James P. Lux

Multiprocessor digital signal processing on Earth orbiting scatterometers

The implementation of a Multi Digital Signal Processor for radar return analysis on a Ku-Band Earth orbiting scatterometer is discussed. Historically, radar signal processing on scatterometers has been implemented with discrete components, Field Programmable Gate Arrays (FPGA)and Application Specific Integrated Circuits (ASIC). These methods are expensive due to long development times, expensive tools, and their lack of modularity. The system presented in this paper uses a radiation tolerant, space qualified version of a commercial general purpose DSP (ADSP-21020) to perform the radar signal processing functions. This approach allows the use of development tools such as compilers, libraries, evaluation boards and emulators. The presented system uses multiple processors interconnected with IEEE-1355 high-speed links to provide the computational power necessary. Operating systems such as Virtuoso provide core capabilities to facilitate scalability, which is important to accommodate changes in functional or performance requirements that inevitably occur late in the development cycle, or even on orbit. A testbed was assembled using a combination of commercial DSP hardware and spaceflight components to evaluate the proposed multiprocessing approaches. Test results of real-time radar echo processing are presented, as well as proposed designs for future investigation.

Ku-band receiver and transmitter for breadboard DSP scatterometer

The design and test results for the Radio Frequency (RF) portion of a breadboard polarimetric scatterometer operating at 13.402 GHz are presented. An integrated breadboard has been developed at the Jet Propulsion Laboratory (JPL) to evaluate the feasibility of a programmable Digital Signal Processing (DSP) approach for a follow-on scatterometer similar to SeaWinds (scheduled for launch in winter 2002). Early breadboards of an integrated system have been identified as valuable assets in developing effective subsystem requirements for the eventual flight instrument. Many compatibility and partitioning issues between the RF and DSP hardware are addressed with empirical results derived from the aforementioned breadboard. The RF portion of the breadboard consists of a dual channel receiver, heterodyning the received signal of 13.402 GHz down to an IF of 37 MHz prior to the analog to digital conversion, and a single channel transmitter, that converts the I/Q baseband transmit waveform up to Ku band. The breadboard makes provision for emulating capabilities such as programmable attenuators, loop-back calibration, and saturation effects in an actual instruments power amplifier. It also provides control interfaces to allow early verification of software control algorithms.

Testbed for development of a DSP-based signal processing subsystem for an earth-orbiting radar scatterometer

A testbed for evaluation of general-purpose digital signal processors in earth-orbiting radar scatterometers is discussed. Because general purpose DSP represents a departure from previous radar signal processing techniques used on scatterometers, there was a need to demonstrate key elements of the system to verify feasibility for potential future scatterometer instruments. Construction of the testbed also facilitated identification of an appropriate software development environment and the skills mix necessary to perform the work. A testbed was constructed with three Astrium MCMDSPs, based on the Temic TSC 21020 general purpose DSP. Commercial data conversion hardware and high-speed serial communication hardware was interfaced to the MCMDSPs to allow demonstration of the key interfaces between subsystem elements: DSP program loading, synchronization and communication between multiple DSPs, interface to the scatterometer radio frequency subsystem, commanding, and science data delivery to the instrument data handling subsystem. A baseline set of requirements for the radar signal processing subsystem was established. From these requirements, signal processing algorithms such as digital filters and FFTs were developed using a combination of standard library functions and custom software. A software framework was developed to coordinate execution of the periodic signal acquisition and processing routines with asynchronous commanding and timekeeping functions. Emphasis was placed on developing modular software that would be applicable to a number of potential future instruments. Performance of the DSP subsystem was evaluated in terms of measured vs. theoretical execution speed, timing accuracy, power consumption, and computational accuracy.

Ground calibration of an Orbiting Spacecraft Transmitter

The SeaWinds Calibration Ground Station (CGS) is a novel Ku-band receiving station that supports the June 1999-launched JPL/NASA SeaWinds scatterometer radar which measures the near-surface wind speeds over 90% of the ice-free oceans. The CGS purpose is to measure very accurately the radar pulse timing, frequency and amplitude characteristics in order to monitor independently the radar and spacecraft platform performance over the course of the mission. These data are critical to maintaining or improving the quality of the end-data product. Results are presented which show the ability to measure spacecraft pulse timing to microsecond detail, spacecraft attitude to better than 0.1 degrees, and output amplitude that will in future work allow estimation of the spacecraft on-orbit antenna pattern. These efforts will improve registration of the scatterometer-derived wind map to the Earths coordinate system and provide an independent assessment of the instrument quality over the life of the mission.

Target & propagation models for the FINDER radar

Finding persons still alive in piles of rubble following an earthquake, a severe storm, or other disaster is a difficult problem. JPL is currently developing a victim detection radar called FINDER (Finding Individuals in Emergency and Response). The subject of this paper is directed toward development of propagation & target models needed for simulation & testing of such a system. These models are both physical (real rubble piles) and numerical. Early results from the numerical modeling phase show spatial and temporal spreading characteristics when signals are passed through a randomly mixed rubble pile.

Radar breadboard for DSP scatterometer

The design and test results for the radio frequency (RF) portion of a breadboard polarimetric scatterometer operating at 13.402 GHz are presented. To evaluate the feasibility of a programmable digital signal processing (DSP) approach for a follow-on scatterometer similar to SeaWinds an integrated breadboard has been developed at the Jet Propulsion Laboratory (JPL). Early breadboards of this type have been identified as valuable assets in developing effective subsystem requirements for the eventual flight instrument. Many compatibility and partitioning issues between the RF and DSP hardware are addressed with empirical results derived from the aforementioned breadboard. The RF portion of the breadboard consists of a dual channel receiver, heterodyning the received signal of 13.402 GHz down to an IF of 37 MHz and a single channel transmitter, that converts the I/Q baseband transmit waveform up to Ku band. The breadboard makes provision for emulating capabilities such as programmable attenuators, loop-back calibration, and saturation effects in an actual instruments power amplifier. It also provides control interfaces to allow early verification of software control algorithms.

Optimizing Systems and Scenarios

System designers and those who study tactics and operational scenarios are often faced with the problem of optimizing a particular outcome by adjusting a combination of conflicting parameters. Often this optimum is searched for by doing a direct computer simulation and trying all combinations of parameters which can easily result in lengthly computations. For example, if there are 10 parameters and 10 values for each parameter there are 10 billion possibilities. A second alternative, which is the main subject of this paper, is to use one of a variety of computerized optimization techniques to find an optimum set of parameters. This approach is computationally many orders of magnitude faster than the direct simulation method. Two examples are discussed using the Fletcher-Powell (F-P) optimization technique.