Satellite-Based, Remote-Sensing Power Aware Computing Application: Satellite-Based, Remote-Sensing Power Aware Computing Application Patrick Shriver, presenter LANL team members Maya Gokhale, Project Lead Scott Briles Michael Cai Kevin McCabe Patrick Shriver USC/ISI team members Steve Crago Dong-In Kang Jinwoo Suh DARPA PAC/C PI Meeting Nov. 27 – 28, 2001


Presentation Outline: Presentation Outline Remote-Sensing Satellite Application FORTÉ satellite mission Ionospheric-dispersed lightning signals Detection and data compression problems Power-Aware, Multiprocessor Solution Power management through algorithm modulation Algorithm Power Experiment Power profiling on JPL’s test bed Test Results Satellite Power Systems System-Level design Trends Current/Future Work


Slide3: Remote-Sensing Satellite Application FORTÉ (Fast On-orbit Recording of Transient Events) Conceptualized, computer-generated graphic


Slide4: FORTÉ Satellite DoE-funded, joint LANL/SNL project Launched August 29, 1997 nearly circular 825 km orbit 7 ft tall, 470 lbs (wet), all-composite structure Payload instrument suite to study optical and RF lightning events study global lightning climatology monitor severe storm activity investigate ionosphere dispersion effects on RF signals application chosen for power-aware, signal processing scheme Remote-Sensing Satellite Application


Slide5: An event creates a wide-band impulsive VHF signal. The signal traverses Earth’s ionosphere. The signal is dispersed (low frequencies are delayed). What is the amount of dispersion? What would be the arrival time if there was no ionosphere? Lightning Signals in the Earth’s Atmosphere Remote-Sensing Satellite Application


Slide6: Remote-Sensing Satellite Application The figures above depict an RF signal dispersed by the ionosphere. The top graph is an illustration in the time domain of the “chirp” signal features. The bottom graph is an illustration of the dispersed signal frequencies vs. the corresponding time-of-arrivals.


Slide7: Remote-Sensing Satellite Application Chirp Waveform RESULT Detection Decision Data Arrays Sub-band Filter’s Center Frequencies {fc} and Time of Detection for Each Sub-band {tod} Chirp Signal Detection Analog Trigger time-frequency domain detection individual filter threshold N of M filters must exceed threshold


Slide8: Remote-Sensing Satellite Application PROBABILITY OF DETECTION versus PROBABILITY OF FALSE ALARM (FALSE ALARM RATE) Operating Curve of Detector (e.g., Analog Trigger box) is Predetermined. Changes in threshold setting allow for ONLY movement along operating curve. For Better Detection we must accept more False Alarms Thresholds Settings Detection Performance


Slide9: The probability of detection is set for the entire system by the first-stage, analog trigger box. Once thresholds are set in the trigger box, probability of detection can not be improved by post processing. However, post processing can reduce the probability of false alarms. Thus, it is desired that the system run at the greatest probability of false alarm that can be permitted by other limiting facts, such as Time available to process data from each trigger-aware event Power available for processing ANALOG TRIGGER FIXES THE PROBABILITY OF DETECTION. POST-TRIGGER PROCESSING CAN REDUCE FALSE ALARMS. Post-Processing Remote-Sensing Satellite Application


Slide10: Remote-Sensing Satellite Application Remote-Sensing Detection Problem Dynamic Environment noise/clutter interference atmospheric distortion varying event rates Detection Triggering threshold settings probabilities of false alarms vs. detection relationship Peak Detection Performance threshold settings fix probabilities of false alarms vs. detection post-processing techniques reduce false alarms on-board, satellite power resources are not always available for post-processing techniques Desired to have a system that is aware of on-board power resources which can adapt on-orbit


Slide11: Remote-Sensing Satellite Application Data Compression Problem Bandwidth Limitation higher gains transmit more data but consume more power lower gains transmit less data but also consume less power limits quantity of information to the end-user without on-board processing, quality of the data can be poor Data Compression reduce the quantity of information improve the quality of data before transmission to end-user on-board, satellite power resources are not always available for post-processing techniques Desired to have a system that is aware of on-board power resources which can adapt on-orbit


Slide12: Desired Existing Remote-Sensing Satellite Application ADC Analog Trigger Ring Buffer Post- Detection Processing Unit Chirp Signal Signal Processing Data Flow


Slide13: Power-Aware, Multiprocessor Solution Analog Trigger Sub-band Filter frequencies and time of arrivals Digitized Waveform Signal Processing Algorithms Chirp Signal


