A two-dimensional receiver OEIC array having an address selector for highly parallel interprocessor networks has been realized. The receiver OEIC array consists of two-dimensionally arranged 1616 (256) optical receiver cells with switching transistors, address selectors (decoders), and a comparator. Each optical receiver comprises a pin PD and a transimpedance-type HBT amplifier. The HBT has an InP passivation structure to suppress the emitter-size effect, which results in the improvement of current gains, especially at low collector current densities. The receiver OEIC array was fabricated on a 3-inch diameter InP substrate with pin/HBT integration technology. Due to the function of address selection, only one cell is activated and the other cells are mute, so the receiver OEIC array shows low crosstalk and low power consumption characteristics. The array also shows a 266-Mb/s data transmission capability. This receiver OEIC array is a most complex InP-based OEIC ever reported. The realization of the two-dimensional receiver OEIC array promises the future interprocessor networks with highly parallel optical interconnections.

The filtered-x algorithm, which is widely applied to active noise control system, requires setting a small step gain. Such a small step gain reduces the noise reduction effect when the alogrithm is implemented by fixed point processing. This paper presents an experimental result that the 'polarized-g' individually normalized least mean square (INLMS) algorithm can provide almost the same noise reduction effect even in the fixed point processing of 16 bits as that in floating point processing.

This paper compares signal classification performance of multilayer neural networks (MLNNs) and linear filters (LFs). The MLNNs are useful for arbitrary waveform signal classification. On the other hand, LFS are useful for the signals, which are specified with frequency components. In this paper, both methods are compared based on frequency selective performance. The signals to be classified contain several frequency components. Furthermore, effects of the number of the signal samples are investigated. In this case, the frequency information may be lost to some extent. This makes the classification problems difficult. From practical viewpoint, computational complexity is also limited to the same level in both methods.IIR and FIR filters are compared. FIR filters with a direct form can save computations, which is independent of the filter order. IIR filters, on the other hand, cannot provide good signal classification deu to their phase distortion, and require a large amount of computations due to their recursive structure. When the number of the input samples is strictly limited, the signal vectors are widely distributed in the multi-dimensional signal space. In this case, signal classification by the LF method cannot provide a good performance. Because, they are designed to extract the frequency components. On the other hand, the MLNN method can form class regions in the signal vector space with high degree of freedom.

In this study we shall put forward a bidirectional synergetic neural network and investigate the crossassociation dynamics in an order parameter space. The present model is substantially based on a top-down formulation of the dynamic rule of an analog neural network in the analogy with the conventional bidirectional associative memory. It is proved that a complete association can be assured up to the same number of the embedded patterns as the number of neurons. In addition, a searching process of a couple of embedded patterns can be also realised by means of controlling attraction parameters as seen in the autoassociative synergetic models.

This paper describes the analysis of integrated optical waveguides using Finite-Difference Time-Domain (FDTD) method, and proposes the design methodology for low loss waveguide components: corner bends and branches. In order to integrate optical waveguides with Si VLSI technologies on a chip, the compact bends or branches are necessary. Since the optical power radiation from a bend or a branch point depends on the waveguide shapes, an accurate analysis of guided wave behavior is required. For the purpose we adopted the FDTD method which can analyze optical waveguides with a large variation of refractive index and arbitrary shape. Proposed design concept is to have all waveguides transmit only the fundamental mode and to design whole waveguides based on the fundamental mode transfer characteristics. For this design concept, waveguide components are required to have not only low radiation loss but also a little mode shift from the fundamental mode. The bend using the double-reflection mirrors and the branch using a slit are proposed for suppressing the mode shift and improving radiation loss. By the FDTD analysis, the following results have been obtained. The radiation loss and mode shift of double reflection bend are 1% and 4%, and those of the slit branch are 2% and 5%, respectively, in 2 µm width waveguide.

A 7-mask self-aligned SiGe base bipolar transistor has been newly developed. This transistor offers several advancements to a super self-aligned selectively grown SiGe base (SSSB) transistor which has a selectively grown SiGe-base layer formed by a cold-wall ultra high vacuum (UHV)/CVD system. The advancements are as follows: (1) a BPSG-filled arbitrarywidth trench isolation on a SOI is formed by a high-uniformity CMP with a hydro-chuck for reducing the number of isolation fabrication steps, (2) polysilicon-plug emitter and collector electrodes are made simultaneously using an in-situ phosphorusdoped polysilicon film to decrease the distance between emitter and collector electrodes and also to reduce the fabrication steps of the elecrodes, (3) a n+-buried collector layer is made by a high-energy phosphorus ion-implantation technique to eliminate collector epitaxial growth, and (4) a germanium profile in the neutral base region is optimized to increase the fT value without increasing leakage current at the base-cellector junction. In the developed transistor, a high performance of 80-GHz fT and mask-steps reduction are simultaneously achieved.

