集成电路可制造性设计理论与统计最优化方法统一模型的研究

文中系统地介绍了集成电路可制造性设计的基本概念，讨论了ＩＣ发展趋势和ＶＬＳＩ的极限，以及功能设计向优化设计过渡的必然性。综合考虑了电路制造的参数成品率极大，最佳容差设计，调整设计和制造费用极小化问题，提出了建立在不可微规划框架上的统计最优化方法的统一模型。     

DCTT统计最优化模型及其性质

本文提出了集成电路最优化中综合考虑成品率极大、最佳容差设计、调整设计和生产效益极大化设计的统计最优化模型(Design Centering,Tolerance and Tuning.简称DCTT模型)。讨论了该模型与广义统计最优化模型的等价性以及其他主要性质。在确定性最优化框架下,该模型为统计最优化和集成电路可制造性设计的进一步发展开辟了新途径。     

集成电路参数中心值和容差的耦合设计方法

基于对集成电路参数成品率中心值设计和容差分配的研究,该文提出了一种参数成品率中心值设计和容差分配耦合求解最优设计值的算法。该算法不需要设计者对电路或工艺的物理结构非常熟悉,从任意初始设计值和任意大小的容差,算法均可收敛到可接受域中的最优设计值。另外,根据工艺线的容差,算法可确定集成电路的最优参数成品率,也可根据实际要求选择适当容差的工艺线,以降低生产成本、提高效益。最后用实例证明了该算法的可行性和实用性,得到了满意的结果。     

低副瓣直线阵的最坏情况容差最优化方法研究

本文以激励系数为设计参量,讨论了直线阵的最坏情况容差最优化问题,给出了直线阵的最坏情况问题,固定容差问题和可变容差问题,研究了它们的新求解方法。固定容差问题将采用正多面体方法来求解,可变容差问题由一个双层迭代算法求解。可变容差问题以实际与理想方向图在容差域中的最大偏差为基准,协调阵列方向图的副(?)电平和激励系数的容差,使阵列方向图的副(?)电平满足所希望的指标要求,同时尽可能地放宽激励系数的容差,文中最后给出了6元和16元直线阵的最坏情况容差最优化的模拟计算结果。     

电路中心值最优化设计系统的研究与实现

本文提出一种容差域可伸缩的重心游移法来改进电路中心值设计的统计最优化方法＾［１］。该方法系统地解决了初始态下，成功点集为空集时的参数中心值设计问题。给出了容差域伸缩策略，并介绍了如何在ＳＰＩＣＥ软件基础上开发电路中心值优化系统的方法。     

最坏情况网络容差的最优化

本文阐述解决连续或离散最坏情况容差分配问题,以及在面向用户的计算机程序中的实现方法,这是应当和选择最适宜的标称设计同时解决的问题。容差问题实质上是在满足性能或别的指标要求的前提下指定误差的问题。我们的方法被认为是微波设计方面的新途径,能够解决各种容差和有关的问题。采用了达金(Dakin)的树形搜索法,新的准牛顿极小化法,以及最小的P次近似法。程序本身是妥善编成的,性能指标和约束条件的附加与删除,成本函数和最优化方法的代换等都容易实现。使用选择方式和拖欠值来扩大灵活性。完整的Fortran程序清单和有关文件将是有用的。     

DCTT统计最优化的进一步策略和方法＊

本文基于半无穷不可微最优化方法的框架和模型,提出了求解电路成品率极大的中心值设计,容差设计,调整设计及电路制造费用极小化为一体的统计最优化(DCTT)的求解方法及其策略。该方法不需要电路函数的凸性要求和构成可按受域的电路性能函数的半光滑假设。在引入抑制约束膨胀的策略后,该方法在确定性优化框架下可解决较大规模的统计最优化问题。为统计最优化的进一步发展开辟了一条新径。     

最佳容差设计问题的研究—正交内切球优化方法

本文提出最佳容差设计方法。该方法在设计中采用初始目标函数和准中心目标函数，使设计时间缩短；在优化中采用正交优化方法，使方法简单、直观；运用求可行域内大内切球半径和球心的方法，同时确定设计元件中心值和容差；为达到最佳设计之目的，本方法在最后进行一次中心值和容差的综合设计。本文给出例证，说明此方法可行，且有优越性。     

解决雷达对抗信号处理中容差问题的新方法

本文分析了雷达对抗信号处理中信号分选、识别的容差问题的特点，以及现有解决容差问题的方法，指出了解决容差问题的可能途径便是采用新的信号处理方法，使容差问题对信号处理不产生影响或基本上不产生影响，并介绍了两种具体的处理方法。对部分问题的计算机模拟表明，新的处理方法是可行的。     

专用集成电路的容差设计

专用集成电路的容差设计电子工业部第１４研究所丁定浩１电路容差设计的目的与意义电路的容差设计，是指在组成电路的元件参数精度等级一定的条件下，能使电路输出响应偏差最小的电路设计技术。它不仅是可靠性设计中保障电路性能参数保持稳定的主要设计手段，而且是提高生...     

