This paper presents a method for obtaining the average number of data packets waiting to be transmitted in Integrated Switched networks via a diffusion approximation. Evaluated in an example network, our method is found to give a good approximation for engineering purposes.

A neural network model is proposed to explain the development of direction selectivity of cortical cells. The model is constructed under the following three hypotheses that are very plausible from recent neurophysiological findings. (1) Direction selectivity is developed by modifiable inhibitory synapses. (2) It results not from the direct convergence of many excitatory inputs from LGN cells but from cortical neural networks. (3) Direction-selective mechanism is independent of orientation-selective mechanism. The model was simulated on a computer for a few kinds of inhibitory connections and initial conditions. The results were consistent with neurophysiological facts not only for normal cats but for cats reared in an abnormal visual environment.

This paper reports the realization of 2 GHz surface-acoustic-wave delay line (SAW DL) oscillator, utilizing 3rd harmonic operation of double interdigital transducers (4F-IDTs). The simple design technique for the SAW DL harmonic oscillators is made clear, based on the state-of-the-art of microfabrication technique. 0.58 µm linewidth aluminum 4F-IDTs of a SAW DL filter have been successfully obtained on the rotated Y-cut quartz by conventional photolithographic contact printing without lift-off process. They consist of 618 lines and 0.4 mm overlap length. On the clarified design technique, a 2 GHz SAW DL oscillator was achieved without external filtering and matching networks. The frequency stability of the experimental 2 GHz SAW DL oscillator is also presented and discussed. The short-term stability of 2 part in 108 at 1 sec. gate time was measured and a nearly parabolic frequency variation with temperature (the turn-over temperature was near 20) was obtained. It was found that SAW DL oscillators can be small, lightweight and stable GHz oscillators in many communication systems.

Single-mode fibers are prospective candidates for future high-capacity and long distance transmission media. In particular, low loss single-mode fibers, operating around 1.27 µm wavelength, where the material dispersion of silica glass falls to zero, are very attractive because of their huge bandwidth capability. Until now, however, the attainment of very low loss in the long wavelength region has been confined to only multimode fibers. Single-mode fibers with GeO2 -SiO2 glass system, which have the broadest low loss window ant the long wavelength region, were fabricated. The minimum loss of 0.5 dB/km at 1.3 µm wavelength, where dispersion of the fiber is negligibly small, was obtained by suitably designing waveguide parameters and reducing OH contamination. Also, practical structural features and the OH ion behavior are clarified in relation to reduction of OH absorption, which is the major problem in reducing loss at the wavelength of interest.

Multiple Markov sequences of interpoint intervals are generated by computer simulation. A kind of rational transition probability density function presented here is found to be satisfactory to simulate wide varieties of neuronal spike trains. Three examples of the Markov sequences are examined and investigated from a viewpoint of point density, two of which are to simulate the actual spike trains obtained from the central neurons of the cat. The third one has exponential type interval histogram. The generated Markov sequences as well as the neuronal spike trains usually show slow fluctuations of point densities. When a point sequence of this kind is smoothed by an appropriate filter, the approximate point density fluctuation is extracted. By an inverse procedure of integral density modulation with this point density as the modulating signal, the point sequence is transformed into another sequence of which the slow component of point density is eliminated. It is found that the transformed point sequence can be approximately regarded as a non-correlative sequence, if a right kind of filter is selected. This shows that both the simulated Markov sequences and the neuronal sequences tested here are almost regarded as the density modulated point sequences of the non-correlative sequences of intervals.

The effects of mode coupling in multimode fibers have been reported using coupled power equations of power flow equations, and the good agreement between theoretical and experimental results provides further evidence that the power flow equation is a useful description of the power distribution in a multimode optical fiber. This paper develops a simple formalism to analyze power flow equations with the help of a variational principle. The results can treat not only any mode coupling mechanism involving the previous works and the experimental results but also any input modal distribution, and the good agreement between theoretical and experimental results shows the validity of our theory. As an example of applications, mode coupling due to random bends of the fiber axis is studied in detail, and the mode coupling-induced losses and the steady-state pulse width are calculated.

The class of codes described in this paper is used for Single b-Adjacent bit-group Error Correction (SbEC). This is especially useful in large memory systems which use integrated semiconductor memory chips, each containing b-output bits. The proposed codes have odd-weight-column H-matrix characteristic and superior error detection capability, and facilitates the construction of modularized parallel high speed encoding / decoding network through rotational coding techniques. For a special case of b1, this class of codes is equivalent to the optimal odd-weight-column SEC-DED (Single Error Correcting - Double Error Detecting) codes. A computation indicates that the odd-weight-column SbEC codes have better error detection capability than the conventional SbEC codes.

This paper presents a method for the study of multivariable positive real rational matrices. The method is based on the following two items. (i) Each port of the network can be represented by a specific element and this element identifies the port conversely. (ii) A Hurwitz polynomial which corresponds to a given multi-variable positive real rational matrix in a one-to-one manner can be found. In short, the followings are shown in this paper. 1) The well defined Hurwitz polynomial corresponds to a given n-variable reactance scattering (or paraunitary) mm matrix in a one-to-one manner. 2) The concept of degeneration (or degree-down)" and degree-up" of immittance functions are clarified. 3) If each element of network is of different kind, no degeneration can occur. 4) For the realization of more than three-variable reactance function, it is generally necessary to degree-up the given immittance function. 5) For the practical realization of two-variable reactance function, it is not always convenient to apply directly the method which is used for the proof of realizability. Here, a new practical realization method for two-variable reactance functions based on junction matrix approach, is presented. The new method needs neither factorization nor degree-up operation.

