GENESYSV8 Examples - Keysightliterature.cdn.keysight.com/litweb/pdf/genesys8/Examples8.pdf · Overview 6 Examples 7.5\Components\Bondwires CompensatedCrossover1.wsp CompensatedCrossover2.wsp

$Page 1: GENESYSV8 Examples - Keysightliterature.cdn.keysight.com/litweb/pdf/genesys8/Examples8.pdf · Overview 6 Examples 7.5\Components\Bondwires CompensatedCrossover1.wsp CompensatedCrossover2.wsp$
GENESYS V8

Examples

Copyright 1986-2001

Eagleware Corporation635 Pinnacle CourtNorcross, GA 30071

Phone: (678) 291-0995FAX: (678) 291-0971E-Mail: [email protected]://www.eagleware.com

Printed in the USA

This page intentionally left blank.

Table of Contents

Chapter 1: Overview ......................................................... 5

Finding Examples ...............................................................6

Chapter 2: Signal Control ............................................... 11

Signal Control\Film Atten.WSP .........................................11 Signal Control\Xfmr Coupler.WSP.....................................14 Signal Control\Edge Coupler.WSP....................................17 Signal Control\ResistiveBroadband10dB.WSP..................18 Signal Control\ResistiveBroadband20dB.WSP..................19 Signal Control\Dual Mode Coupler.WSP...........................21 Signal Control\8 Way.WSP ...............................................24 Signal Control\BSCouplerFinalRecomp.WSP....................27 Signal Control\BSCouplerFinalRecomp.WSP....................30 Signal Control\BSCouplerFinalWhole.WSP.......................33

Chapter 3: Components.................................................. 35

Components\Model Extract.WSP......................................35 Components\Microstrip Line.WSP.....................................37 Components\Stripline Standard.WSP................................42 Components\Spiral Inductor2.WSP...................................45 Components\Box Modes.WSP..........................................48 Components\User Model.WSP .........................................51 Components\BJT NL Model Fit.WSP ................................54

Chapter 4: Amplifiers...................................................... 57

Amplifiers\Stability.WSP ...................................................57 Amplifiers\Balanced Amp.WSP.........................................60 Amplifiers\Amp Feedback.WSP ........................................63 Amplifiers\Amp Noise.WSP...............................................64 Amplifiers\Amplifier Tuned.WSP .......................................66 Amplifiers\Amplifier.WSP..................................................68 Amplifiers\SiGe BFP620 Amp.WSP ..................................70

Chapter 5: Filters ............................................................ 75

Filters\Contiguous Diplexer.WSP......................................75 Filters\ComblineDesign\Final Lossy.WSP .........................77

2

Filters\Coupled Stepped Z.WSP....................................... 83 Filters\Edge Coupled.WSP............................................... 86 ResonanceElimination\ EdgeCoupledWithFictivePorts.WSP89 Filters\Interdigital.WSP..................................................... 92 Filters\Tuned Bandpass.WSP........................................... 95 Filters\Two Level.WSP..................................................... 98 Filters\Xtal Filter.WSP ...................................................... 99 S/FILTER....................................................................... 101

How To Design........................................................101 Equal Termination Example.....................................102 Maximum Realizability Example...............................104 All Series Resonators Example................................106 All Parallel Resonators Example..............................109 Response Symmetry Example.................................112 Equal Inductor Example...........................................115 Physical Symmetry Example....................................116 Termination Coupling Example ................................121 Parametric Bandpass Example................................123 LP All Pole...............................................................127 LP All Pole...............................................................128 Coaxial Resonator Example.....................................130 BP Edge Coupled....................................................132 BP Edge Redundant ................................................134 BP Stub with Inverters .............................................136

Chapter 6: Oscillators ...................................................137

Oscillators\Bipolar Cavity Oscillator.WSP ....................... 137 Oscillators\NegR VCO.WSP........................................... 141 Oscillators\Coaxial OSC.WSP........................................ 143

Chapter 7: Matching ......................................................147

Matching\Synthesis Comparison.wsp ............................. 147 Matching\Ill Behaved Load.WSP .................................... 150 Matching\Power Amp.wsp .............................................. 152 Matching\Unstable Device.wsp....................................... 154 Matching\Match and Stability.WSP................................. 156 Matching\Stability Selection.WSP................................... 158

3

Chapter 8: Detectors..................................................... 161

Detectors\Simple Detector.WSP .....................................161 Detectors\Diode Detector with Co-simulation.WSP .........163

Chapter 9: Mixer............................................................ 165

Mixer\Low Power Mixer.WSP..........................................165

Chapter 10: Antennas..................................................... 167

Antennas\Patch Antenna Impedance.WSP .....................167 Antennas\Array Driver.WSP............................................169 Antennas\Agile Antenna.WSP ........................................171 Antennas\MAmmannPatch.WSP ....................................173 Antennas\Simple Dipole.WSP.........................................175 Antennas\Thin Loop.WSP...............................................176

Chapter 11: Resonance Elimination............................... 179

ResonanceElimination\ LowPassWithFictivePorts.WSP ..179 ResonanceElimination\ MicrostripLineWithFictivePort.WSP181 ResonanceElimination\ MicrostripLineWithFictivePort2.WSP.......................................................................................182 ResonanceElimination\ ResonantStubFictivePort.WSP...183 ResonanceElimination\ TwoLevelBPwithFictivePorts.WSP185


Chapter 1: Overview These examples can be found in the EXAMPLES subdirectory of your Eagleware installation (commonly C:\Eagle or C:\Program Files\GENESYS). The header of each example section is the filename. The required RAM is the value estimated by EMPOWER. They are approximate and are determined by algorithm rather than a test of memory used. The execution times are for a 266 MHz Pentium II with 256Mbytes of RAM operating under Windows 98. In most cases execution time is for the discontinuity mode. When different run options are used during the example the time and RAM listed are for exercise options with the longest time and largest memory requirement.

The following extra examples are provided for your reference, but are not included in the Examples manual. Examples\Benchmarks BulkConductivityTest.wsp LossyGroundPlane.wsp MicrostripStandard.wsp ShortLengthVia.wsp Stripline Standard.wsp Stripline3Ddiscretisation.wsp UltraThinDielectric.wsp ZeroLengthThrough.wsp ZoDefinition.wsp Examples\Components ApertureCoupledMicrostrips.wsp BigPlanarInductor.wsp CrisscrossDecomp.wsp InductorDecmposition.wsp Microstrip Line ND.wsp RfilmInLine.wsp RfilmInTjunction.wsp SpiralInductor.wsp TeeAndCrossModels.wsp TeeModel.wsp ThickResonatorThickLine.wsp ThickResonatorThinLine.wsp TransistorConnection.wsp TwoStubs.wsp ViaTroughGround.wsp

Overview

6

Examples 7.5\Components\Bondwires CompensatedCrossover1.wsp CompensatedCrossover2.wsp Simple3Dbondwire1.wsp Simple3Dbondwire2.wsp SimpleCrossover.wsp Examples\Components\Lines Zdefinition2unsim.wsp Zdefinition3sim.wsp Zdefinition3unsim.wsp Examples\Filters HybridCoplanarBandstop.wsp TwoLevelBPdecomp.wsp TwoLevelBPdecompThinning.wsp

Finding Examples The following indexes have the examples organized by "How To". For a list of examples organized by application, see the Table of Contents.

Linear Simulation

Stability (Amplifier Stability Circles) Balanced Amplifier (Microstrip, NET Component, Layout) Amplifier Feedback Topologies (Multiple Schematics) Amplifier Noise (Microstrip, Noise Circles, Layout) SiGe BFP620 Amplifier (Harbec, Linear Simulation, NonLinear Simulation, User Model, Multiple Schematics, Parameter Sweep) Lower Power Mixer (Harbec, Linear Simulation, Non-Linear Simulation, Data File, Parameter Sweep, Multiple User Models, Multiple Schematics) Simple Detector (Harbec, Linear Simulation, Parameter Sweeps)

Non-Linear Simulation (Harbec)

SiGe BFP620 Amplifier (Harbec, Linear Simulation, NonLinear Simulation, User Model, Multiple Schematics, Parameter Sweep) Lower Power Mixer (Harbec, Linear Simulation, Non-Linear Simulation, Data File, Parameter Sweep, Multiple User

Finding Examples

7

Models, Multiple Schematics) Simple Detector (Harbec, Linear Simulation, Parameter Sweeps) Diode Detector with Co-Simulation (EMPower, Harbec, Co-Simulation)

Data File

Model Extraction (Parameter Extraction, Post-Processing, Optimization, Link to Data File) Lower Power Mixer (Harbec, Linear Simulation, Non-Linear Simulation, Data File, Parameter Sweep, Multiple User Models, Multiple Schematics) Ill Behaved Load (MATCH, Port Impedance Data File)

Electromagnetic (EM) Simulation (EMPower)

EM Box Modes (EMPower, Metal Box with and without Top Cover) Spiral Inductor (EMPower: Decomposition, Co-Simulation: Linear and EM) Microstrip (EMPower: Introduction, Microstrip Impedance, Extrapolating Line Impedance, and Deembedding) Stripline (EMPower: Accuracy vs Cell Size, Extrapolating Line Impedance) Film Attenutor (EMPower, Deembedding) Microstrip 3 dB Directional Coupler (EMPower, Tuning, Decomposition) Resistive 10 dB Attenuator (EMPower) Resistive 20 dB Attenuator (EMPower) Dual Mode Coupler (EMPower: Viewer and Thinning) 8 Way Power Divider (EMPower, Microstrip, Lumped Elements) Transmission Line Amplifier (EMPower, Lumped Elements) Optimized Tuned Transmission Line Amplifier (EMPower, Lumped Elements, Optimization) Coupled Stepped Z (EMPower: Viewer) Edge Coupled (EMPower: Grid Fitting) Interdigital (M/FILTER: EMPower: Wall Grounding) Tuned Bandpass (EMPower, Lumped Elements, Tunable Filter, Co-Simulation: Linear and EM) Two Metal Layer Filter (EMPower)

Overview

8

Diode Detector with Co-Simulation (EMPower, Harbec, Co-Simulation) Patch Antenna Impedance (EMPower, Patch Antenna Impedance and Resonance) Frequency Agile Antenna (EMPower, Linear Simulation, Lumped Elements) Patch Antenna (EMPower) Simple Dipole (EMPower) Thin Loop (EMPower) Low Pass with Fictive Ports (EMPower) Microstrip Line with Fictive Port Wide Width (EMPower) Microstrip Line with Fictive Port Narrow Width (EMPower) Resonant Stub Fictive Port (EMPower) Two Level Bandpass Filter with Fictive Port (EMPower) Edge Coupled Filter with Fictive Port (EMPower)

Parameter Sweep

SiGe BFP620 Amplifier (Harbec, Linear Simulation, NonLinear Simulation, User Model, Multiple Schematics, Parameter Sweep) Lower Power Mixer (Harbec, Linear Simulation, Non-Linear Simulation, Data File, Parameter Sweep, Multiple User Models, Multiple Schematics) Simple Detector (Harbec, Linear Simulation, Parameter Sweeps)

Equations

Transformer Coupler (Lumped Elements, Equations)

User Model

Creating a User Model (User’s Guide) User Model Example (User’s Guide) User Model (Tuning, User Models) SiGe BFP620 Amplifier (Harbec, Linear Simulation, NonLinear Simulation, User Model, Multiple Schematics, Parameter Sweep) Lower Power Mixer (Harbec, Linear Simulation, Non-Linear Simulation, Data File, Parameter Sweep, Multiple User Models, Multiple Schematics)

Finding Examples

9

Optimization

Overview (User’s Guide) Model Extraction (Parameter Extraction, Post-Processing, Optimization, Link to Data File) BJT NL Model Fit (Optimization, SPICE Model, S-Parameters) Optimized Tuned Transmission Line Amplifier (EMPower, Lumped Elements, Optimization) Match and Stability (Match, Stability, Optimization)

Matching

Overview (User’s Guide) Match and Stability (Match, Stability, Optimization) Stability Selection (Stability, Optimization, Stabilization Networks template)

Schematic Symbol

Create a New Schematic Symbol (User’s Guide) Schematic Symbol Example (User’s Guide)

Spice Link

Create a Spice Link (User’s Guide)

Component Footprint

Create a Footprint (User’s Guide) Create a Multiple Part Footprint (User’s Guide) Footprint Example (User’s Guide) Multiple Part Footprint Example (User’s Guide)

Netlist

Overview (User’s Guide) Netlist Example (User’s Guide)

TestLink

Overview (User’s Guide)

Monte Carlo


Overview

10

Yield Optimization


DisCos

Using DisCos (User’s Guide)

Advanced TLINE

Overview (User’s Guide) Walkthrough (User’s Guide) Hairpin 2 Pole (Advanced TLINE, EMPower)

Equalization

Overview (User’s Guide) Equalization Example (User’s Guide)

SFilter

Overview (User’s Guide) Coaxial Resonator Example (SFilter) Equal Inductor Example (SFilter) Equal Termination Example (SFilter) Parametric Bandpass Example (SFilter) Physical Symmetry Example (SFilter) Practical Design Example (SFilter) Response Symmetry Example (SFilter) Termination Coupling Example (SFilter)

Chapter 2: Signal Control

Signal Control\Film Atten.WSP

RAM: 31.7Mbytes Time: 3324s/freq Illustrates: Resistive films, deembedding, benchmark

Given below is a broad bandwidth resistive film 20dB attenuator developed by Res-Net Microwave Inc., of Largo, Florida, which offers a number of high performance resistive microwave products.

This attenuator works well through 12GHz and benchmark data is available through 20GHz. Metal is the darkest color and the mid-tone sections are 50 ohm/square resistive films on a 40mil

Signal Control

12

thick BeO 100 by 200mil substrate with a relative dielectric constant of 6.70.

A much larger grid than 1.25 by 1.25mils could have been used except it was desired to match the dimensions of the attenuator closely. Had an electromagnetic simulation been anticipated the original dimensions could have been fit to a larger grid and the desired DC resistance of the films could still have been achieved. This example would then run quickly.

The launching connectors have a strip width of approximately 25mils while the input and output lines are 70mils. To simulate this discontinuity, the 70 mil lines were not brought to the edge of the board. In addition, deembedding was turned off in this =EMPOWER= run to include stray capacitance from the line to the face of the launching connector. The net capacitance is 0.24pF as computed from the 0.003024mhos 2GHz susceptance reported in the listing file line analysis. These careful precautions would be unnecessary were it not for the desire to compare =EMPOWER= results with measured data to 20GHz.

=EMPOWER= results are given as the traces below and the circles and squares are measured data.

For the resistive film dimensions from the constructed attenuator, the series DC resistance is 43.905 / 53.175 * 50 = 41.283ohms and for the shunt resistors are 10.669 / 27.675 * 50 = 19.276 for each shunt half. This results in a DC attenuation and return loss

Signal Control\Film Atten.WSP

13

of 20.4 and 109 dB. The on-grid =EMPOWER= values are 20.4 and 30 dB.

At low frequencies it is difficult to rationalize the measured 20 dB measured attenuation with the predicted 20.4 dB value other than dimensional errors or film resistivity. The on-grid return loss of 30 dB in combination with the dimensional or resistivity error are potential sources for the disagreement in the return loss.

When ports are not deembedded the normalizing impedance is defined as 1 ohm. Therefore the S-parameter data displayed in the =EMPOWER= log window will not indicate good return loss and 20 dB attenuation during the EM run. The ports can be set at 50 ohms normalizing impedance by adding the option -NI50 in the Cmd Line cell in the =EMPOWER= setup dialog box.

Signal Control

14

Signal Control\Xfmr Coupler.WSP

Illustrates: Transformers/MUI, Equations

Common but puzzling components used in HF through UHF circuits are broadband transformers and couplers. They are hybrid mode devices which operate as magnetic transformers at low frequencies and as coupled transmission lines at high frequencies.

In this example, the coupler is simulated assuming the magnetic transformer mode. The equations are used to tune the coupling value. We specify the primary turns and the equation block calculates the closest integer secondary turns. The analysis provides insight which is elusive without simulation; the optimum unused port termination resistance isn’t equal to Zo and capacitance improves the return loss.

In the figure below, the solid curves are with R1=39 ohms and C1=1.2 pF. Using tune, you will discover the optimum R and C are functions of the coupling value and whether the through or coupled port is optimized.

Signal Control\Xfmr Coupler.WSP

15

Signal Control

16

Signal Control\Edge Coupler.WSP

17

Signal Control\Edge Coupler.WSP

Illustrates: Microstrip, substrates, =LAYOUT=

This example illustrates a 10dB microstrip coupler with three sections for improved bandwidth. Coupled-line couplers in microstrip do not achieve 3 dB coupling without exceptionally close lines. Therefore Wilkinson and branch-line networks are often used when equal splits are needed.

In the layout below, the rubber band lines which connect the three coupled microstrip sections were left unresolved. It is not absolutely neccesary to resolve rubber band lines if the metal is connected by a footprint or by a polygon of metal which you may add. In this case the lumped element capacitors, which improve the directivity, were used to close the natural gap between line sections caused by different line spacing.

Signal Control

18

Signal Control\ResistiveBroadband10dB.WSP

EMPOWER Example: Thin Film Attenuator

Illustrates: Analysis of a structure with thin resistive film and comparison with measured data.

10 dB Broadband Resistive Attenuator, developed by Res-Net Microwave Inc., of Largo, Florida, which offers a number of high performance resistive microwave products.

Simulation details: The structure is simulated with the solid thinning out (because of large areas of metal and film).


19


EMPOWER Example: Thin Film Attenuator

Illustrates: Analysis of a structure with thin resistive film, simulation without de-embedding, and comparison with measured data.

20 dB Broadband Resistive Attenuator, developed by Res-Net Microwave Inc., of Largo, Florida, which offers a number of high performance resistive microwave products.

Simulation details: The structure is simulated without thinning out (the most accurate analysis). Two external ports are marked as No Deembed to take into account actual reactance of the sidewalls. Option -ni50 sets normalization for the EMPOWER simulation.

Signal Control

20

Signal Control\Dual Mode Coupler.WSP

21


RAM: 1.9Mbytes Time: 68s/freq Illustrates: Dual mode couplers, solid thinning, viewer

Given above is a directional coupler laid on a 25 X 25 mil grid centered at 3GHz which a first glance resembles a branchline coupler. Port 1 was placed at the lower left, port 2 at the upper left, port 3 at the lower right and port four at the upper right.

