Parameter Name 
Valid Entries 
Description 


n1.0, n2.0 
The basis set dictates the accuracy of the field representation within each mesh element. The higher the basis set order, the more functions are used, generally resulting in a more accurate result. For more information, see “Basis Set”. 

Direct, Iterative, Automatic 
This parameter controls the type of solver that is used to solve the matrix equation(s). The options are Direct and Iterative. The Direct solver performs an LU factorization of the matrix, and the iterative method uses a variant of the preconditioned conjugate gradient (PCG) method. The Iterative setting will force the use of an iterative technique that does not rely on factorization of the system matrix. One of several iterative methods can be selected, each with distinctive characteristics that make them applicable in different situations. Generally, the direct methods require more memory than the iterative methods but are more reliable and are better for use with multiple right handsides. The iterative techniques can require much less memory than direct methods, but may or may not converge, so some experimentation may be required on any given problem. The Automatic selection is equivalent to Iterative, as this is typically faster than the Direct option. For more information, see “Linear Solver”. 

Automatic, HMLU, MFLU 
Visible if Linear Solver/Method is Direct. Choosing Direct for the linear solver method parameter forces the solver to use one of two methods that rely on an LU factorization of the system matrix – MFLU and HMLU. The Automatic setting allows the solver to choose MFLU or HMLU depending on the project size and computational resources, to obtain the result more quickly. For more information, see “Direct Method”. 

Automatic, Low, Medium, High, Perfect 
The HMLU linear solver produces an inexact LU factorization of the system matrix that is subsequently used in an iterative loop to determine the solution vectors for port/mode excitations. The accuracy of the factor can be controlled via the Factor Accuracy Level solver parameter. Options are Low, Medium, High, Perfect, and Automatic. The Perfect setting yields a factor with an accuracy equivalent to that obtained using the MFLU solver, and will result in the iterative loop converging in a single step. A lower factor accuracy is faster to compute and takes less system memory, as compared to higher accuracy settings, but it can also lead to more iterations being required in the iterative loop. If the Automatic setting is chosen, the solver will adjust the accuracy setting as necessary to minimize runtime and memory, starting with the Low setting. The Automatic setting also allows for failover to a different solver if the default solver cannot solve the problem. 

CG, BiStabCG, CGS, GMR 
Visible if Linear Solver/Method is Iterative. Several iterative linear solver methods are available. The CG option invokes the standard preconditioned conjugate gradient method, BiStabCG invokes the stabilized variant of biconjugate gradient, CGS is the conjugate gradient squared method, and GMR is the generalized minimum residual method. Of these, only GMR has the property that the residual error will reduce every iteration, but CG is generally the most efficient. For more information, see “Iterative Methods”. 

Jacobi, GaussSeidel 
Visible if Linear Solver/Method is Iterative. Use of a matrix preconditioner improves the convergence rate iterative methods. The options are diagonal scaling (Jacobi) and GaussSeidel. Jacobi is more efficient in terms of computer memory. 

Real number > zero 
Visible if Linear Solver/Method is Iterative. When the residual of the matrix equation falls below this value the linear solver considers the result converged. Setting this value too low can cause very long solution times, and setting it too high will result in breakdown of the simulation. 

Integer > zero 
Visible if Linear Solver/Method is Iterative. This is the maximum number of iterations the linear solver will perform before stopping, irrespective of the size of the matrix equation residual. 
Please send email to awr.support@cadence.com if you would like to provide feedback on this article. Please make sure to include the article link in the email.