A small singal variance detection by an analogue technique and misdetection prevention algorithm to check transient variance integrity by digital technique is proposed. Real time acquisition of an one time signal variance is processed by two signal difference stages that make fluctuation as small as possible. Low S/N ratio signal difference is detected without averaging nor smoothing for S/N improvement. Misdetection prevention algorithm compensates analogue circuit's vulnerability to noise, rejecting similar variance more than two times during time window. Test signal generator was made to check one time small signal variance and multiple small signal variance, showing an effectiveness of noise rejection for proposed system. Fluctuation rejection experiments from 1Hz to 5kHz and multiple pulse testified that proposed detection system was effective to work under noisy environment. 0.5% relative signal variation for 5V gate input was detected with 9V_pp fluctuant noise from 1Hz to 100Hz. Total throughput after small signal variance event is lms.

抄録全体を表示

It can be expected that a coplanar waveguide (CPW) bend would present different transmission characteristics from those of a microstrip-line (MSL) bend because, unlike MSL, CPW has a signal conductor and two ground conductors on the same side of the substrate. This paper describes the crosstalk characteristics of two adjacent CPW bends separated by a ground conductor and formed on a printed circuit. Grosstalk characteristics of CPW bends of various line-separations at near and far ends were studied and compared with those of two adjacent straight CPW's, and also with those of MSL bends of the corresponding structures. Excepting the case of close coupling, crosstalk characteristics of CPW bends generally show a maximum at some frequency.

抄録全体を表示

Electromagnetic (EM) noise emission from a printed trace on a digital circuit board is evaluated for about forty samples of high-speed CMOS logic devices that have the same device number and the same package; 74HC00 SOP. The dependence of EM noise spectra on dispersion of switching speed and output resistances of logic gates were investigated. Products of 11 lots from four manufacturers were tested. The EM noise spectra were picked up by a small loop antenna located 20mm distant from the board surface. At frequency higher than 200MHz intensities of the spectra diverge beyond 10dB. The maximum difference of peak values on the spectra was 0.7dB within each lot, and 5.6dB for all the samples. Rise and fall time as well as output resistance of CMOS gates were also measured. As a result, the peak value on the noise spectra was found to be strongly dependent on the fall time and also the the on-resistance Rn of a CMOS gate. Deviation of the measured peak values of the noise spectra is less than 1dB from estimated values based on the on-resistance. This result suggests that the intensity of EM noise spectra can be predicted if the output resistances of CMOS gates are known.

抄録全体を表示

Electro-Magnetic Interference due to electrostatic discharge (ESD) is a serious problem which causes unreliable operation in electronic circuits. Many researchers have studied the subject especially for the ways to protect circuits from ESD. The mechanism of charge accumulation has been studied by measuring the charge quanitity only on the surface. However, this means that the experiments have less reproducibility.
The authors measured the space charge inside some polymer films by the newly developed pulsed electroacoustic method which is not influenced by environmental conditions. The results show that some materials have accumulated charge distributions initially. Thus charge quantity should be measured not only on the surface but also inside the material. In addition, the experimental results show that one of the anti-electrostatic materials can produce the internal space charge electric field by itself when a DC voltage is applied, thereby compensating the external field. These results suggest that space charge measurement is useful in fundamental research of ESD.

抄録全体を表示

It is well known that the voltage transients are caused by the operation of switchgears in a power system. The voltage transients travel along the power line as a traveling-wave, and cause malfunction of electrical devices, communication system, and the operational error of control system.
The voltage transients at the making operation of switchgear were examined to get the fundamental data for considering the immunity of communication system used on a power system. However, it was difficult to observe the voltage transients using the potential transformer and the capacitance potential device, because the transients were very rapid.
Therefore, a measurement method using the coupled transmission line system is proposed to measure the high speed and high voltage transients, and the actual experimental results on the 6.9 kV power system are presented. This method enables the measurement of high speed voltage transients in out of contact with power system.

