Next: Using gd1.pp to analyse
Up: Analysing the real 3D
Previous: Defining Material properties
When we feed the so edited inputfile to gd1 (gd1 
wPlunger02.gdf),
gd1 gives as final table (on a fast machine):
boundary conditions:
xboundary= electric, electric
yboundary= electric, electric
zboundary= electric, electric
--------------------
i freq(i) acc(i) cont(i)
1 405.8365e+06 0.9962102358 1.0000000000 # "grep" for me
2 508.3705e+06 0.0010604778 0.0010350326 # "grep" for me
3 782.2763e+06 0.0000610467 0.0000856254 # "grep" for me
4 785.5886e+06 0.0005086537 0.0008213667 # "grep" for me
5 1.0476e+09 0.0000269777 0.0000532244 # "grep" for me
6 1.0620e+09 0.0002841477 0.0048480407 # "grep" for me
7 1.0930e+09 0.0002171685 0.0038237247 # "grep" for me
8 1.0985e+09 0.0001092116 0.0013357829 # "grep" for me
9 1.1436e+09 0.0000607043 0.0004803485 # "grep" for me
10 1.2149e+09 0.0035331474 0.0303398453 # "grep" for me
11 1.2437e+09 0.0067656728 0.1472070583 # "grep" for me
12 1.2657e+09 0.0690266113 1.0000000000 # "grep" for me
13 1.2844e+09 0.0268558446 0.5452387430 # "grep" for me
14 1.3159e+09 0.0379465992 0.3263500622 # "grep" for me
15 1.3920e+09 0.0940570777 0.8370331159 # "grep" for me
################################
# cpu-seconds for eigenvalues : 496
# start time : 12:26:40
# end time : 12:39:16
# date : 17/09/99
# The computation of the eigenvalues has finished normally..
# Start the postprocessor to look at the results.
stop .. normal end ..
The first mode is garbage, as can be seen from its accuracy.
But also the other modes are not very good.
The reason is: Since we have only a single symmetry plane left (z=0),
we have to compute in a volume that is four times as large as the volume
before. In such a large volume, there are much more modes than 15 in the
frequency range from 0 to estimation (we specified
an estimation of 2GHz).
We could say that we want to have more solutions, but that would
drastically increases our memory consumption.
Since we are using already about 200 MBytes
(our grid contains 1.2 million gridcells), we do not want to do this.
We could try to compute with the single precision version of
gd1, single.gd1, then we could compute 30 modes in 200 MByte.
Instead we adjust our estimation to 1.2 GHz.
The end of our inputfile now is:
-eigenvalues
solutions= 15
estimation= 1.2e9 # the estimated highest frequency
doit
and the final table of a run (gd1 
wPlunger03.gdf) is
boundary conditions:
xboundary= electric, electric
yboundary= electric, electric
zboundary= electric, electric
--------------------
The first 2 solutions seem to be static and are not saved.
i freq(i) acc(i) cont(i)
1 508.3705e+06 0.0000025997 0.0000021336 # "grep" for me
2 782.2764e+06 0.0000000427 0.0000000599 # "grep" for me
3 785.5885e+06 0.0000000052 0.0000000084 # "grep" for me
4 1.0476e+09 0.0000000020 0.0000000040 # "grep" for me
5 1.0620e+09 0.0000000008 0.0000000145 # "grep" for me
6 1.0930e+09 0.0000000229 0.0000004037 # "grep" for me
7 1.0985e+09 0.0000000040 0.0000000495 # "grep" for me
8 1.1436e+09 0.0000000117 0.0000000929 # "grep" for me
9 1.2149e+09 0.0000052841 0.0000453770 # "grep" for me
10 1.2434e+09 0.0000003031 0.0000066280 # "grep" for me
11 1.2587e+09 0.0000017414 0.0000509264 # "grep" for me
12 1.2801e+09 0.0003925427 0.0118239326 # "grep" for me
13 1.2996e+09 0.0001308259 0.0043265902 # "grep" for me
################################
# cpu-seconds for eigenvalues : 743
# start time : 13:10:52
# end time : 13:41:53
# date : 17/09/99
# The computation of the eigenvalues has finished normally..
# Start the postprocessor to look at the results.
stop .. normal end ..
The garbage mode has disappeared, and the accuracies the modes are very good.
Next: Using gd1.pp to analyse
Up: Analysing the real 3D
Previous: Defining Material properties