This paper shows that coherent optimization strategies for multilevel systhesis should rely on a good link between the factorization, the technology mapping and the netlist optimization. Factorization options are shown to play a key role. The technology mapping should optimize both area and critical path and only netlist structure preserving" optimization techniques (buffer insertion, gate replication) should be applied first to preserve the factorization decision. Only in a last step resynthesis of critical areas based on a local view is applied. The approach has been experimented on a set of large combinational benchmarks.

Acceleration techniques have been incorporated into the generalized method of characteristics (GMC) to perform transient analysis of uniform transmission lines, for the special case when the transmission lines are driven by digital signals. These techinques have been proved to improve the simulation speed to a great extent when the analysis is carried out using iterative waveform relaxation method. It has been identified that the load impedance connected to the transmission line has a bearing on the efficiency of one of these acceleration techniques. Examples of an RLCG line terminated with linear loads as well as nonlinear loads are given to illustrate the advantage of incorporating these acceleration techniques.

Test pattern generation is getting much harder as the circuit size becomes larger. One problem is that it tends to take much time and another one is that it is difficult to detect redundant faults. Aiming to cope with these problem, an enhanced unique sensitization technique is proposed in this paper. This powerful global implication reduces the number of backtracks with reasonable computational time. And a fast test pattern generator featuring this unique sensitization demonstrates its performance using large benchmark circuits with over ten thousands of gates. It takes only a minute to detect all testable faults and to identify all redundant faults of 20,000 gates circuit on a workstation.

In this paper, a new method of synthesizing multi-level logic circuits directly from binary decision diagrams (BDDs) is proposed. In the simple multiplexer implementation, the depth of the synthesized circuit was always O (n), where n is the number of input variables. The new synthesis method attempts to reduce the depth of circuits. The depth of the synthesized circuits is O (log n log w) where w is the maximum width of given BDDs. The synthesized circuits are 2-rail-input 2-rail-output logic circuits. The circuits have good testability; it is proved that the circuits are robustly path-delay fault testable and also totally self-checking for single stuck-at faults.

We introduce automatic procedures for generating and verifying sufficient correctness properties of synchronous processors. The targeted circuits are synchronous array processors designed from localized, highly regular data dependency graphs (DDGs). The specification, in the form of a DDG, is viewed as a maximally parallel circuit. The implementation, on the other hand, is a (partially) serialized circuit. Since these circuits are not equivalent from an automata-theoretic viewpoint, we define the correctness of the implementation against the specification to mean that a certain relation (called the β-relation) holds between the two. We use a compositional approach to decouple the verification of the control circuitry from that of the data path, thereby gaining efficiency. An array processor in isolation may not have a definite flow of control, because control may reside in the data stream. Therefore, for the purpose of verification, we construct an auxiliary machine, which keeps a timing reference and generates control signals abstracted from a typical data stream. Sufficient correctness conditions are expressed as past-tense computation tree logic (CTL) formulae and verified by CTL model-checking procedures. Experimental results of the verification of a matrix multiplication array and a Gaussian elimination array are presented.

System and chip synthesis must evaluate candidate Register-Transfer (RT) architectures with respect to finished physical designs. Current RT level cost measures, however, are highly simplified and do not reflect the real physical disign. Complete physical design, on the other hand, is quite costly, and infeasible to be iterated many times. In order to establish a more realistic assessment of layout effects, we propose a new layout model which efficiently accounts for the effects of wiring and floorplanning on the area and performance of RT level designs, before the physical design process. Benchmarking has shown that our model is quite accurate.

Reducing communication complexity is a viable approach to multilevel logic synthesis. A communication complexity based approach was proposed previously. In the previous works, only disjoint input decomposition was considered. However, for certain types of circuits, the circuit size can be reduced by using overlapped decomposition. In this paper, we consider overlapped decompositions. Some design issues for overlapped decompositions such as detecting globals" and deriving subfunctions are addressed. Moreover, the Decomposition Don't Cares (DDC) is considered for improving the decomposed results. By using these techniques together, the area and delay of circuits can be further minimized.

