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.
Four-volume document reviews key technologies for interconnecting in very-large-scale integrated (VLSI) circuits. Document also discusses current and proposed research into novel fabrication techniques for interconnections.
Winner-take-all circuit selects best-match stored pattern. Prototype cascadable very-large-scale integrated (VLSI) circuit chips built and tested to demonstrate concept of electronic associative pattern recognition. Based on low-power, sub-threshold analog complementary oxide/semiconductor (CMOS) VLSI circuitry, each chip can store 128 sets (vectors) of 16 analog values (vector components), vectors representing known patterns as diverse as spectra, histograms, graphs, or brightnesses of pixels in images. Chips exploit parallel nature of vector quantization architecture to implement highly parallel processing in relatively simple computational cells. Through collective action, cells classify input pattern in fraction of microsecond while consuming power of few microwatts.
It is shown that the use of redundant on-chip circuitry improves the testability of an entire VLSI circuit. In the study described here, five techniques applied to a two-bit ripple carry adder are compared. The techniques considered are self-oscillation, self-comparison, partition, scan path, and built-in logic block observer. It is noted that both classical stuck-at faults and nonclassical faults, such as bridging faults (shorts), stuck-on x faults where x may be 0, 1, or vary between the two, and parasitic flip-flop faults occur in IC structures. To simplify the analysis of the testing techniques, however, a stuck-at fault model is assumed.
An electronic neurocomputer which implements a radial basis function neural network (RBFNN) is described. The RBFNN is a network that utilizes a radial basis function as the transfer function. The key advantages of RBFNNs over existing neural network architectures include reduced learning time and the ease of VLSI implementation. This neurocomputer is based on an analog/digital hybrid design and has been constructed with both custom analog VLSI circuits and a commercially available digital signal processor. The hybrid architecture is selected because it offers high computational performance while compensating for analog inaccuracies, and it features the ability to model large problems.
Accurate fault models are required to conduct the experiments defined in validation methodologies for highly reliable fault-tolerant computers (e.g., computers with a probability of failure of 10 to the -9 for a 10-hour mission). Described is a technique by which a researcher can evaluate the capability of the pin-level stuck-at fault model to simulate true error behavior symptoms in very large scale integrated (VLSI) digital circuits. The technique is based on a statistical comparison of the error behavior resulting from faults applied at the pin-level of and internal to a VLSI circuit. As an example of an application of the technique, the error behavior of a microprocessor simulation subjected to internal stuck-at faults is compared with the error behavior which results from pin-level stuck-at faults. The error behavior is characterized by the time between errors and the duration of errors. Based on this example data, the pin-level stuck-at fault model is found to deliver less than ideal performance. However, with respect to the class of faults which cause a system crash, the pin-level, stuck-at fault model is found to provide a good modeling capability.
Feedback shift registers are efficient periodic binary sequence generators. Polynomials of degree r over a Galois field characteristic 2(GF(2)) characterize the behavior of shift registers with linear logic feedback. The algorithmic determination of the trinomial of lowest degree, when it exists, that contains a given irreducible polynomial over GF(2) as a factor is presented. This corresponds to embedding the behavior of an r-stage shift register with linear logic feedback into that of an n-stage shift register with a single two-input modulo 2 summer (i.e., Exclusive-OR gate) in its feedback. This leads to Very Large Scale Integrated (VLSI) circuit architecture of maximal regularity (i.e., identical cells) with intercell communications serialized to a maximal degree.
Report analyzes costs of developing and producing low-production-volume, customized VLSI (very large-scale, integrated) circuits. Relationship is developed between IC cost and electronic system cost using IC cost models based on design/fabrication approach. Emphasis is on development of understanding between cost and volume for custom circuits to be used by NASA. Reliability is major cost component in models. Report is divided into five sections and includes four appendices with useful reference literature.
The Big Viterbi Decoder (BVD) is a powerful error-correcting hardware device for the Deep Space Network (DSN), in support of the Galileo and Comet Rendezvous Asteroid Flyby (CRAF)/Cassini Missions. Recently, a prototype was completed and run successfully at 400,000 or more decoded bits per second. This prototype is a complex digital system whose core arithmetic unit consists of 256 identical very large scale integration (VLSI) gate-array chips, 16 on each of 16 identical boards which are connected through a 28-layer, printed-circuit backplane using 4416 wires. Special techniques were developed for debugging, testing, and locating faults inside individual chips, on boards, and within the entire decoder. The methods are based upon hierarchical structure in the decoder, and require that chips or boards be wired themselves as Viterbi decoders. The basic procedure consists of sending a small set of known, very noisy channel symbols through a decoder, and matching observables against values computed by a software simulation. Also, tests were devised for finding open and short-circuited wires which connect VLSI chips on the boards and through the backplane.
