Advanced Model Order Reduction Techniques in VLSI Design

2.2: Moment Computation in MNA Formulation

2.2 Moment Computation in MNA Formulation

In this section, we present the method for efficient computation of moments using classic circuit analysis techniques. We start with the modified nodal analysis (MNA) formulation of the general RLC linear circuits and derive the recursive moment computation formula, which is the most critical step for all the moment-matching methods and Krylov subspace projection-based model order reduction methods.

2.2.1 Recursive moment computation

For a general linear network, we can apply modified nodal analysis to formulate it in the state space equation form


where G and C are the conductive and storage element matrices; B and L are the input and output positions matrices; and state variables x can be nodal voltage or branch currents of the linear circuit.

Upon applying the Laplace transformation of the state equation, we have the state equations in the s domain


Assuming the initial condition is zero, X(0)=0, and the impulse response ( U(s)= 1) is applied, the state equation will become


Expanding X(s) using Taylor s series at s=0, we obtain


We then obtain the state moment computation formula in a recursive form


Notice that G ?1 here means we solve G x= b and G ?1 is used for solving for all the moments. Numerically, we only need to perform one LU decomposition of G= LU and then use the L and U matrices to solve for all moments...

UNLIMITED FREE
ACCESS
TO THE WORLD'S BEST IDEAS

SUBMIT
Already a GlobalSpec user? Log in.

This is embarrasing...

An error occurred while processing the form. Please try again in a few minutes.

Customize Your GlobalSpec Experience

Category: Inertial Sensor Chips
Finish!
Privacy Policy

This is embarrasing...

An error occurred while processing the form. Please try again in a few minutes.