We present algorithms for the optimization of sequential synchronous digital circuits using structural model, i.e. interconnections of combinational logic gates and synchronous registers. This approach contrasts traditional methods using state diagrams or transition tables and leveraging state minimization and encoding techniques. In particular, we model circuits by synchronous logic networks, that are weighted multigraphs representing interconnections of gates implementing scalar combinational functions. With this modeling style, area and path delays are explicit and their variation is easy to compute when circuit transformations are applied. Sequential logic optimization may target cycle-time or area minimization, possibly under area or cycle-time constraints. Optimization is performed by a sequence of transformations, directed to the desired goal. This paper describes the fundamental mechansms for transformations applicable to sequential circuits. We review first retiming and peripheral retiming techniques. The former method optimizes the position of the registers, while the latter repositions the registers to enlarge maximally the combinational region where combinational restructuring algorithms can be applied. We consider then synchronous algebraic and Boolean transformations, that blend combinational transformations with local retiming. Both classes of transformations require the representation of circuits by means of logic expressions with labeled variables, the labels representing discrete time-points. Algebraic transformations entail manipulation of time-labeled expressions with algebraic techniques. Boolean transformations exploit the properties of Boolean algebra and benefit from the knowledge of don't care conditions in the search for the best implementation of local functions. Expressing don't care conditions for sequential circuits is harder than for combinational circuits, because of the interaction of variables with different time labels. In addition, the feasibility of replacing a local function with another one may not always be verified by checking for the inclusion of the induced perturbation in local explicit don't care set. Indeed, the behavior of sequential circuits, that can be described appropriately by the relation between input and output traces, may require relational models to express don't care conditions. We describe a general formalism for sequential optimization by Boolean transformations, where the don't care conditions are expressed implicitly by synchronous recurrence equations. We present then an optimization method for this model, that can exploit degrees of freedom in optimization not possible for other methods, and hence providing solutions of possible superior quality. We conclude by summarizing the major features and limitations of optimization methods using structural models.

This paper describes the hierarchical behavioral description language celled SFL and its processing system. This integrated CAD system called PARTHENON is used for designs of the leading ASICs in the NTT Systems Labs. This paper shows, therefore, the effectiveness of PARTHENON as a practical high-lelel synthesis system through real design experience. SFL was developed to aid in the design of the hardware functions and behaviors of ASICs composed solely of clocksynchronized circuits. The main features of SFL are as follows: (1) It is not mixed with connection description, but employs only behavioral description (like procedual description in program language), and it provides hierarchical expression of behavioral description. (2) It permits the description of parallel processing operations by adopting a new hardware task concept. And, (3) it is linked with the behavioral simulator, logic synthesizer, and other components of the processing system. After describing SFL in some detail, a brief explanation of its synthesizer and other processing components is provided, along with its application results in the real design of some leading ASICs at the NTT Systems Laboratories.

The Princeton University Behavioral Synthesis System (PUBSS) performs high-level synthesis on communicating processes. The compiler accepts models written in a subset of VHDL, but performs synthesis using a more specialized model, the behavior FSMs (BFSMs), for synthesis. The simulation semantics of VHDL presents challenges in describing behavior without overly constraining that behavior solely to make the simulation work. This paper describes mismatch between the simulation semantics provided by VHDL and the synthesis semantics required for high-level synthesis and describes how we solved these problems in PUBSS.

Since the time required for testing logic circuits is proportional to the number of test vectors, the size of test sets as well as test generation time is one of the most important factors to be considered in test generation. The size of test sets becomes an essential issue, especially for scan designed circuits, because of the need to shift a test vector serially into the scan path. In this paper, we propose new methods of generating compact test sets to detect al the irredundant single stuck-at faults in combinational circuits. The proposed algorithms calculate a test function for each fault which corresponds to the set of all test vectors for the fault and generate a compact test set by analyzing the test functions. The analysis is based on finding a test vector which detects the largest number of remaining faults. Since our methods select a test vector among all the test vectors, represented by a test function, for a target fault, smaller test sets can be generated, in general, than that by conventional test set compaction methods. The experimental results show that the size of test sets generated by our method is about one-third as large as that without compaction.