Estimation is done for the offset, tilt and end separation loss of grade index fiber connectors with arbitrary modal power distribution using the modal evaluation technique instead of geometrical treatment as reported by D. Gloge. Comparison of the results with experiments shows that it provides a reasonable description and is useful in designing optical fiber connectors and splices.

The upper bound of periods realized in a system of 1-dimensional uniform neural networks with refractory period is analyzed for each transition function. As for one function whose refractory period is 2, the period is not to be bounded. For the other functions with the refractory period R the upper bound equals 1 or R1. They contrast with the fact that the period is bounded by 4 for every case without refractory period. The results are obtained by constructing directly the periodic states of every cell from a boundary cell to the other one after another.

A simple expression is derived for the total noise in PCM encoding without a holding circuit. Simulation results and measurements on an experimental codec show that the expression provides a good approximation.

Usually, for the definitions on n-variable positive real functions, the properties on 2n-dimentional domain are used as the conditions of the definition, as examplified below. On the domain D {(p1, p2, , pn)|Re p10, Re p20, , Re pn0)}, |Re w(p1, p2, , pn) must be positive." In the present paper, the approach is simplified by the following two items. (i) Fixing all variables except for one on the axis (real axis or imaginary axis), the problems on multi-variables are reduced to the one's on singlevariable. (ii) The problems are discussed only on one-dimentional domain, that is, on the axes. Simplifying the problems concerning with multi-variable positive real function by the two items stated above, the definitions of several classes of positive real function are presented. Clearly, these definitions may simplify extremely the verification of positive realness of any given function, and may give an effective means for the studies of multi-variable positive real function.

A New carrier recovery subsystem for binary PSK systems using synchronized push-pull oscillator is described. A property that the push-pull oscillator has two stable states of which the phase difference is π, can be well applied for this purpose. This new subsystem does not include phase-locked loop (PLL), therefore pull-in time and pull-in range are not limited by the loop delay of PLL. The circuit of this push-pull oscillator is very simple. Yet, it is proven to achieve rather good performance, if a limiter which normalizes the input amplitude, when practically deformed on account of the band-limiting, is attached. Our experiments have had good agreement with the theory and computer simulation.

Experimental results and theoretical consideration of the slant path rain attenuation characteristics and site diversity effects are described, together with the prediction method. The slant path rain attenuation has been measured by using 17 and 27 GHz sun-tracking radiometers at Yokosuka Electrical Communication Laboratory in Japan for 3 years. At a distance of 33 km from the laboratory, the atmospheric noise was also measured by using 18 GHz passive radiometer for the above-mentioned period to know the site diversity operation characteristics. It is shown that calculated values of slant path rain attenuation and site diversity effects agree well with measured values. For example, calculated and measured values of the rain attenuation, at 17 GHz for 0.1% of time within the time interval when the elevation angle of the sun is from 30 to 60, are 5.5 and 5.1 dB, respectively. Experimental results using 18 GHz passive radiometers fixed at 20 and 55 elevation angles indicate that the rain attenuation has the elevation angle dependence.

In this paper we derive several results on the dependent 2-unit system with single repair facility. The dependency means that it is allowed for the two units to fail at the same time. First, for the dependent 2-unit warm standby system, we obtain the following results: Laplace transform of the system reliability by using the conditional transform, limit theorems for the system reliability and mean time to the system failure (MTSF), distribution of the number of repairs completed during (0, t] and Laplace transform of the system reliability and MTSF in the case that the repair facility is capable of only n repairs. Second we discuss on the effects of repair policies to the availability of the dependent 2-unit hot standby system.

The error-rate performance of digital FM with discriminator-detection under the fast Rayleigh fading environment is theoretically analyzed in the presence of Gaussian noise and co-channel interference. While the effect of intersymbol interference caused by band width lmitations and time-delay spread is neglected, the fading spectrum effect is taken into consideration. The error-rate including not only the Rayleigh envelope fading effect but also the random FM noise effect is given in a simple closed form.

A narrow-bandwidth optical wave filter was constructed by a stacked directional coupler with two slab waveguides of different materials. The matched phase constant β0 defines the center wave-length of the filter, whereas the different dispersion relations and the inhomogeneity of the guides define the filter bandwidth. The narrowest 3 dB bandwidth was 0.2 nm at the center wavelength of 616.8 nm.

On the transient wave form of the oscillation at 4416 Å in a He-Cd II hollow cathode laser excited by the square wave discharge, a sharp peak is found in the initial region where the peak value of the discharge current is high. The excitation mechanism for the peak is due to the direct collision between electrons and Cd atoms.

In this paper, an analysis was made on the variation of oscillating mode numbers of a semiconductor laser at high speed pulse modulation by solving multi-mode rate equations which relate a quasi-Fermi level to photon numbers. The laser gain was assumed to be homogeneous and has a dependency of (), where is ωEg (relative photon energy). Single mode oscillation conditions for short pulse modulation were obtained associated with following two cases. A) One is the criterion how wide the mode separation must be made as of a short cavity laser and of an integrated twin-guide (ITG) laser. B) The other is the condition how large we must give an excess loss to the unwanted modes such as of the ITG laser, of a distributed feedback (DFB) laser, and of a distributed Bragg reflector (DBR) laser.

The present paper deals with the realization of a class of homogeneous positive real function, that is, the class of function with denominator composed of linear factors only. The theories are stated by the following steps. (a) Firstly, two basic theorems on the homogeneous positive real functions are presented. (b) Secondly, the theory of partial fraction expansion of the given function with denominator composed of linear factors, is presented. Here, any partial fraction is a homogeneous positive real function of degree two. (c) Lastly, the realization procedures of the partial fractions are presented.