Slide14: Power-Aware, Multiprocessor Solution Estimate total electron content (tec) and TOA using Least Mean Squares (LMS) Fit and trigger box report Reduce influence of erroneous triggers Maximum-likelihood fit Calculate Spectrogram on digitized waveform data multiple short FFTs. Making a more refined trigger box through software Compute Bank of Matched Filters on digitized waveform data Searching for closest correlation to known parameters Vary number of processors based on power availability Compute Adaptive Match Filter on digitized waveform data Iterative processing to find filter with the best match. Non-Deterministic Computation Increasing Accuracy Decreasing Prob. of False Alarm


Power-Aware, Multiprocessor Solution: Power-Aware, Multiprocessor Solution Power Aware Control Node 0 Inter- Connection Network ith event (i+1)th event (i+2)th event Node 1 Node 2 Function Control Vector LMS ML ST MF AF Hand- shaking/ Control Vector Data Bus to All Nodes Trigger Node 3 Key Concept: Choose signal processing algorithm based on power availability; accuracy of event detection proportional to amount of available power PAMA-based Parallel Processing Ring Buffer


Algorithm Power Experiment: Algorithm Power Experiment JPL Test bed Wind River PPC750 266MHz processor VxWorks operating system Tektronix TDS 7104 Digital Phosphor Oscilloscope provides current vs. time measurements


Algorithm Power Experiment: Algorithm Power Experiment Software FFT Trigger Experiment Testing Least-Mean Squares (LMS) Maximum Likelihood (ML) Software FFT Trigger (ST) Matched Filter (MF)


Algorithm Power Experiment: Algorithm Power Experiment Dynamic Energy Range: 100,000 Power Test Results


Slide19: Power Spacecraft Experiment Payload Spacecraft Bus Sensors * Attitude Determination & Control System ** Command & Data Handling generation devices storage components control & distribution unit Satellite Power Systems


Satellite Power Systems: Satellite Power Systems Power subsystem is critical satellite size failure of power system results in loss of the mission power generation, utilization, and dissipation issues Power environment is dynamic generation devices (solar arrays, RTGs) degrade over time storage devices (batteries) have limits on charge capacity & recharge cycles orbit characteristics Power control and distribution system design power budgets are strictly allocated power is distributed in a digital manner design for worst-case scenarios


Slide21: Satellite Power Systems Current & Future Trends in Satellites “Faster, better, cheaper” -Dan Goldin, former director of NASA Increase in capability (better) more on-board computational processing increase in power Reduction in cost (cheaper) smaller size decrease in power ? improved power-efficient devices smarter power management


Slide22: Satellite Power Systems Power-Aware, Remote-Sensing Satellite Application Benefits Enabled on-board processing increase capability of satellites Increased detection accuracy post-processing decreases probability of false alarms Enhanced power aware computing power management through multiprocessor algorithm modulation Developed adaptive spacecraft power distribution smarter power management


Current/Future Work: Current/Future Work Current Status Power Aware Computing Application FORTÉ satellite mission power management through multiprocessor algorithm modulation solves detection and data compression problems Signal Processing Algorithms coded 4 routines (LMS, ML, ST, MF) 6 order of magnitude algorithm energy spectrum


Current/Future Work: Current/Future Work Near-Term Tasks Adaptive Filter creation and testing integration with the other signal processing routines Main (Control) Routine utilize Linux MPI library calls determine function control vector decisions demonstration with 4 processor nodes


Current/Future Work: Current/Future Work Longer-Term Tasks RAD750 Power Performance Testing Advanced Signal Processing sensor fusion wavelet transformation adaptive clutter rejection Next-Generation FORTÉ Systems conventional power budget allocations fixed power consumption with varying event rates Spacecraft Power Management Scheme power/thermal system-level simulations gradient power management


PAC/C Satellite ApplicationPaper: PAC/C Satellite Application Paper Entitled “A Power-Aware, Satellite-Based Parallel Signal Processing Scheme” Author list consists of the LANL & USC/ISI team members Published in Power Aware Computing book edited by Robert Graybill and Rami Melhem Kluwer Academic Publishers Summarizes work performed thus far for PAC/C phase 1 Describes remote-sensing application Discusses detection problem Presents power test results


PAC/C Satellite ApplicationPaper: PAC/C Satellite Application Paper References: FORTÉ mission and operations “The FORTÉ receiver and sub-band triggering unit,” by Enemark and Shipley, 8th Annual AIAA/Utah State University Conference on Small Satellites, 1994. “Four years of operations and results with FORTÉ,” by Roussel-Dupre et. al., AIAA, 2001. http://www.ngs.noaa.gov/GRD/GPS/Projects/TEC Signal Processing and Computing Fundamentals of statistical signal processing: Estimation theory by Kay, Prentice Hall, 1993. Discrete-Time Signal Processing by Oppenheim and Schafer, Prentice Hall, 1989. Numerical Recipes in C by Press et. al., Cambridge University Press, 1992. Spacecraft Systems Space Mission Analysis and Design by Larson and Wertz, Kluwer Academic Press, 1992.