For reasons discussed momentarily we refer to this device as a dual mode coupler. The sides are roughly half wavelength. This results in undesirably large size at lower frequencies, but at the higher microwave frequencies the smaller size of a branch line coupler introduces difficulties. 3GHz is normally the lowest frequency indicated for the dual mode coupler [Kawai and Ohta, 1994].

This version was designed for equal splits at ports 3 and 4. Transmission and reflection responses including losses as computed by =EMPOWER= and displayed by GENESYS are given below.

Signal Control

22

Notice that unlike the conventional branch line coupler the dual mode coupler is solid in the center. The large metal area requires solid thinning which was specified in the =EMPOWER= Properties Dialog box. Fortunately, the dual mode coupler has symmetry in both the XZ and YZ planes which compensates for the slower execution of solid thinning.

In conventional pseudo-TEM mode microstrip, current is maximum at the edges of the strip. In this coupler energy launched at port 1 soon transitions to TE10 in the large patch area, thus the term dual mode. This is evident in the Viewer


23

screen given below. Notice that the maximum current is located at the center of the structure. The Viewer is particularly useful in designing structures of this type. In a preliminary run, it was observed in the Viewer that the signal transferred to port 4 was significantly less than signal at port 3. The idea of placing an obstruction (notch) in the line between port 1 and port 3 immediately bore fruit. As an exercise, use the Viewer to observe the dynamic behavior of this coupler.

The previously cited reference gives additional information on these couplers including broadbanding techniques.

Signal Control

24

Signal Control\8 Way.WSP

RAM: 124.3MB Time: 11067s/freq Illustrates: Power dividers, large =EMPOWER= runs, lumped elements

The task is an 8 way power divider. It will be based on the cascade of Wilkinson equal dividers. This divider was tested first to verify and optimize the performance of this building block before the much larger, and slower executing, 8 way divider is tested.

The input is on the left and the outputs are on the top and bottom walls. The Wilkinson operates by transforming the two 50ohm terminations to 100ohms through quarter wavelength sections of narrower 70.7ohm lines. The two 100ohm impedances in parallel at the junction provide a 50ohm match to the short 50ohm input line. After the test and optimization of this single Wilkinson they were assembled in the 8 way unit shown below.

Signal Control\8 Way.WSP

25

This entire assemblage of 7 Wilkinsons and connecting lines was then run in =EMPOWER=. Initial runs indicated significant non-flatness near the center frequency due to box modes. The ports were spaced closer and the box size was reduced. The final design has what are believed to be box modes above and below the band of interest. Some of the initial runs were lossless to conserve execution time. The final runs were lossy and the required memory and time are for the lossy case.

Signal Control

26

The final transmission and selected isolation magnitudes are given here. Because =EMPOWER= generates all port data, the graphs in GENESYS may be modified to display data for any set of ports without rerunning =EMPOWER=.

Signal Control\BSCouplerFinalRecomp.WSP

27


Illustrates: =EMPOWER= ML, Decomposition

=EMPOWER= Example: Development of a microstrip surface mount 3-dB directional coupler for Mike Giacalone from Americal Technical Ceramics. This example uses advanced =EMPOWER= techniques to allow easy tuning and optimization of circuit performance. Notice in the design that at the center of the coupler, the lines are separated for a small portion. Adjusting the length of this section allows fine adjustment of the coupling.

There are three files in this example:

6. BSCouplerDevelopment.wsp - Implements coupler using tuned capacitors and decomposition. This allows a very quick analysis of the file, plus the coupler can be tuned.

7. BSCouplerFinalRecomp.wsp - The coupler's capacitors have been replaced with patches. This file still uses recomposition, allowing line length adjustments.

8. BSCouplerFinalWhole.wsp - In this final file, the coupler is simulated as a whole. This file confirms the validity of the decompositional simulation in the previous file.

The simulation results from the final coupler:

Signal Control

28

Specifications: Coupling: -3dB (S21 and S31) Return loss (S11): less then -20 dB. Isolation (S41): less then -20dB Simulation details: All elements are simulated with the solid thinning out technique to increase accuracy. CoupledLine: A segment of broadside coupled microstrip lines (conductors are in different levels) is simulated and optimized to get Ze=120, Zo=20 (see listing file). Inputs 1 and 2 are coupled as well as inputs 3 and 4. The program automatically goes to the mixed analysis mode (line+discontinuity). 50toCoupledJunction: A tee-like connection of two 50-Ohm microstrips in three-layered media with the coupled microstrip line (design CoupledLine). The inputs 3 and 4 on the right side form two-mode coupled input. CPJunction: A segment of coupled microstrip lines (design CoupledLine) is connected with a segment of a similar line with the same conductor widths but less coupled (to reduce overall coupling). The reference plane of the coupled inputs are shifted in the center. Recomposition: The coupler is assembled from two tee-like junctions of 50-Ohm microstrip lines and coupled line segments (design 50toCoupledJunction), two coupled microstrip line segments (line empower.l3 from the design 50toCoupledJunction, length L1), two step-like junctions with less coupled line (design CPJunction), and the less coupled line segment in the middle (line empower.l2 from the design


29

CPJunction, length L2). Descriptors of all segments must be precalculated. Run simulations EM2, EM3 to simulate the components. Goal coupling -3 dB (S21 and S31) and return loss and isolation below -20 dB could be achieved by tuning lengths (L1 and L2) of coupled line segments and capacitances AuxCap of capacitors Tuning Cap.

Design is not finished because capacitors are only circuit level solution and must be implemented physically (see possible solution in BSCouplerFinalRecomp.wsp).

Signal Control

30


EMPOWER Example: Microstrip surface mount 3-dB directional coupler designed for Mike Giacalone from American Technical Ceramics. Illustrates: An advanced multilevel example with decomposition. Parts of the coupler are simulated with EMPOWER in one work space, then recomposed (coupled sections in the mode space) and optimized to achieve desired performances. Specifications: Coupling: -3dB (S21 and S31) Return loss (S11): less then -20 dB. Isolation (S41): less then -20dB Simulation details: All elements are simulated with the solid thinning out technique to increase accuracy. CoupledLine: A segment of coupled microstrip lines (conductors are in different levels) is simulated and optimized to get Ze=120, Zo=20 (see listing file). Inputs 1 and 2 are coupled as well as inputs 3 and 4. The program automatically goes to the mixed analysis mode (line+discontinuity). 50toCoupledJunction: A tee-like connection of two 50-Ohm microstrips in three-layered media with the coupled microstrip line (design CoupledLine). The inputs 3 and 4 on the right side form two-mode coupled input. A rectangular patch was added to achieve 3-dB coupling, and return loss and isolation below -20dB. The dimensions of the patch were estimated using a couple of auxiliary capacitors hooked up at inputs 2 and 4 of the preliminary structure of the coupler (see BSCouplerDevelopment.wsp) CPJunction: A segment of coupled microstrip lines (design CoupledLine) is connected with a segment of a similar line with the same conductor widths but less coupled (to reduce coupling). The reference plane of the coupled inputs are shifted in the center. Recomposition: The coupler is assembled from two tee-like junctions of 50-Ohm microstrip lines and coupled line segments (design 50toCoupledJunction), two coupled microstrip line


31

segments (line empower.l3 from the design 50toCoupledJunction, length L1), two step-like junctions with less coupled line (design CPJunction), and the less coupled line segment in the middle (line empower.l2 from the design CPJunction, length L2). Descriptors of all segments must be precalculated. Run simulations EM2, EM3 to simulate the components. Goal coupling -3 dB (S21 and S31) and return loss and isolation below -20 dB could be achieved by tuning lengths (L1 and L2) of coupled line segments.

Simulation details: Solid thinning out is used to increase accuracy. Option -O2 is used to reduce line simulation time. Simulation EM2VIEW contains data to visualize current flow at the central frequency.

For the final EM simulation of the whole structure, optimized with decomposition see BSCouplerFinalWhole.wsp).

Signal Control

32

Signal Control\BSCouplerFinalWhole.WSP

33

Signal Control\BSCouplerFinalWhole.WSP

EMPOWER Example: Microstrip surface mount 3-dB directional coupler designed for Mike Giacalone from American Technical Ceramics. Illustrates: An advanced multilevel example

Here is the final EM simulation of the whole structure, optimized with decomposition in the example BSCouplerFinalRecomp.wsp. Specifications: Coupling: -3dB (S21 and S31) Return loss (S11): less then -20 dB. Isolation (S41): less then -20dB

Simulation details: Solid thinning out is used to increase accuracy. Option -O2 is used to reduce line simulation time. Simulation EM2VIEW contains data to visualize current flow at the central frequency.

Signal Control

34

Chapter 3: Components

Components\Model Extract.WSP

Illustrates: Post-Processing, Parameter extraction, optimization, simulation/data overrides, Link to Data File.

This is a model for a VHF bipolar transistor with package parasitics. The parameters in the BIP model and the package parasitics are optimized to match the model to measured S-parameter data. This is often refered to as “parameter extraction”. The schematic for the model is shown below.

The device data is loaded using a "Link to Data File". This example uses post processing, described in the Equation section of the Reference manual. The global equations subtract the device measured data (defined in the data file link "Data") from the model (defined in "BipModel", using sweep "Linear1"):

Using Linear1.bipModel Diff=abs(.RECT[S] - Data.Data.RECT[S]) errorS11=diff[1,1] errorS12=diff[1,2] errorS21=diff[2,1]

Components

36

errorS22=diff[2,2] The variable "diff" contains a 2x2 matrix with each element containing the magnitude of the complex difference between the measured and calculated S-Parameters. The optimization targets specify that the difference of the s-parameters should be zero:

It was empirically determined that S21 needed a smaller weight, allowing the optimizer to work harder on the other parameters. This example also illustrates the ability to plot different networks on one graph (note the Smith chart/Polar Chart properties which plot, for example, "S11" [from Linear1.BipModel] and "Data.Data.S11"):

Components\Microstrip Line.WSP

37


RAM: 7.7Mbytes Time: 253s/freq Illustrates: Microstrip impedance, EMPOWER introductory concepts, effect of cell size, required wall and cover spacing for microstrip, extrapolating line impedance and deembedding.

This example is simply a microstrip line in a box which is 400 mils long (in the x direction, the direction of propagation along the microstrip) by 855 mils wide ( y direction) with a cover 1000 mils above the microstrip metal layer. The line is 95 mils wide on a 31 mil thick substrate with a relative dielectric constant of 2.21. The characteristic impedance reported by =T/LINE= and the circuit simulator =SuperStar= which uses analytical expressions from Jansen and Kirschning is 50.1 ohms at frequencies up to 6 GHz. Shown below is a plot from the =LAYOUT= module where the problem was defined.

The grid spacing in x=20 mils and in y=11.875 mils. There are 8 grids across the microstrip line for a width of 95 mils. The box

Components

38

contains 20 cells in the x direction and 72 cells in y. Also defined in the General tab of the Preferences dialog box are the x and y coordinates of the origin, units, and the =LAYOUT= drawing snap angle. Default dimensions for line widths, drill diameter, and drawing port sizes are also specified. You may select the General Layer and =EMPOWER= Layer tabs to view the layer setup for this problem:

The return loss is generally good, suggesting a characteristic line impedance near 50 ohms. However, when a transmission line is 180o long the input impedance equals the terminating impedance regardless of the characteristic impedance of the line. To determine the line Zo, select Show Listing File from the right-click menu of EM1. Near the end of this file is a section labeled "Line parameters at 2.000e+009 Hz" where Zo is given as 47.746776 ohms. This is an error of 4.7% with respect to the analytical results of 50.1 ohms.

Zo vs. Cell Size If the "Grid Spacing Y" in the =LAYOUT= Preferences dialog box is changed from 11.875 to 5.9375 mils (increasing the number of cells across the line to 16), then recalculating =EMPOWER= yields a Zo of 48.810311 ohms.

Extrapolating Zo As the number of cells is increased the accuracy of the electromagnetic simulation improves but the simulation time increases. 128 cells only requires 7.8 Mbytes in this simple problem. But such a large number of cells for typical problems such as filters would require an unreasonable amount of computer memory and execution time. EMPOWER is


39

asymptotically well behaved with respect to the number of cells and the following formula by Richardson allows for accurate extrapolation of Zo.

where Z(Nw) is the impedance calculated by =EMPOWER= with Nw cells across the line. For example, in the previous runs with 16 and 8 cells across the line, Z(16)=48.810 and Z(8)=47.747. From this we compute for Zo=49.873.

Given in the table below are the direct and extrapolated line impedances for this example with Nw=2 to 128.

Nw Size (mils) Z(Nw) Zo

2 47.5 41.292 -

4 23.75 45.595 49.898

8 11.875 47.747 49.899

16 5.9375 48.810 49.873

32 2.96875 49.339 49.868

64 1.484375 49.602 49.865

128 0.7421875 49.733 49.864

Notice the well behaved convergence of this data. This is an inherent advantage of Method of lines electromagnetic simulation. From this table we conclude that the correct line impedance at 2 GHz is very close to 49.864 ohms.

Generalized S-Parameters One of the important features of =EMPOWER= and =SuperStar= is the support of generalized S-parameters. Many electromagnetic simulators write S-parameter data with a fixed reference impedance, typically 50 ohms. In the above example we see that accurate line impedances require a large number of cells. The 95 mil wide line has a Zo of 49.864 ohms, very close to 50 ohms. But electromagnetic simulation with 4 cells yields a Zo of 45.595 ohms, an error of approximately 8.6%.

EMPOWER writes and GENESYS reads generalized S-parameter data. Generalized S-parameter data is normalized not to a fixed impedance such as 50 ohms, but to an impedance which matches the deembedded line impedance of the electromagnetic simulation. This reference impedance is

Components

40

generally a function of frequency. =EMPOWER= writes files with the extension .R1,.R2, etc., which GENESYS reads when Generalized is checked in the =EMPOWER= simulation properties dialog.

This technique substantially improves the accuracy of simulations with a minimum number of cells, vastly improving the economy of electromagnetic simulation.

The following example illustrates this point. In the =LAYOUT= Preferences dialog box, change the Grid Spacing Y to 23.75 mils which is 4 cells across the line. Next, right-click EM1 in the workspace window and select Properties. Click the "Generalized" check box and press "Recalculate Now". This uses generalized S-parameter data from the =EMPOWER= run. The results using conventional and generalized S-parameter data are off the bottom of the graph, at around -160 dB.

When circuit subsections are added together the generalized S-parameter technique will result in significantly improved accuracy. To understand why this is so, consider the case of simply cascading true 50 ohm lines. If the lines are under discretized, the resulting lower impedances may strongly interact at quarter wavelength frequencies and cause significant error, even though all lines are equal width and no discontinuities exist.

De-Embedding When EMPOWER computes the electromagnetic fields of the microstrip line in a box, the simulation emulates a true physical configuration including side walls, end walls and the cover. Many of the circuits that EMPOWER simulates will be used within other circuits away from walls. The sidewalls and cover may be set at an adequate distance such that they do not significantly impact the results. However, the lines must approach the end walls to be launched. For this reason, =EMPOWER= includes deembedding routines to remove the effect of the end walls before reporting results. This is the normal operating mode. On the other hand, if the circuit being simulated is the entire circuit in a box and the effects on end walls are desired, then deembedding may be turned off. To do this, double-click on all ports of the circuit in the =LAYOUT= module and select "No Deembed" in the Port Type option.

Given below are the return loss responses of the line with 16 cells across. The line impedance at 2GHz with deembedded ports is 48.8103 ohms. The plots below are with a normalizing impedance of 48.8103 ohms and the trace which dips to over


41

100 dB return loss at 2 GHz is the deembedded line. The second trace is the same line but with non-deembedded ports. As can be seen the capacitance to the end walls degrades the return loss. Although it may not be not obvious through 6GHz, at higher frequencies this problem becomes more severe.

Zo vs. Wall Spacing and Cover Height Given below are Z(16), Z(8) and Zo for various box widths and cover heights. The original box width of 855 mils is (855-95)/2 or 380 mils spacing between the edge of the microstrip and the sidewalls of the box.

When the cover is well removed, Zo drops 0.5% at a box width of 475 mils, and with well removed sidewalls Zo drops 0.5% at a cover spacing of 310 mils. These spacings represent an edge to wall spacing of 6X the substrate thickness and a cover spacing of 10X substrate thickness.

Width Cover Z(16) Z(8) Zo

855 1000 48.810 47.747 49.874

475 1000 48.537 47.473 49.602

285 1000 47.712 46.641 48.783

166.25 1000 44.304 43.134 45.475

855 310 48.554 47.491 49.617

855 133 47.573 46.512 48.634

475 310 48.537 47.473 49.601

Components

42

Components\Stripline Standard.WSP

RAM: 18.2Mbytes Time: 776s/freq Illustrates: Benchmark, extrapolating line impedance, accuracy vs. cell size.

It would seem that benchmarking the accuracy of electromagnetic simulation is simply a matter of comparing measured results to computed results. Several examples of this form of benchmarking are given in other examples. However, the difficulty here is the identification and extraction of measurement errors.

An effective method of eliminating this difficulty is to compare electromagnetic simulation results for a system for which an exact solution is known. This is the basis of a benchmark proposed by Rautio [1994]. It consists of a zero metal thickness stripline of width 1.4423896mm in a box with a ground to ground spacing of 1mm and length in the direction of propagation (x) of 4.99654097mm which is 90 degrees at 15GHz. The filling material has a relative dielectric constant of exactly 1 (the relative dielectric constant of air is 1.0006). The sidewall spacing is 10X the strip width which is sufficiently removed to not impact the line Zo to several significant digits.

Even this "simple" test configuration is not a pure first principle system. To launch a signal the line must approach the end walls which increases the capacitance of the line to ground. The process of removing this effect is deembedding and in the following results the automatic deembedding routines in =EMPOWER= have been utilized.

The line Zo and transmission phase are taken from the =EMPOWER= listing file. In each case the cell size in the x direction is 0.156142mm which is 128 per wavelength at 15GHz. The following table gives the computed Zo and transmission phase vs. the number of cells across the line.

Components\Stripline Standard.WSP

43

Ny Zo(ohms) Error(%) ANG[S21] ( o) Time (sec)

2 39.969 20.06 90.0094 <1

4 45.059 9.88 90.0094 <1

8 47.575 4.85 90.0094 1

16 48.806 2.39 90.0094 7

32 49.413 1.17 90.0094 39

64 49.715 0.57 90.0094 202

128 49.865 0.27 90.0094 1455

To enhance accuracy, thinning is turned off. Execution time includes line and deembed modes on a 266 MHz Pentium II with 256 Mbytes of RAM.

