SEARCH · Engineering Papers
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
An approach for algorithm specifications and development is described for SMAP's radar onboard processor with multi-stage demodulation and decimation bandpass digital filter. Point target simulation is used to verify and validate the filter design with the usual radar performance parameters. Preliminary FPGA implementation is also discussed.
Recurrent Neural Networks (RNN) is widely applied to temporal sequence analysis, where real-time performance is usually in demand. However, RNN suffers a heavy computational workload as the model comes with a large weight matrix. To alleviate the pain, model compression (pruning) schemes have been proposed for RNN that pruning the redundant (near-zero) weight-values. On the one hand, the non-structured pruning methods achieve a considerable pruning rate while bringing the computational irregularity, which is un-friendly to parallel-hardware. On the other hand, the existing structured pruning methods consider the hardware parallelism; However, they suffer a poor pruning rate due to the restrict constraints on pruning structure. This paper presents CSB-RNN, an optimized full-stack RNN framework with the novel compressed structured block (CSB) technique. The CSB-pruned RNN model comes with both fine-granularity that benefits the pruning rate and regular structure that facilitates the hardware-parallelism. Further, we propose a novel hardware architecture for inferencing the CSB-pruned model. Different from conventional parallel hardware, this architecture solves the block-workload imbalance issue and achieves an over 95% hardware utilization. With the experiments on 10 RNN models in 5 application domains, the CSB-RNN realizes 7×-20× lossless compression and up to 50× acceptable lossy-compression, which is 2×-7× to the prior art. With the addition of the novel hardware, the compressed-RNN inference reaches a super real-time latency of 10-400µs with FPGA implementation.
In high energy physics experiment, as the luminosity increases, pile-up issues on detectors such as calorimeters become non-negligible. Deconvolution approaches with mathematic pre-assumptions such as Sparse Representation are developed for data analysis stage. For online computation tasks such as for trigger primitive creation, signal availability is significantly different as in offline data analysis stage, and therefore, different (yet simpler) algorithms should be explored. In this document, several approaches of deconvolution suitable for FPGA implementation are discussed.
Modern hyperspectral imaging systems are able to acquire far more data than can be downlinked from a spacecraft. Onboard data compression helps to alleviate this problem, but requires a system capable of power efficiency and high throughput. Software solutions have limited throughput performance and are power-hungry. Dedicated hardware solutions can provide both high throughput and power efficiency, while taking the load off of the main processor. Thus a hardware compression system was developed. The implementation uses a field-programmable gate array (FPGA). The implementation is based on the fast lossless (FL) compression algorithm reported in Fast Lossless Compression of Multispectral-Image Data (NPO-42517), NASA Tech Briefs, Vol. 30, No. 8 (August 2006), page 26, which achieves excellent compression performance and has low complexity. This algorithm performs predictive compression using an adaptive filtering method, and uses adaptive Golomb coding. The implementation also packetizes the coded data. The FL algorithm is well suited for implementation in hardware. In the FPGA implementation, one sample is compressed every clock cycle, which makes for a fast and practical realtime solution for space applications. Benefits of this implementation are: 1) The underlying algorithm achieves a combination of low complexity and compression effectiveness that exceeds that of techniques currently in use. 2) The algorithm requires no training data or other specific information about the nature of the spectral bands for a fixed instrument dynamic range. 3) Hardware acceleration provides a throughput improvement of 10 to 100 times vs. the software implementation. A prototype of the compressor is available in software, but it runs at a speed that does not meet spacecraft requirements. The hardware implementation targets the Xilinx Virtex IV FPGAs, and makes the use of this compressor practical for Earth satellites as well as beyond-Earth missions with hyperspectral instruments.
In high energy physics experiment trigger systems, block memories are utilized for various purposes, especially in indexed searching algorithms. It is often demanded to globally reset all memory locations between different events which is a feature not supported in regular block memories. Another common demand is to be able to update the contents in any memory location in a single clock cycle. These two demands can be fulfilled with registers but the cost of using registers for large memory is unaffordable. In this paper, a register-like block memory design scheme is described, which allows updating memory locations in single clock cycle and effectively refreshing entire memory within a single clock. The implementation and test results are presented.
One approach to parallel digital signal processing decomposes a high bandwidth signal into multiple lower bandwidth (rate) signals by an analysis bank. After processing, the subband signals are recombined into a fullband output signal by a synthesis bank. This paper describes an implementation of the analysis and synthesis banks using (Field Programmable Gate Arrays) FPGAs.
