Sei sulla pagina 1di 40

Filter Design and Tuning

using CST Studio Suite


Abstract: Bandpass filters with small fractional bandwidths are quite challenging to design and
tune. After reviewing various tuning techniques using CST Studio Suite a further approach to
compute coupling factor bandwiths simultaneously is demonstrated. A combination of 3D and circuit
models team up with new built in optimizers to efficiently tune bandpass filters. The new System
and Assembly Modelling (SAM) in CST Studio Suite enables the interdisciplinary usage of coupled EMThermal- Structural and Sensitivity analyses within parameter-sweeps and optimization loops and is
demonstrated on a simple test vehicle. Some accuracy considerations with respect to various
meshing techniques conclude this presentation.
Franz Hirtenfelder
Applications Engineer
CST Branch Office Munich
Elsenheimer Strasse 55
D-80687 Mnchen (Munich)
Germany
Tel: +49 89 2420 828 101, Mob: +49 170 9160 110, Fax: +49 89 2420 828 198
Email: franz.hirtenfelder@cst.com, Web: www.cst.com
CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Agenda

Introduction
Tuning Methods
Solver selections, accuracy
System Assembly and Modeling (SAM)
Guidelines and Summary

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Introduction: Flow Chart


Typical Flow Chart of the Filter design and Tuning process
Dimensioning

Specifications

3D EM Simulation

Circuit Design
Analytical models

Empirical adjustments
on the structure

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Measurements

Introduction: from equations to a 3D model

3D
CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Introduction: General Optimizer Setup


Parameterization

Definition of Goal function Optimizer Choice

1. Define structure parameters

2. Define parameter ranges

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Introduction:

Optimizers

Classic Powell
Interpolated Quasi Newton
Trust Region Framework
Nelder-Mead Simplex Algorithm
CMA Evolutionary Strategy
Particle Swarm Optimization
Genetic Algorithm
global

local
Initial parameters already
give a good estimate of the
optimum, parameter ranges
are small

Example:
Waveguide Corner Goal:
Minimize S11

Initial parameters give a


poor estimate of the
optimum, parameter
ranges are large

x
y

x
CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Introduction:

Classic Powell:
Particle Swarm

Quasi Newton:
Genetic Algorithm

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Optimizers

Nelder Mead
Trust Region Framework
(TRF)

Introduction:

Convergence Speed
Nelder Mead: 70 Iterations

TRF: 15 Iterations

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Introduction: Summary
To summarize
In general many parameters to consider
Wide parameter ranges necessary in order not to
miss an optimum
-> global optimizer strategies required
A possible solution to speed up the tuning effort
is pretuning:
Advantage:
local optimizers can be used in addition
CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Agenda

Introduction
Tuning Methods
Solver selections, accuracy
System Assembly and Modeling (SAM)
Guidelines and Summary

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:

Introduction

Classification of Filters

LP-Prototype

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:
Coupling Bandwidth,
Group delay

Group Delay
Coupling-Coefficients and Td-Values
computations are available via Macro
GroupDelay-Macros and
ResultsTemplates available for CSTMWS and CST-DS

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:

Group Delay

Tchebychev Filter
===================
Order
= 4
Bandwidth
= 25 MHz (rel. BW=2.3%)
Center Frequency = 1100 MHz
Passband ripple = 0,01 dB (1,100747 VSWR)
Return loss
= -26,3828 dB
Normed g values:
------------------------------------------g1 = 0,7129
g2 = 1,2004
g3 = 1,3213
g4 = 0,6476
g5 = 1,1008
Corresponding coupling coefficients in MHz / (rel):
------------------------------------------k_E = 35,07
(0,0318809)
k1_2 = 27,03
(0,0245688)
k2_3 = 19,85
(0,0180464)
k3_4 = 27,03
(0,0245688)
k_out = 35,07
(0,0318809)
Group Delay Time
---------------t_d1 = 18,153 ns
t_d2 = 30,566 ns
t_d3 = 51,798 ns
t_d4 = 47,057 ns
t_d5 = 71,78 ns

Only two variables at a time!!


CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

ke=
Qe =

2
td1

t d 1. . f o
2

Tuning Methods:

Group Delay Examples


Iris Coupled Cavity
Filter

Tuning of a Dual Mode Filter

Short

Hairpin
Filter

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:

Field Strengths

Port 1 is excited

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:

Field Strengths

The E-max is recorded vs. frequency. What can be observed is that the
peaks of E-max coinside with the peaks of the transmission-groupdelay,
as expected. The groupdelay may act here as a probability function of
energy: High fields when there is a high chance of energy concentration

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:

Pin-Probes and Eigenmodes

Even-Mode F1

CBW

_ corrected

1 S1
S
=

+ 2
2 S2
S1

Odd-Mode F2

F 2 F 2
2
1
2

F 1 + F 22

S 2 S 22

12
2
S1 + S 2

S1, S2 single cavity modes

Pin Probes:

Add two small discrete ports or


Face ports to excite the modes
CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

CBW

Tuning Methods:

Port-Tuning
3D results

Discrete Ports or FacePorts are assigned at


the Resonators

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:

Port-Tuning

Use Tuner for C3 and C4 to


manually adjust a good filter
response (resonance-tuning)

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:

Port-Tuning

Coupling between resonators are designed as negative Cs (act as TLs 90 deg)

1. Use Tuner for C34


and C45 to manually
adjust a good filter
response

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:

Port-Tuning

Correlation to the real Geometry (Space Mapping):


Once the optimum for a certain capacitor has been found, we need to find out their relation to the real 3D geometry.
This is done by slightly changing the 3D model for the given resonator where the capacitor is attached to. Then we
need to retune the circuit again to account for this geometrical change. Now we have two positions in 3D and two
values for the capacitance C of the circuit. With a linear extrapolation we are able to find out the desired
mechanical change to set the capacitance to zero, meaning that the lumped capacitance has no impact anymore and
can be eliminated in the circuit.

