We present a transformation method for reducing the number of states of bypassed Knuth LR(k) parsers. Bypassed Knuth LR(k) parsers behave like Knuth LR(k) parsers except that they perform syntactic analysis without reducing productions of the form X::Y (X and Y are grammar symbols) that have empty semantic actions. Using a characterization of correct bypassed LR(k) parsing obtained by parsing contexts approach of this paper, we show that the transformation of merging of states having same core preserves the validity of the parsers provided that the parsers are free from parsing action conflicts. (Note that in general this property cannot be guaranteed in other type of Bypassed LR(k) parsers than Bypassed Knuth LR(k) parsers.) Using this transformation we can construct a faster LR(k) parser having reasonable number of states. This result is a significant reduction in the size of bypassed LR(k) parsers.

A mirror-type optical switch using a z-cut lithium niobate crystal (LiNbO₃) was fabricated, and a two-by-two transverse type optical switch of the Michelson interferometer with two mirror type LiNbO₃ crystals and two Wollaston prisms (WP's) was constructed. An optical filter, and a one-by-four optical switch using a combination of an acoustooptic tunable filter (AOTF) employing a Tellurium oxide crystal (TeO₂) and this mirror-type LiNbO₃ crystal are proposed for applications.

A cylindrical PZT around which some length of a fiber is wound is demonstrated as a component of optical phase shifter and modulator in a single-mode fiber interferometer. Its sensitivity under the operation of DC input mode is 0.18 π rad/V/meter in phase change at the optical wavelength of 6328 . Some characteristics of amplitude response linearity and frequency response uniformity for the AC drive voltage applied to the PZT cylinder are also discussed.

Store-and-forward communication networks with no effective means of flow control are subject to congestion under heavy load conditions. In this paper a distributed congestion control policy, called PF-C, with nodal buffer management strategy based on traffic priorities is proposed and analyzed. The priority of a packet is depended upon the number of hops it has traveled and the number ob hops it has to travel to reach its destination. More importantly, packets are identified into classes according to their priorities. The limited number of buffers (limit value) at any node to be occupied by each class of packets are assigned. The mechanism of this policy is that, a packet which arrived at a given node is rejected if the total number of allocated buffers exceeds a limit value corresponding to its class. Rejected packets are dropped from the network and considered as lost. This policy is analyzed in the context of symmetrical networks, whereby a queuing model is developed. Finally, network throughput which is an example of system performance, is formulated in terms of offered load, total number of buffers in a node and the limit values. Numerical applications to loop networks clearly demonstrate the good behavior of this policy to achieve the maximum network throughput.

A planar-circuit-type 3 dB hybrid consisting of a four-port disk-shaped resonator is reinvestigated. A great improvement in the characteristics is obtained by rearranging the four ports upon consideration of the higher order resonator modes without loss in simplicity of the structure.

A method is proposed to design synchronous mod N counters (logical type) by using many-valued JK flip-flops. Then costs of the counters are derived.

An improved algorithm of autocorrelation pitch detection is presented. Preliminary experiments show that the algorithm can considerably reduce the errors caused by the ordinary autocorrelation pitch detector.