The paper proposes a method of digital parallel divider circuit module and its one-chip integration. The multiplication and division are performed very frequently in computers, electronic control systems, signal processing for sound, and image processing. So far, we have proposed a new module of multiplier and divider, which is easily cascadable to perform arithmetic for any word length, without the use of external circuitry. As the multipliers are used more frequently, many fast multipliers with high precision have been proposed and integrated. But there are few one-chip dividers, because they are available only within the limited word length of dividend. So that, there are few merits of making one-chip integrated dividers. The proposed parallel divider circuit module is cascadable on dividend for any word length, without the use of external circuitry. The soundness of the circuit was confirmed by computer simulation. We can design a system without the restriction of the bit length of dividend by using this divider. The proposed method makes it possible to implement one-chip integrated dividers suitable for fast division for any word length.
In order to realize a general purpose magnetic field sensor having a detection sensitivity of the order of 10-12 T/√Hz in the room temperature, a problem of removing the effect of earth magnetism from the sensor has to be solved, because its magnitude is 4.5×104 times larger than that in the range to be detectable. For this purpose, a well-known approach of exciting the sensor in the opposite direction to the earth magnetism to cancel out the value has been adopted. This paper aims at realizing a magnetic field sensor free from the earth magnetism by introducing a method called “zero magnetic field”. In practice, we developed such a sensor and the tested results confirmed that the detection performance was maintained without losing the high sensitivity.
There are some papers which present a design method of a model reference adaptive control system (MRACS) by state feedback. But it is necessary to consider the performance of the estimation time and error for the MRACS. This paper presents a design method of a discrete-time MRACS by learning control method. Since proposed construction uses an approximately inverse functional compensator, it is not necessary to estimate state variables or identify plant parameters. A system designed by proposed method can decrease the control error between the output of the plant and the reference model without the estimation of the state variables or the identification of the plant parameters. Therefore, exactmodel matching (EMM) can be realized very simply. The approximately inverse functional compensator is given by the inverse function of a reference model and has inverse property only in frequency zone of signals passing through the plant and the reference model.
This paper describes a method of integrating symbolic and numeric facilities for multivariable controller design. Our controller design system is embodied in REDUCE, a symbolic manipulation software system. In this system, controller design is interactively performed using functions, in symbolic manipulation phase, which enable to manipulate polynomial matrices and FORTRAN programs, in numeric manipulation phase, for solving parameter optimization problems like a linear programming one and simulating designed control systems. Getting output from the symbolic manipulation phase, one can go to the numeric manipulation phase through interface functions newly added to REDUCE, and, after necessary numeric manipulations, go back to the symbolic manipulation phase. In this paper, two types of interfacing are considered: 1)linking to a FORTRAN software package for solving linear programming problems and 2)generation of FORTRAN codes for simulators.