Circuit Simulation Optimization2016-11-19T00:28:03+00:00

case study

Industry: Electronics, Consumer Products, Office Products

Circuit Simulation Optimization

FOCUS: Optimize the design of analog circuit using simulation and DOE Matrices


Analog circuits are often designed and simulated using simulation software such as Spice, or one of its derivatives. Optimization techniques in the past have been somewhat limited, or based on the experience of the engineer. The engineer desired a more effective way to optimize the circuit prior to building prototypes.


The engineer built the initial model as usual, using pSpice. A full factorial design matrix was chosen at 3 levels per factor. Four factors were chosen, including resistor values (ohms), capacitor values (microfarads).  There were multiple responses, the most important of which was a rise time of a certain signal.  The full factorial design was chosen for many reasons. It provides analysis of the n-way interactions. It also provides full orthogonality of the quadratic terms. A second choice would have been a Central Composite (Box-Wilson) design. The number of runs in the full factorial design (34 = 81) was acceptable in this case, since the simulator provided these values almost immediately. Had the simulator taken much more time to provide a response value, a different design probably would have been chosen. Since the simulator was deterministic, there was no need to run additional runs. As such, the full regression model was used, without dropping any terms from the model (in the case of deterministic simulators, there is no experimental noise, so p-values represent only magnitude of effects). Multiple-response optimization was then used to optimize the design.  Next closest standard component values were chosen, and modeled with the regression model.  Monte Carlo simulation was performed for expected value analysis. Computer-based parameter design was used to shift the component values for robustness. Computer-based Tolerance Analysis was used in parts tolerance decisions.


Circuit was designed faster and optimized significantly better than past designs, with significantly lower prototype costs.

* Due to nondisclosure agreements, the organization referenced in this example cannot be disclosed.