Study On Design And Optimization Methods Of Analog Integrated Circuits

With the development of integrated circuit manufacturing technology,the feature size continues to decrease,leading to the emergence of non-ideal effects of analog integrated circuits,and the increasing difficulty of design.Meanwhile,it is hard to find mature analog integrated circuit design automation tools on the market.The research conducted on analog integrated circuit design and optimization methods aims to improve design efficiency and shorten the design cycleDesign of analog integrated circuits includes topology selection,circuit parameter tuning,layout design,manufacturing and testing,etc.This dissertation focuses on the first two steps.For circuit topology,designers generally select an appropriate topology from existing topology libraries that can meet design specifications most of the time.When there are performance requirements that can not be met,circuit designers can modify circuit topology and optimize circuit parameters.For an average circuit designer,manual parameter tuning work can be a repetitive process,especially for complex circuits due to the high nonlinearity relation between circuit performances and design variables.Therefore,it is important to study automated optimization method of analog circuit parameters in order to improve design efficiency.Starting from the design and optimization of complex band-pass filters,this dissertation analyzes the cause for in-band ripple and proposed a compensation scheme for traditional active-RC complex band-pass filters.For an analog circuit parameter optimizer,there could be different selections for both optimization algorithms as well as performance evaluators.This dissertation studied several algorithms including global algorithms and the local minimum search(LMS)algorithms.For circuit performance evaluation,this dissertation tries to solve the evaluation efficiency problem by introducing a model-based evaluator.The main contents are summarized as follows:1.To reduce the passband ripple of active-RC complex band-pass filter(CBPF),this dissertation derives the formula for the process of shifting a low-pass filter to a CBPF.It is found that the limited gain-bandwidth product of the operational amplifiers(Op-Amp)is the main reason for the nonlinear spectrum shift from low-pass filter to CBPF.Based on this,this dissertation adjusts the values of cross-coupling resistors and adds compensation capacitors to the topology of the traditional active-RC CBPF,and reduces the passband ripple of the active-RC CBPF.Based on this compensation method,an active-RC CBPF is implemented in TSMC 0.13 ?m CMOS technology with a center frequency of 12.24 MHz,a bandwidth of 9 MHz,and a passband ripple less than 1 dB.The simulation results and tape test results verify the effectiveness of the new passive compensation method2.To improve the efficiency of the simulation-based global optimization,this dissertation proposes a new analytical model and simulation-assisted analog parameter automatic optimization method.It first performs an exhaustive global search using an analytical model to locate a few most promising regions,and then cast simulation-based LMS for those regions to find the best design point.The system combines the efficiency of the analytical model and the accuracy of the SPICE simulations,provides a possible solution to solve the long convergence time and insufficient search space coverage associated with general simulation-based global optimization methods.This dissertation derives the analytical model of the 5th-order CBPF and uses this method to automatically optimize the parameters of the active-RC CBPF with traditional topology.Compared with the simulation-based classic method,this approach consumes less time while achieving similar results.3.To improve the efficiency of SPICE-based global and local combined optimization,this paper presents a new analog circuit optimization system for automated sizing of analog integrated circuits.It consists of a GA-based global optimization engine and a machine learning-based local optimization engine.The proposed method adopts local ANN models instead of global models to address the problem of the large training data sets needed by the global models.The key new idea is to use parallel computation to train ANN models for design space neighborhoods thus the LMS can have a much faster search speed.For the GA-based global optimization,circuit performances are calculated by parallel SPICE simulations.For the LMS,circuit performance data is derived from ANN model predictions instead of SPICE simulations.Since most time for an ANN-based LMS is spent on SPICE calls which can be run in parallel,the LMS process can also exploit the multiple core configuration of a modern computational server in addition to the GA global search and speed up the optimization while maintaining good solution accuracy4.This dissertation realizes the parameter automatic optimization of a two-stage rail-to-rail operational amplifier,a 5th-order active-RC Chebyshev CBPF,and a three-stage operational amplifier using the proposed ANN assisted optimization method(GSLA)and two reference methods which are simulation-based GA optimization method(GS)and simulation-based GA combined with simulation-based LMS(GSLS).Compared with the GS method,the proposed GSLA method provides better optimization results;compared with the GSLS method,the proposed GSLA method provides about more than three times speed enhancement and comparable results.The main contributions are:1.This dissertation proposes a new passive compensation method for the active-RC CBPF and reduces the passband ripple of the active-RC complex band-pass filter.2.This dissertation proposes a new analytical model and simulation-assisted analog parameter automated optimization method.It reduces the convergence time and improves search space coverage associated with general simulation-based global optimization methods.3.This dissertation proposes an automated analog circuit optimization system,which consists of a GA-based global optimization engine and an ANN model-based local optimization engine.It realizes a fully parallelized optimization system and improves the efficiency of a global and local combined optimization.