A model for a large network with an unidirectional linear respone (ULR) is proposed in this letter. This deterministic system has powerful computing properties in very close correspondence with earlier stochastic model based on McCulloch-Pitts neurons and graded neuron model based on sigmoid input-output relation. The exclusive OR problems and other digital computation properties of the earlier models also are present in the ULR model. Furthermore, many analog and continuous signal processing can also be performed using the simple ULR neural network. Several examples of the ULR neural networks for analog and continuous signal processing are presented and show extemely promising results in terms of performance, density and potential for analog and continuous signal processing. An algorithm for the ULR neural network is also developed and used to train the ULR network for many digital and analog as well as continuous problems successfully.

Most of actual environmental systems show a complicated fluctuation pattern of non-Gaussian type, owing to various kinds of factors. In the actual measurement, the fluctuation of random signal is usually contaminated by an external noise. Furthermore, it is very often that the reliable observation value can be obtained only within a definite fluctuating amplitude domain, because many of measuring equipments have their proper dynamic range and original random wave form is unreliable at the end of amplitude fluctuation. It becomes very important to establish a new signal detection method applicable to such an actual situation. This paper newly describes a dynamical state estimation algorithm for a successive observation contaminated by the external noise of an arbitrary distribution type, when the observation value is measured through a finite dynamic range of measurement. On the basis of the Bayes' theorem, this method is derived in the form of a wide sense digital filter, which is applicable to the non-Gaussian properties of the fluctuations, the actual observation in a finite amplitude domain and the existence of external noise. Differing from the well-known Kalman's filter and its improvement, the proposed state estimation method is newly derived especially by paying our attention to the statistical information on the observation value behind the saturation function instead of that on the resultant noisy observation. Finally, the proposed method is experimentally confirmed too by applying it to the actual problem for a reverberation time measurement from saturated noisy observations in room acoustics.

A technique for reducing fiber four-wave mixing (FWM) in multichannel transmissions is proposed. Birefringent elements are inserted on the way of transmission lines. Due to the effect of birefringent elements on the polarization states, the effective crosstalk due to FWM is expected to be 3/16 of that in the worst case in conventional systems.

This paper characterizes a class of optimal fixed-to-fixed length data compression codes for memoryless Gaussian sources that achieve asymptotically the rate-distortion bound under squared-error criterion. Any source output of blocklength n is encoded by two steps, i.e., 1) to quantize in gain by scholar quantizers and 2) to quantize in shape by pointsets on n-dimensional hyperspheres. To show the asymptotic optimality of the proposed codes, rate-distortion properties of the codes are analyzed in detail by using a random coding argument on the n-dimensional unit hypersphere. It is shown that asymptotic behaviors of the proposed codes are mainly determined by the choice of scalar quantizer of the gain. As a results, deep insights into not only the class of asymptotically optimal codes but also the rate-distortion bound itself are obtained.

This paper reviews fiber optic link techniques from the microwave and millimeter-wave transmission point of view. Several architectures of fiber optic links are reviewed. The application of MMIC technologies to the optical receivers are discussed and 26-GHz subcarrier transmission experimental works are described. Novel fiber optic links which utilize both optical device nonlinearities and microwave functional circuits are also reviewed. A system concept of millimeter-wave cellular radio using fiber optic links is finally discussed.