Increasing the number of y cells improves accuracy at the expense of execution time. A sufficient number of cells for good accuracy is practical in this small problem. For larger systems, a small cell size results in unacceptable execution time. There are several methods available for improving accuracy while using moderately large cell size.

Richardson's Extrapolation This technique was introduced in the Microstrip Line example. The results for this benchmark are given below. This significantly improves accuracy and large cell size provides small errors. Successful application of Richardson's extrapolation requires well behaved asymptotic results from the EM simulator, which is one of the advantages of =EMPOWER='s MOL technique. While this technique may be used to determine more accurately the Zo of a line, it requires two runs and it has little direct impact on the precision of an EM analysis of a compound circuit.

Ny Zo(ohms) Error(%) Zo (ohms) Error (%)

2 39.969 20.06 - -

4 45.059 9.88 50.149 0.30

8 47.575 4.85 50.091 0.18

16 48.806 2.39 50.037 0.074

32 49.413 1.17 50.020 0.040

64 49.715 0.57 50.017 0.034

128 49.865 0.27 50.015 0.030

Components

44

Edge Positioning Singularities exist at the edge of metal strips. When the EM solution is well behaved as it is with the MOL, improved accuracy can be obtained by modifying the position of the edge [Schulz, 1981]. Significantly improved accuracy will result if the line is analyzed with the edge moved toward the center of the line by 25% of the cell size. Alternatively, the system can be analyzed with a given line width and later constructed with edges moved away from the line center by 25% of the cell size.

In the table below the line was analyzed with reduced width lines. The accuracy is significantly improved for relatively large cell size. In fact, increasingly smaller cell size results in little improvement and may even increase the error although the error at this point is small. The difficulty with this method is that adjusting cell size and line widths is tedious for circuits with arbitrary metal. Although it is not mathematically pleasing, the circuit may be analyzed and when the layout artwork is generated from =LAYOUT= an etch factor is entered to widen the lines by 25% of the cell size. This method works well and is simple to implement when the x and y cell sizes are similar.

Ny y cell size Zo(ohms) Error (%)

2 0.5408961 50.749 1.50

4 0.3155227 50.074 0.15

8 0.1690300 50.014 .03

16 0.087332218 50.013 .03

32 0.044370383 50.014 .03

64 0.022361265 50.015 .03

128 0.011224651 50.015 .03

Generalized S-Parameters Generalized S-parameters are perhaps the most user transparent method for improving accuracy while using larger cell size. It is easily implemented using =EMPOWER= and GENESYS Version 7.0 and later because they support generalized parameters. Few competitive EM/circuit simulator environments support these parameters. For additional details on the theory please refer to the Deembed chapter of the =EMPOWER= manual and for an illustration refer to the Microstrip Line example.

Components\Spiral Inductor2.WSP

45


RAM: 12.9Mbytes Time: 380s/freq Illustrates: Spiral inductors, lumped elements, accuracy vs. cell size, co-simulation of circuit and EM simulation

Shown below is the layout on a 5 X 5 mil grid of a spiral inductor in series with the transmission system. The small squares near the center and on the left are landing pads for a bond wire to connect to the center of the spiral. The WIRE model in =SuperStar= was used for this bond wire and in =LAYOUT= the BONDWIRE 100 X 20 mil footprint was selected. The bond wire is longer because it is a loop that must clear the spiral. This total system consists of an input microstrip line, the bond wire, the spiral, and an output microstrip line on a 62 mil thick substrate with a relative dielectric constant of 4.8.

The graph below shows the S21 responses from GENESYS circuit theory (top trace) and S21 from =EMPOWER=. Metal loss was included in the =EMPOWER= run.

Components

46

On the next graph are input reactances, IM[I1], using =EMPOWER= (peak trace) and circuit theory (smaller excursions). At 100 MHz, the circuit theory reactance of 13.07 ohms corresponds to an inductance of 20.8nH and an =EMPOWER= reactance of 15.52 ohms corresponds to 24.7nH.

Spirals are notoriously difficult to model in circuit theory simulators and the =EMPOWER= data is assumed more


47

accurate. Notice that near resonance (1400MHz) the reactances are more divergent. In general, the accuracy of spirals in a circuit theory simulator is useful at low frequencies, but expecting these models to be accurate at higher frequencies where the capacitance is significant is risky at best. Fortunately, =EMPOWER= runs for spirals generally do not require more than one or two cells across the lines and simulations are relatively fast.

To illustrate this given next are the same responses with only one cell across the line for the =EMPOWER= run. You may duplicate these results by running =EMPOWER= after changing the cell size from 5 mils to 10 mils. Notice that the reactance at 100 MHz is now 14.19 ohms corresponding to an inductance of 22.58nH. This is within 13% of the value with 2 cells across each line. Considering the excellent convergence properties of MOL electromagnetic simulation, we can assume the results with 2 cells are more accurate and is closer to the actual inductance than the 8.6% difference determined with 1 versus 2 cells. Those of you with sufficient memory (about 100Mbytes) could run this problem with 4 cells and compare results.

The inductances reported above are the effective inductances of the entire systems including the two microstrip lines, the bond wire and the spiral. The inductance of the microstrip lines could be removed by moving the reference planes in =LAYOUT=. The inductance of the bond wire could be manually removed by finding the inductance of just a bond wire in =SuperStar=.

Components

48

Components\Box Modes.WSP

RAM: 0.35Mbytes Time: <1s/freq Illustrates: Box modes, surface modes

Have you ever designed an amplifier, paying careful attention to the stability factor, only to have it oscillate despite all your precautions? Have you fought poor ultimate rejection in filter stopbands? Have you been unable to obtain flat gain in your amplifier? Have you fought large spurious signals in your system? A common cause of these problems is often box modes. These effects are often poorly understood but they are easily studied using =EMPOWER=.

Shown below is the simple system of a 3.6" by 1.96" box with a cover 0.250 inches above 50 ohm microstrip lines 1 inch long on a 62 mil thick substrate with a dielectric constant of 4.5. The grid is 40 X 40mils. The input and output lines are broken by a gap of 1.6 inches, far too large for any significant transfer from input to output. So it would seem.

Given next is the forward transmission, dB[S21], for this system with the box cover on (the top trace) and the top cover off (the bottom trace).

Components\Box Modes.WSP

49

Notice three peaks of transmission in the covered case at 3150, 4125 and 5250MHz. The peaks at 3150 and 5250MHz illustrate almost no attenuation from the input and output! These transmission peaks occur at the resonant frequencies of the box acting as a cavity. The first peak is the dominant (k101) mode of a 3.6 by 1.96 inch cavity partially loaded with the substrate dielectric. The remaining peaks result from higher order modes and additional peaks would exist if the sweep were extended. The transmission zero near 4500MHz is the result of two equal magnitude components from different modes vector adding out of phase.

How is a transfer attenuation of nearly zero decibels possible? Consider this LC circuit:

Components

50

Without the two parallel LC components the effective series capacitance is 0.005pF and the transmission in a 50 ohm system would be approximately -40 dB at 3150MHz. The addition of the resonant lossless LC pair reduces the attenuation to zero decibels. The ends of the microstrip lines in the original box radiate sufficiently to couple fully into the resonant cavity.

Clearly, the consequences of packaging circuitry in an enclosure are significant. First of all, even at a frequency as low as 1500MHz, it is unreasonable to expect stopband performance to exceed 60dB rejection. At higher frequencies, stopband rejection is totally destroyed. Poor stopband performance can devastate filter performance. When an amplifier is placed in a box, feedback near resonant frequencies may result in oscillation. When subsystems are enclosed in a box, spurious signals generated at one location may appear at another location with little attenuation, completely bypassing a filter designed to remove the spurs.

The lower response trace in the responses displayed by is with the cover removed. Actually, a layer with 377 ohms impedance was used for the cover. Sufficiently thick absorbent material on the cover would approach this condition. The Q is destroyed by energy lost from the cavity and fortunately the resonant peaks are removed. However, notice that the stop band performance remains less than -60dB across the sweep range and it worsens with increasing frequency.

The most effective weapon against these problems is a smaller box. Shielding of subsections within one enclosure is notoriously ineffective and careers have been wrecked by confident attempts.

For additional information on Box modes please refer to Chapter 8 of the =EMPOWER= Manual, Box Modes.

Components\User Model.WSP

51


Illustrates: Tuning, User Models

The crystal oscillator shown above was synthesized using the =OSCILLATOR= program. To pull the oscillation frequency the varactor network to the right of the crystal was added. We created a model for a varactor diode so that the frequency versus tuning voltage can be studied.

The capacitance of the varactor versus the tuning voltage is given by the expression: Cv=Co/(1+Vt/0.7)^Gamma

Vt is the varactor reverse tuning voltage. Co is the capacitance at Vt=0. Gamma is the power of the C vs V curve which is controlled by the varactor doping profile. It is typically 0.5 for abrupt varactors.

There are parasitics associated with the varactor. The schematic shown below depicts the series loss resistance which results in finite varactor Q. Package capacitance and package inductance are shown as Cp and Lp.

Components

52

To create a new user model:

2. Right-click on Designs in the Workspace Window and select "Add User Model (Schematic)". Enter the global name for the model ("Varactor" for our example.. You can choose to save this model into the MODEL directory.

3. Enter the parameters which will be passed to the model:

y The opening screen looks like a regular =SCHEMAX= screen. Use the same techniques you would use to draw a circuit schematic to draw a schematic of the model. If a component value is a constant, simply enter the value in that component’s dialog box. If a value is computed by equations or is to be passed as a parameter in the model, choose and enter a variable name in the dialog box field.

y If equations are used in a model, right-click the model icon in the Workspace Window and select "Edit Model Equations". For the varactor example, enter equations in the Equation Text Editor window shown here:

Cv=Co/(1+Vt/0.7)^Gamma C4=Co/(1+4/0.7)^Gamma Rs=1/(3.1416E8*C4*1E-12*Q)

Here the series resistance, Rs, is computed from the varactor Q which is typically specified at 4 volts bias and 50 MHz. In this way the model finds the effective Q at the desired tuning voltage and operating frequencies. The oscillator open loop cascade gain and phase are shown in the sample GENESYS screen for a tuning voltage of 2.64 volts.


53

Components

54

Components\BJT NL Model Fit.WSP

Illustrates: How to compare and optimize a nonlinear SPICE model to measured S-parameters.

As nonlinear models are complex, it is always a good practice to compare S-parameters derived from the nonlinear model to actual measured S-parameters. In this way, you can be sure that you have good parameters and that they have been entered properly. In addition, nonlinear models characterize devices over all frequencies and all bias levels. To improve accuracy, it is usually a good idea to optimize the parameters of the model for the particular bias condition a designer has selected for a given application.

In this workspace are three schematics, "Meas MMBR," "Test MMBR," and "MMBR Before Opt" Meas MMBR is a simple circuit that uses measured S-parameters for a Motorola MMBR941. Test MMBR contains the SPICE model for the intrinsic device and a set of parasitic inductors and capacitors, the values of which have already been optimized to fit the measured data. MMBR Before Opt contains the nonlinear SPICE model, before optimization has been run on the parasitic elements.

Components\BJT NL Model Fit.WSP

55

In the Equations window, the difference between Meas MMBR and Test MMBR is calculated. Then, optimization targets were defined that drives the differences to zero. Note that the S21 target weighting is set to 0.1 so that minimizing the errors of this large number will not overwhelm the errors in match. Weighting can be adjusted depending on a given design's requirements. The initial model for the device is shown in MMBR Before Opt. Inductances and capacitances were included to bond wires inside the package and lead inductances from the package itself. Capacitances were added between each junction. Initial values were estimates based on the modeling engineer's experience. As the optimization was run, values for several of the parasitics became very small. So small, in fact, that inductor L6 and capacitors C3, C5, and C7 were removed in the final model. As the plots of match and transmission show, the simpler model, using the optimized values, results in a much better fit.

Components

56

To experiment with this example, tune some of the parasitic values to move the curves from ideal. then use Actions/Optimize/Automatic to have the optimizer adjust the values back to improve the performance.

Chapter 4: Amplifiers

Amplifiers\Stability.WSP

This example illustrates stability circles and designing an amplifier for stability. The first step is to examine the stability characteristics of the selected active device before adding additional circuitry. Stability should be examined over as broad a frequency range as possible, and not just over the range desired for the amplifier.

Shown above are the input and output plane stability circles for an HP/Avantek AT41586 bipolar transistor biased at 8 volts and 25 mA. The shaded regions of the Smith chart represent regions of instability. To insure stability, the impedance presented to the device at its input terminal should avoid the shaded region of the input plane stability circles. Similar conditions should be satisfied at the output. In this case, since the circles above represent the lowest frequency and since the top half of the Smith chart is inductive, stability is enhanced by insuring that the device is capacitively terminated at low frequencies. Therefore, we will use a series capacitor at the input and output with the smallest value which does not disturb the desired amplifier. To further enhance stability, resistors to RF ground are added at the input and output. These will also be a part of the bias scheme.

Amplifiers

58

These capacitors and resistors are evident in the schematic shown above. The microstrip tee and transmission line models are added to account for the physical structure which is necessary to add the resistors to the amplifier. The remaining microstrip models comprise the matching networks which were optimized for 10dB of gain and best flatness.

Amplifiers\Stability.WSP

59

The results after optimization of the lengths of lines in the input and output matching networks are shown in the GENESYS screen shown above. Notice that the entire Smith chart region, which represents any possible passive load, is stable for both the input and output. Also notice that the sweep range for the amplifier gain and match is from 2000 to 2800 MHz, but the sweep range for the stability analysis is from 100 to 6000 MHz, the entire range for which S-parameter data was available. The layout after resolution of the rubber band lines is given below.

Amplifiers

60

Amplifiers\Balanced Amp.WSP

Illustrates: Microstrip design, the branch line coupler, the NET component, using two different networks in a workspace, and the =LAYOUT= module

This example is 2100-2900 MHz balanced amplifier. The single-ended (SE) amplifier used in this balanced circuit is shown at below. It is given the name AMP at its input. The return loss of the SE amp is shown on the Smith chart below. Notice the poor return loss.

Amplifiers\Balanced Amp.WSP

61

The balanced amplifier shown below is built using branch line couplers to split the input signal and later to combine the signals. The SE amp is duplicated in the balanced amplifier using the NET component. NET is given the designator AMP. As components in the SE amp are optimized they effect both amps in the balanced circuit. The branch line couplers deliver reflected signals to the terminating resistors so the return loss of the balanced circuit is improved, as shown in the response below.

Shown below is a finished layout of the balanced amplifier. This layout was created by selecting “Edit Layout” from the Layout menu in =SCHEMAX=. Footprints for the lumped elements and dimensioned metals are automatically placed on the layout page. You then select objects and snap nodes together to resolve rubber band lines.

Amplifiers

62

To duplicate the single ended amplifier in the layout, a few extra steps are required:

y Initially, the single-ended amplifier components and two NET objects are placed in the layout. First, return to =SCHEMAX= and double-click on each NET object. In the dialog box, select the LAYOUT button and then choose Replace with Open Circuit. This removes NET objects from the layout.

y Next, finish laying out the one SE amp. Draw a box around the SE amp portion of the layout and select Copy and Paste from the Edit menu in =LAYOUT=. Move these duplicated components to the desired position.

Amplifiers\Amp Feedback.WSP

63

Amplifiers\Amp Feedback.WSP

This example illustrates four feedback amplifier topologies. It is interesting to compare their gain, reverse isolation and match. Properly placed inductors and capacitors can typically improve the responses. This is also a good example of using multiple schematics on multiple graphs.

Amplifiers

64

Amplifiers\Amp Noise.WSP

Illustrates: Noise Figure, Microstrip, Circles, Layout of Microwave Circuits

This example of low-noise amplifier design is based on an article by Rob Lefebvre published in the March/April 1997 issue of Applied Microwave and Wireless magazine. It is a 9.5 to 10 GHz LNA using an HP/Avantek 10135 GaAs FET.

The amplifier schematic includes an extra FET with only the viaholes to ground the FET source leads. This portion of the schematic was added to display the noise circles of the FET alone. The center of the device noise circles is the impedance which should be presented to the device to achieve the best noise figure for the amplifier. This is the impedance seen looking toward the source at the input to the device.

Narrowband low-noise amplifier design is more straightforward than broadband design: 1) The device is stabilized with source inductance and or shunt resistors at the device input and output, 2) the input network is designed to present the correct

Amplifiers\Amp Noise.WSP

65

impedance to the device and 3) the output network is designed for maximum gain.

For broadband design the concept is the same. However, presenting the correct impedance to the device across the band and a flat gain requires balancing multiple goals. This is best accomplished using a modern simulator such as GENESYS to optimize all of the requirements simultaneously. The short arc inside the first noise circle is the locus of impedances versus frequency which should be presented to the FET. For even broader bandwidth, the =MATCH= synthesis program can be used to find a network which presents near optimum impedance to the device over the entire band.

Shown on the left above are the noise circles of the amplifier with the input network present. Notice that the center of the optimum noise arc passes through the center of the Smith chart. This indicates that the input network has been optimized so that at the middle of the frequency band a 50 ohm source will provide the optimum noise performance. This is verified by examining the noise figure versus frequency plot on the left in the =SuperStar= output screen. The gain flatness was acheived by optimization of the output matching network. Better output return loss could have been acheived by optimizing for match instead of gain flatness. The match at the input falls where it must because the input network is optimized for best noise and not best match.

The layout below was generated for the completed amplifier. The microstrip lines and discontinuities were automatically generated in =LAYOUT=.

Amplifiers

66

Amplifiers\Amplifier Tuned.WSP

Illustrates: Automatic integration of lumped elements into the electromagnetic simulation and combined EM and circuit co-simulation.

This example is an optimized variant of Amplifier.wsp, a 2.3-2.5 GHz Bipolar Common Emitter amplifier.

Simulation details: EM frequency sweep has four points, co-simulation sweep is the same as the circuit one. External ports are de-embedded to remove sidewall reactances and normalized to 50 Ohm. Other parameters are default, including the following: Planar internal ports are used to hook up capacitor and resistors. Tree Z-directed internal ports are used to simulate transistor connection.

Amplifiers\Amplifier Tuned.WSP

67

Amplifiers

68

Amplifiers\Amplifier.WSP

Illustrates: Automatic integration of lumped elements into the electromagnetic simulation and combined EM and circuit co-simulation.

This example is a 2.3-2.5 GHz Bipolar Common Emitter amplifier.

