The degradation of ultra-thin SiO2 films accompanied by the hole direct tunneling is investigated using a substrate hot hole (SHH) injection technique. Hot holes from the substrate as well as cold holes in the inversion layer are injected into the gate oxides in p-channel MOSFETs with pp+ poly-Si gates, while the gate bias is kept low enough to avoid the simultaneous electron injection from the gate. During the SHH stress, in contrast to the case of thicker oxide films, a strong correlation is observed between the oxide film dagradation and the injected hole energy, whereas no degradation occures due to the hole direct tunneling from the inversion layer. These experimental findings indicate the existence of threshold energy for trap creation process, which has been predicted by the theoretical study of hole-injection-induced structural transformastion of oxygen vacancy in SiO2.
Schottky-source/drain MOS transistors were fabricated on a highly resistive SOI (Silicon On Insulator) substrate with either chromium or nickel gate electrode. Device isolation was realized by LOCOS process with the consideration for the planarization prior to the gate patterning by lift-off. Source and drain silicidation with titanium disilicide was done adopting the self-alignment process utilizing the gate electrode as a mask, without introducing the covering film on the gate sidewall. Submicron-channel devices with the gate length as shortt as 0.1μm were fabricated with chromium and nickel gate electrodes on the same substrate and satisfactory FET characteristics were observed both for p-type operation with chromium- and nickel- gate devices and for n-type operation with the chromium-gate device. Thus, p-type and n-type devices can be complementarily integrated in the same chip only by changing the gate metal, without the need of well structure in the substrate, This is favorable for higher-density integration.
A sequential test of stress-induced-migration and electromigration was employed to evaluate Al interconnect reliability. We investigated electromigration after storage testing for stress-induced failure and found a deterioration in electromigration lifetime of quarter-micron Al interconnects caused by slit-like voids generated during the storage test.
Recently, ultra-small Si devices utilizing quantum-size structures have received much attention to open a new Si device stream. However, such ultra-small Si structures might cause unintended behaviors in the device operation. In this paper, we evaluate the influence of shrinking of a Si nanowire channel through Si nanowire device characteristics. In the fabricated 15-10-nm-wide Si nanowire device, we confirm that the Si nanodevice expectedly works as a single electron or few electron memory. However, in the other 5-10-nm-wide Si nanowire device, we observe an unintended Coulomb blockade effect due to the fluctuation of the Si nanowire width. To suppress the unintended Coulomb blockade effect, we reveal the usefulness of the self-limiting oxidation effect to Si nanowires, for the first time, and improvement (39%) in uniformity of the Si nanowire is demonstrated. It is shown, therefore, that self-limiting oxidation is promising for fabrication of Si nanodevies.
A new technique to form the silicon on nothing (SON) structure is presented as an alternative to the silicon on insulator (SOI) structure. A large plate-shaped empty space in silicon (ESS) below the surface of the silicon substrate can be fabricated by connecting the spherical empty spaces, which are formed by the surface migration of silicon on patterned silicon substrate. ESS technique has a potential to change the micro-process for the fabrication oflarge-scale integrated circuits (LSI) and it can be applied to various manufacturing technologies.
In this paper, a discrete-time CNN using 1-dimensional cell circuits with S stable points (S=2, 3, 4, ... ) is proposed. The circuit structure and the behavior of the proposed CNN is simple by exploiting discrete-time 1-dimensional circuits as the cell circuits. Since the proposed cell circuit takes as inuch as S of states, the proposed CNN can demonstrate the multiple-valued traveling wave phenomena. Furthermore, the proposed cell circuit is suitable for integration thanks to the circuit design by using switched-current (SI) techniques. The computer simulations are performed concerning a 2-dimensional CNN which is constructed with the proposed cell circuits. The 1-dimensional cell circuit with S stable points is designed by a standard CMOS process. Furthermore, the proposed cell circuit is built with commercially-available IC's. The experimental result is in close agreement with the simulation result. The proposed CNN is integrable by a standard CMOS technology.
