Pavan Gunupudi
Project Area: CAD for VLSI
Project Title: Sensitivity computation of parameters in nonlinear DAEs
Project Description: Parameters such as resistances, capacitances, inductances, transistor dimensions etc. influence the operation of circuits. The effects of variation in these parameters greatly change circuit responses and are of serious concern during the design and fabrication stages of the circuit. The sensitivity of circuit responses with respect to these parameters ie. the effect of fluctuations of parameters on system responses identify which parameters have the greatest impact on circuit performance. This information is crucial in designing systems that are robust to parameter variations.
A brute force way of computing time-domain sensitivities of nonlinear circuits is to derive a linear time-varying differential algebraic equation and solve it using transient analysis. Though the sensitivity of all circuit responses can be found using this method it is highly inefficient as it leads to solution of a large dense matrix unknown called the sensitivity matrix.
Sometimes the sensitivity of all circuit responses need not be known. It is sufficient to compute the sensitivity of few of the circuit responses. A selection vector is chosen to select only a few responses out of the sensitivity matrix. This method is known as adjoint sensitivity. Students will explore a technique called TADsens to compute the sensitivity of selected time domain responses using an adjoint sensitivity analysis.
Desired Technical Skills:
- Forming the MNA matrix given a circuit
- Finding the solution of time-varying nonlinear algebraic differential equations using MATLAB
Suitable for:
- ELEC 4907 – Electrical Engineering
- ELEC 4908 – Engineering Physics
- SREE 4907 – Sustainable and Renewable Energy Engineering