Photonic integrated circuits (PICs) are required for future optical communication systems, because various optical components need to be compactly integrated in one-chip configurations with a small number of optical alignment points. Bandgap energy controlled selective metal organic vapor phase epitaxy (MOVPE) is a breakthrough technique for the fabrication of PICs because this technique enables the simultaneous formation of waveguides for various optical components in one-step growth. Directly formed waveguides on a mask-patterned substrate can be obtained without using conventional mesa-etching of the semiconductor layers. The waveguide width is precisely controlled by the mask pattern. Therefore, high device uniformity and yield are expected. Since we proposed and demonstrated this technique in 1991, various PICs have been reported. Using electroabsorption modulator integrated distributed feedback laser diodes, 2.5 Gb/s-550 km transmission experiments have been successfully conducted. Another advantage of the selective MOVPE technique is the capability to form narrow waveguide layers. We have demonstrated a polarization-insensitive semiconductor optical amplifier that consists of a selectively formed narrow (less than 1 µm wide) bulk active layer. For a four-channel array, a chip gain of more than 20 dB and a gain difference between TE and TM inputs of less than 1 dB were obtained. We have also reported an optical switch matrix and an optical transceiver PIC for access optical networks. By using a low-loss optical waveguide, a 0 dB fiber-to-fiber gain for the 14 switch matrix and 0 dBm fiber output power from the 1.3 µm transceiver PIC were obtained. In this paper, the selective MOVPE technique and its applications to various kinds of PICs are discussed.

The usage of a diagram, which we call a state fence diagram (SFD), for analyzing discrete event systems such as reactive systems, is presented. This diagram is useful for event concurrent response and scenario analysis by using its three description styles.

The defects in the cold rolled strips have textural characteristics, which are nonuniform due to its irregularities and deformities in geometrical appearance. In order to handle the textural characteristics of images with defects, this paper proposes a surface inspection method based on textural feature extraction using the wavelet transform. The wavelet transform is employed to extract local features from textural images with defects both in the frequency and in the spatial domain. To extract features effectively, an adaptive wavelet packet scheme is developed, in which the optimum number of features are produced automatically through subband coding gain. The energies for all subbands of the optimal quadtree of the adaptive wavelet packet algorithm and four entropy features in the level one LL subband, which correspond to the local features in the spatial domain, are extracted. A neural network is used to classify the defects of these features. Experiments with real image data show good training and generalization performances of the proposed method.

Approximate maximum likelihood (ML) detection implemented by a reduced state Viterbi algorithm (VA), called the reduced state Viterbi coherent detection (RSVCD) algorithm in this paper, is described for the reception of uncoded M-ary PSK (MPSK) signals transmitted over additive white Gaussian noise (AWGN) channels. An M-state trellis, each state representing one of M signal constellation points, is used. The RSVCD algorithm performs parallel channel estimation based on the per-survivor processing principle (PSPP). Simple decision feedback CD (DFCD) is deduced as a special case of RSVCD. Unified BER expressions are derived for RSVCD, DFCD, and approximate ML detection implemented as an ML-state Viterbi algorithm (referred to as VACD) [6] as well as ideal CD and differential detection (DD). Computer simulation results are also presented and compared with theoretical results.

A new numerical technique, termed the method of matrix-order reduction (MMOR), is developed for handling electromagnetic problems in this paper, in which the matrix equation resulted from a method-of-moments analysis is converted either to an eigenvalue equation or to another matrix equation with the matrix order in both cases being much reduced, and also, the accuracy of solution obtained by solving either of above equations is improved by means of a newly proposed generalized Jacobian iteration. As a result, this technique enjoys the advantages of less computational expenses and a relatively good solution accuracy as well. To testify this new technique, a number of wire antennas are examined and the calculated results are compared with those obtained by using the method of moments.

We frequently use a polynomial to approximate a complex function. This study shows a method which determines the optimum coefficients and the number of terms of the polynomial, and the error of the polynomial is estimated.

A modified error correction/detection scheme based on the scheme by Yi and Lee is proposed. Algebraic decoding is used to perform error correction. Error detection is performed by an absolute value test. It is shown that the proposed scheme bridges the performance gap between Yi and Lee's scheme and Forney's optimal scheme.