Simulation details: EM frequency sweep has four points, co-simulation sweep is the same as the circuit one. External ports are de-embedded to remove sidewall reactances and normalized to 50 Ohm. Other parameters are default, including the following: Planar internal ports are used to hook up capacitor and resistors. Tree Z-directed internal ports are used to simulate transistor connection.

Amplifiers\Amplifier.WSP

69

Amplifiers

70

Amplifiers\SiGe BFP620 Amp.WSP

Illustrates: The analysis of an amplifier designed, built, and measured by Gerard Weaver using Infineon's new BFP620 Silicon-Germanium bipolar transistor. This device is a high performance, low cost device housed in a 4-lead ultra miniature SOT-343 surface mount package. With a transition frequency (Ft) in excess of 70 GHz, the device is ideal for high performance applications such as portable wireless communications products.

The amplifier was designed to operate between 1930 to 1990 MHz. Design goals included 15dB of gain, unconditional stability, and an input referenced third order intercept of +10dBm. Low part count and PCB area were also desired.

The schematics "Dev NL Test" and "Dev Lin Test" were used to verify the nonlinear model. The nonlinear device was biased at 2V and 8mA, and S-parameters of the device were simulated. The outputs "Dev Refl" and "Dev Trans" compare the S-parameters of the nonlinear device to the measured S-parameters. Excellent correlation is seen, indicating that the model was well extracted for the device (model parameters were supplied directly from Infineon).

Once the model parameters were verified, the device was entered as a User Model in "BFP620," allowing for easy reuse in future designs. The user model was then placed in the overall design found in schematic "Amp NL." Mr. Weaver took extra care in his simulation model to include parasitics as part of his lumped components. He also included transmission line segments to model the parasitics of the PC board. The S-parameters of the overall amplifier were simulated and graphed on output "S-params." At 1960MHz, the gain is 15.4 dB, slightly better than the design goal. The overall response shape was very close to the measured results (available from Infineon).


71

Large signal analysis was then completed on the circuit. "Spect 1-tone" shows the output spectrum of the amplifier when driven with -20 dBm of available power. Note that the power measured at Port 1 (P1) is less than -20 dBm due to the mismatch of the port. The output spectrum is quite clean.

Amplifiers

72

"Compression" plots the gain compression of the amplifier as it is driven from -20 dBm to 0 dBm. The blue trace shows the gain compression. The 1dB compression point is approximately -5 dBm. The red and green traces show the output and input power versus drive level.

"Wave 1-tone" shows the output waveforms at each point in the power sweep.


73

"Spect 2-tone" shows the output spectrum around the center frequency when the amplifier is driven with two -20 dBm signals. This spectrum was used to calculate the input referenced TOI level, shown to be +9.3 dBm, very close to the 10.3 dBm measured on the built circuit.

Convergence controls were adjusted for this circuit. Full Jacobian calculation was set to "always" which speeds convergence as is typical when analyzing BJT circuits. In addition, the number of harmonics in "HB 1-tone" was increased to 16 so that the signals were adequately sampled during the power sweep.


Chapter 5: Filters

Filters\Contiguous Diplexer.WSP

This example illustrates the design of a contiguous diplex filter. It was designed by the following steps:

1. Design a 7th order singly-terminated Butterworth highpass filter using the synthesis program =FILTER= and write a file named CONTIGUS.SCH.

2. Return to =FILTER=. Design a 7th order singly-terminated Butterworth lowpass and write a file named LP.SCH. Then run GENESYS and display the lowpass response.

3. Open the schematic and draw a box around the entire lowpass schematic. Selecting “Cut” and “Paste” from the Edit menu places the lowpass schematic in the buffer and back in the schematic.

4. Next, load CONTIGUS.SCH, open the schematic, and select Paste from the Edit menu which drops the lowpass over the highpass. Drag the lowpass schematic off of the highpass schematic using the mouse.

5. Connect together the two filters at the singly-terminated, zero-impedance ends and modify the port impedances and graph properties to display the desired information. The resulting schematic is shown below:

Filters

76

The responses, isolation, and return loss are shown on the graph below. Notice that the RL is excellent through the entire crossover region. This is a natural and desirable consequence of designing diplexers by connecting together singly-terminated filters with identical cutoff frequencies (contiguous). With lossless, ideal components, the RL is theoretically infinite at all frequencies. Similar results are acheived using Chebyshev filters with contiguous 3 dB corner frequencies.

Filters\ComblineDesign\Final Lossy.WSP

77


Illustrates: EMPOWER Example - Design of combline filter.

This is a multi-step example and starts with ComblineDesign\Step 1.wsp.

Filter specifications: Type: 5-resonator combline on a PTFE substrate. Center frequency: 1800 MHz Passband: 1700-1900 MHz. Step1: Optimization of the synthesized filter - Step 1.wsp Simulation details: The filter was synthesized with the M/FILTER. Capacitors values and resonator lengths where specified as tunable variables and optimized to get proper passband and return loss. There is no EM simulation yet, because the layout of the filter fits only a grid with excessively fine grid (1x1 mils).

Filters

78

Step 2: Initial EM simulation - Step 2.wsp Features: Fitting results of the synthesis to a grid and subsequent interactive optimization. Filter specifications: Type: 5-resonator combline on a PTFE substrate. Center frequency: 1800 MHz Passband: 1700-1900 MHz.

Simulation details: The filter optimized in the Step 1 example is adjusted to a grid with cell size 6x6 mils. All microstrip line widths and spacings are rounded off to be multiples of 6 mils. It made the passband of the circuit theory model a bit wider and shifted it a bit. The values of the capacitances adjusted to get the center frequency back.

EM analysis results: The passband is shifted to lower frequencies about 8% and 10% wider then expected. The return loss are also no good.


79

Step 3: Shifting the center frequency - Step 3.wsp Features: Fitting results of the synthesis to a grid and subsequent interactive optimization. Simulation details: The filter adjusted to the grid in the Step 2 example. The lengths of the coupled line segments are made shorter by direct shifting them up and down in the layout. The input line segments are completed by tapered line segments. Results: The center frequency is shifted.

Filters

80

Step 4: Narrowing down the passband - Step 4.wsp This step illustrates the fitting results of the synthesis to a grid and subsequent interactiveoptimization. To make the passband narrower, the spacings of the coupled line segments are increased. Results: The passband is adjusted.

Step 5: Improving matching - Step 5.wsp Features: Fitting results of the synthesis to a grid and subsequent interactive optimization. Simulation details: The tap point is moved up a little bit to improve matching. Results: The return losses is improved.


81

Last Step: Final simulation and tuning with losses - Final Lossy.wsp Features: Fitting results of the synthesis to a grid and subsequent interactive optimization.

Simulation details: Only the capacitances where adjusted a little bit to tune passband of the filter.

Filters

82

Notes: The structure is very sensitive to the capacitances values. So, it will be necessary to tune a fabricated filter.

Filters\Coupled Stepped Z.WSP

83


RAM: 1.0Mbytes Time: 7s/freq Illustrates: Filters, =EMPOWER= thinning, stopbands, multiple =EMPOWER= runs

This filter illustrates filter design techniques which would not be feasible using conventional circuit theory simulation because it involves coupling between complex shaped resonators. The final filter design is shown below.

The filter is a bandpass centered at 2950MHz which utilizes three stepped impedance resonators (SIR). SIRs are physically shorter than uniform impedance resonators. The inductance is increased in the narrow sections near the viahole ground and the capacitance is increased at the open end. Both effects decrease the resonant frequency for a given physical length. With uniform impedance resonators this structure would be interdigital.

The filter was originally designed also using =EMPOWER= and the general technique described beginning on page 379 of the book "HF Filter Design and Computer Simulation". The resonant frequency and external Q vs. tap point were first determined using =EMPOWER= with a single resonator. Next, the internal coupling coefficient vs. resonator spacing was determined using two resonators. The filter was designed from this data and the technique described in the reference. The input and output taps reduce the frequency of the end resonators. The center resonator frequency is reduced by the addition of loading metal at the open end. This compensation was determined empirically.

Filters

84

Adequate resolution in the passband in combination with a wide sweep to observe the stopband requires a large number of frequencies. To conserve execution time, =EMPOWER= was first run from 2500 to 3400MHz with 31 points ("EM Passband"). Next, an =EMPOWER= simulation ("EM Stopband") was run from 1150 to 10500 MHz with 61 points. The final GENESYS screen after this second run is given below. The critical frequency was set at 4000 MHz for accuracy through the passband but precision was not required in the stopband. A critical frequency of 4GHz allowed thinning with accurate passband results.

To further reduce execution time, the above data was taken without Viewer data. One simulation was then modified, requesting three frequencies from 1000 to 5000 with "Generate Viewer Data" checked. The top view of the current magnitude at 3000 MHz is given below.


85

Notice that the lines leading to the input and output taps appear as thin lines in the Viewer. This line was one cell across and the edges were not aligned with the edges of the cell. It mapped and displayed onto one grid line. =EMPOWER= calculates current along lines and not the space between lines. In this case, the results are useful but improved precision would have resulted from a more careful grid alignment. The input and output lines also are not exactly aligned with the grid. This inexact alignment has caused a nonsymmetrical mapping. This is due to the edge of the line being exactly centered in a cell.

Filters

86

Filters\Edge Coupled.WSP

Files: EdgeCoupledOpened.WSP and EdgeCoupledBenchmark.WSP RAM: 2.0Mbytes Time: 15s/freq Illustrates: Benchmark, filter loss, phase, fitting grid to existing design, reference planes

This 3-section edge-coupled 12 GHz bandpass filter is analyzed in =EMPOWER= and compared to measured results [Wolff and Gronau, 1989]. This paper presented a practical method for deembedding the effect of connectors from measured data and it presented measured results for the 3-section edge-coupled bandpass filter of interest in this example. Since the connectors were deembedded from the presented data we will run =EMPOWER= with the automatic deembedding routines invoked.

This paper was selected because the authors have an record of reliable work. The circuit of interest was only a portion of the paper and a description of this circuit was brief. We chose appropriate box dimensions to complete the description for the =EMPOWER= run.

The filter dimensions were fit to a grid of dx=0.105 and dy=0.105mm. For a definition of these dimensions please refer to the original paper. The original and on-grid values are given here to illustrate the dimensional errors introduced by placing the filter on grid.

Parameter Original On Grid

L1 4.612 4.620

L2 4.519 4.515

L3 4.502 4.515

L4 4.599 4.620

S1 0.100 0.105

S2 0.340 0.315

S3 0.325 0.315

S4 0.100 0.105

H 0.508 0.508

Filters\Edge Coupled.WSP

87

Wa1 0.815 0.840

Wb1 0.331 0.315

Wa2 0.387 0.420

Wb2 0.381 0.420

Wa3 0.380 0.420

Wb3 0.393 0.420

Wa4 0.329 0.315

Wb4 0.806 0.840

Hbox 5.080 5.080

The largest percentage errors are with the microstrip widths. In bandpass filter structures there is a large sensitivity of the responses to the resonant frequency of the resonators. This is predominantly determined by the resonator length and the percentage errors in length associated with placing the metal on the grid were small. The width of resonators has only a small impact on the resonant frequency but they impact filter bandwidth.

In this example the reference planes were shifted from the edge of the box to the edge of the input and output coupling lines. They are shifted by selecting the port and dragging the reference line in the desired direction. The shifted reference lines may be observed in the =LAYOUT= module for this example.

Given below are plots of S21 and S11 linear magnitudes and phases as computed by =EMPOWER= and displayed by GENESYS.

Filters

88

Excellent agreement with measured data (not shown, see the reference above for measured data) was achieved. As stated earlier, the choice of grids introduced small percentage errors in the length of metal but width errors were significant. While the center frequency appears to be slightly higher in the simulation this is possibly the result of bandwidth shrinkage with most of the shift occurring on the lower side of the passband. An interesting exercise for those with adequate memory would be to run this filter with gridding closer to the original dimensions.

Edge coupled filters are susceptible to radiation loss. Given below are responses of the previous this filter in an enclosure with the height (y direction) doubled to 18.9mm and with the cover removed. Notice that the insertion loss on the low side of the passband has increased. This is due to radiation. Can you predict why radiation increases the loss primarily below the center frequency? Notice the non-monotonic response in the lower transition region. This is probably due to box modes. Would you expect radiation to occur in the original width enclosure with the cover removed? Answers to these questions are given in Section 5.9 of HF Filter Design and Computer Simulation.

ResonanceElimination\ EdgeCoupledWithFictivePorts.WSP

89


EMPOWER Example: Edge-Coupled Filter - Resonance Elimination Technique.

Illustrates: EM simulation of a filter with edge-coupled microstrip resonators and comparison with published results of simulation and measurement [1]. Also demonstrates a resonance elimination technique that uses fictive loaded internal ports to expand and interpolate S-matrices of the filter and then reduces the expanded descriptor matrix to the initial one. Description of the problem: The three-resonator edge coupled filter is described in [1]. Transverse dimensions of the coupled line sections were rounded to the nearest integer number of discretisation distances. The discretisation distances along and across lines were 0.105 mm. Dimensions of the Grounau and Wolff filter and (rounded) values are:

l1 = 4.612 (4.62) l2 = 4.519 (4.515) l3 = 4.502 (4.515) l4 = 4.599 (4.62) Wa1 = 0.815 (0.840) Wb1 = 0.331 (0.315) Wa2 = 0.387 (0.420) Wb2 = 0.381 (0.420) Wa3 = 0.380 (0.420) Wb3 = 0.393 (0.420) Wa4 = 0.329 (0.315) Wb4 = 0.806 (0.840) s1 = 0.100 (0.105) s2 = 0.340 (0.315) s3 = 0.325 (0.315) s4 = 0.101 (0.105) w = 1.580 (1.575)

Filters

90

Simulation details: Design Layout1 describes the whole filter that is simulated in EM1 at 49 frequency points (about 20 min per frequency point). S-parameters are generalized. Design plus4ports is the same filter but with 4 additional internal ports breaking resonators in halves and eliminating resonances when they are 50-Ohms loaded. Corresponding simulation EM2 is done only for 7 frequency points in the same frequency range as the initial filter (S-matrix is very smooth). Design short4ports takes 6-port data from the simulation EM2 and sets proper boundary conditions in fictive port areas (short circuit conditions). The external ports are normalized to impedances calculated in the EM2 simulation to have generalized parameters of the filter. Thinning out is off for both simulations. Option -O4 is used in all simulations to enforce de-embedding line segment to be shorter then default and to reduce de-embedding simulation time. Simulation results: The simulation results obtained by the resonance elimination technique is almost on the top of the en-block simulation results. 7 times acceleration is achieved with simple interpolation technique and without considerable loss of accuracy.


91

Reference: 1) Gronau G., Wolff I. A simple broad-band device de-embedding method using an automatic network analyzer with time-domain option. - IEEE Trans., v. MTT-37, 1989, N 3, pp. 479-483. The filter shown in Fig.11 of the paper.

Filters

92

Filters\Interdigital.WSP

RAM: 19.6Mbytes Time: 1227s/freq Illustrates: Low loss filters, non-resonant simulation, wall grounding

The requirement for this exercise is a low loss printed 5th order bandpass filter with a center frequency of 1178MHz and a bandwidth of approximately 50MHz. This is a bandwidth of 4.2% and a loaded Q of 23.6. Low insertion loss dictates a high unloaded Q so a suspended stripline structure with a large ground to ground spacing of 500 mils was selected. The metal is etched on a 32mil thick teflon PWB with a relative dielectric constant of 2.2. The metal is centered between the grounds and the teflon is below the metal. The layering setup may be viewed in the layer tabs of the =LAYOUT= Properties dialog box. The metal pattern is shown below.

The lines terminate in ground at the sidewalls rather than through via holes. The filter may be designed using the technique referred to in the Coupled Stepped Z example, or in this case by designing a stripline filter in =M/FILTER= and entering a relative dielectric constant based on a filling factor, a dielectric constant of 1 + (32/500) * (Er-1)=1.08. This serves as a reasonable starting point.

The small 4.2% bandwidth increases the sensitivity of the response to small errors in the analysis. Therefore, a fine grid is necessary. In this case a grid of 24mils in x and 18mils in y was used. This results in 13 cells across the width of the lines which is generally more than necessary. However, this cell size was selected because the spacing to the input/output coupling lines is

Filters\Interdigital.WSP

93

small. The small grid in y is required to satisfy the need in narrowband filters for precise control of resonator frequencies.

This problem contains 225 cells in x and 136 cells in y for a total of 30,600 cells. This in combination with the need for several frequencies in the passband and the probable need for several runs to iterate the design would result in a tedious and time consuming effort were it not for a number of tricks employed with this filter. Listed in order of importance they are:

TIP 1: The interdigital filter behaves simultaneously as multiple coupled lines and resonators. It is the resonant process which is frequency selective. The frequency dependence of the coupling coefficients is slow and monotonic. The lines were terminated in ports rather than ground so that they are nonresonant. GENESYS then reads this data and terminates all ports except the input and output with a direct connection to ground. This manifests in resonance and the filter bandpass behavior is observed. Because =EMPOWER= is analyzing a monotonic nonresonant structure only 5 frequencies were used. During the interactive design process 3 frequencies were used and then for the final plot given here a final =EMPOWER= run with 5 frequencies was used. The results for 3 and 5 frequencies in this example agree within hundredths of a decibel at every frequency in the 101 frequency GENESYS plot. The schematic to ground the nodes is shown below along with the circuit response. Notice the use of the NPO8 block with the filename "WSP:Simulations\EM1\EMPOWER.SS". This instructs the NPO to get it's data from the Current workspace (WSP:), "Simulations" Folder, Simulation "EM1", file EMPOWER.SS. This is the method to get raw =EMPOWER= data from a simulation. For more details on this method, see Appendix B in your =EMPOWER= manual.

Filters

94

TIP 2: Once the resonator grounds are replaced with ports it is possible to add a line model from the port to ground. If these line models are also stripline with a similar substrate description then the line lengths may be tuned or optimized with immediate GENESYS display of the results. Resonator length corrections for a new =EMPOWER= run are thus found quickly.

TIP 3: A similar circuit can be analyzed entirely as a circuit theory file in GENESYS to estimate the source of response errors in the =EMPOWER= run. For example, if GENESYS shows that end spacings which are narrowed match =EMPOWER= results then the spacings are increased for the next =EMPOWER= trial.