GaSb/InAs hot electron transistors (HETs) composed of a type-II misaligned quantum well operate at room temperature. The collector current is well described by the thermionic emission from the emitter. In order to get insight of the electron transport in the HET, the base width was varied or the collector barrier was modulated. The emitter's barrier height for the thermionic emission decreases with decreasing base width. This is caused by the increase of the quantum confinement energy in the InAs base with decreasing base width. Among HETs with a GaSb collector, a GaInSb abrupt layer, or a GaInSb graded layer at the collector edge, the latter type has the largest collector current. This indicates that collector grading reduces not only the collector barrier height, but also the quantum mechanical reflection of electrons. Collector-graded HETs with a 5 nm-thick base show a current gain of 8. The sheet resistance of InAs base is one order of magnitude less than bulk InAs without doping. This reduction is partly due to the accumulation of electrons transferred from the GaSb valence band to the InAs conduction band.

A complete analysis for the quantization noises and the reconstruction noises of the wavelet pyramid coding system is given. It is shown that in the (orthonormal) wavelet image coding system, there exists a simple and exact formula to compute the reconstruction mean-square-error (MSE) for any kind of quantization errors. Based on the noise analysis, an optimal bit allocation scheme which minimizes the system reconstruction distortion at a given rate is developed. The reconstruction distortion of a wavelet pyramid system is proved to be directly proportional to 2-2, where is a given bit rate. It is shown that, when the optimal bit allocation scheme is adopted, the reconstruction noises can be approximated to white noises. Particularly, it is shown that with only one known quantization MSE of a wavelet decomposition at any layer of the wavelet pyramid, all of the reconstruction MSE's and the quantization MSE's of the coding system can be easily calculated. When uniform quantizers are used, it is shown that at two successive layers of the wavelet pyramid, the optimal quantization step size is a half of its predecessor, which coincides with the resolution version of the wavelet pyramid decomposition. A comparison between wavelet-based image coding and some well-known traditional image coding methods is made by simulations, and the reasons why the wavelet-based image coding is superior to the traditional image coding are explained.

We have developed a new HIGFET structure achieving an extremely high K-value of 13.3 S/Vcm with a gate length of 0.15 µm. Self-aligned ion implantation is excluded to suppress a short-channel effect. An i-GaAs cap layer and an n+-GaAs contact layer are employed to reduce source resistance. The threshold voltage shift is as small as 50 mV when the gate length is reduced from 1.5 µm to 0.15 µm. Source resistance is estimated to be 53 mΩcm. We have also developed a new fabrication process that can achieve a shorter gate length than the minimum size of lithography. This process utilizes an SiO2 sidewall formed on the n+-GaAs contact layer to reduce the gate length. A gate length of 0.15 µm can be achieved using 0.35 µm lithography.

This paper describes IC-oriented high-performance AlGaAs/GaAs heterojunction bipolar transistors that were fabricated to demonstrate their great potential in applications to high-speed integrated circuits. A collector structure of ballistic collection transistors with a launcher (LBCTs) shortens the intrinsic delay time of the transistors. A novel and simple self-aligned fabrication process, which features an base-metal-overlaid structure (BMO), reduces emitter- and base-resistances and collector capacitance. The combination of the thin-collector LBCT layer structure and the BMO self-alignment technology raises the average value of cutoff frequency, fT, to 160 GHz with a standard deviation as small as 4.3 GHz. By modifying collector thickness and using Pt/Ti/Pt/Au as the base ohmic contact metal in BMO-LBCTs, the maximum oscillation frequency, fmax, reaches 148 GHz with a 114 GHz fT. A 2:1 multiplexer with retiming D-type flip-flops (DFFs) at input/output stages fabricated on a wafer with the thin-collector LBCT structure operates at 19 Gbit/s. A monolithic preamplifier fabricated on the same wafer has a transimpedance of 52 dBΩ with a 3-dB-down bandwidth of 18.5 GHz and a gain S21 OF 21 dB with a 3-dB-down bandwidth of 19 GHz. Finally, a 40 Gbit/s selector IC and a 50 GHz dynamic frequency divider that were successfully fabricated using the 148-GHz fmax technologies are described.