Design Optimization Of CMOS Operational Amplifier Based On Gm/Id Method

Optimization based design of analog integrated circuit is an important part of analog EDA(Electronic Design Automation),which has been paid considerable attentions to for many years.At present,the optimization of analog circuits is carried out by different kinds of optimization algorithms,which use the device sizes as the optimization variables.In order to obtain accurate and global optimal solution,global optimization algorithms based on the transistor level circuit simulation are employed,which often consume a lot of computations,so optimization based circuit design has not been widely applied in practice.This thesis focuses on the optimization of CMOS operational amplifiers based on the g_m/I_dmethodology.By changing the traditional optimization variables from the device sizes to the g_m/I_dand bias currents which are constrained by the KVL and KCL,the number of optimization variables and the associated optimization space can be significantly reduced.By utilizing the human experiences or knowledge about the circuit under design(the range of input common voltage and the output swing,current distribution rules of the circuits),the number of optimization variables can be further reduced.For some commonly used operational amplifiers,the dimension of the design space can be reduced to 3～5.The thesis first gives the method to select the topologically independent g_m/I_dand I_dvariables using graph theory algorithms.Then,the multi-objective genetic algorithm NSGA-II is used to carry out the optimization for some commonly used single and multistage amplifiers,in which the dimensions are further reduced by the utilization of the prior knowledges on these circuits.The Pareto fronts of performance versus area are obtained under different power consumption levels.By comparison of the optimization results before and after dimension reduction,it is shown that the same optimization results can be achieved,and the computational amounts can be greatly reduced.Since there are only 3～5 independent design variables after dimension reduction for these circuits,and the dimension can be further reduced by the constraint of the phase margin specification for muti-stage amplifiers,the DSE(Design Space Exploration)of the circuits can be carried out by the simple parameter sweeping.The DSE provides an effective means to observe the comprehensive affections of different design variables on the circuit performance.The design spaces of the folded and cycle folded common-source common-gate amplifiers,two-stage Miller-compensated Op Aamps,are explored by the parameter sweeping,and the effects of different g_m/I_dvariables on the circuit performance are fully demonstrated.The DSE results provide effective and clear guidance for the selection of the g_m/I_dparameters in the practical design of these amplifiers.