抄録全体を表示

The earth potential difference (EPD) has been observed continuously for monitoring artificial and natural earth current at three observatory locations including one urban area. The EPD signals have a complicated waveform because of many kinds of earth current sources. In this paper the micro variation waveforms in EPD signals observed at each location are described and analyzied qualitatively concerning with earth currrent sources. To investigate the relationship between EPD signals and atmospheric electric phenomena the most suitable location is choosen at Sado island in the Japan Sea. It is revealed that induced electro-static field by charged clouds generates earth current when the field exceeds the threshold value. Its value is determined to be 100kV/m from the observed EPD signals.

抄録全体を表示

The misconvergence of three beams on the screen in a color cathode ray tube (CRT) is the main factor to estimate the performance of CRT. The magnetic plates should be used to convert three beams on the screen. We have estimated the magnetic field made by the deflection yoke (DY) in CRT with the surface magnetic charge method. This method can calculate the magnetic field precisely in the open boundary and the computer memory size is small because of the division of only the surface on the magnetic materials. But, as the magnetic plates should be very thin, the surfaces of the magnetic plates should be divided into many small elements to guarantee the exact results. Therefore much CPU time and large memory size are demanded. To overcome this difficuties, we newly developed the hybrid method which has two kinds of parameters; the surface magnetic charge on the ferrite core in DY and the scalar potential on the magnetic plate. The calculation is iterated to take into accout the mutual effeciency till the unknown values are converted. This paper describes the newly developed method and the results of the comparison with the measurements to confirm the efieciency of the developed method. The calculated magnetic fields and the changes of misconvergence made by the magnetic plates give good agreement with the measurements.

抄録全体を表示

Cell layout in analog LSI chip designs entails a lot of strict constraints, and its complete automation is known to be a very difficult problem to solve. The layout in analog LSI designs usually has the same relative position of elements as in the circuit schematic. Given this fact, in this paper, we consider the placement problem to minimize the overall net length and the used region area with the constraints of neighbourhoodrelation of elements. We propose two ideas based on genetic algorithms; (1) a method to scan the placement tasks of small blocks in overlapping stages, (2) a method to line up several elements along the x-direction gradually by stochastic adjustment of y-coordinates. Computational experiments show the efficiency of these methods.

抄録全体を表示

A new three-layer pattern classification neural network_Adaptive Fuzzy Classification Neural Network (AFC) is introduced. AFC allows hidden neurons to grow to meet the demands of the problem during training in order to overcome the difficulty in predetermining the number of neurons. In AFC, each hidden neuron is regarded as a fuzzy cluster representing a shape of hypersphere in n-dimensional pattern space. Furthermore, in AFC's learning algorithm, neuron merge and fission operations are proposed which are quite useful in reducing redundant neurons. Effectiveness of AFC is shown by comparing the simulation results with conventional neural networks through experiments.

抄録全体を表示

Described in this paper is the use of multiple digital-signal processors (DSP's) to implement a real-time digital controller. The authors have already proposed a new configuration of multi-DSP system with the parallel processing scheme which can shorten both the throughput time and the sampling period in the digital controller. In this time, an optimal scheduling problem is considered on the multi-DSP system so that the sampling period of the digital controller will be minimized. And then, an optimization algorithm based on a branch-and-bound approach is proposed. The optimization algorithm proposed here invokes a heuristic search method that weighs the performance of a multiply-add module installed in each DSP.

抄録全体を表示

This paper proposes a new method of relative words retrieval. In this method, firstly, the conceptual space, onto which a lot of words in the database are mapped by using the technique of the multidimensional quantification IV, is utilized for the preliminary retrieval of words. Next, the retrieved words are sorted to sophisticate the result based on conceptual intention, which can estimate relevance between words.
The features of this method are 1) it can retrieve apposite words without conisdering the dimensions of conceptual space, 2) it can retrieve appropriate words based on low-dimensional conceptual space that can be reconstructed easily. 3) it can reduce a size of memory which is used to store the eieenvectors is reduced

抄録全体を表示