Session
Pre-Conference Workshop Session 2: Next on the Pad - Research & Academia
Location
Utah State University, Logan, UT
Abstract
An arms race in the last decade between computing performance and power-reduction has led to accelerated progress in both hardware and software, provisioning fast, efficient algorithms and small, capable, low-power hardware devices. Meanwhile, small satellites have become increasingly important in defense and commercial missions, due to their low resource consumption in terms of power, space, and cost. Improvements in computer vision algorithms and commercial off-the-shelf (COTS) edge computing introduce a means of addressing challenges posed for spacecraft in the realm of autonomy and perception. The University of Georgia’s Small Satellite Research Laboratory has been working to address these obstacles in developing an interface to merge Graphics Processing Unit (GPU) computations into the standard PC104+ satellite stack. This paper presents the Core GPU Interface (CORGI), a hardware solution which integrates the NVIDIA Jetson TX2/TX2i module into the cube satellite stack. The CORGI can be used as a standalone flight computer or as a co-processor with a designated onboard computer (OBC). This board will fly on the Multi-view Onboard Computational Imager (MOCI), a 6U satellite scheduled for launch into low earth orbit (LEO) in 2022, and will serve NASA and the Air Force Research Laboratories’ (AFRL) efforts to demonstrate autonomy and high-performance computing on small satellites. The CORGI provides input/output capability for command and telemetry and development interfaces for increased usability, specifically DisplayPort and two USB 3.0 type-A interfaces. Additionally, a UART umbilical provides an interface between an off-PCB computer through the PC104+ stack, enabling the TX2i to be externally triggered for science data handoff by a radiation-tolerant onboard computer. The CORGI is designed as a payload processor, and thus utilizes the TX2i module’s USB 3.0 hub to connect one or more satellite imagers over USB. The CORGI provides other standard development access points, including general-purpose input-output (GPIOs), from both the designated onboard computer and TX2i module. Standard headers also provide a means of testing the TX2i line voltages for discharge monitoring and power status. An SD card provides flash memory for logging science data and telemetry. This paper presents 1) PCB design and manufacturing specifications to assist teams in designing hardware for space applications. 2) Spice simulations to validate power management circuit design and evaluation for relevant cases. 3) Electrical test results demonstrate the nominal operation of both power management and regulation circuitry and the validation of correct discharge behavior during a power outage. 4) We present tests that demonstrate CORGI’s operation and performance characteristics under load.
Design Simulation and Testing of a Custom CoProcessor for Cubesatellites in LEO
Utah State University, Logan, UT
An arms race in the last decade between computing performance and power-reduction has led to accelerated progress in both hardware and software, provisioning fast, efficient algorithms and small, capable, low-power hardware devices. Meanwhile, small satellites have become increasingly important in defense and commercial missions, due to their low resource consumption in terms of power, space, and cost. Improvements in computer vision algorithms and commercial off-the-shelf (COTS) edge computing introduce a means of addressing challenges posed for spacecraft in the realm of autonomy and perception. The University of Georgia’s Small Satellite Research Laboratory has been working to address these obstacles in developing an interface to merge Graphics Processing Unit (GPU) computations into the standard PC104+ satellite stack. This paper presents the Core GPU Interface (CORGI), a hardware solution which integrates the NVIDIA Jetson TX2/TX2i module into the cube satellite stack. The CORGI can be used as a standalone flight computer or as a co-processor with a designated onboard computer (OBC). This board will fly on the Multi-view Onboard Computational Imager (MOCI), a 6U satellite scheduled for launch into low earth orbit (LEO) in 2022, and will serve NASA and the Air Force Research Laboratories’ (AFRL) efforts to demonstrate autonomy and high-performance computing on small satellites. The CORGI provides input/output capability for command and telemetry and development interfaces for increased usability, specifically DisplayPort and two USB 3.0 type-A interfaces. Additionally, a UART umbilical provides an interface between an off-PCB computer through the PC104+ stack, enabling the TX2i to be externally triggered for science data handoff by a radiation-tolerant onboard computer. The CORGI is designed as a payload processor, and thus utilizes the TX2i module’s USB 3.0 hub to connect one or more satellite imagers over USB. The CORGI provides other standard development access points, including general-purpose input-output (GPIOs), from both the designated onboard computer and TX2i module. Standard headers also provide a means of testing the TX2i line voltages for discharge monitoring and power status. An SD card provides flash memory for logging science data and telemetry. This paper presents 1) PCB design and manufacturing specifications to assist teams in designing hardware for space applications. 2) Spice simulations to validate power management circuit design and evaluation for relevant cases. 3) Electrical test results demonstrate the nominal operation of both power management and regulation circuitry and the validation of correct discharge behavior during a power outage. 4) We present tests that demonstrate CORGI’s operation and performance characteristics under load.