A logic design has been conceived for a field-programmable gate array (FPGA) that would implement a complex system of multiple digital state-space filters. The main innovative aspect of this design lies in providing for reuse of parts of the FPGA hardware to perform different parts of the filter computations at different times, in such a manner as to enable the timely performance of all required computations in the face of limitations on available FPGA hardware resources. The implementation of the digital state-space filter involves matrix vector multiplications, which, in the absence of the present innovation, would ordinarily necessitate some multiplexing of vector elements and/or routing of data flows along multiple paths. The design concept calls for implementing vector registers as shift registers to simplify operand access to multipliers and accumulators, obviating both multiplexing and routing of data along multiple paths. Each vector register would be reused for different parts of a calculation. Outputs would always be drawn from the same register, and inputs would always be loaded into the same register. A simple state machine would control each filter. The output of a given filter would be passed to the next filter, accompanied by a "valid" signal, which would start the state machine of the next filter. Multiple filter modules would share a multiplication/accumulation arithmetic unit. The filter computations would be timed by use of a clock having a frequency high enough, relative to the input and output data rate, to provide enough cycles for matrix and vector arithmetic operations. This design concept could prove beneficial in numerous applications in which digital filters are used and/or vectors are multiplied by coefficient matrices. Examples of such applications include general signal processing, filtering of signals in control systems, processing of geophysical measurements, and medical imaging. For these and other applications, it could be advantageous to combine compact FPGA digital filter implementations with other application-specific logic implementations on single integrated-circuit chips. An FPGA could readily be tailored to implement a variety of filters because the filter coefficients would be loaded into memory at startup.
Topics covered include: Instrument for Measuring Thermal Conductivity of Materials at Low Temperatures; Multi-Axis Accelerometer Calibration System; Pupil Alignment Measuring Technique and Alignment Reference for Instruments or Optical Systems; Autonomous System for Monitoring the Integrity of Composite Fan Housings; A Safe, Self-Calibrating, Wireless System for Measuring Volume of Any Fuel at Non-Horizontal Orientation; Adaptation of the Camera Link Interface for Flight-Instrument Applications; High-Performance CCSDS Encapsulation Service Implementation in FPGA; High-Performance CCSDS AOS Protocol Implementation in FPGA; Advanced Flip Chips in Extreme Temperature Environments; Diffuse-Illumination Systems for Growing Plants; Microwave Plasma Hydrogen Recovery System; Producing Hydrogen by Plasma Pyrolysis of Methane; Self-Deployable Membrane Structures; Reactivation of a Tin-Oxide-Containing Catalys; Functionalization of Single-Wall Carbon Nanotubes by Photo-Oxidation; Miniature Piezoelectric Macro-Mass Balance; Acoustic Liner for Turbomachinery Applications; Metering Gas Strut for Separating Rocket Stages; Large-Flow-Area Flow-Selective Liquid/Gas Separator; Counterflowing Jet Subsystem Design; Water Tank with Capillary Air/Liquid Separation; True Shear Parallel Plate Viscometer; Focusing Diffraction Grating Element with Aberration Control; Universal Millimeter-Wave Radar Front End; Mode Selection for a Single-Frequency Fiber Laser; Qualification and Selection of Flight Diode Lasers for Space Applications; Plenoptic Imager for Automated Surface Navigation; Maglev Facility for Simulating Variable Gravity; Hybrid AlGaN-SiC Avalanche Photodiode for Deep-UV Photon Detection; High-Speed Operation of Interband Cascade Lasers; 3D GeoWall Analysis System for Shuttle External Tank Foreign Object Debris Events; Charge-Spot Model for Electrostatic Forces in Simulation of Fine Particulates; Hidden Statistics Approach to Quantum Simulations; Reconstituted Three-Dimensional Interactive Imaging; Determining Atmospheric-Density Profile of Titan; Digital Microfluidics Sample Analyzer; Radiation Protection Using Carbon Nanotube Derivatives; Process to Selectively Distinguish Viable from Non-Viable Bacterial Cells; and TEAMS Model Analyzer.
The Fermilab Main Injector (MI) and Recycler Ring (RR) share a common beam loss monitor (BLM) system, making loss events difficult to attribute to their source machine when beam is present in both simultaneously. The Real-time Edge AI for Distributed Systems (READS) project addresses this by deblending BLM readings in real time using machine learning (ML). The current FPGA based implementation meets the sub-3 ms latency requirement but carries a resource intensive hls4ml development cycle, motivating exploration of GPU based deployment. This paper characterizes inference latency on an NVIDIA Jetson Orin Nano and introduces a packet organization scheme for assembling synchronized event frames from seven distributed BLM DAQ streams. Using a Python based DAQ simulation with injected timing jitter in place of unavailable live beam data, the pipeline achieved an average end to end latency of 0.456 ms (σ = 0.122 ms) across 167,000 test frames, comfortably meeting the timing constraint. Early outliers were attributed to TensorRT warm-up rather than steady state limitations, suggesting GPU based inference is a viable alternative to the existing FPGA implementation.
