Modelling cavity filter
temperature drift in CST MWS
Dr Novak Petrovic
September 2009 Abridged
Presentation Outline
• Company background
• Role of CST MWS
– Conventional solutions
– HTS solutions
– Ceramic solutions
• Practical Problems and Current Work
– Definition
– CST MWS solution
• Q&A, further discussion
September 2009 Abridged
Company Background
• Founded in 2002
• “Microwave and Materials Designs”
• New materials for microwave applications in the wireless cellular market
• 16 staff with knowledge of RF and materials technologies
• Venture capital backed
• Pursuing OEM & operator relationships
September 2009 Abridged
Solutions overview
• Cellular wireless infrastructure
• Spectrum re-farming (U900)
• Interference mitigation (Custom)
• Co-siting & co-location (Sharing)
• Low noise solutions (Cryogenic)
• LTE – filtering & range extension
September 2009 Abridged
HTS solutions
September 2009 Abridged
Conventional solutions
September 2009 Abridged
Materials solutions
Novel ceramic materials made in house
Abridged
Application of CST MWS
Abridged
Design and Characterisation
Coupling matrix
optimisation
Design
Tuning
Characterisation
Matlab / Java
Abridged
Problem 1
Filter response change due to temperature change?
Thermal expansion
of metals
Change of dimensions
Change of resonant
frequency
Change of filter
response
Abridged
Change in performance
Abridged
Change in performance
Keep the drift to
less than 1 GSM ch
Operational temperature
specification: -40 to +85 deg C
Abridged
Current approach
1tan2 CZfResonance condition:
fringingplatescrew CCCC
diameterresonator_
metercavity_dialn60coaxZ
Abridged
Approximating capacitance
fringingplatescrew CCCC
1tan2 CZf
Partly empirical
Abridged
Temperature expansion
Thermal expansion coefficient:Linear:
αL << 1
small ΔT
Area:
αA ≈ 2αL
isotropic
materials
Vol:
αV ≈ 3αL
isotropic
materials
T
L
L
0
1
TV
V
V
0
TA
A
A
0
TL
L
L
0
Abridged
Automation of current approach
Abridged
Automation of current approach
Only works for
“static” analytic cases,
known impedance
September 2009 Abridged
CST MWS solution
Formulate parameters
as a function
of temperature
September 2009 Abridged
CST MWS solution
Abridged
Practical caseSpreadsheet calculation: -0.762 MHz drift, -20.2 ppm/˚C
CST MWS: -0.9 MHz drift, -23.8 ppm/˚C
C ppm/deg 106
1
12drift
Tf
fff
Abridged
Measured results
Total temperature drift of -0.9 MHz was measured
(+40 ˚C temperature change, rate of -0.023 ˚C/MHz)
CST MWS works
Abridged
Extension to arbitrary shapes
αA ≈ 2αL
isotropic
materials
TA
A
A
0
Abridged
Scaling check
Reference objects to cavity
(or fixing screw location)
Area of scaled objects
does check out.
Abridged
Other possibilities
Parametrise all features
Abridged
Accuracy
Criterion: Establish logical convergence (experiments),
by examining change in error, and referenced to mesh.
35 LPW: 141,120 meshcells, f1 = 946.4 MHz f2 = 945.5 MHz
Also confirmed convergence on basic coaxial case.
25 LPW: 54,400 meshcells, f1 = 941.7 MHz f2 = 940.8 MHz
15 LPW: 17,248 meshcells, f1 = 934.5 MHz f2 = 933.6 MHz
45 LPW: 269,500 meshcells, f1 = 946.9 MHz f2 = 946.0 MHz
September 2009 Abridged
Problem 2
• Current work: Design ceramic filter
– Calculate temperature drift and compensate
• Previous methodology applied
• However, a lot more CST MWS simulations
September 2009 Abridged
General design notes
More precise simulations.
Ceramics changes in addition to metal.
