Design of a multi-band RF front-end between 2-12 GHz.

Examensarbete

LITH-ITN-ED-EX--06/016--SE

Design of a multi-band RF

front-end between 2-12 GHz.

Sang Yu

2006-04-07

LITH-ITN-ED-EX--06/016--SE

Design of a multi-band RF

front-end between 2-12 GHz.

Examensarbete utfört i Elektronikdesign

vid Linköpings Tekniska Högskola, Campus

Norrköping

Sang Yu

Handledare Pär Håkansson

Handledare Allan Huynh

Examinator Shaofang Gong

Rapporttyp Report category Examensarbete B-uppsats C-uppsats D-uppsats _ ________________ Språk Language Svenska/Swedish Engelska/English _ ________________ Titel Title Författare Author Sammanfattning Abstract ISBN _____________________________________________________ ISRN _________________________________________________________________ Serietitel och serienummer ISSN

Title of series, numbering ___________________________________

Nyckelord

Datum

Date

URL för elektronisk version

Avdelning, Institution

Division, Department

Institutionen för teknik och naturvetenskap

Department of Science and Technology

2006-04-07

x

x

LITH-ITN-ED-EX--06/016--SE

Design of a multi-band RF front-end between 2-12 GHz.

Sang Yu

The state-of-the-art standard for short range wireless data transmission is the Ultra Wide Band radio (UWB) standard. UWB provides the maximum data rate of 480 Mbit/s. However, for some data transmission applications this is far from sufficient. Since there already are a lot of research and development activities in wireless technology, the demands on even higher data rates is considered as necessary.

The research group of Communication Electronics at ITN has a new concept to design a radio frequency front-end for wireless and high-speed data transmission using a multi-band technique. With help of this technique wideband signals can be transferred simultaneously and then be divided into multiple of sub-bands for down conversion. Compare to other technologies such as Bluetooth and WLAN, only a single narrow band is used, which gives much lower data rate.

The assignment is to design one wide-band frequency demultiplexer with eleven sub-bands, 200 MHz each. The principle of the frequency demultiplexer is based on the properties of quarter-wavelength at a certain frequency. This was done with help of input matching networks using microstrip lines. The theoretical maximum data rate is calculated to 6.42 Gbit/s.

Upphovsrätt

Detta dokument hålls tillgängligt på Internet – eller dess framtida ersättare –

under en längre tid från publiceringsdatum under förutsättning att inga

extra-ordinära omständigheter uppstår.

Tillgång till dokumentet innebär tillstånd för var och en att läsa, ladda ner,

skriva ut enstaka kopior för enskilt bruk och att använda det oförändrat för

ickekommersiell forskning och för undervisning. Överföring av upphovsrätten

vid en senare tidpunkt kan inte upphäva detta tillstånd. All annan användning av

dokumentet kräver upphovsmannens medgivande. För att garantera äktheten,

säkerheten och tillgängligheten finns det lösningar av teknisk och administrativ

art.

Upphovsmannens ideella rätt innefattar rätt att bli nämnd som upphovsman i

den omfattning som god sed kräver vid användning av dokumentet på ovan

beskrivna sätt samt skydd mot att dokumentet ändras eller presenteras i sådan

form eller i sådant sammanhang som är kränkande för upphovsmannens litterära

eller konstnärliga anseende eller egenart.

För ytterligare information om Linköping University Electronic Press se

förlagets hemsida

Copyright

The publishers will keep this document online on the Internet - or its possible

replacement - for a considerable time from the date of publication barring

exceptional circumstances.

The online availability of the document implies a permanent permission for

anyone to read, to download, to print out single copies for your own use and to

use it unchanged for any non-commercial research and educational purpose.

Subsequent transfers of copyright cannot revoke this permission. All other uses

of the document are conditional on the consent of the copyright owner. The

publisher has taken technical and administrative measures to assure authenticity,

security and accessibility.