The multiple registration schemes (MRSs) proposed here are classified into 3 cases by combining five registration schemes which are power up registration scheme (PURS), power down registration scheme (PDRS), zone based registration scheme (ZBRS), distance based registration scheme (DBRS), and implicit registration scheme (IRS) as follows: the first is MRS1 which covers PURS, PDRS, and ZBRS; the second is MRS2 which covers PURS, PDRS, and DBRS; the third is MRS3 which covers PURS, PDRS, IRS, and DBRS. The three proposed schemes are compared each other by analyzing their combined signaling traffic of paging and registration with considering various parameters of a mobile station behavior (unencumbered call duration, power up and down rate, velocity, etc.). Also, we derive allowable location areas from which the optimal location area is obtained. Numerical results show that MRS3 yields better performance than ZBRS, DBRS, MRS1, and MRS2 in most cases of a mobile station behavior, and it has an advantage of distributing the load of signaling traffic into every cell, which is important in personal communication system.

Multipath interconnection networks can support higher bandwidth than those of nonblocking networks by passing multiple packets to the same output simultaneously and these packets are buffered in the output buffer. The delay-throughput performance of the output buffer in multipath networks is closely related to output traffic distribution, packet arrival process at each output link connected to a given output buffer. The output traffic distributions are different according to the various input traffic patterns. Focusing on nonuniform output traffic distributions, this paper develops a new, general analytic model of the output buffer in multipath networks, which enables us to investigate the delay-throughput performance of the output buffer under various input traffic patterns. This paper also introduces Multipath Crossbar network as a representative multipath network which is the base architecture of our analysis. It is shown that the output buffer performances such as packet loss probability and delay improve as nonuniformity of the output traffic distribution becomes larger.

In this paper, we consider the following node-to-set disjoint paths problem: given a node s and a set T = {t1,...,tk} of k nodes in a k-connected graph G, find k node-disjoint paths s ti, 1 i k. We give an O(n2) time algorithm for the node-to-set disjoint paths problem in n-dimensional star graphs Gn which are (n - 1)-connected. The algorithm finds the n - 1 node-disjoint paths of length at most d(Gn) + 1 for n 4,6 and at most d(Gn) + 2 for n = 4,6, where d(Gn) = 3(n-1)/2 is the diameter of Gn. d(Gn) + 1 and d(Gn) + 2 are also the lower bounds on the length of the paths for the above problem in Gn for n 4,6 and n = 4,6, respectively.

A highly reliable network which can restore itself from network failures is one important concept for the future high capacity broadband network. In such self-healing network, flooding based failure-restoration algorithm is used to locate new routes and then to reroute failure traffic to that routes automatically when network failures such as link or node failures occur. Since the speed of this algorithm is degraded by the large amount of restoration messages produced by the process, such large volume messages should be reduced. In this paper, the scheme will be proposed, which reduces the large volume messages and efficiently selects alternative routes. In this scheme, the Message Wall will be used to filter useless restoration messages at the tandem nodes and Multi-Message Selecting method will be used to rapidly select a group of link-disjointed alternative routes from the feasible ones in each Flooding Wave sequence. The simulation results show that restoration messages are dramatically reduced and adequate alternative routes can be quickly found out.

In high performance compilers to process pointer-handling programs, precise pointer alias analysis is useful for the compilers to generate efficient object code. It is well known that most compiler techniques such as data flow analysis, dependence analysis, side effect analysis and optimizations are related to the alias problem. However, without data structure information, there is a limit on the precision of the alias analysis. Even though the automatic data structure detection problem is complex, when pointer manipulation satisfies some restrictions, some data structures can be detected automatically by compilers with some knowledge of aliases. In this paper, we propose an automatic data structure detection method for Pascal and Fortran 90. Linear list, tree and dag data structures are detected. Detected data structure information can be used not only for raising the precision of alias analysis but also for some optimizing techniques for pointer handling programs directly.

The number of DRAMs that have adopted hierarchical word-line architecture has increased as developed DRAM memory capacity has increased to more than 64 Mb. Use of the architecture enhances many kinds of DRAM performances, such as access time and fabrication process margin. However, the architecture does cause some problems. This paper describes some kinds of hierarchical word-line circuitries that have been proposed. It also describes a partial subarray activation scheme that is combined with hierarchical word-line and data-line architectures and discusses their potential and required specifications for future multi-giga bit DRAMs.

This paper deals with an orthogonal functional expansion of a non-linear stochastic functional of a stationary binary sequence taking 1 with equal probability. Several mathematical formulas, such as multi-variate orthogonal polynomials, recurrence formula and generating function, are given in explicit form. A simple example of orthogonal functional expansion and stationary random seqence generated by the stationary binary sequence are discussed.