Competitive co-evolution algorithm can adaptively acquire solutions of a problem. However, in a problem which does not have the optimal solution, it needs to decide a set of effective solutions as the best solution. In this paper, we propose a competitive co-evolution algorithm with a packaging solutions to solve the problem. Our algorithm has two characteristics. The one is minimization of the number of individuals in the set by extraction of the complemental solutions. The other is evaluating solutions in some continued generations by setting a life-time to an individual. We apply the proposal method to the Game in order to investigate its effectiveness. Furthermore, we analyze the process of the set formation. In the simulation results, our method can acquire the complemental strategies and shows a better performance than a conventional method.
In this paper, we propose a motion planning method for manipulation of an object by multiple mobile robots. In the previous cooperative works by multiple mobile robots, manipulation technique based on force-control has been proposed. However, mobile robots are moving by position-control, and motion errors can often arise. Then, we build the manipulation technique, which is suitable for mobile robots by position-control. We propose the manipulation method without using sensor information, and consider the motion errors of mobile robots and the indefinite element of environment from the planning stage. We compute the contact position to the object where the object is stable. And the order of operation is generated from the result of these analyses. After that, we compute the sticks' paths and each robot's motion by considering the limitation of robot motion. We verify the effectiveness of our proposed method through simulations.
This paper experimentally implements tactile sensing of shapes of unknown surfaces with sliding motion. Sliding motion is described by an algorithm and realized by a hybrid position and force scheme. A miniature finger-tip tactile sensor is fixed at the end of a two-link finger to measure the contact force and moment. With the contact force, moment and joint angle information, the contact position, surface normal are calcu-lated. The surface normal is utilized to restrain the deviation of finger from the surface during the surface following. Experimental results show the tactile sensing with sliding motion works well for a line, a concave arc, and a convex arc. Experimental results also show the sliding motion works well at different force levels and different speeds. Therefore tactile sensing with sliding motion can perceive a surface whose shape is not known a priori.
It has become possible to integrate many circuit blocks as a system on chip. However it is difficult to design all circuit blocks from the beginning. On such trends, VSI is being expected as a new distributing commodities of the circulation, and as a reusable core for the designing of a large scale ASIC. But there are some serious problems for the standardization, and ease of use of the core. In this paper, we propose a method of hardware modularizing which perform just as the data processing device in an object of the object-oriented technique. The user doesn't need to know the contents of an abstracted part called VC. It can get a result by sending a message with some parameters if we want to use the module. Therefore, it doesn't need to prepare for any programs and circuits newly. The hardware design method proposed here with the object-oriented technique is just for the VSI.
Base station equipment in a W-CDMA system was claimed communication quality depending on RF block dynamic range. It is possible that Direct-conversion modulator enhances the system dynamic range by eliminating the intermediate frequency (IF) block. Base-band signals are directly quadrature-modulated at radio frequency(RF), and the dynamic range depends on RF characteristic of quadrature modulator. We have developed the type of modulator to obtain modulating accuracy and low adjacent-channel power ratio (ACPR).
A neuro-PID control is applied to the position control of a hard disk drive. The PID gains are tuned using a neural network in such a way that the tracking error is reduced while avoiding the stimulation of a mechanical resonance. A nonlinear performance criterion is used for updating the connection weights of the neural network so that unwanted motions of the magnetic head are reduced when the tracking error is sufficiently small. The effectiveness of the proposed controller is demonstrated by simulations.
In this paper, we present an architecture of a neuro-hardware that can be realized on by far a small-scaled circuit compared to the conventional approach. In order to reduce the scale of the circuits, the architecture employs a new method of computing the membrane potential and the sigmoidal function by encapsulating the probability properties into relative delay between two pulses. Proposed architecture enables to integrate more than one hundred of neurons on a latest FPGA chip, which means thirteen-fold miniaturization compared to conventional architecture.
This study proposes a new method for detecting characteristic points (CPs), the Q and S points, in elec-trocardiogram (ECG) using a multichannel ART-based neural network (MART). The method integrates the previous two methods: the slope detection techniques and neural networks. The slope detection techniques are able to locate CPs exactly. However, it is not robust to noise. On the other hand. the method of neural networks locates approximate locations of the Cps and self-organizes in response to newly input patterns. This self-organizing ability makes the method robust. The MART integrates these two methods to implement a reliable CPs detection. For the CPs detection. ECG is divided into cardiac cycles by preprocessor, and each cardiac cycles is input to the channel one of the MART. A rectangle is made from each cardiac cycle and input to channel two of the MART. Patterns of the two channels are transmitted to the F3 layer of the MART, and then the winner node of the F3 layer recalls template patterns to the channels in the Fl layer. When the pattern recognition carried out by the MART. the template locates CPs in the ECG. The method were evaluated using MIT/BIH arrhythmia database. The standards deviation between detected CPs and CPs estimated by referee are within the limit of the SDs recommended by the CSE committee.