The Fermilab Main Injector (MI) and Recycler Ring (RR) share a common beam loss monitor (BLM) system, making loss events difficult to attribute to their source machine when beam is present in both simultaneously. The Real-time Edge AI for Distributed Systems (READS) project addresses this by deblending BLM readings in real time using machine learning (ML). The current FPGA based implementation meets the sub-3 ms latency requirement but carries a resource intensive hls4ml development cycle, motivating exploration of GPU based deployment. This paper characterizes inference latency on an NVIDIA Jetson Orin Nano and introduces a packet organization scheme for assembling synchronized event frames from seven distributed BLM DAQ streams. Using a Python based DAQ simulation with injected timing jitter in place of unavailable live beam data, the pipeline achieved an average end to end latency of 0.456 ms (σ = 0.122 ms) across 167,000 test frames, comfortably meeting the timing constraint. Early outliers were attributed to TensorRT warm-up rather than steady state limitations, suggesting GPU based inference is a viable alternative to the existing FPGA implementation.
Emerging technology is enabling the design community to consistently expand the amount of functionality that can be implemented within Integrated Circuits (ICs). As the number of gates placed within an FPGA increases, the complexity of the design can grow exponentially. Consequently, the ability to create reliable circuits has become an incredibly difficult task. In order to ease the complexity of design completion, the commercial design community has developed a very rigid (but effective) design methodology based on synchronous circuit techniques. In order to create faster, smaller and lower power circuits, transistor geometries and core voltages have decreased. In environments that contain ionizing energy, such a combination will increase the probability of Single Event Upsets (SEUs) and will consequently affect the state space of a circuit. In order to combat the effects of radiation, the aerospace community has developed several "Hardened by Design" (fault tolerant) design schemes. This paper will address design mitigation schemes targeted for SRAM Based FPGA CMOS devices. Because some mitigation schemes may be over zealous (too much power, area, complexity, etc.. . .), the designer should be conscious that system requirements can ease the amount of mitigation necessary for acceptable operation. Therefore, various degrees of Fault Tolerance will be demonstrated along with an analysis of its effectiveness.
A sequencer for a radar altimeter provides accurate attitude information for a reliable soft landing of the Mars Science Laboratory (MSL). This is a field-programmable- gate-array (FPGA)-only implementation. A table loaded externally into the FPGA controls timing, processing, and decision structures. Radar is memory-less and does not use previous acquisitions to assist in the current acquisition. All cycles complete in exactly 50 milliseconds, regardless of range or whether a target was found. A RAM (random access memory) within the FPGA holds instructions for up to 15 sets. For each set, timing is run, echoes are processed, and a comparison is made. If a target is seen, more detailed processing is run on that set. If no target is seen, the next set is tried. When all sets have been run, the FPGA terminates and waits for the next 50-millisecond event. This setup simplifies testing and improves reliability. A single vertex chip does the work of an entire assembly. Output products require minor processing to become range and velocity. This technology is the heart of the Terminal Descent Sensor, which is an integral part of the Entry Decent and Landing system for MSL. In addition, it is a strong candidate for manned landings on Mars or the Moon.
This paper presented a non-uniform multiphase (NUMP) time-to-digital converter (TDC) implemented in a field-programmable gate array (FPGA) with real-time automatic temperature compensation. NUMP-TDC is a novel, low-cost, high-performance TDC that has achieved an excellent performance in Altera Cyclone V FPGA. The root mean square (RMS) for the intrinsic timing resolution was 2.3 ps. However, the propagation delays in the delay chain of some FPGAs (for example, the Altera Cyclone 10 LP) vary significantly as the temperature changes. Thus, the timing performances of NUMP-TDCs implemented in those FPGAs are significantly impacted by temperature fluctuations. In this study, a simple method was developed to monitor variations in propagation delays using two registers deployed at both ends of the delay chain and compensate for changes in propagation delay using a look-up table (LUT). When the variations exceeded a certain threshold, the LUT for the delay correction was updated, and a bin-by-bin correction was launched. Using this correction approach, a resolution of 8.8 ps RMS over a wide temperature range (5 °C to 80 °C) had been achieved in a NUMP-TDC implemented in a Cyclone 10 LP FPGA.
