Please use this identifier to cite or link to this item:
|Title:||Time-domain optimization of amplifiers based on distributed genetic algorithms|
|Authors:||Tavares, Rui Manuel Leitão Santos|
|Publisher:||Faculdade de Ciências e Tecnologia|
|Abstract:||The work presented in this thesis addresses the task of circuit optimization, helping the designer facing the high performance and high efficiency circuits demands of the market and technology evolution. A novel framework is introduced, based on time-domain analysis, genetic algorithm optimization, and distributed processing. The time-domain optimization methodology is based on the step response of the amplifier. The main advantage of this new time-domain methodology is that, when a given settling-error is reached within the desired settling-time, it is automatically guaranteed that the amplifier has enough open-loop gain, AOL, output-swing (OS), slew-rate (SR), closed loop bandwidth and closed loop stability. Thus, this simplification of the circuit‟s evaluation helps the optimization process to converge faster. The method used to calculate the step response expression of the circuit is based on the inverse Laplace transform applied to the transfer function, symbolically, multiplied by 1/s (which represents the unity input step). Furthermore, may be applied to transfer functions of circuits with unlimited number of zeros/poles, without approximation in order to keep accuracy. Thus, complex circuit, with several design/optimization degrees of freedom can also be considered. The expression of the step response, from the proposed methodology, is based on the DC bias operating point of the devices of the circuit. For this, complex and accurate device models (e.g. BSIM3v3) are integrated. During the optimization process, the time-domain evaluation of the amplifier is used by the genetic algorithm, in the classification of the genetic individuals. The time-domain evaluator is integrated into the developed optimization platform, as independent library, coded using C programming language. The genetic algorithms have demonstrated to be a good approach for optimization since they are flexible and independent from the optimization-objective. Different levels of abstraction can be optimized either system level or circuit level. Optimization of any new block is basically carried-out by simply providing additional configuration files, e.g. chromosome format, in text format; and the circuit library where the fitness value of each individual of the genetic algorithm is computed. Distributed processing is also employed to address the increasing processing time demanded by the complex circuit analysis, and the accurate models of the circuit devices. The communication by remote processing nodes is based on Message Passing interface (MPI). It is demonstrated that the distributed processing reduced the optimization run-time by more than one order of magnitude. Platform assessment is carried by several examples of two-stage amplifiers, which have been optimized and successfully used, embedded, in larger systems, such as data converters. A dedicated example of an inverter-based self-biased two-stage amplifier has been designed, laid-out and fabricated as a stand-alone circuit and experimentally evaluated. The measured results are a direct demonstration of the effectiveness of the proposed time-domain optimization methodology.|
|Description:||Thesis presented in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the subject of Electrical and Computer Engineering|
|Appears in Collections:||FCT: DEE - Teses de Doutoramento|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.