Defense program funding [2]. A series of internal prototype (DoD) selecting SpaceCube, which is replacing an older and demonstrations to NASA officials showcased the u.s. Government work not protected by U.S. copyright computational power and its inherent reconfigurable architectures are very hard to design and intrinsically advantages over typical space processors. NASA expensive to change such that they are portable to multiple recognized the clear potential of the technology, and missions, dynamic functional requirements, or new post provided the funding needed to increase the technology launch mission objectives or corrections. readiness level (TRL) for space flight applications. Specifically, the Hubble Space Telescope Servicing Mission A new approach is needed to meet the increasing challenges 4 management team infused SpaceCube as the main required by space processing systems. A hybrid computing avionics for an experimental payload called Relative system that combines multiple processors, reconfigurable Navigation Sensors (RNS) [3]. The use of SpaceCube FPGAs, flexible interface options, with a modular within the RNS system will be described in detail later in architecture is the solution that will bridge the gap between this paper. today's avionics requirements and yesterday's typical stand alone sequential processing architecture. A hybrid The version of the SpaceCube that was initially developed computing architecture is able to retain the function of a in the 2006-2009 timeframe is known as SpaceCube v1.0. multi-purpose computer that runs typical C&DH and Follow-on versions have been developed [1]; however the G&NC. However, in addition to these types of tasks, it has design and use of Space Cube v1.0 will be the focus of this the advantage of supporting computationally complex tasks paper. that require FPGA co-processors to handle math such as FFT, matrix manipulation, parallel floating point operations, 2. HYBRID FLIGHT COMPUTING or implementing an advanced interface such as CameraL ink, Spacewire, gigabit Ethernet, or support the implementation There is a growing need for higher performance processing of a custom interface. systems for space. Instrument precision and speed capabilities are rapidly evolving which levies tougher The modularity of such a system allows for the quick electrical interfacing and data bandwidth requirements on adaptation to changing avionics requirements. A modular the computing node of the system. In addition, on-board system, for example, can support adding a bulk memory processing of the data products, in some cases in real-time, card, a custom electrical interface, or expand the VO is now a common requirement. bandwidth required. A modular and reconfigurable system yields a high probability of using the same basic avionics On-board processing improves system efficiency and package for different mission applications, or follow-on functionality in two areas. First, by allowing the spacecraft missions, even if interface and computing requirements are to preprocess data products on board, a smaller or drastically different. compressed data volume per data set can be sent to ground, which increases the amount of time an instrument can be SpaceCube fits the need of a hybrid, reconfigurable, turned on and collecting data. It is typical for high data rate modular space processing system. This paper will show science instruments to constrain their data collection to 10- how cost and schedule can be reduced by reusing the same 20% of the mission time to fit within the limited downlink basic system for new missions. Reuse of hardware bandwidth. This problem continues to grow as instrument architecture greatly reduces the amount of up front Non capabilities increase. Second, it enables for applications on Recurring Engineering (NRE) costs and time associated board the spacecraft to make autonomous decisions on the with building a new system with new requirements from the processed data products. This ability opens up a much more ground up. challenging range of mission objectives that can be targeted for space applications. 3. SPACE CUBE v1.0 DESCRIPTION Typical space processing systems generally consist of a The SpaceCube v1.0 system is a compact, modular, low single radiation hardened processor such as the BAE power, reconfigurable multiprocessing platform for space RAD750, Aeroflex LEON3FT, BroadReach