A probabilistic design optimization for MEMS tunable capacitors

This paper presents a design optimization method for MEMS parallel-plate capacitors under fabrication uncertainties. The objective of the optimization problem is to maximize the production yield considering the fabrication tolerances. The method utilizes aspects of the advanced first-order second-moment (AFOSM) reliability method in probabilistic design to find a linearized feasible region for performance functions and uses an analytical double-bounded-probability distribution function (DB-PDF) to approximate the distribution of random variables. Then, it attempts to place the tolerance box in such a way that the portions of the box with higher yield lies in the feasible region. The yield is directly estimated using the joint cumulative distribution function (CDF) over the tolerance box requiring no numerical integration and saving considerable computational complexity for multidimensional problems. For this reason, any arbitrary distribution can be considered for random parameters and the problem is not restricted to normality assumptions. Numerical examples, verified by Monte-Carlo simulations, demonstrate that optimal designs significantly increase the yield. The advantage of the proposed design optimization method is that the yield can be maximized in early design stages without tightening tolerances or increasing the fabrication cost and complexity. The application of the presented method is not limited to tunable capacitors and can be extended to other MEMS devices.     

An optimization design method for chemically amplified resistprocess control

A novel optimization method called chemically-amplified resist process optimization design (CARPOD) applicable to the chemically amplified resist (CAR) process development is described. The method finds the optimal process conditions and the design center (maximum process tolerance space) of a CAR process with minimum experimental runs. First a modified response surface method is used to form the numerical response surface of a CAR, and its most sensitive point, which is the minimum requirement of X-ray dose, is located as an optimal process condition by an optimization method called POSM under the constraint of the contrast of the photoresist. Second, the design center is found to maximize the process tolerance space around the optimal process condition. Third, verifications are made on the optimal design as well as the design center. The process optimization of AZ PF-514 has been used as an example to show that the CARPOD method can identify the optimal process condition as well as the maximum tolerable parameter space with minimum experimental runs     

一种自动选择参数的雷达辐射源SVC分选方法

雷达信号分选的容差问题一直是雷达对抗情报处理中的难题,而支持向量聚类法（SVC）是一种能够有效避免容差问题的聚类方法,但现有参数搜索方法不能快速准确地确定SVC最优聚类结构,从而限制了支持向量聚类法的广泛应用。针对这一问题,提出了一种可以自动选择参数的SVC聚类方法。它通过采用一种综合的参数搜索方法,自动选择惩罚因子和核函数宽度两个参数,从而确定最优的聚类结构。仿真实验表明,此方法可在较少的迭代次数下获得最优的聚类结构,提高了雷达信号的分选正确率。     

LFM-PC复合调制信号优化设计算法

针对相位编码(Phase Coded,PC)信号在高速运动平台和高速目标探测中运用存在多普勒敏感问题,研究了一种基于线性调频(Linear Frequency Modulation,LFM)信号和相位编码信号复合调制的宽多普勒容限信号形式——LFM-PC复合调制信号.本文推导了LFM-PC信号的模糊函数,在多普勒容限内...     

Global Optimum Design of Uniform FIR Filter Bank With Magnitude Constraints

The optimum design of a uniform finite impulse response filter bank can be formulated as a nonlinear semi-infinite optimization problem. However, this optimization problem is nonconvex with infinitely many inequality constraints. In this paper, we propose a new hybrid approach for solving this highly challenging nonlinear, nonconvex semi-infinite optimization problem. In this approach, a gradient-based method is used in conjunction with a filled function method to determine a global minimum of the problem. This new hybrid approach finds an optimal result independent of the initial guess of the solution. The method is applied to some existing examples. The results obtained are superior to those obtained by other existing methods.      

遗传算法在计算机网络优化设计中的应用

计算机网络优化是一个多约束条件的非线性 0— 1问题。针对该问题的特点 ,建立了该问题的数学模型 ,给出了一个解决此问题的遗传寻优算法。通过计算机网络优化设计实例的计算结果表明 ,该方法能快速求出问题的全局近似最优解 ,且具有较高的计算精度     

计算机通信网中路由选择和容量分配问题的遗传算法求解

计算机通信网中，对路由选择和容量分配问题进行综合优化设计时的数学模型，是一个多约束条件的非线性０－１规划。本文以遗传算法的基本思想为基础，设计了一个求解该问题的遗传寻优算法，对计算机通信网优化设计实例的计算表明，这一方法能够迅速求出问题的全局近似最优解，并具有高的计算精确度。     

Channel allocation algorithms for WLANs using distributed optimization

The performance of a wireless local area network (WLAN) depends on the channel assignments among interfering access points (APs). Due to the limited number of non-overlapping channels, severe interference scenarios may arise if no appropriated spectrum planning is employed. In our study we focus on WLANs scenarios where APs may belong to different administrative domains, which is actually a very common situation in dense urban deployments. In such cases the use of centralized algorithms is not feasible and the already proposed distributed methods does not guarantee optimal channel assignment. In this paper, therefore, we formalize the channel allocation as a distributed constraint optimization problem (DCOP) and propose a new distributed channel assignment algorithm named DCAA-O, which can find the optimal solution to the channel assignment problem for a group of APs. A suboptimal strategy denoted DCAA-S is also investigated, which aims at reducing the number of control messages to be exchanged between APs, while still achieving a suboptimal solution which is very close to the optimal one. The simulation results show that the proposed algorithms are able to outperform the best known techniques both in terms of solution quality and number of exchanged messages.