September 2009 Abridged
Manufacturer info
τf
Adjustable
September 2009 Abridged
Standard cavity: temperature
D = 0.5’’
H = 0.2’’
L = 1.05’’
C = 1.5’’
f0 ≈ 4 GHz
T
f
ff
0
0
1
September 2009 Abridged
Composite coefficient
T
1
T
L
LC
1
T
D
DT
H
HL
11
T
f
ff
0
0
1
Dielectric constant
Cavity
Resonator (dimensions)
ppm/deg C
September 2009 Abridged
Extract τε
f0 at amb = 3959.39 MHz
f0 at amb + 60˚ C = function of τε (change of ε with T)
Parametrise:
-dielectric const. change
-resonator dim. change
-cavity dim. change
-safe to ignore holder
Change τε to get quoted τf
T
1
τf = 0
September 2009 Abridged
Conversion
Change τε, calc f0 at
amb & temp, calc τf
T εnomr,newr, 1
C deg 60T
September 2009 Abridged
Extract τε: check
C ppm/deg 3.15ε
1,05.0 ,1 ,2
1 CBA
CLεf4
3 CBA
CLf
ε3
4
CB
A
31
r
2
rr
0
4
553.8
LD
f
September 2009 Abridged
Drift of the design
Representative cavity vs entire filter
Metallic resonator methodology
f0 @ amb = 913.15 MHz
f0 @ amb + 60 ˚C = 913.67 MHz
Sweep τε until design stabilises
C ppm/deg 5.9design
C ppm/deg 0 want We design
MWS
September 2009 Abridged
Sweep desired τε
Work out required τε so dopants can be selected
Change until f0 @ +0 ˚C and f0 @ +60 ˚C are the same
C ppm/deg 7.3
September 2009 Abridged
Convert back to τf in std cavity
C ppm/deg 7.3Want
C ppm/deg 8 into Translates f
C ppm/deg 6 toup Rounded f
C ppm/deg 5.1 Therefore design ORDER !
MWS
f0 @ amb = 913.15 MHz
f0 @ amb + 60 ˚C = 913.23 MHz
C ppm/deg 7.14 Have
September 2009 Abridged
Standard cavity: frequency
September 2009 Abridged
Summary
• Work out tau_epsilon from specs
• Work out expected design drift
• Work out required tau_epsilon
• Translate to tau_f in standard cavity
• Translate to standard cavity frequency
• Order
• Keep working...
September 2009 Abridged
AccuracyCase f0
at amb
MHz
f0
at amb
+60 deg C
τf
ppm/deg C
Δf
MHz
40 LPW
606,528
3953.604 3953.658 0.23 0.054
30 LPW
262,236
3956.217 3956.277 0.25 0.06
20 LPW
75,816
3962.602 3962.675 0.31 0.073
15 LPW
30,400
3972.118 3972.207 0.37 0.089
Used τε = -15.3 ppm/deg C
September 2009 Abridged
Keep working...
Entire filter,
all parametrised...
Spurious coupling
(coupling sign)
Temperature drift
Verification
September 2009 Abridged
Tune and measureTuned with the aid of Mesaplexx Filter Design Tool
September 2009 Abridged
Tuning
Extract coupling matrix
by comparing to model
Note difference to desired,
change (screws) manually,
5 – 50 times (?)
September 2009 Abridged
Accuracy
SPW = 6, MIN = 6
Tetrahedrons: 156,411
Adaptation: 1 @ 897.7 MHz
Accuracy = 1e-6
delta S = 0.01 (both)
SPW = 4, MIN = 4
Tetrahedrons: 41,872
Adaptation: None
Accuracy = 1e-4
delta S = 0.01 ~ 7 min
~ 31 min
September 2009 Abridged
Drift illustration-1.5 ppm/deg C
896.92 MHz
vs
897 MHz
September 2009 Abridged
Relevance
• Turn-around time (12 weeks)
• Qualification of custom-made ceramics
– Q, εr, τε (the difficult trio)
– nonlinear coefficients, complex ε
– cryogenic temperatures (non-linear)
• Tight specification (GSM channels)
• New tuners, dual-mode filters, frequency
scaling ...
September 2009 Abridged
Unexpected twistsDielectric supports contribute as well!
September 2009 Abridged
Parameter extraction
Extraction of material properties by numerical simulation.
September 2009 Abridged
Future work
• Power handling in comblines
– Line integral to calculate voltage
• Thermal runaway in ceramic filters
– Power handling
• Printed filters
• Lumped element extraction
• More macros
Abridged
2008 - EDN Innovation AwardsWinner – Best application of RF design
Winner – Best overall project
Thank you!
Contact:
Novak Petrovic