Filters\Tuned Bandpass.WSP

95


RAM: 5.3Mbytes Time: 92s/freq Illustrates: Tunable filters, non-resonant simulation, lumped elements, correlation of lumped/EM simulation

Filters may be electronically tuned using voltage variable capacitors (varactors). While varactors may adjust the center frequency, the remaining element values are generally not optimum for the new center frequency. Adhering to the following principles mitigates these difficulties:

1. Use like-kind resonators for each section. Avoid the classic BP with alternating series and shunt resonators.

2. For constant bandwidth vs. frequency, internal coupling between resonators should decrease with increasing frequency. Series coupling capacitors are common in filters but are a very poor choice for tuned filters.

3. As above, external coupling should decrease with frequency. Series coupling inductors may be used for both internal and external coupling.

The filter is a 2-section microstrip combline bandpass. Combline structures are grounded on one end and are capacitively loaded at the open end. If the lines are a quarter wavelength long, magnetic coupling near the grounded ends and electrostatic coupling at the open ends are equal in magnitude but opposite in phase, resulting in no coupling (an all-stop structure)! With capacitive loading at the open ends, the electrostatic coupling sections are shortened and coupling is predominantly magnetic.

Filters

96

In the case of a tunable bandpass, the length of the combline is selected such that the coupling between resonators is reduced at the proper rate as the frequency is increased. A similar technique is used for external coupling. The outside line sections are not resonators but provide external coupling which decreases with frequency.

The responses on the left are circuit theory simulations by GENESYS and on the right are =EMPOWER= results. The traces are with the varactors at 0.55pF. You can tune the varactors to shift the frequency.

Notice the transmission ripple and return loss predicted by =EMPOWER= are somewhat higher. The circuit theory simulation does not include the capacitance of the varactor landing pad footprints. This stray capacitance is a higher percentage of the total capacitance at higher frequencies and the resulting frequency differences are more significant at higher frequencies. An inherent advantage of electromagnetic simulation is that more of these strays, not only capacitance but path length and others, are incorporated in the simulation thus resulting in improved accuracy.

This example illustrates an important concept which can save significant execution time. The above responses were generated from =EMPOWER= data for only only 4 frequencies! Furthermore, =EMPOWER= data does not need to be retaken to tune this filter using the varactors! The length of the coupled lines are less than 90o long and they are not resonant. Resonance is achieved with capacitive loading. The =EMPOWER= data is taken without the varactors present and the line impedances and couplings change very slowly with frequency. EM data is only required for a few frequencies. This technique can be directly applied to all combline filters, thus


97

saving significant execution time. It can even be applied to interdigital filters by performing the EM analysis with the via holes replaced with ports. After the EM run, the ports are replaced with circuit theory via hole models to achieve resonance. This technique is should not be overlooked!

Filters

98

Filters\Two Level.WSP

=EMPOWER= Example: Two level microstrip bandpass filter. Features simulation of a filter with coupled microstrip resonators in two metallization levels and comparison with published results of simulation and measurement [1].

Three side-coupled microstrip resonators are on a substrate. An additional resonator is placed on an inverted substrate and over the three resonators to achieve coupling with the input resonators (through one). Both substrates have the same height 0.51 mm and dielectric constant 2.33. They are separated by a foam dielectric with height 3.3 mm and dielectric constant 1.07. Microstrips are 1.49 mm wide (rounded off to 1.5mm). Whole structure is in a box 40 mm by 20 mm.

Reference:

1) A.A. Melcon, J.R. Mosig, M.Guglielmi, Efficient CAD of boxed microwave circuits based on arbitrary rectangular elements, IEEE Trans., v MTT-47,1999, N 7, p.1045-1058

Filters\Xtal Filter.WSP

99

Filters\Xtal Filter.WSP

Note: This crystal filter requires very high precision components. We strongly recommend changing the "Digits right of decimal" in Options from the Tools menu to 6. You may want to change it back after you are done with this example.

A 4th order Chebyshev filter with a center frequency of 9.001 MHz, a bandwidth of 3 KHz and 300 ohms terminating impedance is designed using a crystal with the following parameters:

Rs = 31 ohms Lm = 24.54 millihenries (24.54E6 nH) Co = 4.18 pF Cm = 0.0127429 pF, which resonates with Lm at the crystal series frequency

=FILTER= was used to design the shunt-C coupled bandpass, specifying 24.54E6 nH for the inductor, and then writing the GENESYS file. The shunt-C coupled bandpass filter topology is similar to ladder crystal bandpass filters. Since the shunt-C filter allows specifying the series inductance, designing ladder crystal filters is straightforward. Each series inductor-capacitor pair is converted to a crystal XTL model as shown in the circuit above. The response is shown in the figure below.

Filters

100

The 2.4 pF crystal parallel capacitance causes the high side selectivity to be greater. This places an upper limit on the bandwidth of this type of ladder crystal filter. Placing an inductor in parallel with the crystal to resonate out Co may allow a wider bandwidth.

S/FILTER

101

S/FILTER

How To Design

This chapter utilizes examples to illustrate "How To Design" filters that meet important and practical design goals. For example, how to design filters with equal input and output terminations or how to design filters with all series resonators. These requirements often arise in the development of filters for specific applications. For example, when quartz crystals or transmission lines realize the final filter, all series or all shunt resonators are typically required. Direct synthesis creates filters with maximum economy for specific responses but achieving certain desirable topologies must be directed.

One strength of S/FILTER over other synthesis programs is that it provides the user with a rich set of tools for directing the synthesis process. This chapter uses examples to illustrate how to use these tools. Once mastered, these tools are easily applied to solving your particular requirements.

Note: Each example in this section lists two filenames. The .WSP file contains the schematic and any associated layouts or optimizations. The .SF$ file contains the S/FILTER settings used to design any filters used in the example.

Filters

102

Equal Termination Example

Files:

1. Filters\SFILTER\Equal Termination.WSP

2. Filters\SFILTER\Equal Termination.SF$

This example illustrates how to design filters for equal input and output termination resistance. These techniques may be used to also design filters with specific, unequal, termination resistance such as 50 Ω input and 75 Ω output. Given here is the Design tab of S/FILTER for a 50 Ω, 10.7 MHz IF filter with 300 KHz bandwidth.

The Table in the Extractions tab is Customized to show the TRF Ratio, the number of Inductors and the Permutation (extraction sequence). A transformer ratio of 1:1 would result in a 50 Ω output. The Table was sorted by the transformer ratio by clicking on the TRF Ratio column header. Next the table is scrolled to find a transformer ratio of 1. Although this often occurs, in this case the closest ratios are 0.59 and 1.69 and a transform is required to remove the transformer. The Permutation sequence ∞ 10.4 ∞ DC 11 ∞ is selected as shown here.

S/FILTER

103

The resulting schematic is

Next, the transformer is removed while keeping a 50 Ω output by selecting L2 and applying a Norton Series transform with the Calculate N option. The following schematic results

Filters

104

Maximum Realizability Example

Files:

1. Filters\SFILTER\Maximum Realizability.WSP

2. Filters\SFILTER\Maximum Realizability.SF$

The final filter schematic in the Equal Termination example has six inductors, a maximum to minimum inductor ratio of 318 and a minimum inductor value of 4.72 nH which is very low for a 10.7 MHz filter. Narrow bandpass filters often have significant realization issues but could we do better in this case? This example illustrates methods for improving the realizability of filters.

To maximize realizability and resonator Q for a 10.7 MHz bandpass, we desire the minimum number of inductors, inductance between 1000 and 100,000 nH, equal 50 W input and output terminations, and no transformer.

The Extraction table can be configured to display Permutations sorted by an error based on departure from user goals. Given here is an Extration Goals setup dialog box.

Next the extractions were sorted first by Error by clicking on the Error column header and then by number of Inductors. The resulting Extraction tab is given here.

S/FILTER

105

The red entries in the table indicate inexact permutations.

Entry #4 with the Permutation ∞ 10.4 11 ∞ ∞ DC is selected. The minimum inductor is lower than desired but other 4-inductor extractions are less desirable. If 112.75 nH is deemed too small then the 5-inductor extraction 11 ∞ 10.4 ∞ DC ∞ is an alternative.

The 4 inductor permutation has a transformer. It is removed by clicking the Remove Transformer button. Fortunately this reduces the maximum inductor value. The resulting schematic is

Filters

106

All Series Resonators Example

Files:

1. Filters\SFILTER\Series Resonators.WSP

2. Filters\SFILTER\Series Resonators.SF$

This example illustrates how to design bandpass filters with all series resonators. A three-section 800 Hz bandwidth communications receiver IF filter centered at 9 MHz is realized using quartz-crystal resonators. The quartz-crystal parameters are Lm = 24E6 nH (24 mH), Cm = 0.01303 pF (13.03 fF, Co = 3.6 pF and Rm = 31 Ω. The motional inductance and capacitance given here series resonate at 9.0 MHz. In the final filter, each crystal may series resonate at slightly different frequencies as determined by the final Cm in the schematic. The filter will be designed with all inductors equal to precisely 24 mH.

The Specifications tab is given here. The extraction sequence ∞ DC ∞ ∞ ∞ ∞ with a series element first is chosen because it results in a series L-C at the input. The source resistance is tuned until the first series inductor is 24 mH. The required resistance is 101.5 Ω. The load resistance is set equal to 101.5 Ω.

The initial schematic is given here. The number of digits to the right of the decimal in element values has been set to 8 to view

S/FILTER

107

the precise values associated with high-Q quartz-crystal resonators.

Next two Norton Shunt transforms are applied to the shunt coupling capacitors to convert the remaining two series inductors into series L-C resonators to conform to the motional element branch of the equivalent circuit model for quartz-crystals. C2 is selected, Apply Norton Shunt is selected and "Choose the transformer ratio (N)" is selected. We desire L2=L1 so a transformer ratio equal to the square root of L2/L1 = 7.587542e-5 is entered.

The schematic after applying the same transform to the second shunt capacitor is given here. The filter is symmetric, after the transforms the input and output resistances are equal, and S/FILTER automatically removed the transformer. If a transformer remains with a ratio near but not exactly 1:1 the transformer may be manually deleted.

Next, the design is standardized by changing the terminating resistance at ports 1 and 2 from 101.5 to 100 Ω, setting the shunt coupling capacitors to 180 pF, adding 31 Ω series resistors and adding the crystal 3.6 pF static capacitors. The inductors are set one at a time to precisely 24 mH and tuning its series capacitor to correct the response. Finally, the series capacitors are optimized to clean up the tuning. The final response and schematic are given here.

Filters

108

S/FILTER

109

All Parallel Resonators Example

Files:

y Filters\SFILTER\Parallel Resonators.WSP

y Filters\SFILTER\Parallel Resonators.SF$

This example illustrates how to design bandpass filters with all parallel resonators. The Specifications tab of S/FILTER shows the filter specifications.

The Extraction tab indicates there are 4 unique extraction sequences. Notice that Series Element First option is not selected so that the first element will be shunt. The table has been customized to show only the extraction sequence and the transformer ratio. Sequence 3 is selected since the transformer ratio is essentially 1 providing equal input and output termination resistance without requiring a transformer.

Filters

110

The initial schematic for extraction sequence 3 is given here. To create a topology with all parallel resonators the circuit is transformed to place capacitors in parallel with the shunt input and output inductors.

First, C1 is selected and a Norton Series transform is applied. We require a positive capacitor on the left and the resulting negative capacitor on the right will be absorbed by C2. The option is selected "Choose the transformer ratio (N)". To maximize realizability N=SQR(L2/L1) or 0.2003249 is entered. This shifts the impedance of the filter to the right of C1 up an amount which causes L2=L1. The Simplify Circuit transform next combines the two transformers into one transformer at the output.

S/FILTER

111

Next we will apply a Norton Series to the remaining series capacitor to place a capacitor in parallel with the shunt output inductor. This time we want the positive capacitor on the right. Since L3>L1 we choose N=SQR(L3/L1) or 4.991998. Finally, because the transformer ratio is near unity it is absorbed into the load. The History tab with a list of the transforms and the resulting schematic are given here.

Filters

112

Response Symmetry Example

Files:

y Filters\SFILTER\Response Symmetry.WSP

y Filters\SFILTER\Response Symmetry.SF$

A symmetrical transmission amplitude response is often desired, particularly for IF filters. Conventional exact transform bandpass filters have a symmetrical response when plotted on a logarithmic frequency scale (geometric symmetry). It is arithmetic symmetry (plots on a linear frequency scale) that is desired in IF filters. Arithmetic symmetry also results in symmetry in the group delay (equal peak group-delay values near the lower and upper cutoffs). None of the popular lowpass to bandpass transforms possess arithmetic symmetry. In 1989 Eagleware developed a lowpass to bandpass transform that results in arithmetic symmetry. You may refer to pages 165 to 167 of HF Filter Design and Computer Simulation for more information on this transform. However, this transform is approximate and is available for all-pole (no finite transmission zeros) filters only.

The direct synthesis routines in S/FILTER offer a more elegant solution to this problem. This example illustrates how to exactly design bandpass filters with arithmetic symmetry. Carassa [Band-Pass Filters Having Quasi-Symmetrical Attenuation and Group-Delay Characteristics, Alta Frequenza, July 1961, p. 488] proved that if the number of transmission zeros at infinity is 3 times the number at DC, the response possesses arithmetic symmetry. This can be maintained even with the addition of transmission zeros at finite frequencies.

The Specifications tab for a 70 MHz IF filter with 30 MHz bandwidth is given here. Notice that the quantity of zeros at infinity are 3 times the quantity at DC. The finite zeros were tuned to 38.7 and 103.2 to achieve a minimum attenuation in the stop band of 42 dB.

S/FILTER

113

The extraction sequence ∞ 103.2 38.7 ∞ ∞ DC without any transforms results in the schematic here. This sequence resulted in the lowest inductor count (4) and a low ratio of inductors (6.48888) but a transformer ratio of 0.39288.

To remove the transformer the following transforms were applied.

1. C4 was swapped with L3/C5

2. L4 was swapped with C6

3. A Norton Series was applied to C6 using the option "Calculate N to remove existing transformer"

Filters

114

4. Simplify Circuit was clicked

The final schematic is given here.

Notice that the transformer is removed and the inductor ratio was reduced to 2.56 (L2/L1). The response for this remarkable filter is given here. Notice the excellent transmission and amplitude and group-delay symmetry.

S/FILTER

115

Equal Inductor Example

Files:

1. Filters\SFILTER\Equal Inductors.WSP

2. Filters\SFILTER\Equal Inductors.SF$

Modern chip capacitors are small, inexpensive and have high Q. Inductors are more expensive, larger, more susceptible to parasitics and have lower Q. Because filter realizability is largely a function of the ratio of component values, many consider the ultimate realizability is filters with all equal inductors. This example illustrates how to design these filters. All transforms are selected to add capacitors and keep the number of inductors at a minimum.

We will start with the IF filter created in the previous example and add Norton transforms until all inductors equal the first inductor.

The following transforms are required:

1. A Norton Shunt is applied to C1 with N = SQR(L2/L1) = 1.60109 followed by Simplify

2. A Norton Series is applied to C4 with N = SQR(L3/L1) = 0.85837 followed by Simplify

3. A Norton Shunt is applied to C9 with N = SQR(L4/L1) = 0.728207 followed by Simplify

At this point the output transformer turns ratio is near unity and the transformer is removed by applying the Absorb TRF Into Load transform.

The final schematic is shown below.

Filters

116

Physical Symmetry Example

Files:

1. Filters\SFILTER\Physical Symmetry.WSP

2. Filters\SFILTER\Physical Symmetry.SF$

Symmetry is an element of beauty. It is both appealing and practical. In a previous example we illustrated how to create filters with response symmetry. In this example we illustrate how to create filters with element value symmetry (physical symmetry). Physical symmetry reduces the number of unique element values which must be modeled, designed, purchased, stocked and picked for assembly, thus saving both design and manufacturing effort. The GENESYS electromagnetic simulator =EMPOWER= automatically detects physical symmetry. When symmetric filters are realized as distributed structures they execute as much as 16X faster and require 16X less memory in =EMPOWER=. EM simulation of large filters might not be feasible without symmetry.

Symmetry results naturally, without additional user effort, in filter types that are:

1. All Butterworth

2. Odd-order Chebyshev

3. All Chebyshev coupled-resonator bandpass in =FILTER=

4. Lowpass filters with finite transmission zero pairings and an odd quantity of zeros at infinity

5. Bandpass filters with finite transmission zero pairings above or below (not both) the passband and odd plus equal quantities of zeros at DC and infinity

Symmetry may be forced:

1. By optimizing the response while forcing symmetry

The design of types 1 through 3 is straightforward using =FILTER=. These are restricted to all-pole filters. Types 4 and 5 with finite transmission zeros require the direct synthesis techniques of =S/FILTER=. Types 4 through 6 benefit from additional explanation and are the subject of this section.

Consider the Specification tab of a type 4 2300 MHz lowpass filter shown below. The placement of zeros conforms to the rules

S/FILTER

117

of type 4: the number of zeros at infinity is odd (1) and finite zeros are paired at 3800 MHz.

The schematic for the lowpass and the transmission and reflection responses are given here. Notice symmetry of the component values mirrored about the center 4.89 nH inductor.

Filters

118

Next consider case 6 where symmetry are other objectives are forced during optimization. Consider a 10.7 MHz IF bandpass filter with 400 kHz bandwidth. The following Specification tab defines the design.

The extraction sequence ∞ DC DC DC ∞ ∞ ∞ ∞ ∞ ∞ DC DC results in the schematic given here.

S/FILTER

119

Using techniques described in the All Series example, 5 Norton transforms are next applied to the shunt-coupling elements to create the series resonator filter given here. A transformer ratio equal to the square root of adjacent capacitors is used is applied to shunt inductors and a ratio equal to the square root of adjacent inductors is applied to shunt capacitors. It is tempting to select the transform option that allows choosing the new inductor right of L2 to equal the original L1, but the transform also modifies L1.

This filter is approximately symmetrical but forcing values to be physically symmetric disturbs the response. The transformer is deleted and a set of optimization goals equal to the original response is added. Next physical symmetry is forced in the Equations folder by setting element values on the right and left side of the filter equal to each other and by setting all inductors values equal. Optimization is launched to adjust the values marked with "?" to correct the disturbance introduced when the values where changed. The schematic and equate variables are given below.

Also the unloaded Q of all inductors are set at 160. The response after optimization illustrates the insertion loss introduced by finite inductor Q.

Filters

120

S/FILTER

121

Termination Coupling Example

Files:

1. Filters\SFILTER\Termination Coupling.WSP

2. Filters\SFILTER\Termination Coupling.SF$

Consider the final schematic in the earlier All Parallel Resonators example 800 to 1000 MHz bandpass filter. The shunt inductors are 1.93 nH. These are practical values if the parallel resonators are to be converted to transmission line resonators but for an L-C filter 1.93 nH is small. This example illustrates how to use the flexibility of the GENESYS environment to make any desired change to a schematic. We will use series capacitors as approximate impedance transformers at the input and output of the filter to increase the design impedance thus increasing the shunt inductors.

