淀积条件对GDa—SiNx薄膜电学特性的影响

本文讨论了用辉光放电法制备氮化硅薄膜时衬底温度、射频功率和气体流量比对薄膜的电导率、介电常数和击穿强度的影响。通过优化生长条件，制备了优质非晶氮化硅薄膜，其介电常数为７．５、击穿强度为５．５ＭＶ／ｃｍ、电导率为１０－１３（Ωｃｍ）－１。     

铝液夹杂物检测DMIRT－Ⅲ法的试验研究

提出了一种新的铝液夹杂物快速检测方法。导出了该法的检测原理方程。分析并试验研究了影响检测重复性和分辨率的因素。试验选择了较好的检测参数，并获得了较好的检测结果。     

Simulation study of peak junction temperature and power limitationof AlGaAs/GaAs HBTs under pulsed and CW operation

The authors have used a 3-D transmission-line matrix (TLM) modeling method to study the junction temperature distribution and power limitation of device geometries with multiple embedded heat sources. Peak values of the junction temperature against the dissipated power density under both pulsed and CW operation are presented for a typical power AlGaAs/GaAs HBT structure. These data should facilitate the rapid determination of junction temperature for a given output power, which is of paramount importance in power device design     

Two-level optimal output feedback stabilizer design

The paper addresses the design of two-level power system stabilizers using an optimal reduced order model whose state variables are torque angles and speeds. The reduced order model retains their physical meaning and is used to design a two-level linear feedback controller that takes into account the realities and constraints of electrical power systems. The two-level control strategy is used, and a global control signal is generated from the output variables to minimize the effect of interactions. The effectiveness of this controller is evaluated and a multimachine system is given as an example to illustrate the advantages of the proposed method. Responses of the system with a two-level scheme and an optimal reduced order scheme are included for comparative analysis.     

Gate and drain currents in off-state buried-type p-channel LDDMOSFETs

The buried-type p-channel LDD MOSFETs biased at high positive gate voltage exhibit novel characteristics: (1) the ratio of the drain to gate currents is about 1×10^-3 to 5×10^-3; and (2) the gate and drain currents both are functions of only the gate voltage minus the n-well bias. Such characteristics are addressed based on the formation of the surface n ⁺ inversion layer due to the punchthrough of the buried channel to the underlying shallow p-n junction. The measured gate current is due to the Fowler-Nordheim tunneling of electrons from this inversion layer surface and the holes generated within the high-field oxide constitute the drain current. The n⁺ inversion layer surface potential is found to be equal to the n-well bias plus 0.55 V. As a result, both the oxide field and the gate and drain currents are independent of drain voltage     

Eigenvalue-based algorithms for testing positive realness of SISO systems

A computational test for scalar positive real and strictly positive real functions is proposed in this paper. The input for this algorithm is state space representation, not necessarily minimal. It is derived for the discrete time problem, based on numerical stable eigenvalue computations. The test can be used in the continuous counterpart, which is in fact an improvement of an algorithm posed by Bai and Freund. Differences between the discrete and continuous time problems are also pointed out.     

由呼出气中的氡浓度估计铀矿工体内~(225)Ra的负荷量

本文测量了某铀矿不同工种的56名矿工下矿井作业前的呼出气中的~(222)Rn浓度,并由此估算了他们体内的~(226)Ra 负荷量。按不同工种的人数加权,呼出气中的~(222)Rn 浓度平均值为34.1Bqm~(-3),体内~(226)Ra 负荷量的均值为1.95kBq。同时还对9名休息了2天的和2名离开井下作业3—4年的矿工以及12名来下过矿井的本矿工人呼出气中氡浓度进行了测量,由此估算出其体内~(226)Ra 负荷量的均值分别为1.11、0.94和0.17kBq。     

On-line learning delivery decision support system for highlyproduct mixed semiconductor foundry

A production learning system (PLS) based on the tool model was constructed as a decision support and real-time information update system to forecast the cycle time. A tool model includes a waiting model and a processing model. Each of the waiting and processing models uses a backpropagation neural network to establish the relationship between the input and output (time) of the model. Hence, cycle time estimation, tool group move and confirm line item performance (CLIP) value can be obtained based on the memory stored in the neural network. The result shows that the forecasting ability of the PLS has an error rate below 8% on average