Radio frequency interference (RFI) impairs L-band radiometry outside the protected 20 MHz frequency band around 1413 MHz. However, bandwidths of 100 MHz or more are desired for certain remote sensing applications as well as certain astronomy applications. Because much of the RFI in this band is from radars with pulse lengths on the order of microseconds, traditional radiometers (i.e., those which directly measure total power or power spectral density integrated over time scales of milliseconds or greater) are poorly-suited to this task. Simply reducing integration time and discarding contaminated outputs may not be a practical answer due to the wide variety of modulations and pulse lengths observed in L-band RFI signals, the dynamic and complex nature of the associated propagation channels, and the logistical effort associated with post-measurement data editing. This motivates the design and development of radiometers capable of coherent sampling and adaptive, real-time mitigation of interference. Such a radiometer will be described in this presentation. This design is capable of coherently-sampling up to 100 MHz bandwidth at L-band. RFI mitigation is implemented in FPGA components so that real-time suppression is achieved. The system currently uses a cascade of basic time- and frequency-domain detection and blanking techniques; more advanced algorithms are under consideration. The modular FPGA-based architecture provides other benefits, such as the ability to implement extremely stable digital filters and the ability to reconfigure the system "on the fly". An overview of the basic design along with on-the-air results from an initial implementation will be provided in the presentation. Related L-band RFI surveys will be described to illustrate the relevance of this approach in a variety of operating conditions.
Radio frequency interference (RFI) impairs L-band radiometry outside the protected 20 MHz frequency band around 1413 MHz. However, bandwidths of 100 MHz or more are desired for certain remote sensing applications as well as certain astronomy applications. Because much of the RFI in this band is from radars with pulse lengths on the order of microseconds, traditional radiometers (i.e., those which directly measure total power or power spectral density integrated over time scales of milliseconds or greater) are poorly-suited to this task. Simply reducing integration time and discarding contaminated outputs may not be a practical answer due to the wide variety of modulations and pulse lengths observed in L-band RFI signals, the dynamic and complex nature of the associated propagation channels, and the logistical effort associated with post-measurement data editing. This motivates the design and development of radiometers capable of coherent sampling and adaptive, real-time mitigation of interference. Such a radiometer will be described in this presentation. This design is capable of coherently-sampling up to 100 MHz bandwidth at L-band. RFI mitigation is implemented in FPGA components so that real-time suppression is achieved. The system currently uses a cascade of basic time- and frequency- domain detection and blanking techniques; more advanced algorithms are un- der consideration. The modular FPGA-based architecture provides other benefits, such as the ability to implement extremely stable digital filters and the ability to reconfigure the system "on the fly". An overview of the basic design along with on-the-air results from an initial implementation will be provided in the presentation. Related L-band RFI surveys will be described to illustrate the relevance of this approach in a variety of operating conditions.
This paper compares two waveform implementations generating the same RF signal using the same SDR development system. Both waveforms implement a satellite modem using QPSK modulation at 1M BPS data rates with one half rate convolutional encoding. Both waveforms are partitioned the same across the general purpose processor (GPP) and the field programmable gate array (FPGA). Both waveforms implement the same equivalent set of radio functions on the GPP and FPGA. The GPP implements the majority of the radio functions and the FPGA implements the final digital RF modulator stage. One waveform is implemented directly on the SDR development system and the second waveform is implemented using the JTRS/SCA model. This paper contrasts the amount of resources to implement both waveforms and demonstrates the importance of waveform partitioning across the SDR development system.
The RHIC interaction rate at sPHENIX will reach around 3 MHz in pp collisions and requires the detector readout to reject events by a factor of over 200 to fit the DAQ bandwidth of 15 kHz. Some critical measurements, such as heavy flavor production in pp collisions, often require the analysis of particles produced at low momentum. This prohibits adopting the traditional approach, where data rates are reduced through triggering on rare high momentum probes. We explore a new approach based on real-time AI technology, adopt an FPGA-based implementation using a custom designed FELIX-712 board with the Xilinx Kintex Ultrascale FPGA, and deploy the system in the detector readout electronics loop for real-time trigger decision.
The RHIC interaction rate at sPHENIX will reach around 3 MHz in pp collisions and requires the detector readout to reject events by a factor of over 200 to fit the DAQ bandwidth of 15 kHz. Some critical measurements, such as heavy flavor production in pp collisions, often require the analysis of particles produced at low momentum. This prohibits adopting the traditional approach, where data rates are reduced through triggering on rare high momentum probes. We explore a new approach based on real-time AI technology, adopt an FPGA-based implementation using a custom designed FELIX-712 board with the Xilinx Kintex Ultrascale FPGA, and deploy the system in the detector readout electronics loop for real-time trigger decision.