First the filter is designed with termination resistance higher so that the shunt inductors are 10 nH. This requires a termination resistance of 50*(10/1.93)=259.067 Ω. Next the same procedures illustrated in the All Parallel example are used to create a filter with all parallel resonators and series coupling capacitors.

Next, series capacitors are manually added to the schematic at the input and output. These series capacitors step the final termination resistance, Rs=50 Ω, up to the design impedance of the filter, Rp=259.067 Ω. The capacitor value required is given by

where fo is the geometric center frequency equal to SQR(Lower Cutoff * Upper Cutoff). In this case Cs=1.736 pF resulting in a residual capacitance, Cp, that is effectively in parallel with the adjacent resonator.

This capacitance in this case is 1.403 pF. This capacitance is subtracted from the initial parallel capacitance of the first and last resonators. The schematic after these manual modifications is given here.

Filters

122

This impedance transformation is exact only at fo but for narrow bandwidth filters this process works well. The bandwidth of this filter is 200/SQR(Fl*Fu)=0.224=22.4% which begins to stress the accuracy of the transform. The response of the filter is given as the dashed responses in the plot given here.

Optimization is then applied to correct the response as given by the solid traces.

S/FILTER

123

Parametric Ba ndpass Example

Files:

1. Filters\SFILTER\Parametric Bandpass.WSP

2. Filters\SFILTER\Parametric Bandpass.SF$

In the All Parallel Resonators example we illustrated how to create all-pole filters with all parallel resonators. This is useful when L-C filters will be converted to structures using transmission line or ceramic resonators. In this example a similar process is illustrated for filters with transmission zeros at finite frequencies.

A 850 to 950 MHz bandpass with singular zeros at DC, infinity, 751 MHz and 1068 MHz is specified. A passband ripple of 0.177 dB is chosen. Again we start with extractions with a shunt element first. The extraction sequence 751 MHz, DC, 1068 MHz and ∞ was chosen. Next a Norton Series transform was applied to C1 and L2 with a turns ratio of 0.36 followed by a Simply circuit transform. Next a Norton Series was applied to C4 and L4 with the ratio automatically selected to remove the transformer. After simplification the schematic is:

The first transform ratio of 0.36 was found empirically to give somewhat equal inductors. The inductor values are somewhat small for an L-C filter, particularly if high Q is desired in the inductors. If this filter is to be constructed with L-C elements the approximate impedance transformation illustrated in the Termination Coupling example could be employed. In this case we plan to realize the parallel resonators using ceramic resonators. Therefore we desire equal shunt inductors in these resonators. This filter is found using optimization. The shunt inductors are set at 1.0 nH and the remaining element values are optimized to achieve the original response. The final element values are given in the following schematic.

Filters

124

The ratio of maximum to minimum inductor value is 2.49. The ratio for a conventional cookbook realization of this filter is over 7. The parametric bandpass has all parallel resonators as well as an improved inductor ratio. These advantages come at the expense of an additional inductor and capacitor. The response after optimization is given here.

Next, the conversion of the L-C resonators in this parametric bandpass to ceramic resonators is illustrated. Using the parallel L-C to quarter-wave transmission line equivalent formula given earlier, the three quarter-wave line resonators from left to right have the following parameters

Zo=4.624 Ω Fo=937.033 MHz

Zo=4.432 Ω Fo=898.149 MHz

Zo=4.266 Ω Fo=864.424 MHz

The low line impedance is consistent with the high dielectric constant of ceramic resonators. For other characteristic

S/FILTER

125

impedance the original L-C filter is designed with the appropriate shunt inductance. After replacing the L-C resonators with quarter-wave line resonators the responses are given below as dashed traces.

Recall the L-C/transmission line equivalences are accurate at the resonant frequency only. Notice the passband return loss and transmission are close to the original L-C filter. However, further from the passband the rejection is less than the original filter in lower stopband and greater than the original filter in the upper stopband.

The solid traces are optimization of element values in an attempt to achieve the original stopband rejection. The stopband frequencies were shifted lower to accommodate the effect of the line resonators. Also, for practical reasons, before optimization the characteristic impedance of all three resonators were set equal at 4.6 Ω. The final schematic after optimization is given here.

Filters

126

S/FILTER

127

LP All Pole

Files:

1. Filters\SFILTER\LP All Pole.wsp

2. Filters\SFILTER\LP All Pole.SF$

This is a very straightforward design in S/FILTER using 3 zeroes at infinity and two unit elements. The element values are quite practical, and the response is shown below.

Filters

128

LP All Pole

Files:

1. Filters\SFILTER\LP Elliptic.wsp

2. Filters\SFILTER\LP Elliptic.SF$

This example illustrates using S/FILTER to create an exact distributed elliptic lowpass filter.

The transmission zero specifications are:

9. # Infinity = 1

10. # Finite = 1,2,3...N

11. # UE= 2 x # Finite

The extraction sequence is then UE,FZ,UE,UE,FZ,UE,UE,FZ.......UE,UE,FZ,Infinity,UE. This example uses # Finite = 2 and # UE = 4 and the sequence UE,FZ,UE,UE,FZ,Inf,UE.

Next "Kuroda Wire Line Transfers:Full:Series Shorted (Right or Left)" are applied to each series transmission line and wire line to create transmission lines and stubs. Finally, "TLines:Stepped Resonators:Convert to Two Step Resonator" are applied to the shorted and open wire lines to ground.

Transmission line impedances are typically practical for this structure.

S/FILTER

129

To make this structure more physically realizable, the 24 ohm transmission line was split into two 48 ohm lines. S/FILTER was then closed, and one of the 48 ohm lines was moved up above the other 48 ohm line. Advanced T/LINE was then used to convert the structure to microstrip, yielding this schematic:

Filters

130

Coaxial Resonator Example

Files:

1. Filters\SFILTER\top-c resonator.wsp

2. Filters\SFILTER\top-c resonator.SF$

This example extends the concepts presented in the "All Parallel Resonators Example". Using different specifications, it extends the design process to the use of square coaxial resonators. The example shows the power of the transforms from lumped to distributed elements and the use of "Advanced TLINE" to change electrical line models into physical models.

A fourth order bandpass filter with a bandwidth of 40 MHz centered about 1.0 Ghz was designed with the S/FILTER synthesis program. The initial design in terms of lumped elements (L,C) resulted is shown below (see the SF$ file).

The next step was to split the shunt capacitor into a 15 pf element to combine with the 1.24 nH inductor. Use the "Lumped to distributed equivalent" transform with the "parallel LC to ground to grounded stub". This generates an approximately equivalent transmission line. This electrical transmission line will be converted to a physical model using Advanced TLINE.

For this example, ceramic coaxial resonators will be used. A common configuration is a square shape with circular inner conductor. Specific parameters used were: dielectric constant = 90.5, inner diameter = 1 - 2 mm, and outer side = 4.0 mm. The length will be set by the resonant frequency of the LC network. To convert to a physical line type, select "Convert Using Advanced TLINE" from "Schematic" on the toolbar. Then under "Process" select "Square Coax (Round Conductor)" and "outside" equal to 4 mm. The inner conductor diameter and length are computed automatically. The inner diameter can be set to one of the commercially available sizes. The overall response will be relatively insensitive to this value.

S/FILTER

131

Since the transforms from lumped to distributed elements and from electrical to physical lines involves some approximations, a final optimization is desirable. The capacitors were gang-tuned to preserve symmetry. The optimization goals were:

S21 < -48 db for f < 920 MHz, S21 < -45 db for f > 1080 MHz, S11 < -18 db for ( 980 MHz < f < 1020 MHz )

The resulting responses and component values after optimization are shown in the figures below (see the wsp file). Clearly the requirements were met with reasonable component values. As a final step, the capacitor values were "tuned" to standard values. From the tuning window, "standard" was selected and the "value" set at 1%. The values of each of the ganged sets were adjusted manually while observing the frequency response with respect to the optimization goals.

Filters

132

BP Edge Coupled

Files:

1. Filters\SFILTER\BP Edge Coupled.wsp

2. Filters\SFILTER\BP Edge Coupled.SF$

This example illustrates using S/FILTER to create an exact distributed edge-coupled bandpass filter.

NOTE: Although this is a bandpass, the Highpass type option is selected in the Specification tab because this design procedure utilizes the reentrant mode of transmission lines.


• # Infinity = 0

• # Finite = 0

• # DC = 1

• # UE= 2,4,6...2xN

The extraction sequence is then UE,UE...DC...UE,UE with a series element first.

This example uses # DC = 1 and # UE = 4 and the sequence UE,UE,DC,UE,UE.

Next the following transforms are applied:

1. Basic Operations:Split Series Element is applied to the open wire line at the center

2. Kuroda Wire Line Transfer:Equal:Series Open Right is applied to the UE left of center and each UE to the left.

S/FILTER

133

3. Kuroda Wire Line Transfer:Equal:Series Open Left is applied to the UE right of center and each UE to the right.

4. Simplify Schematic

5. Basic Operations:Split Series Element is applied to each internal open wireline. Do not apply this transform to the end two open wire lines.

6. Coupled Lines:Interdigital Lines:Open,Open is applied to all UE.


Transmission line impedances are practical only for Maximally Flat responses or low ripple Chebyshev filters of bandwidth from 20 to 40 %. For a technique to improve realizability of this filter for other parameters please refer to BP Edge Redundant.SF$.

Filters

134

BP Edge Redundant

Files:

1. Filters\SFILTER\BP Edge Redundant.wsp

2. Filters\SFILTER\BP Edge Redundant.SF$

This example illustrates using S/FILTER to create an exact distributed edge-coupled bandpass filter for narrower bandwidth. Realizability is improved by adding redundant transmission lines at the input and output. These lines have a characteristic impedance of 50 ohms so they have no effect on the response. Because transforms are later applied to these redundant elements they must be equal in length to the other lines in the filter. The design process is similar to the BP Edge Coupled topology without redundant lines.

Note: Although this is a bandpass, the Highpass type option is selected in the Specification tab because this design procedure utilizes the reentrant mode of transmission lines.


• # Infinity = 0, # Finite = 0, # DC = 1

• # UE= 2,4,6...2xN

The extraction sequence is then UE,UE...DC...UE,UE with a series element first.

This example has a Cutoff of 950 MHz, an equiripple of 0.07 dB and a 1/4 Wave Freq of the transmission lines of 1000 MHz. It uses uses # DC = 1 and # UE = 4 and the sequence UE,UE,DC,UE,UE.


1. Basic Operations:Insert Element with lines = 50 ohms and length 75mm are applied to the input and output.

S/FILTER

135

2. Basic Operations:Split Series Element is applied to the open wire line at the center

3. Kuroda Wire Line Transfer:Equal:Series Open Right is applied to the UE left of center and each UE to the left including the redundant input line.

4. Kuroda Wire Line Transfer:Equal:Series Open Left is applied to the UE right of center and each UE to the right including the redundant output line.


6. Basic Operations:Split Series Element is applied to each internal open wireline. Do not apply this transform to the end two open wire lines.

7. Coupled Lines:Interdigital Lines:Open,Open is applied to all UE.


Transmission line impedances are practical for bandwidths narrower than the edge coupled filter without redundant lines. Bandwidths from 20% to a few percent are practical. The practical limit is due to insertion loss in the final filter due to PWB losses. Below 10% bandwidth evanescent modes may propagate, circuit theory models become inaccurate and electromagnetic simulation using EMPOWER is suggested.

Filters

136

BP Stub with Inverters

Files:

1. Filters\SFILTER\BP Stub with Inverters.wsp

2. Filters\SFILTER\BP Stub with Inverters.SF$

This example illustrates using S/FILTER to create distributed bandpass filter with resonating shorted stubs connected by approximate inverters.

NOTE: Although this is a bandpass, the Highpass type option is selected in the Specification tab because this design procedure utilizes the reentrant mode of transmission lines.


6. # Infinity = 0

7. # Finite = 0

8. # DC = N

9. # UE= 0

This example has a Cutoff of 750 MHz (the mirror upper cutoff is then 1250 MHz), an equiripple of 0.07 dB and a 1/4 Wave Freq of the transmission lines of 1000 MHz. It uses uses # DC = 5.


1. Inverters:Replace Element with Inverter is applied to each open wire line

2. Replace Inverter with: Stub to Ground + Series TLine + Stub to Ground is applied to all ideal inverters


Realizability is acceptible for 30 to 70% bandwidth with the stubs becoming too low in impedance for narrower bandwidth.

Chapter 6: Oscillators

Oscillators\Bipolar Cavity Oscillator.WSP

Illustrates: How to complete nonlinear oscillator analysis in GENESYS using HARBEC. The oscillator simulated is a cavity oscillator taken from "Oscillator Design and Computer Simulation" by Randy Rhea, Noble Publishing, pages 225-231. A different device is used which results in slightly different component values.

The first step in the analysis is to examine the open loop Bode response. The schematic "NL Open Loop" is constructed. Port 1 is the input port and port 2 the output port. In final configuration, port 1 will connect with port 2. The device was built with a nonlinear device model, allowing the DC bias points to be verified and displayed on the schematic.

The frequency is swept over a narrow range as this is a particularly high Q oscillator. L1 was adjusted so that the gain peaked at the same frequency that the phase crosses zero degrees. This frequency, approximately 1100 MHz, should be

Oscillators

138

near the frequency of oscillation.

Next, the loop was closed, as shown in the schematic "NL Closed Loop." In addition, a source was attached to the loop, passing through a high-Q LC tank. The purpose of this source is to stimulate the circuit to incite oscillation; the purpose of the tank is to allow the signal to be fed at the fundamental but to otherwise isolate the source from loading the circuit.

Oscillators\Bipolar Cavity Oscillator.WSP

139

Then the frequency and amplitude of the source are adjusted until no current is being supplied through the LC tank. At that frequency and amplitude. In that state, the source will not be affecting the circuit, indicating that the state of oscillation has been found.

To get the best starting point, a guess is made for the amplitude and the frequency of the source. The frequency is taken from the open loop analysis; the amplitude is set to 1 Volt as the voltage tap is base of the transistor. Then the a harmonic balance sweep is completed for both frequency and amplitude. The current supplied by the source is plotted for both sweeps ("Freq Sweep" and "Level Sweep"), showing a minimum current at about 1102 MHz and 1.02 Volts.

Oscillators

140

With this starting point, the optimizer can be used to find the exact amplitude and frequency. Right click on "Opt1" and select "Optimize" to start the process. Open "Error Current" to watch the error current reduce as the frequency and voltage are adjusted.

This same process can be used with any oscillator. It provides a very good estimate of the output amplitude frequency that can be combined with the phase noise calculated by using Lesson's Rule with loaded Q calculated from open loop analysis. (See CoaxialOSC.WSP for an example.)

Oscillators\NegR VCO.WSP

141

Oscillators\NegR VCO.WSP

Illustrates: Post processing, negative resistance

This is a UHF negative resistance oscillator designed by the synthesis program =OSCILLATOR=. It tunes from 750 to 950 MHz. The graph on the left is the resistance and reactance looking through the transmission line/varactor resonator. If the resistance is negative, when the left side of the resonator is grounded, the circuit will oscillate at the frequency where the net reactance is zero.

Oscillators

142

The graph on the upper right plots 1/S11 on a Smith chart. Post processing was used as shown on the lower-right above to calculate 1/S11. The first line, "USING Linear1.NegR" sets the default Simulation/Data for the equation. The next line gets the reciprocal of S11 in complex (rectangular) form. Two important things to notice about this equation:

1. The period (.) before RECT is absolutely necessary. All post-processed measurements must have a period, and even though there is a USING statement, there must still be a period in front of the operator. If there was no USING statement, the line "InverseS11 = 1/Linear1.NegR.Rect[S11]" would have been used.

2. The operator RECT is required.

The solid traces are for a tuning varactor capacitance of 1.65 pF (oscillation at 950 MHz) and the dashed traces are with a capacitance of 4.5 pF (oscillation at 750 MHz).

Please refer to Oscillator Design and Computer Simulation for additional information on oscillator design.

Oscillators\Coaxial OSC.WSP

143


Illustrates: Post-Processing, Loaded-Q, Phase Noise.

The circuit above is a coaxial resonator oscillator, designed to be tunable between 800 and 900 MHz. The desired oscillation frequency for this example is 850 MHz. In this example, we will use post-processing to calculate the phase noise of the oscillator. The equations we will use for this are:

Cv=?7.343 Fo=850E6 PdBm=?7 kT=4E-21 NFdB=?6 F=10^(NFdB/10) Flicker=?10000 Ps=1E-3*10^(PdBm/10) 'The function of the next line is not obvious: It gets the frequencies of the baseband simulation. ssbfreq=1e6*freq+0*baseband.loop.db[s11] 'Get the loaded q of the loop at the oscillation frequency. Q=GETVALUEAT(linear1.loop.QL[S21],Fo) 'Calculate Leeson's equation to determine the noise.

Oscillators

144

ssb=10*log(0.5*(Flicker/ssbfreq+1)*((Fo/ssbfreq/2/Q)^2+1)*(kT*F/Ps))

There are a few lines which are of special interest:

1. ssbfreq=1e6*freq+0*baseband.loop.db[s11] : This line is a trick used to create a new sweep for post-processing. The linear simulation "Baseband" was created solely to provide a sweep for post-processing. This line gets S11 from the Baseband sweep, then promptly multiplies it by zero. This clears the values, but it will retain the independent frequency data. Next, by adding it to FREQ, this creates a "unity" sweep, where each point in the sweep contains the frequency in Hz (orignally in MHz, then multiplied by 1e6).

2. Q=GETVALUEAT(linear1.loop.QL[S21],Fo) : The GetValueAt function takes a measurement as the first parameter and the frequency as the second parameter. This statement gets the value of linear1.loop.QL[S21] at frequency Fo (850 MHz).

3. ssb=10*log(0.5*( Flicker/ssbfreq+1)*((Fo/ssbfreq/2/Q)^2+1)*(kT*F/Ps)) : Calculates Leeson's equation. Notice that since ssbfreq contains swept data, the result of this calculation is swept data. The results of this equation are shown in the graph below:

This example shows the power of post-processing in GENESYS. Additional things that could be done include:

1. Optimization of SSB and circuit simulation simultaneously.

2. A function could be written to determine the actual oscillation frequency when the capacitor is tuned. Fo


145

would be set to this value, and then the phase noise calculations would be at this frequency.

