MP ME ST PR PRN DS DSS <> <> EH <> PP <> EME MFS

*Method*Mode-extraction method to be used for the modal analysis.

**LANB**— Block Lanczos

**LANPCG**— PCG Lanczos

**SNODE**— Supernode modal solver

**REDUC**— Householder (reduced)

**UNSYM**— Unsymmetric matrix

**DAMP**— Damped system

**QRDAMP**— Damped system using QR algorithm

**VT**— Solve with the Variational Technology method of ANSYS DesignXplorer

*NMODE*The number of modes to extract. The value can depend on the value supplied for

*Method*. Defaults to the number of master DOFs when*Method*= REDUC. For the other methods,*NMODE*has no default and must be specified. If*Method*= LANB, LANPCG, or SNODE, the number of modes that can be extracted can equal the DOFs in the model after the application of all boundary conditions.Recommendation:

When *Method*= REDUC,*NMODE*should be less than half of the number of master DOFs.When *Method*= LANPCG,*NMODE*should be less than 100 to be computationally efficient.When *Method*= SNODE,*NMODE*should be greater than 100 for 2-D plane or 3-D shell/beam models and greater than 250 for 3-D solid elements to be computationally efficient.*FREQB*The beginning, or lower end, of the frequency range of interest.

For

*Method*= LANB, UNSYM, DAMP, and QRDAMP,*FREQB*also represents the first shift point for the eigenvalue iterations. If values for UNSYM or DAMP are zero or blank, the default value is -1.0. For the other methods, the default is internally computed. Eigenvalue extraction is most accurate near the shift point; multiple shift points are used internally in the LANB and QRDAMP methods. For LANB, LANPCG, UNSYM, DAMP, and QRDAMP methods with a positive*FREQB*, eigenvalues are output beginning at the shift point and increase in magnitude. For UNSYM and DAMP methods with a negative*FREQB*value, eigenvalues are output beginning at zero magnitude and increase.Choosing higher

*FREQB*values with the LANPCG and SNODE methods may lead to inefficient solution times because these methods will find all eigenvalues between zero and*FREQB*before finding the requested modes between*FREQB*and*FREQE*.*FREQE*The ending, or upper end, of the frequency range of interest (in Hz). The default for

*Method*= SNODE is described below. The default for all other methods is to calculate all modes, regardless of their maximum frequency.The default is 100 Hz for

*Method*= SNODE. To maintain solution efficiency, you should not set the*FREQE*value too high; for example, not higher than 5000 Hz for an industrial problem. The higher the*FREQE*value used for the SNODE method, the more solution time it will take and the more eigenvalues it could produce. For example, if*FREQE*is set to 1e8, it will cause the underlying supernodal structures to find all the possible eigenvalues of each group of supernodes; hence, it will take an excessive amount of solution time.*Cpxmod/PRMODE**CPXMOD*(Valid only when*Method*= QRDAMP).**ON**— Calculate complex eigenmode shapes.

**OFF**— Do not calculate complex eigenmode shapes. This is the default.

*PRMODE*The number of reduced modes to print. Valid only when

*Method*= REDUC.*Nrmkey*Mode shape normalization key:

**OFF**— Normalize the mode shapes to the mass matrix (default).

**ON**— Normalize the mode shapes to unity instead of to the mass matrix. If a subsequent spectrum or mode superposition analysis is planned, the mode shapes should be normalized to the mass matrix (

*Nrmkey*= OFF).*--*Unused field.

*BlockSize*The block vector size to be used with the Block Lanczos eigensolver (used only when

*Method*= LANB).*BlockSize*must be an integer value between 0 and 16. When BlockSize = zero or blank, the code decides the block size internally (normally, a value of 8 is used). Typically, higher*BlockSize*values are more efficient under each of the following conditions:When running in out-of-core mode and there is not enough physical memory to buffer all of the files written by the Block Lanczos eigensolver (and thus, the time spent doing I/O is considerable).

Many modes are requested (>100).

Higher-order solid elements dominate the model.

The memory usage only slightly increases as

*BlockSize*is increased. It is recommended that you use a value divisible by 4 (4, 8, 12, or 16).*--*Unused field.

*Scalekey*Matrices scaling key for acoustic-structural interaction:

**OFF**— Do not scale the matrices (default).

**ON**— Scale the matrices.

Specifies modal analysis (**ANTYPE**,MODAL) options.
Additional options used only for the Supernode (SNODE) eigensolver
are specified by the **SNOPTION** command. If *Method* = LANPCG, ANSYS automatically switches to the
PCG solver internally for this modal analysis. You can further control
the efficiency of the PCG solver with the **PCGOPT** and **EQSLV** commands.

The Block Lanczos method is strongly recommended for high-frequency
magnetic eigenvalue problems. The initial frequency guess is not critical.
The ratio of *FREQE* to *FREQB* can be up to 1e6. The PCG Lanczos method is not supported for high-frequency
magnetic eigenvalue problems.

For models that involve a non-symmetric element stiffness matrix,
as in the case of a contact element with frictional contact, the QR
damp eigensolver (**MODOPT**, QRDAMP) extracts modes
in the modal subspace formed by the eigenmodes from the symmetrized
eigenproblem. The QR damp eigensolver symmetrizes the element stiffness
matrix on the first pass of the eigensolution, and in the second pass,
eigenmodes are extracted in the modal subspace of the first eigensolution
pass. For such non-symmetric eigenproblems, you should verify the
eigenvalue and eigenmode results using the non-symmetric matrix eigensolver
(**MODOPT**, UNSYM ).

The UNSYM, DAMP, and QRDAMP options cannot be followed by a subsequent spectrum analysis.

This command is also valid in PREP7.

**Distributed ANSYS Restriction. **All extraction methods,
except VT, are supported within Distributed ANSYS. However, PCG Lanczos, UNSYM,
and DAMP are the only distributed eigensolvers that will run a fully
distributed solution. The Block Lanczos and Supernode eigensolvers
are not distributed eigensolvers; therefore, you will not see the
full performance improvements with these methods that you would with
a fully distributed solution. The REDUC and QRDAMP methods are supported,
but do not use distributed memory parallelism within Distributed ANSYS.

Command Option Method | Available Products |

LANB | MP ME ST PR PRN DS DSS <> <> EH <> PP <> EME MFS [1] |

LANPCG | MP ME ST PR PRN DS DSS <> <> <> <> PP <> EME MFS [1] |

SNODE | MP ME ST PR PRN <> <> <> <> EH <> PP <> EME MFS |

REDUC | MP ME ST PR PRN DS DSS <> <> <> <> PP <> EME MFS [1] |

UNSYM | MP ME ST <> <> <> <> <> <> <> <> PP <> EME MFS |

DAMP | MP ME ST <> <> <> <> <> <> <> <> PP <> EME MFS |

QRDAMP | MP ME ST <> <> <> <> <> <> <> <> PP <> EME MFS |

VT | <> <> <> <> <> <> <> <> <> <> <> <> VT <> <> |