For an ellipsoidal distribution, we didn't derived manually the fields expressions, and we will use equations
6.26-
6.28, that uses a form factor, function of the dimensions of the ellipsoid

,

,

.
For the comparison between this theory and our simulations, we generated different bunches of different dimensions. All permutations of values of

,

,

presented in table
7.1were used, representing 896 distributions.
The difference between
and
is anedoctical. As a symmetry was expected between x and y, it appeard unnecessary to use the same values for
and
. We chose to use values close to the expected range that can be found at a0 :
,
. But after all, we decided to study the symetry between
and
, so we extended the values of
.
Our stand alone algorithm has been run on all this distributions, for a grid of
. Maximum fields values and rms values have been extracted from the simulation. Combined with theoretical values, a database has been established. Its structure is described in table 7.2.
In table 7.2, DEX stands for the relative difference :
 |
(7.25) |
As a results of this DEX, DEY, DEZ translate the percentage of difference between the simulation and the theory. We suggest here, to look at the influence of

and

on this three parameters, for a given

mm. These results are plotted on figure
7.3. The best agreement is for

. One explanation can be that the theory developped by Lapostolle as an extension of the case

, is less valid if the bunch does not have revolution symmetry. Nevertheless, we don't reject the hypothesis that the 3D Poisson solver has less validity in this situation also.
Figure 7.3:
Relative difference between theory and simulation for different ellipsoids for the three components of the field. Light color represent a percentage of error of less than 5 %. The three figures show the relative difference between the theory and the simulation : (a) DEZ(
), (b) DEX(
) ,(c) DY(
). The best agreement is for
|