The relationship of integrated circuit (IC) cost to electronic system cost is developed using models for integrated circuit cost which are based on design/fabrication approach. Emphasis is on understanding the relationship between cost and volume for custom circuits suitable for NASA applications. In this report, reliability is a major consideration in the models developed. Results are given for several typical IC designs using off the shelf, full custom, and semicustom IC's with single and double level metallization.
Simulation of circuits having more than 2000 active devices requires the largest, fastest computers available. A vector computer, such as the CYBER 205, can yield great speed and cost advantages if efforts are made to adapt the simulation program to the strengths of the computer. ASPEC and SPICE (1), two widely used circuit simulation programs, are discussed. ASPECV and VAMOS (5) are respectively vector adaptations of these two simulators. They demonstrate the substantial performance enhancements possible for this class of algorithm on the CYBER 205.
Two silicon CMOS application specific integrated circuits (ASICs), a data generation chip, and a data checker chip were designed. The conversion of the data generator circuitry into a pair of CMOS ASIC chips using the 1.2 micron standard cell library is documented. The logic design of the data checker is discussed. The functions of the control circuitry is described. An accurate estimate of timing relationships is essential to make sure that the logic design performs correctly under practical conditions. Timing and delay information are examined.
The complexity of GaAs field effect transistor (FET) very large scale integration (VLSI) circuits is limited by the maximum power dissipation while the uniformity of the device parameters determines the functional yield. In this work, digital GaAs FET circuits are presented that eliminate the DC power dissipation and reduce the area to 50% of that of the conventional static circuits. Its larger tolerance to device parameter variations results in higher functional yield.
A new basic VLSI circuit element is presented that can be used to realize pulse mode asynchronous sequential circuits. A synthesis procedure is developed along with an unconventional state assignment procedure. Level input asynchronous sequential circuits can be realized by converting a regular flow table into a differential mode flow table, thereby allowing the new synthesis technique to be general. The new circuits tolerate 1-1 crossovers. This circuit also provides a means for state sequence detection and real time fault detection.
This paper introduces an improved method for designing the class of CMOS VLSI asynchronous sequential circuits introduced in the paper by Sterling R. Whitaker and Gary K. Maki, 'Self Arbitrated VLSI Asynchronous Circuits.' Of main interest here is the simple design by inspection rules that arise from these circuits. This paper presents a variation on these circuits which reduces the number of transistors required.
Proposed universal noiseless coder (UNC) compresses stream of data signals for efficient transmission in channel of limited bandwidth. Noiseless in sense original data completely recoverable from output code. System built as very-large-scale integrated (VLSI) circuit, compressing data in real time at input rates as high as 24 Mb/s, and possibly faster, depending on specific design. Approach yields small, lightweight system operating reliably and consuming little power. Constructed as single, compact, low-power VLSI circuit chip. Design of VLSI circuit chip made specific to code algorithms. Entire UNC fabricated in single chip, worst-case power dissipation less than 1 W.
Scheme for efficient interconnection makes VLSI design feasible. Concept for fast Viterbi decoder provides for processing of convolutional codes of constraint length K up to 15 and rates of 1/2 to 1/6. Fully parallel (but bit-serial) architecture developed for decoder of K = 7 implemented in single dedicated VLSI circuit chip. Contains six major functional blocks. VLSI circuits perform branch metric computations, add-compare-select operations, and then store decisions in traceback memory. Traceback processor reads appropriate memory locations and puts out decoded bits. Used as building block for decoders of larger K.
A fully systolic architecture for the implementation of digital sequence correlator/accumulators is described. These devices consist of a two-dimensional array of processing elements that are conceived for efficient fabrication in Very Large Scale Integrated (VLSI) circuits. A custom VLSI chip that was implemented using these concepts is described. The chip, which contains a four-lag three-level sequence correlator and four bits of accumulation with overflow detection, was designed using the Integrated UNIX-Based Computer Aided Design (CAD) System. Applications of such devices include the synchronization of coded telemetry data, alignment of both real time and non-real time Very Large Baseline Interferometry (VLBI) signals, and the implementation of digital filters and processes of many types.
Papers from the symposium are presented from the following sessions: (1) featured presentations 1; (2) very large scale integration (VLSI) circuit design; (3) VLSI architecture 1; (4) featured presentations 2; (5) neural networks; (6) VLSI architectures 2; (7) featured presentations 3; (8) verification 1; (9) analog design; (10) verification 2; (11) design innovations 1; (12) asynchronous design; and (13) design innovations 2.