Mech. Position p1 Ca= x1 F


Mech. Position p2 Ca= x2 F

Ca
Ca@p1 =x1F

p2 p1
p x = p1 + x1
x1 x2

Ca@p2=x2F
0

p1

p2

px

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:Tuning via Slope Susceptance


To compute the coupling factors k12,k23,
etc, the Input and Output converters are
separated from the main part of the filter.
Ports of 50 Ohms are assigned and connected
to each resonator. S-Parameter and YParameter are computed via a DS task. The
imaginary parts of the Y-matrix are used to
compute the coupling factors simultaneously.

k12

3
CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

k34

k23

Tuning Methods:Tuning via Slope Susceptance


Templates:

k34

kij =

2.Bij
dBii dB jj
df df

k12

k23

= 2.im(Yij ) / sqr ((im(Yii _ deriv).(im(Y jj _ deriv))

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:Tuning via Slope Susceptance


Circuit setup for Coupling Coefficients
To get the coupling coefficients simultaneously the circuit is modified such that each resonator is
attached to a e.g. 50 Ohm port. The resulting Y-Matrix is used to compute all couplings:

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:Inverse Chirp Z


The chirp Z-Transformation can be used as a
more flexible means to calculate discrete Fourier
transforms. In particular, the unit circle version
(known as chirp-transform) can be used to create
a high-quality zoom function.
Golden (reference) Filter required

Inverse Chirp-Z response

S-Parameter
fo

1
ICZ-Bandwidth

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Tuning Methods:Inverse Chirp Z

Tuning of 1st resonator

Tuned to a min.dip
2

Tuning of 2nd resonator

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Example of a Diplexer:

Groupdelay

RX

TX
Rx part is considered,
but not shown here,
posts are short out

Green Curve: tuned Groupdelay


Red curve: Inital groupdelay for the
first two posts next to the common
port after the separate tuning of TX
and RX

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Example of a Diplexer:
InvChirpZ transformation is applied to the
open filter and it can clearly be seen that
some of the resonators are not tuned to
their center frequency f1 and/or f2. This
is indicated by the dips of the response.
The coupling seems to be rather ok,
repesented by the time-delay.
The advantage of this method is that the
individual mistuned resonators can be
identified!

2nd resonator mistuned


1D Results > invChiprZ_TX

1D Results > invChiprZ_RX

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

InvChirpZ

Agenda

Introduction
Tuning Methods
Solver selections, accuracy
System Assembly and Modeling (SAM)
Guidelines and Summary

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Solver Selection
Time Domain (TD)

Frequency Domain (FD)


General Purpose

Eigenmode (E)

Resonant: Fast S-parameter (MOR)

Modal (Resonant Fields)


Lossy/Lossless
Resonant: S-Parameter, Fields (Modal Analysis)

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Accuracy vs. Meshdensity

10/10

20/20

30/30

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Accuracy vs. Meshdensity


Variable / Mesh

Coarse 10/10

Medium
20/20

Fine 30/30

40/40

60/60

Coupl_tuner_23

7.5 mm

6.35

6.35

6.35

6.35

Ke_offset

5.68

5.6

5.6

5.45

5.45

Re_tuner_L_1

6.107

5.8

5.85

5.722

5.731

Re_tuner_L_2

5.165

4.94

4.97

4.924

4.942

k23

18.2

18.2

18.2

18.22

18.22

CPU Time (HEX MOR)

26s

129s

485s

18m

60m

Coupl_tuner_23

k12
k23

Re_tuner_L_2
Re_tuner_L_1
Ke_offset
CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

FD General Purpose Tet-Mesh

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

(6 steps/lambda)

Agenda

Introduction
Tuning Methods
Solver selections, accuracy
System Assembly and Modeling (SAM)
Guidelines and Summary

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

SAM: Multiphysics Simulation

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

SAM: Multiphysics Simulation


Coupling between Simulation Projects

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

SAM: Multiphysics Simulation


Parameter Sweep / Optimization
S-Parameter (pure EM)

S-Parameter
(including Therm./mech.
Deformation)

apt_height
CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Agenda

Introduction
Tuning Methods
Solver selections, accuracy
System Assembly and Modeling (SAM)
Guidelines and Summary

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Some Guidelines
Start out with a rather coarse mesh models

Dont use meshadaptation


Pretuning is helpful to apply local optimizer
strategies
Select the appropriate solver
Refine the mesh and retune again

(Parameter ranges are a lot smaller)


until parameter changes are smaller than a
given manufactoring tolerance

CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Summary

CAD Modeler easy to use with respect to


parameterization

CST Complete Technology: TD, FD, E, Th, Mech


Various optimizer strategies
Optimization and parameterization control via
complex post processing templates
Various meshing techniques available
Flexible link to circuit simulator CST- DESIGN
STUDIO including CST- MICROWAVE STUDIO
submodels
Various tuning procedures available for a successful
tuning
CST COMPUTER SIMULATION TECHNOLOGY | www.cst.com | May-12

Potrebbero piacerti anche