The floorplan design problem is a combinatorial optimization problem, and it is difficult to find an exactly optimal solution in practical applications. A variety of genetic algorithms for floorplan design have been reported in the literature, however in most of these algorithms, some kind of floorplans can not be represented by means of chromosomes, and consequently, we can never obtain such kind of floorplans using these algorithms. In this paper, a new genetic algorithm for floorplan design is proposed. First, the problem is formulated as a rectangle-packing problem, and a “sequence-pair” is used for genotype coding in order to represent any possible floorplan. Next, properties of process of decoding from sequence-pair to floorplan are investigated in detail, and two guidelines to inherit the “local floorplan” can be derived from them. On the basis of these guidelines a new crossover operator can be designed for sequence-pair, and a new genetic algorithm can be constructed. Moreover, our algorithm is implemented and applied to a MCNC benchmark, and it is shown that each of the two guidelines is effective and that the resultt obtained by our algorithm is better than that by the existing algorithm.
In this paper, a technique to acquire modulation pulses for simultaneously working multiple supersonic wave sonars so as to decrease their errorneous measurement is proposed. Because the structure of the sonar can be implemented simply and cheaply, the multiple sonars are usually equipped on an autonomous mobile robot which moves the inside/outside of a building. Therefore, it is necessary to compose of the modula-tion pulses to meet the requirement to decrease the errorneous measurement between the sonars. Besides, a technique to minimize the length of the signal of the modulation pulses by using the genetic algorithm (GA), known as one of the optimization algorithms, is proposed in this paper. As a result, the more suitable modulation pulses can be obtained.
The observed phenomena in the actual magnetic and sound field environment are inevitably contaminated by the background noise of arbitrary distribution type. Therefore, in order to evaluate the wave motion type environment, it is necessary to establish some estimation methods to remove the undesirable effect of the background noise. In this paper, a digital filter for estimating a specific signal under the existence of a background noise with non-Gaussian distribution form is proposed. By applying the well-known least mean squared method for the moment statistics with several orders, a practical estimation method including Kalman filter as a special case is derived. The effectiveness of the proposed theoretical method is experimentally confirmed by applying it to the actual estimation problem in magnetic and sound field environment.
In speech recognition under noisy environment, it is necessary to construct the system which reduces the noise and enhances the speech. Then it is effective to imitate the human auditory system which has an excellent analytical mechanism of spectrum for speech enhancement. In this paper, we propose an adaptive method using the auditory mechanism which is called lateral inhibition. This method first estimates the intensity of the noise intensity by neural network, then adjusts adaptively both the coefficients of the lateral inhibition and the adjusting coefficient of amplitude component according to the intensity of noise for each input frame. It is confirmed that this method is effective for the speech degraded not only by white noise but also by colored noise, judging from the spectral distortion measurement.
Control of the backing movement of a trailer-truck system is very difficult because the dynamics are nonlin-ear. In this paper, a new control method using neuro-controllers (NCs) developed using a genetic algorithm (GA) is presented. We use a 3-5-1 neural network to control the steering angle. In the GA process, a simple evaluation function is used. We use only final configurations of the trailer-truck system after control trials in the evaluation function. We apply the control method not only to simulations but also to experiments. The results of both show that the control method is very effective. For verifying the effectiveness of this control method, we employ another control method, linear quadratic regulator (LQR). The results show that the control method using NCs evolved by GA exhibits better control performance than the LQR.
This paper describes a method for developing object-oriented frameworks based on reference models. Ref-erence models are a reuse technique to represent directly the logical structure of a domain. A framework embodies both an abstract design that is common to all applications from the domain and hot spots that allow the framework to be customized to application-specific requirements. It is difficult for developers to extract the objects in a domain. The existing object-oriented methods do not have the guide of object ex-traction. Moreover, developers must specify the flexible parts of applications by hot-spot analysis. There is the difference of the abstraction level between domain models and object models of frameworks. Reference models can show the guide of object extraction and mediate between domain models and object models of frameworks. We apply our approach to develop the framework of an industrial system. We propose MVCP (Model-View-Controller-Proxy) model for monitoring domains. By applying this model to this system, we can limit the scope of the objects in relation to hot spots. We discuss the effects in our approach.
