Changes to the thesis of Hasan Khaled Rouf
N.B.: Because of the change of texts, page numbers have been slightly changed
C1 – p. 35. Explicitly state that E and H fields are computed at same time. Done
C2 – p. 35. “the resulting method ..... has high accuracy and low anisotropy”. Better accuracy and lower anisotropy than Yee FDTD? than ADI-FDTD? The reader expects a brief comment. Dropped the phrase ‘has high accuracy and low anisotropy’ here and clarified this at the bottom of p.37 - ‘On the other hand …. are also higher’ with the reference.
C3 – p. 36. In (2.45) and (2.46) the matrices are 6x6 matrices (involving continuous spatial derivatives of fields). Just below (2.46) it is stated “(2.46) involves large sparse matrix”. This is contradictory. In fact, discretisation on space of (2.46) will yield a large matrix (6 nodes x 6 nodes). Add a sentence to say this. Done
C4 – p. 47-48. Explicitly state that the grid where are placed the nodes is the Yee grid (that permits the spatial derivatives to be second order in accuracy). Added (P.47)
C5 – p. 53 (or in appendix). State about e: whether it is assumed as frequency dependant or not in the Mur equations. Stated in appendix (P.203)
C6 – p. 53-84. Add a comment explaining why the Mur boundary condition is used in different ways to the right and to the left of the domain. Added (P.79-80)
C7 – p. 86. The statement that the Crank-Nicholson method is “unconditionally stable” needs to be modified, especially given the results of Chapter 5. The text needs to explain how it is more stable than the explicit method. Some references to the literature can justify that the method is stable under certain conditions and this can be related to the situations examined in the thesis. Modified
C8 – If possible, add the skin depth and the space step in table p. 93. Added in table (p.96) and text added (P.93)
C9 – How is set the initial solution with iterative solvers (the current E values?). Stated (p.124)
C10 – For the 2D ADI, where are the nodes on the grid? Are there 3 components on the boundary, requiring then use of Mur ABC for the three components? A figure explaining the answer should be added to the text. Added (figures - 7.1 and 7.2 and their references in the text P.152-155,157-159)
C11 – In the caption to Figure 6.6 the reference to the scaling of E_z should be removed since this is a detail of the implementation of the code and not of the physics being modelled. Similarly, in Figure 6.8, the Max values should be correctly scaled by this factor of CFLN. Done
C12 - p. 182. In Table 8.4 wall clock timings are given. These timings should be supplemented by a column, which gives the speed up on 32 cores. Some analysis in the text should comment on the different in speedup between CFLN 1 and 3 and suggest how the relationship between speedup and CFLN could be further explored. Done (p.185)
C13 – Minor corrections to spelling and grammar All done
Page 30 3 lines from end "either" not "neither"
p37 line 4 "over the last decade" not "over the last one decade"
p39 "Grossly discretizing" does not seem to make sense in this sentence. What is the author trying to say?
p41 last para "Yee's"
p101 Section 4.5 line 2 suggest replacing "unbounded" with "without bound"
p123 line from the end: "lower" not "lowered"
p126 line 1 "the problems concerning simulation" not "the concerned problem of simulation"
p128 line 9 "on mice" not "on the mice"
p134 line 2 from bottom, "up to" not "upto"