Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Studies on surface roughness for rectangular waveguide structures
Christian Lehmann
Ferdinand-Braun-Institut
Leibniz-Institut für Höchstfrequenztechnik
Berlin
Contents
2
• Modeling and numerical problems
• Overview of analytical approaches
• Simulated surfaces
• Results
• Conclusion
Activity 1 : Losses in waveguides – Simulation Difficulties
3
• Modeling and numerical problems
• Problems with the propagation constant in FD.
• Propagation constant from Time Domain.
• All other results best from FD.
• But hexahedral mesh necessary (large memory resource needed).
• Modeling the roughness has to account for very large ratios between smallest and
largest cell size.
• Difficulties in convergence.
• Problems with eigenmode solvers (bad matrix condition).
• Common similar discretization scheme for all roughness shapes.
• Delay in project.
• Presently first results with promising outcome.
Activity 1 : Losses in waveguides – Analytical approaches
4
G. Gold, K. Helmreich 2012: A Physical Model for Skin Effect in Rough Surfaces
Conductivity Profile
Calibrate profile parameters with
measurements
† taken from J. Phillips and G. Edwards (2010):
Transmission Line Modelling for Multi-Gigabit Serial Interfaces
†
Hall, Heck: Advanced Signal Integrity
for High-Speed Digital Designs
E. Hammerstad and O. Jensen: Accurate models
of computer aided microstrip design
†
Power Loss Method
Basic Maxwell
Phase constant β is not influenced by the decay α
Fitting Parameters for α
1D periodic structures (Hammerstad & Jensen)
2D periodic structures (Hall & Heck)
Activity 1 : Losses in waveguides – Simulated Surfaces
5
roughnessmod
el
not-to-scaleplot hex. meshcells
(FD)
0=none
31k
1=layer
44k
2=rampT
250k
3=stepT
238k
4=rampL
236k
5=stepL
231k
6=tower
1.3M
7=tower_inv
1.3M
Meshing
cell resolution in different directions
and on the surface
Model type (WL-86)
flat, ramp, step, layered
repeating in
longitudinal/transversal direction or
both
Model dependent parameters
filling rate (steps/towers)
conductivity profile (layer)
Surface parameters
period length
roughness
border thickness
Dimensions
waveguide width-height
border thickness
Material and models
H&J model or regular Maxwell
conductivity
solver (freq/time)
6
Activity 1 : Losses in waveguides – Results
2200 2400 2600 2800 3000 3200 3400
1,1
1,2
1,3
1,4
1,5
1,6
1,7
1,8
vp
h/c
0
frequency in GHz
flat 1d
layer(k=0) 1d
ramp T 1d
step T 1d
ramp L 1d
step L 1d
tower 1d
inv tower 1d
layer(k=0) 10d
ramp T 10d
step T 10d
ramp L 10d
step L 10d
tower 10d
inv tower 10d
2300 2320 2340 2360 2380 2400
1,45
1,46
1,47
1,48
1,49
1,50
1,51
1,52
1,53
structures with d= 10
vp
h/c
0
frequency in GHz
flat 1d
layer(k=0) 1d
ramp T 1d
step T 1d
ramp L 1d
step L 1d
tower 1d
inv tower 1d
layer(k=0) 10d
ramp T 10d
step T 10d
ramp L 10d
step L 10d
tower 10d
inv tower 10d
flat and all structures with d= 1
7
Activity 1 : Losses in waveguides – Results
2000 2200 2400 2600 2800 3000 3200 3400
0
4
8
12
16
20
de
ca
y in
dB
/cm
frequency in GHz
model 0
model 1
model 2
model 3
model 4
model 5
3150 3200 3250 3300 3350
3
4
5
6
de
ca
y in
dB
/cm
frequency in GHz
model 2 rms=110 nm
model 2 rms=220 nm
model 2 rms=330 nm
model 2 rms=440 nm
model 2 rms=550 nm
8
Activity 1 : Losses in waveguides – Results
Roughness effects 2
Frequency domain simulation
Different profile structures yield different frequency behaviour.
3150 3200 3250 3300 3350
3,5
4,0
4,5
5,0
5,5
6,0
6,5
de
ca
y in
dB
/cm
frequency in GHz
model 4 rms=110 nm
model 4 rms=220 nm
model 4 rms=330 nm
model 4 rms=440 nm
model 4 rms=550 nm
9
Activity 1 : Losses in waveguides – Results
Roughness effects 3
@3240 GHz
Comparison to Hammerstad & Jensen (model 2) under research
model 2 model 4 0 100 200 300 400 500 600
4,0
4,5
5,0
5,5
6,0
6,5
7,0
de
ca
y in
dB
/cm
rms in nm
model 2
model 4
E. Hammerstad and O. Jensen: Accurate models
of computer aided microstrip design
†
10
Activity 1 : Losses in waveguides – Results
1.8e7 2.5e7 3.3e7
11
Activity 1 : Losses in waveguides – Results
2000 2200 2400 2600 2800 3000 3200 3400
3
4
5
6
7
8
9
10
11
de
ca
y in
dB
/cm
frequency in GHz
rms = 120 nm, kappa_0 = 0
rms = 120 nm, kappa_0 = 1.45e7
rms = 120 nm, kappa_0 = 2.90e7
rms = 120 nm, kappa_0 = 4.35e7
rms = 120 nm, kappa_0 = 5.80e7
model 2 rms=110 nm
model 4 rms=110 nm
12
Activity 1 : Conclusion
Difficult simulation conditions in solver
Dense discretization and shape independent distribution.
Time domain only for propagation constant.
Frequency best for 3D structures.
Present results promising
Some more results can verify the analytical models Layer model by Gold & Helmreich with best chances