In civil engineering, in order to improve weak foundations, many sand pillars are frequently struck into underground using a casing pipe. In that case, a smooth construction will be achieved if the sinking velocity of the sand level in the pipe is measured on-line accurately. The measurement of the level is reduced to that of a length of the air pillar in the pipe. The authors thus previously proposed, as the first step, an on-line accurate pipe length measurement system for straight pipes, which applied a bank of Kalman filters and Bayesian theorem to a linear dynamic model of the stationary wave in the pipe. However, it was found that the pre-proposed system could not be directly applied to the sand level mea-surement in the casing pipe, because several patterns of stationary waves were simultaneously formed in the pipe due to its complicated inner structure. This paper thus presents an on-line sand level measurement system which introduces a new dynamic model to take into consideration of the two dominant patterns of the stationary waves in the pipe. The criterion to obtain the optimal modes of the stationary wave is also presented to realize a much higher measurement accuracy.
This paper presents a FIR-type neural network to identify nonlinear systems as a nonlinear identification model. We derive a learning algorithm of this neural network based on the gradient descent method and the conjugate gradient method. In this learning algorithm, we derive the algorithm which updates both weight coefficients contained in FIR-synapses and gains of sigmoid functions in each unit of each layer. Also, we present a generalized sigmoid function as the nonlinear function contained in each unit in order to designate the upper and lower limits of the function. It is shown that the accuracy of the identification model based on FIR-type neural networks is superior to that of the identification model based on ordinary backpropagation neural networks and the conjugate gradient algorithm is superior to the gradient descent algorithm in point of the convergence and the learning cycle.
This paper presents structure of a neural network for analysis and knowledge extraction from continuous valued problems and a training method for the structure. The proposed neural network consists of two types of hidden units. One type of hidden units has weights connected to only a group of input units. Another one has weights connected to all input units. The former type of hidden units allows to analyze each relation between a certain input data and corresponding output data. The latter type of hidden units ensures performance of the neural network as same as conventional neural networks. While the neural network is been training, needless hidden units are pruned automatically by the superposed energy function, the structure learning with forgetting, and the compact structuring algorithm. Therefore, it is easy to analyze the neural network and extract knowledge. The effectiveness of the proposed method is shown by function approximation, peak load forecasting of electric power and water flow forecasting into the dam.
We proposed two selection methods, i) selection method by using phonemic environmental index (PER method), ii) selection method by searching minimal connective distortion index path (MLD method) for small scale speech synthesis system. This paper argued the relation between phonemic environmental index and connective distortion index. The followings were clarified, a) VCV sequence selected by PER method ranks automatically the supreme 10% of all VCV sequences with respect to connective distortion index criterion. b) VCV sequence selected by MLD method ranks automatically the supreme 10% of all VCV sequences with respect to phonemic environmental index criterion. Hence, these two selection criteria have a very strong relation each other.
We study the radio communication system in oil storage tank, where many reflection waves exist. In order to decrease the reflection waves, we present a communication system using higher order TE and TM modes. Electromagnetic wave propagation in the tank is simulated numerically, so that it turns out DU ratio above 12dB can be realized.
The direct tunneling (DT) conduction in the low voltages for the n+poly-Si/SiO2/p-Si (100) with a thickness of 2.7_??_3.3nm is examined based on WKB method from a viewpoint of the inelastic scattering in the p-Si(100) substrate considering the influence of the electron effective mass in the n+poly-Si grains. It is shown that these considerations reproduce the measured DT currents in the low voltages.
This paper describes the direction of looking aside for the driver. The system detects driver's looking aside while situation of ancontact and nonrestriction by use of image processing technology. From facial images, an eye region and eyelids is obtained. The eye direction is detected from the position of eyelids and pupil.
This paper presents a designing method of a robust controller which can be reduced within a prescribed degradation of a control performance. The method uses p-synthesis to take each variation of a plant and a controller into account. A numerical example is presented and its result is analyzed by using the Gap Metric.