3. The entire phase-noise calculation could be put into a function, stored into the model directory, then used for any circuit.


Chapter 7: Matching

Matching\Synthesis Comparison.wsp

This example includes a few circuits to illustrate the use of MATCH and some basic matching concepts. The example deals with matching a 50 ohm source to a 200 ohm load from 100 to 200 MHz using a single network. Bandwidth and network complexity are considered.

In the first attempt, a Tee network is tried. To do this, we right-clicked on Synthesis in the workspace tree and added Match. You can double-click the Tee icon to see this setup. The circuit file and response are given below. The exact component values will depend on how long MATCH has optimized. Because minimum Q was specified in MATCH and the terminations are resistive, the tee network contains two components. (A third element was created but was an insignificantly small inductor which was deleted.) The response is typical of a narrowband matching network. In this example, the 15 dB return loss bandwidth is approximately 127 to 162 MHz, or 24%.

Matching

148

Our second attempt is a second order L-C pseudo lowpass network. Because the pseudo lowpass is synthesized as a bandpass transfer function, the second order network has four components, two inductors and two capacitors. You can double-click the LCPseudoLowpass icon to see this setup.

The results are as given. Notice the return loss is now 15 dB or better across nearly the entire 100 to 200 MHz range. The increased performance results from a four component network, as opposed to the previous two components, and from an efficient synthesis algorithm.

Our third attempt is a stepped impedance transmission line is used for the matching network. The 3rd order stepped Zo network has three cascaded transmission lines. The results are given below. Notice the change in the return loss scale to 5 dB per division.

Although there are three elements as opposed to four in the L-C pseudo lowpass, the match is better than 30 dB across the entire band. Each transmission line segment in the stepped Zo

Matching\Synthesis Comparison.wsp

149

matching network is approximately as effective as two lumped elements. However, keep in mind transmission lines are larger than lumped elements.

Matching

150

Matching\Ill Behaved Load.WSP

In this example, we attempt to match a 50 ohm source to an antenna represented by an R-X data file, ANTENNA.RX. The antenna load is highly frequency dependent. The resistance is approximately 33 ohms and increases slowly with frequency. The antenna resonates between 14 and 14.05 MHz, and the reactance increases rapidly with frequency. This data is typical for narrow band antennas. The frequencies, resistances and reactances in this data file are:

Frequency (MHz) Resistance (ohms) Reactance (ohms)

14.00 31.6 -6.6

14.05 32.0 4.7

14.10 32.4 16.0

14.15 32.7 27.2

14.20 33.1 38.4

14.25 33.5 49.5

14.30 33.9 60.7

14.35 34.3 71.7

Matching resistive terminations is fundamentally easier than matching between terminations where either or both have a reactive component. Theoretically it is possible to match purely resistive terminations over an arbitrarily wide bandwidth, although it may take a very large number of components and be impractical. When either or both terminations are reactive, there are fundamental relationships between the nature of the terminations, the required bandwidth and the required quality of match.

The matching networks were designed using MATCH. You can view the setups by double-clicking on the Tee and LC Bandpass icons under the synthesis folder.

Matching\Ill Behaved Load.WSP

151

First we attempted to match the antenna with a Tee network, but the worse case S11 at the band edges was -6dB. We then used a 4th order LC Bandpass network, resulting in the network shown above and the response shown below. The ill-behaved nature of this load is well handled by MATCH even though the element values are certainly difficult.

Matching resistive terminations is fundamentally easier than matching between terminations where either or both have a reactive component. Theoretically it is possible to match purely resistive terminations over an arbitrarily wide bandwidth, although it may take a very large number of components and be impractical.

When either or both terminations are reactive, there are fundamental relationships between the nature of the terminations, the required bandwidth and the required quality of match.

As an exercise try a Monte Carlo analysis of this circuit with the default 5% tolerance on components. Clearly this circuit would need tuning after construction. Tuning all components would be undesirable. As a second exercise, try placing the capacitors on the nearest standard value and optimizing the three inductors for a return loss of -18 dB.

Matching

152

Matching\Power Amp.wsp

Here we illustrate power amplifier design. Because of the popularity and power of S-parameter based computer programs, much effort has been expended in recent years on measuring high-power S-parameters and using this data for power amplifier design. Unfortunately, if a high-power DUT accidentally oscillates, S-parameter measurement equipment is easily damaged.

A second concern of basing power amplifier design on high-power S-parameters is the following; does matching to the device high-power S-parameters yield the best power output, efficiency and gain?

Historically, power amplifier design has used the following procedures:

1. Tuners are placed at the device input and output

2. The tuners are adjusted for best power operation

3. The device is removed and the tuners are measured

4. The manufacturer publishes the device input and output impedances which should be matched

In this example, we’ll design a 800 to 950 MHz 0.75 watt output Motorola MRF559 amplifier with Vcc = 12.5. We’ll design the output matching network. The input network would be designed using the same procedure. The Motorola data sheet for the MRF 559 recommends matching to these device resistances and reactances versus frequency for the output:

800 23.4 -37.7 850 23.7 -36.8

Matching\Power Amp.wsp

153

900 23.9 -36.0 950 24.5 -35.6

We drew the "Custom" schematic, consisting of a series line, shunt capacitor, series line and a shorted stub. Our network has fewer parts than the data sheet network, but the resulting match is excellent. Power is delivered to the MRF559 via a shorted stub as in the data sheet. (We specified a 50 ohm source and the load is the file MRF559OU.RX, so the network is oriented backwards.)

MATCH was then run, which in this example is a shortcut for setting up the simulation, graphs, and optimization blocks. You can see the match setup "PowerAmp" by double-clicking it on the tree. Notice how we specified Custom for the network type and chose the schematic circuit Custom. MATCH then put the terminations on the PowerAmp schematic.

Next, the schematic was copied to a new schematic called "Physical". The terminations were adjusted to be the same as the PowerAmp schematic. "Convert using Advanced TLINE" was then selected from the Schematic menu, and the schematic was converted to microstrip. After tuning, the results are as shown below:

Matching

154

Matching\Unstable Device.wsp

Matching an unstable device, particularly if it is potentially unstable in the passband is not good practice. The appropriate technique is to stabilize the device before proceeding with the match. The device is stabilized by adding one or more networks around the device.

In this topic, we will create a multistage 8 to 12 GHz amplifier, matching to an unstable device. We will use two NE71083.330 GaAsFET transistors with a series R-L stabilizing network connected from the drain to the gate.

Finally we write S-parameter for the device and its stabilization network to the file NE71083.STA using the “Write S-data” feature from the SuperStar File menu. This is the device specified in the MATCH setup.

The basic steps that were followed are:

1. Discover that the device you want to match is unstable. You can easily do this in several ways: Read the data sheet, run MATCH and see the warnings. run our Stability template, or run our "Stabilize Device" template.

2. Select New from template, choosing the "Stabilize Device" template.

3. Change the device as required.

4. Optimize to make the device stable

5. Choose one of the stabilizing configurations.

6. If desired, delete the stability test networks and add the stabilizing network to your device schematic.

7. Run MATCH and create the matched amplifier. This example has a 2nd order TRL pseudo lowpass for each of the three matching networks

8. Use Advanced TLINE to create a physical representation.

Select Start Level 1 from the Run menu. At this point, MATCH checks the NE71083.330 transistor S-parameter data files and reports that they are potentially unstabile. At this point, Level 1 run would normally be aborted and the device stabilized. Example 7 is a device stabilization example.

Matching\Unstable Device.wsp

155

We have already stabilized the NE71083.330 by placing a series R-L feedback network from the drain back to the gate consisting of a 330 ohm resistor and a 5 nH inductor. S-parameter data for this stabilized device was stored in the file NE71083.STA.

To continue this example, select the Network menu and change each NE71083.330 S-parameter data file to NE71083.STA. Next, select Start Level 1 from the Run menu, make a few Level 2 passes and finish with Level 3 passes. The return losses will be approximately 7 dB.

Next, run the amplifier in SuperStar. The SuperStar circuit file and our results are given on the following page as the solid curves. The gain ranges from approximately 24 dB at the low end to 20 dB at the high end. Next add the optimization block in the file below to flatten the gain to 20 dB. The result after many rounds is shown as the dotted curves.

Flattening the gain response of an amplifier is a process of finding a good match at frequencies of low gain and a poorer match where the gain is higher than the norm. The processes in MATCH and SuperStar are quite different, but together they form a powerful amplifier design technique. MATCH attempts to find the best match across the band and SuperStar “frequency shapes” this match for a flat response. MATCH finds an excellent starting point for SuperStar optimization.

Because active device gain is typically higher at low frequencies, a flattened amplifier often has a poor match at low frequencies. When flat gain and good match are required, an isolator is added or a balanced amplifier configuration is used. Another technique of achieving both match and flat gain is using resistive or transformer feedback. This sacrifices gain and is less common at microwave frequencies where gain is difficult to come by.

In the antenna example, the objective was to obtain a match to a highly reactive, frequency dependent termination. These conditions typically require high order networks. With amplifiers, the match is sacrificed to flatten the gain. High order networks with excellent initial match are overkill. Network selection is based on topologies with amplifier stability and gain flattening capability

Matching

156

Matching\Match and Stability.WSP

This example shows that a high frequency GaAsFET (3-18 GHZ) can be stabilized with a feedback network and then matched using the broadband matching program called MATCH.

First, the stability and matching can be observed for the FET by itself. Notice the stability factor K is less than 1 for all frequencies below about 10 GHz. Consequently, the device is potentially unstable and should be stablized before proceeding to match the device.

The FET was stabilized using a "Stabilization Networks.wsp" template. See the "Stability Selection.wsp" workspace to see how the stabilization network was selected for this example.

Amplifier after stabilization from the "Stability Selection.wsp":

The stability factors K and B1, and the matching of the stabilized device can be seen in the respective schematic, table, and graph.

Finally, a MATCH is added to the Synthesis folder and a matching circuit can be found for the stabilized device. A second order "TRL Pseudo Lowpass" matching network is used on the input side of the FET and a third order "TRL Pseudo Lowpass" matching network is used on the output side of the FET. The Min TRL Zo is set to 35 ohms and the Max TRL Zo has been set to 60 ohms. The optimize button on the MATCH dialog box can be used to optimize the input and output match. The gain of the circuit can also be added to the optimization so that the resultant gain is greater than the target gain of 6 dB.

After the circuit has been optimized this electrical circuit can be converted to a physical implementation such as microstrip or stripline by using Advanced TLINE and adding a layout. Furthermore, once the physical implementation has been

Matching\Match and Stability.WSP

157

created optimization targets can be placed on the physical implementation to optimize the layout for something that can be easily manufactured.

Optimized performance:

Final electrical implementation:

Matching

158

Matching\Stability Selection.WSP

This workspace is used to assist in choosing a network and the component values to stabilize the device used in the "Match and Stability.wsp" example.

The "Stabilization Networks.wsp" template (selected "New from Template..." from the "File" menu) is used stabilize an active device. Basically, the workspace contains 4 feedback topologies, Resistive, Shunt RL, Series L, and Shunt C.

Resistive:

Shunt RL:

Series L:

Matching\Stability Selection.WSP

159

Shunt C:

The active device was set to the NE71083.330 GaAs FET in the Unstabilized Device schematic.

The Start and Stop frequencies of the Linear simulation were changed to 3000 and 18000 respectively, to accommodate the frequency range of this device.

Graphs and optimization targets exist for each of the feedback topologies. Once optimization is launched the simulator will changed the lumped element values until the best stability and maximum gain is achieved.

The resulting optimization for the Shunt RL Feedback is shown to be:

Matching

160

Notice that both the Series L and Shunt C Feedback topologies produced stability K values less than 1 for some frequencies. Definitely, we should try to avoid these feedback topologies for this active device. The Resistive Feedback topology also had good performance and could also be used.

The results of this stabilization selection was used in the "Match and Stability.wsp" example.

Chapter 8: Detectors

Detectors\Simple Detector.WSP

Illustrates: A simple diode detector circuit.

A blocking capacitor separates the applied DC bias from the input signal; a inductive choke separates the DC supply from the RF supply. The diode is used to peak detect, the capacitor holds the charge, and the resistor sets the bandwidth of the detector.

Shown in the graphs are the spectra and waveforms at the input and at the detector.

Detectors

162

"Vout vs. Pin" shows the DC output voltage as a function of the input power level. Note that the detected voltage shows the typical square-law shape.

The final graph shows the detected voltage as a function of the detection capacitance. Note that the voltage is largely independent of the capacitance as long as the RC time constant is sufficiently below the input 1GHz signal.

Detectors\Diode Detector with Co-simulation.WSP

163

Detectors\Diode Detector with Co-simulation.WSP

Illustrates: EM-harmonic balance co-simulation to analyze a simple diode detector including the effects of the metal traces that interconnect the ports and the lumped components.

The circuit is shown in the schematic "Detector."

The layout is in "Layout1." The circuit is driven by two signals that form the complex waveform shown in "Time Input." In addition to detecting the DC level of the signal, the nonlinear character of the diode results in a mixing--a down-conversion--of the two signals.

The "Time Output" shows the detected output, showing a DC level proportional to the level of the input waveform and a ripple that is the product of the two input signals. Higher order feed-through terms are also visible riding on the waveform.

Detectors

164

The spectra of the input and output waveforms are shown in "Sch Spectrum" (for the circuit based results) and "EM Spectrum" (for the EM-circuit co-simulation results.

Note that the time waveforms are nearly identical in the detected signal. Since the inputs to the diode are relatively low in frequency, the parasitic effects are small.

Chapter 9: Mixer

Mixer\Low Power Mixer.WSP

Example: Low Power Mixer

Illustrates: The design of a single transistor mixer designed, built, and measured by Ken Payne of Artetronics. It is a typical design used in low cost, low power applications such as pagers and wireless remotes.

The mixer is barely biased on, which operates the transistor in a very nonlinear region. Then, a relatively low power LO is lightly coupled onto the base, together with the RF signal. The transistor then effectively mixes these two signals together.

The Smith charts in this show the port matches of the mixer as calculated using linear and nonlinear device models. Measured results are also shown. Note that the RF and IF ports are well matched at their operating frequencies and mismatched at the other mixing frequencies. The LO port is intentionally mismatched so that the LO will only lightly couple into the mixer.

Mixer

166

Graphs are included that show the spectra and waveforms in the mixer. Two spectral graphs are shown where the Maximum Mixer Order was set at 5 and 10, resulting in 31 and 61 frequencies, respectively. This test was run to make sure that a sufficient number of harmonics were specified to model the circuit. Since the data moved so little, Maximum Mixer Order of 5 is sufficient.

Two power sweeps were completed. The first shows the effect of LO drive level on conversion gain. The second shows gain compression resulting from RF drive level.

Chapter 10: Antennas

Antennas\Patch Antenna Impedance.WSP

RAM: 16.0Mbytes Time: 236s/freq Illustrates: Patch antenna impedance and resonance, solid model thinning, benchmark

Shown below is a microstrip fed patch antenna originally investigated by Wolff and Gao [1998]. The substrate is 0.79 mm thick with a relative dielectric constant of 2.2. The original dimensions and dimensions after placement on a 0.18 mm grid are:

Parameter Original (mm) On-Grid (mm)

Substrate Height 20 20.16

Substrate Width 20 20.16

Patch Height 7.2 7.2

Patch Width 3.59 3.6

Microstrip Feedline Width 0.715 0.62

Feed below top 2.9 2.88 (top of microstrip line)

Antennas

168

=EMPOWER= was run in a 1.0mm tall box with an open cover to simulate free space. Given below are the real and imaginary components of the input impedance of the patch antenna with a reference plane at the junction of the patch and the microstrip feedline. The reported operating frequency of the antenna is 13.12GHz. The frequency of maximum resistance found by =EMPOWER= is 13GHz.

This problem involves metal with a large number of cells in both x and y. The default wire thinning mode of =EMPOWER= is stressed by this class of problem and =EMPOWER= reports a maximum radiation resistance at 12.35GHz. Therefore solid thinning was specified in the =EMPOWER= properties dialog box. Solid thinning runs faster and requires less memory than no thinning and accuracy is compromised very little if at all. Solid thinning requires significantly more memory and execution time than wire thinning. Wire thinning generally works well with filters and other structures which are a composite of narrow and medium width lines. When wide lines or large patches of metal are involved solid thinning is suggested.

Antennas\Array Driver.WSP

169

Antennas\Array Driver.WSP

Shown below is the schematic for a transmission line match and phasing network to drive a 3-element phased array of loop antennas for receiving. Each loop has a terminal impedance of 560 ohms in series with 4700 nH. Although =SuperStar= handles complex terminations, in this case we simply placed 4700 nH inductors at the output of the driver.

In order to produce low sidelobes, it is desired to drive the array with a binomial amplitude distribution; the center element should be driven with twice the amplitude of the end elements. For maximum broadside gain all elements should be driven in phase. Results after optimization are given below. Note that the phases are within 20 degrees and the gain and reflections are good.

Antennas

170

Antennas\Agile Antenna.WSP

171

Antennas\Agile Antenna.WSP

A microstrip patch antenna loaded with two varactor diodes.

Illustrates: combined EM and circuit theory simulation of a frequency agile antenna, manual connection of lumped element in =EMPOWER=, comparison with measured data.

The design contains one common layout with one external and two internal z-directed ports. Schematic "WithParasitics" (shown below) takes data from the EM1 analysis and hooks up two models of the varactor diodes.

Schematic "WithoutParasitics" (not shown) simulates the same structure but the parasitic circuit elements which took the diode case into account have been removed. The result is closer to the

Antennas

172

experimental (2.35 GHz) [1], which can be explained by the fact that z-directed ports themselves have physical dimensions similar to the diode case, so the corresponding inductance and capacitance which must be taken into account in an equivalent circuit of a lumped element are already inherently included in the =EMPOWER= simulation.

Reference: R. Gillard, S. Dauquet, J. Citerne, Correction procedures for the numerical parasitic elements in global electromagnetic simulators. - IEEE Trans., v. MTT-46, 1998, N 9, p. 1298-1306.

Antennas\MAmmannPatch.WSP

173

Antennas\MAmmannPatch.WSP

Illustrates: Simulation of different kind of patch antenna feed and comparison with published measured data [1].

This example is a patch antenna simulation.

Problems: EM1(Layout1) - patch antenna 36.85 x 36.85 mm x mm on RT Duroid 5870 substrate fed by a coaxial prob. Antenna is made of 1 oz. copper. Substrate height is 3.18 mm, dielectric constant is 2.33.

