User manual MICROSIM PSPICE OPTIMIZER

[. . . ] MicroSim PSpice Optimizer Analog Performance Optimization Software User's Guide MicroSim Corporation 20 Fairbanks Irvine, California 92618 (714) 770-3022 Version 8. 0, June, 1997. Printed in the United States of America. TradeMarks Referenced herein are the trademarks used by MicroSim Corporation to identify its products. MicroSim Corporation is the exclusive owners of "MicroSim, " "PSpice, " "PLogic, " "PLSyn. " Additional marks of MicroSim include: "StmEd, " "Stimulus Editor, " "Probe, " "Parts, " "Monte Carlo, " "Analog Behavioral Modeling, " "Device Equations, " "Digital Simulation, " "Digital Files, " "Filter Designer, " "Schematics, " "PLogic, " "PCBoards, " "PSpice Optimizer, " and "PLSyn" and variations theron (collectively the "Trademarks") are used in connection with computer programs. MicroSim owns various trademark registrations for these marks in the United States and other countries. [. . . ] In many cases, there are several legitimate ways to describe the design. Example: Assume you want to design a resistive terminator that produces an output voltage of 3. 75 V (± 0. 1 V) at the junction of the two resistors, and the Thevenin equivalent resistance of the resistor combination must equal 100 (±1 ). Your objective is to find the best resistor values to meet these two specifications: · · output voltage of 3. 75 V (Ve) equivalent resistance of 100 (Re) Figure 4-1 Resistive Terminator Circuit You can manually solve this problem using the following simultaneous equations: 5R 2 ------------------ = 3. 75 R1 + R2 1 2 and ------------------ = 100 R +R 1 2 R R These equations solve to R1 = 133. 3 and R2 = 400 , an exact solution. Constrained Optimization 4-3 When using the PSpice Optimizer, you can set up this problem in one of three ways: · · Consider Ve and Re as equally important; set up both as goals. Consider Ve as the most important requirement to meet, even at the expense of Re; set up Ve as a constraint and Re as a goal. Consider Re as the most important requirement to meet, even at the expense of Ve; set up Re as a constraint and Ve as a goal. · Note Because at least one optimization goal is necessary, the case where both Ve and Re are constraints is excluded. If the problem, like this one, has a solution, the PSpice Optimizer might arrive at the same answer for all three methods. However, most problems do not have a single, exact solution as this one does. For most designs, the result is a compromise that minimizes the goals while not violating the constraints. Constrained Optimization Many problems in analog circuit optimization are naturally expressed as the minimization of a function representing a goal (e. g. , power consumption) which is subject to one or more constraints (e. g. , bandwidth). Constraints are typically complicated nonlinear functions of the parameters of the problem, so manual optimization is a difficult task. Most other analog circuit optimizers implement only unconstrained optimization of a single goal or a sum-of-squares of several goals. To tackle a problem like the problem outlined above, other optimizers must combine the functions for the goals and constraints and then optimize the combination. Unfortunately, this scheme does not differentiate between reduction of the goals and violation of the constraints. In 4-4 Understanding Optimization Principles and Options general, constraints are given much greater weight than the goals. In particular: · · · If a very large value is used for the weight of the constraints, numerical problems occur. If a more reasonable value is used, the result is not a true solution of the original problem. Using a sequence of weights, and performing a series of minimizations can lead to the true solution, but at the expense of a large increase in optimization time (because of all of the extra evaluations required to solve the intermediate problems). The PSpice Optimizer implements both constrained and unconstrained minimization algorithms. This means that the optimizer: · · Tackles constrained problems directly and efficiently. Calculates Lagrange multipliers for the solution, which provide valuable insight to design tradeoffs. Types of Constraints Constraints are restrictions placed on potential solutions to optimization problems. The simplest constraints are bound constraints--simple limits on the ranges of the parameters (e. g. , a resistor whose value has to be at least 100 ). More challenging constraints that frequently arise in analog circuit optimization have dependencies on other characteristics of the design. Example: Consider optimizing a MOS amplifier cell which must satisfy these specifications: · · Reduce power consumption. Make sure gain-bandwidth product of the cell is greater than or equal to some minimum value. Constrained Optimization 4-5 The power consumption of the cell is the goal (the characteristic to be minimized) and the gain-bandwidth product is the inequality constraint. To continue the example, consider the dependence of power consumption and gain-bandwidth product on bias current in one of the amplifier stages. [. . . ] Create a separate file containing the optimization parameter definitions. Using the PSpice Optimizer, complete the setup and run the optimization as usual. Refer to your PSpice user's guide for any questions you might have on circuit-file syntax. The remaining sections explain what you need to do once your design is defined as a circuit file. Setting Up the Circuit File C-3 Setting Up the Circuit File To use the PSpice Optimizer to optimize a design defined as a netlist 1 2 3 4 Decide on the design parameters you want the optimizer to vary. Place parameter definitions (. PARAM) in a separate parameters file (. par). [. . . ]