QUESTION:
The inductance differ greatly by ANSYS7.0 and ANSYS8.1.
Which result is right?
ANSWER:
The 8.1 result is probably the correct one. I know of 2 bugs (descriptions appear below) that have been corrected that may be responsible for the difference you see at releases 7.0 and 8.1. If the model you see the difference in is 3D DSP, then it is likely that the second defect below is responsible for the difference you see.
1) Defect 14461: Error in LMATRIX with bfa
LMATRIX was not developed for solid model loads.
To allow LMATRIX work for the user problem with BFA load,
an undocumented BFASCALE command only applied in LMATRIX is developed. Note, this solution is not general and fully tested; it resolves the issue of this error report and let customers use LMATRIX with BFA load. We should come back to this issue later at development and address it with our rigorous verification procedures.
2) Defect 20637: DSP / Lmatrix incorrect with linear permeability
Incorrect inductance coefficients may be provided
by the LMATRIX command macro if both of the
the following conditions are satisfied:
 LMATRIX is applied after an operating (nominal) solution is computed
by the magnetic differential scalar potential (DSP) method
 the relative permeability of the magnetic region
is high (mur > 50,000).
The operating (nominal) solution computed by DSP method is correct LMATRIX provides correct numbers for mur < 50,000 after DSP method or if the operation solution is calculated by magnetic vector potential (MVP) methods (nodal or edge) regardless of mur values.
The following source changes were needed to implement
the new incremental DSP method:
scmgfl.F
el96.F
magma5.F
gtacti.F
LMATRIX.MAC
These tests have been added to the set to verify correct resolution: ev9665s ev9666
Results obtained after the code change look great. The 3D MSP inductances now agree very well with those obtained for 2D MVP and inductance asymptotically approaches a constant v
QUESTION: The inductance differ greatly by ANSYS7.0 and ANSYS8.1. Which result is right? ANSWER: The 8.1 result is probably the correct one. I know of 2 bugs (descriptions appear below) that have been corrected that may be responsible for the difference you see at releases 7.0 and 8.1. If the model you see the difference in is 3D DSP, then it is likely that the second defect below is responsible for the difference you see. 1) Defect 14461: Error in LMATRIX with bfa LMATRIX was not developed for solid model loads. To allow LMATRIX work for the user problem with BFA load, an undocumented BFASCALE command only applied in LMATRIX is developed. Note, this solution is not general and fully tested; it resolves the issue of this error report and let customers use LMATRIX with BFA load. We should come back to this issue later at development and address it with our rigorous verification procedures. 2) Defect 20637: DSP / Lmatrix incorrect with linear permeability Incorrect inductance coefficients may be provided by the LMATRIX command macro if both of the the following conditions are satisfied:  LMATRIX is applied after an operating (nominal) solution is computed by the magnetic differential scalar potential (DSP) method  the relative permeability of the magnetic region is high (mur > 50,000). The operating (nominal) solution computed by DSP method is correct LMATRIX provides correct numbers for mur < 50,000 after DSP method or if the operation solution is calculated by magnetic vector potential (MVP) methods (nodal or edge) regardless of mur values. The following source changes were needed to implement the new incremental DSP method: scmgfl.F el96.F magma5.F gtacti.F LMATRIX.MAC These tests have been added to the set to verify correct resolution: ev9665s ev9666 Results obtained after the code change look great. The 3D MSP inductances now agree very well with those obtained for 2D MVP and inductance asymptotically approaches a constant value as relative permeability becomes semiinfinite (all reluctance is dropped in the airgap in this model). 