EM2(Layout2) - patch antenna 33.7 x 33.7 mm x mm on two layer substrate composed from 3.18 mm RT Duroid 5870 and a substrate with dielectric constant 4.5 and height 1.54 mm. The antenna is fed by a 2.9 mm microstrip line deposited on the interface of the two dielectrics. The metal parameters are the same as in the prob-fed antenna.

Simulation details: The key parameters that must be chosen properly for an antenna simulation are box size and position and boundary conditions at the top cover of the box. A rule of a thumb is to chose the box size as 4-5 of a critical patch size and to remove the top cover on at least one size of the critical size of the patch and put 377 Ohm boundary conditions their. Usually it works like here. The solid thinning out is recommended for the patch simulations. The grid cell size was chosen to make the patch smaller on a quarter

Antennas

174

of the cell size, that increase accuracy of the simulation according to U. Schulz. The problem is the prob-fed patch is simulated as z-directed internal port with size 2.8x2.8 mm. It made thicker to reduce the internal inductance that roughly corresponds to the inductance of a viahole with the same size or to an inductance of the central conductor of the feed coaxial.

Simulation results: The prob-fed patch has central frequency 2480 MHz that is about 1% higher than in the experiment [1], the bandwidth of lossless patch is 2.8% (3.1% in the experiment). The microstrip-fed patch has central frequency 2400 MHz that is 2% lower than in the experiment [1].

Reference: 1) M. Ammann, Design of rectangular microstrip patch antennas for the 2.4 GHz band, Applied MicrowaveWireless, Nov./Dec., 1997, p. 24-34.

Antennas\Simple Dipole.WSP

175

Antennas\Simple Dipole.WSP

=EMPOWER= Example: Simple Dipole Antenna in Open Space. This example features simulation of open space boundary conditions, internal port, and description of the electromagnetic analysis results as multiport for a co-simulation. The layout is a thin dipole 58 mm long and 1 mm wide in open space.

Simulation details: Cell size 1x1mm, Box size 150x151 mm in the dipole plane and two surfaces with resistance 377 are placed 50 mm off the dipole plane to absorb EM fields. An internal port with current direction along X is placed in the center of the dipole to simulate excitation. The simulation results are shown below:

Antennas

176

Antennas\Thin Loop.WSP

=EMPOWER= Example: Thin Loop Antenna in Open Space

Features simulation of open space boundary conditions, internal port. The antenna is a thin square loop is 32x32 mm made of 1 mm wide strip line in open space.

Simulation details: Cell size 1x1mm, Box size 150x150 mm in the dipole plane and Two surfaces with resistance 377 are placed 50 mm off the dipole plane to absorb EM fields. An internal port with current direction along X is placed in the center of the dipole to simulate excitation. The input impedance and reflection are shown in the graphs below.

Antennas\Thin Loop.WSP

177


Chapter 11: Resonance Elimination

ResonanceElimination\ LowPassWithFictivePorts.WSP

EMPOWER Example: Microstrip Low Pass Filter - resonance elimination technique

Illustrates: Comparison of the EM simulation with measured data and data obtained by the method of moments [1]. Demonstrates also effectiveness of the resonance elimination technique.

Description of the problem: The filter consists of six microstrip stubs connected symmetrically on both sides along a microstrip line. Box size: 335 x 365 mil. Substrate height: 25 mil Dielectric constant: 9.6 Input and output microstrip line widths: 25 mil. Distance from the microstrip plane to the box top cover: 250 mil. The problem is lossless. Cell size: 5x5 mil. The other dimensions see in the Layout1.

Resonance Elimination

180

Simulation details: Design Layout1 describes the whole filter that is simulated in EM1 in 181 frequency points to catch narrow resonances. Design plus8port is the same as Layout1 but with 8 additional internal ports created to break resonances in the structure. The simulation of it needs only 19 frequency points over the frequency range from 2 to 20 GHz to get the resonances back in the design fictiveShort. S-matrices are generalized for all problems.

Simulation results: The en-block simulation of the filter is very close to the simulation and experimental data obtained in [1]. The resonance elimination technique gives 10-times acceleration in comparison with the en-block simulation on the equidistant frequency grid.

Reference:

1) J.E. Bracken, D.-K. Sun, Z.J. Cendes S-domain method for simultaneous time and frequency characterisation of electromagnetic devices. - IEEE Trans., v. MTT-46, 1998, N9, p.1277-1290.

ResonanceElimination\ MicrostripLineWithFictivePort.WSP

181

ResonanceElimination\ MicrostripLineWithFictivePort.WSP

EMPOWER Example: One-cell long internal input investigation

Illustrates: That an one-cell long internal port does not affect calculated characteristics of a microstrip line segment when short circuited in a co-simulation. The port width is equal to the line width.


182

ResonanceElimination\ MicrostripLineWithFictivePort2.WSP

EMPOWER Example: One-cell long internal input investigation

Illustrates: That an one-cell long internal port does not affect calculated characteristics of a microstrip line segment when short circuited in a co-simulation. The port width is equal to the line width.

ResonanceElimination\ ResonantStubFictivePort.WSP

183

ResonanceElimination\ ResonantStubFictivePort.WSP

EMPOWER Example: A resonant stub in a microstrip line - resonance elimination technique

Illustrates: Comparison of the EM simulation with experimental data [1] and data obtained by the method of moments[2]. Demonstrates also effectiveness of the resonance elimination technique.

Description of the problem: The microstrip line is 0.23 mm wide is on a 0.254 mm thick substrate with dielectric constant 9.9. The microstrip stub is 0.51 mm wide and 1.5 mm long (1.53 in the papers). The structure is in a 2.8028 mm by 3.105 mm box. The top cover is 24.5 mm from the signal layer.

Simulation details: Design Layout1 describes the whole structure that is simulated in EM1. Design plus1port is the same as Layout1 but with an additional internal port created to break resonance in the structure. The simulation of it needs only 7 frequency points to get the resonance back in the design fictiveShort.

The cell size is 0.0637 x 0.0575 mm and S-matrices are generalized for all problems.


184

Reference: 1) F. Giannini, G. Bartolucci, M. Ruggieri, Equivalent circuit models for computer-aided design of microstrip rectangular structures. - IEEE Trans., v. MTT-40, 1992, N2, p.378-388.

2) J.E. Bracken, D.-K. Sun, Z.J. Cendes S-domain method for simultaneous time and frequency characterization of electromagnetic devices. - IEEE Trans., v. MTT-46, 1998, N9, p.1277-1290.

ResonanceElimination\ TwoLevelBPwithFictivePorts.WSP

185

ResonanceElimination\ TwoLevelBPwithFictivePorts.WSP

EMPOWER Example: Two level microstrip bandpass filter - resonance elimination technique

Illustrates: Simulation of a filter with coupled microstrip resonators in two metallization levels and comparison with published results of simulation and measurement [1]. Also demonstrates a resonance elimination technique that uses fictive loaded internal ports to expand S-matrices, to interpolate it and then reduces the expanded descriptor to the initial one.

Description of the problem: Two side-coupled microstrip resonators are on a substrate together with an input and output line segments. An additional resonator is placed on an inverted substrate and over the two resonators to achieve coupling with the input and output resonators (through one). Both substrates have the same height 0.51 mm and dielectric constant 2.33. They are separated by a foam dielectric with height 3.3 mm and dielectric constant 1.07. Microstrips are 1.49 mm wide (rounded off to 1.5mm). Whole structure is in a box 40 mm by 20 mm.

Simulation details: Design Layout1 describes the whole filter that is simulated in EM1 at 101 frequency points (3 min 40 sec per frequency point). S-parameters are generalized. Design plus3ports is the same filter but with 3 additional internal ports breaking resonators in halves and eliminating resonances when they are 50-Ohms loaded. Corresonding simulation EM2 is done only for 5 frequency points in the same frequency range as the initial filter (S-matrix is very smooth). Design fictiveShort takes 5-port data from the simulation EM2 and sets proper boundary conditions in fictive port areas (short circuit conditions). The external ports are normalized to impedances calculated in the EM2 simulation to have generalized parameters of the filter. Thinning out type for both simulations is wire with coefficient 5.


186

Simulation results: The simulation results obtained by the resonance elimination technique is almost on the top of the en-block simulation results. 20 times acceleration is achieved with simple interpolation technique ( without any loss of accuracy.)

Reference: 1) A.A. Melcon, J.R. Mosig, M.Guglielmi, Efficient CAD of boxed microwave circuits based on arbitrary rectangular elements, IEEE Trans., v MTT-47, 1999, N 7, p.1045-1058

Index

1 1/S11, 141

2 2-section microstrip combline ,

95

3 3-section edge-coupled , 86

4 4-inductor extractions , 104

8 8 Way.WSP, 24

A Agile Antenna.WSP , 171 All Parallel Resonators

Example , 109, 123 All Series Resonators Example ,

106 All-pole , 116 Alta Frequenza , 112 Amp Feedback.WSP , 63 Amp Noise.WSP , 64 Amplier , 60 Amplifier , 66, 68, 70, 152 Amplifier.wsp , 66 Amplifiers/Amplifier

Tuned.WSP , 66 Amplifiers/Amplifier.WSP , 68 Amplifiers/SiGe BFP620

Amp.WSP , 70 Antenna , 171, 173, 175 Antenna load , 150 ANTENNA.RX , 150

Antennas/MAmmannPatch.WSP, 173

Array Driver.WSP , 169 Attenuator , 11, 18, 19

B Balanced Amp.WSP , 60 Benchmarking , 42 BIP, 35 BipModel , 35 Box , 176 Box Modes , 24, 86 Box Modes.WSP , 48 Box Width , 37 Branch-line , 17 BSCouplerDevelopment.wsp ,

27 BSCouplerFinalRecomp.wsp ,

27 BSCouplerFinalWhole.wsp , 27 Butterworth , 116

C Calculate N , 102, 112

remove existing transformer , 112

Calculate N To Remove Existing Transformers , 112

Capacitors , 121, 123 Carassa , 112 Cell Size , 37, 176 Chebyshev , 75, 116 Chebyshev coupled-resonator ,

116 Circles , 57, 64 Circuit co-simulation , 68 Coaxial OSC.WSP , 143 Coaxial Resonator Example ,

130 Combline , 95 Combline filter , 77

Design , 77 Components/BJT NL Model

Fit.WSP , 54

188

Compression , 70 Computer Simulation , 112 Connectors , 86 Contiguous Diplexer.WSP , 75 Convergence properties , 45 Convert Using Advanced

TLINE, 130, 152 Cosimulation , 45 Co-simulation , 163, 175 Coupled lines , 86 Coupled Microstrip Lines , 27 Coupled Stepped Z.WSP , 83 Coupled-line , 17 Cover Height , 37 Crystal , 99, 101 Customized , 102

D Data File , 150 DC bias points , 137 Decomposition , 27, 33 Deembedding , 37, 42, 86 Design , 77

combline filter , 77 Designs , 123 Detector , 161, 163 Detectors/Simple

Detector.WSP , 161 Dielectric , 123 Dielectric Constant , 123 DIODE, 161 Diode detector , 163 Diode detector circuit , 161 Diplexers , 75 Dipole antenna , 175 Discontinuities , 64 Dual Mode Coupler.WSP , 21

E Eagleware , 112 Edge Coupled , 86 Edge Coupler.WSP , 17 Edge Positioning , 42 Edge-Coupled Filter , 89 EdgeCoupledBenchmark.WSP ,

86

EdgeCoupledOpened.WSP , 86 Edit Menu , 60 Electromagnetic , 116 Electromagnetic simulation , 68 EM, 68 EMPOWER, 18, 19, 21, 24, 27,

30, 33, 77, 83, 89, 98, 175, 176, 179, 181, 182, 183, 185

Equal Inductor Example , 115 Equal Termination Example ,

102 Equations , 14, 35, 51, 116 Etch Factor , 42 EXAMPLES subdirectory , 5 Extraction sequence , 112, 123 Extractions , 102, 112, 116, 123 Extractions Tab , 104 Extrapolation , 37 Extration Goals , 104

F Fictive port , 89, 185 fictive short , 183 Film , 18 Film Atten.WSP , 11 FILTER, 116 Filters/ComblineDesign/Final

Lossy.WSP , 77 Finite Zeros , 112, 116 Force Symmetry , 116

G Generalized , 37 Generalized S-Parameters , 42 Generate Viewer Data , 83 GETVALUEAT , 143 Grid , 37 Group-delay , 112 Group-Delay Characteristics ,

112

H HARBEC , 6, 70, 137 Harmonic Balance , 137 Harmonics , 70

189

HF Filter Design , 112 High-Q quartz-crystal

resonators , 106 How To Design , 101

I Ill Behaved Load.WSP , 150 Impedances , 123 Inductor Q , 116 Inductors equal , 115 Infinity , 112, 123 Instability , 57 Interdigital , 83, 95 Interdigital.WSP , 92 Internal Ports , 175, 176

L Layout Menu , 60 L-C pseudo lowpass network ,

147 Linear , 6 Lower Stopband , 123 Lowpass , 112, 147

M M/FILTER, 6 MATCH, 6, 147, 150, 152 Matching concepts , 147 Matching resistive

terminations , 150 Matching/Ill Behaved

Load.WSP , 150 Maximum Realizability

Example , 104 Measured data , 18, 54, 173, 179 Measured S-parameters , 54 Measurements , 141, 143 Microstrip , 17, 33, 37, 57, 64, 77,

98, 173, 179, 181, 182, 185 Microstrip Line.WSP , 37 Microwave Circuits , 64 MIXER, 165 Mixer/Low Power Mixer.WSP ,

165 Model , 51

Model Extract , 35 MOL electromagnetic

simulation , 45 Monte Carlo , 6

N Negative Resistance , 141 NegR VCO.WSP, 141 NET, 60 Noise Circles , 64 Noise Figure , 64 Non-deembedded , 37 NonLinear , 6 Nonlinear model , 70 Nonlinear models , 54 Non-resonant simulation , 92, 95 Nonsymmetrical , 83 Norton , 112, 123 Norton Series , 102, 109, 112,

115, 123 Norton Shunt , 106, 115 Norton transforms , 115, 116

O Odd-order Chebyshev , 116 Operators , 141 Optimization , 6, 35, 54, 64, 77,

116, 121, 123 Optimization Targets , 35 Optimize , 137 OSCILLATOR , 137 Oscillators/Bipolar Cavity

Oscillator.WSP , 137 Output matching , 64

P Parameter extraction , 35 Parameter Sweep , 6 parameter sweeps , 161 Parameters , 123 Parametric Bandpass Example ,

123 Parasitics , 35 Passband , 116, 123 Passband ripple , 123

190

Patch antenna , 173 Patch Antenna

Impedance.WSP , 167 Permutation , 102, 104 Phase Noise , 137, 143 Physical Symmetry , 116 Physical Symmetry Example ,

116 Poor stopband performance , 48 Port Impedance , 75 Post-processing , 35, 141, 143 Power Amp , 152 PowerAmp , 152 Prob-fed antenna , 173 Pseudo lowpass , 147

Q Quarter-wave , 123 Quartz-crystal resonators , 106

R RAM , 5 Reference Plane , 86, 167 Remove existing transformer ,

112 Calculate N , 112

Remove sidewall reactances de-embedded , 68

Remove Transformer , 104 Resonance , 179, 183 Resonance Elimination

Technique , 89, 179, 183 ResonanceElimination/

EdgeCoupledWith FictivePorts.WSP , 89

ResonanceElimination/ LowPassWithFictivePorts.WSP, 179

ResonanceElimination/ MicrostripLineWithFictivePort.WSP, 181

ResonanceElimination/ MicrostripLineWithFictivePort2.WSP, 182

ResonanceElimination/ ResonantStubFictivePort.WSP, 183

ResonanceElimination/ TwoLevelBPwithFictivePorts.WSP, 185

Resonant stub , 183 Resonator Q , 104 Resonators , 83, 92, 98, 101,

104, 116, 121, 123, 130, 185 Response Symmetry Example ,

112 Richardson , 37 Richardson's Extrapolation , 42 Rubber Bands , 17, 60

S Series Element First , 106, 109 Shunt-coupling , 116 Sidelobes , 169 Signal Control/

BSCouplerFinalRecomp.WSP, 30

Signal Control/ BSCouplerFinalWhole.WSP , 33

Signal Control/ ResistiveBroadband 20dB.WSP , 19

Simple Dipole.WSP , 175 Simplify , 112, 115 Simplify Circuit , 109, 112 Single-ended , 60 Singly-terminated , 75 SIR, 83 Smith Chart , 64, 141 Smith charts , 165 Snap Angle , 37 Solid Thinning , 18, 27 S-parameter , 35, 57, 152

difference , 35 S-parameter measurement , 152 S-Parameters , 6, 54, 89 Specifications , 109 Specifications Tab , 112, 116 SPICE, 6, 54 Spiral Inductor2.WSP , 45

191

Stability , 48, 57, 156 Stability.WSP , 57 Stabilization Networks.wsp , 158 Stabilize , 158 Stopband , 123 Stopband performance , 48 Stripline Standard.WSP , 42 Subsections , 37, 48 Symmetrical , 112, 116

T Termination Coupling Example ,

121 Termination Coupling.SF$ , 121 Terminations , 102, 104, 109,

121, 123, 150, 169 TESTLINK , 6 Thin Film Attenuator , 18 Thin Loop Antenna , 176 Thin Loop.WSP , 176 Top Cover , 48 Transformations , 121, 123 Transformer , 104, 112, 116, 121,

123 Transformer Ratio , 112, 116 Transformers , 14, 115 Transforms , 112, 115, 116, 121 Transistor , 35 Transmission lines , 101, 121,

123 Transmission lines realize , 101 Tuned Bandpass , 95 Two Level.WSP , 98

U Unstable Device , 154

Matching , 154 Upper Cutoff , 121 Upper Stopband , 123 User Model , 6 User Model.WSP , 51

V Varactor , 51, 95, 171 Viaholes , 64 Viewer , 21

W Weights , 35 Wilkinson , 17, 24 WIRE, 45

X Xfmr Coupler.WSP , 14 Xtal Filter.WSP , 99

Y Yield , 6

Z Zo, 37, 42

Extrapolating , 37

This page intentionally left blank.

Documents

GENESYSV8 Examples - Keysightliterature.cdn.keysight.com/litweb/pdf/genesys8/Examples8.pdf · Overview 6 Examples 7.5\Components\Bondwires CompensatedCrossover1.wsp CompensatedCrossover2.wsp