<html>
<body>
Hello Christophe,<br><br>
<blockquote type=cite class=cite cite>Did you try to solve the linear
system with a direct solver<br>
(Algorithm=11 in solver.par)?</blockquote><br>
Thank you very much for this advice. It finaly led me to some
results.<br>
Here you can see what I have done:<br><br>
1) first trial:<br><br>
Algorithm=11 and sparse matrix, and<br>
I received the message:<br><br>
<font face="Courier New, Courier">GetDP : Error :
Unknown algorithm for sparse matrix solver<br><br>
</font>2) second trial<br><br>
Algorithm=11 and full matrix and this time I obtained<br>
the solution which seems to be OK, for coarse mesh and low
frequency.<br>
However, for a dense mesh, there are problems with the memory
(obviously<br>
due to the full storage), and I cannot solve the problem at all.
<br><br>
3) other trials with iterative algorithms<br><br>
For a relatively dense mesh (which I cannot solve with LU due to memory
limitation),<br>
I have tried all the other available iterative algoritms. The only<br>
one that converged was algorithm=10, but the convergence was extremely
slow.<br>
I improved its convergence by using scaling=4. <br><br>
<font face="Courier New, Courier">
Matrix_Format
1<br>
Matrix_Printing
0<br>
Matrix_Storage
0<br>
Scaling
4<br>
Renumbering_Technique
1<br>
Preconditioner
2<br>
Preconditioner_Position
2<br>
Nb_Fill
20<br>
Permutation_Tolerance
0.05<br>
Dropping_Tolerance
0<br>
Diagonal_Compensation
0<br>
Re_Use_ILU
0<br>
Algorithm
10<br>
Krylov_Size
40<br>
IC_Acceleration
1<br>
Re_Use_LU
0<br>
Iterative_Improvement
0<br>
Nb_Iter_Max 5000<br>
Stopping_Test 1e-10<br><br>
</font>In this case, for 100Hz it takes 154 iterations, but, if I want to
increase<br>
the frequency (e.g. 10 KHz) it takes 817 iterations.<br><br>
So, I wonder if:<br>
- Can I use a direct solver for sparse matrices?<br>
- Is there any possibility to improve the convergence of the iterative
method that<br>
worked for this problem (algorithm=10) so that I would be able to work
with high<br>
frequencies (up to GHz maybe...)?<br><br>
I also noticed that, when using several frequencies in the same
resolution, the<br>
number of iterations is greater than the maximum number of iterations of
the<br>
independent resolutions for each frequency. I assume that this is due to
different<br>
initializations. So, what do you reccommend? Is it better (faster) to
solve for <br>
each frequency separately?<br><br>
Thanks a lot for your helpful support.<br><br>
With kind regards,<br>
Florin<br><br>
<br>
<x-sigsep><p></x-sigsep>
Florin Ciuprina, Ph.D.<br>
Associate Professor<br>
----------------------------------------------------------------------------<br>
POLITEHNICA University of Bucharest<br>
Electrical Engineering Faculty<br>
Electrical Materials Laboratory<br>
Spl.Independentei 313, 77206, Bucharest, Romania<br>
Tel: +40.21.402 92 91<br>
Fax: +40.21.410 43 55<br>
e-mail: florin@elmat.pub.ro<br>
----------------------------------------------------------------------------</body>
</html>