According to intellectual property law the author has the right to be

mentioned when his/her work is accessed as described above and to be protected

Design of a multi-band front-end between 2-12 GHz

by:

Sang Yu

20th February 2006

Supervisors: Allan Huynh

Ph.D. student, ITN

Pär Håkansson

Ph.D. student, ITN Examiner: Professor Shaofang Gong

Acknowledgements

This Master Thesis report is the result of work done at the Department of Science and Technology (ITN) at Linköping University. The report is the final step towards a Master of Science degree in Electronics Design Engineering at Linköping University, Campus Norrköping.

I would like to thank everyone who helped and encouraged me during the work with this thesis. Especially, I would like to thank:

• My supervisors Allan Huynh and Pär Håkansson at the Department of Science and Technology, ITN of Linköping University for all support and assistance during the thesis work.

• My examiner professor Shaofang Gong at the Department of Science and Technology, ITN of Linköping University for all assistance and valuable comments during the thesis work.

• Lars Klooster at MTT Components & Systems AB for all help with the CAD-program Microwave Office.

• My opponent Erik Ottosson for his valuable comments and feedback on my thesis report.

Sang Yu February 2006

Abstract

This report is the result of a Master Thesis work done at the Department of Science and Technology (ITN) at Linköping University between September 2005 and February 2006. The main goal for this thesis is to design and verify the concept of a multi-band radio frequency front-end between 2-12 GHz. The work is done by using two different types of CAD tools for RF electronics, Microwave Office by Applied Wave Research Inc. and ADS by Agilent Technologies Inc. Another goal of this project was to introduce Microwave Office to the research group of Communication Electronics at ITN.

The assignment is to design one wide-band frequency demultiplexer with eleven sub-bands, 200 MHz each. The principle of the frequency demultiplexer is based on the properties of quarter-wavelength at a certain frequency. This was done with help of input matching networks using microstrip lines. The theoretical maximum data rate is calculated to 6.42 Gbit/s.

The result confirms the principle of a frequency demultiplexer using quarter-wavelength microstrip line theory. A guard-band is needed between each sub-band when using Coupled Line Filter of order between3≤ N ≤6.

Substrate thickness equal to 0.254 mm works fine in relatively low frequencies. At frequencies higher than 4.3 GHz the thickness gets too large, which affects the small mircostrip lengths in the matching networks. The problem is solved by using a smaller substrate thickness.

Starting the frequency demultiplexer at 7 GHz, a theoretical active region can be up to 10.5 GHz. Since eleven 200 MHz sub-bands with 50 MHz wide guard-bands are needed, the frequency band satisfies the requirement.

Sammanfattning

Denna rapport är resultatet av ett examensarbete gjort på institutionen för teknik och naturkunskap, ITN Linköpings universitet mellan september 2005 och februari 2006. Huvudmålet för detta arbete är att designa och verifiera det nya konceptet av en multiband radiofrekvens front-end mellan 2-12 GHz. Arbetet utförs med hjälp av två stycken CAD-verktyg för RF elektronik, Microwave Office av Applied Wave Research Inc. och ADS av Agilent Technologies Inc. Ett delmål för arbetet var också att introducera Microwave Office för forskningsgruppen på ITN.

Uppgiften är att designa en funktionell bredbandig frekvensdemultiplexer med elva subband, 200 MHz vardera. Frekvensdemultiplexerns princip är baserad på kvartsvåglängd egenskaperna vid en bestämd frekvens. Det gjordes med hjälp av input matching networks för mikrostrip ledare. Den teoretiska maximala datahastigheten beräknas vara 6,42 GHz/s.

Resultatet i denna rapport bekräftar att principen av en frekvensdemultiplexer fungerar genom att använda sig utav kvartsvåglängdsteori för mikrostrip ledare. Ett spärrband behöver användas mellan varje subband när man använder sig utav Coupled Line Filter av order mellan3≤ N ≤6.

Substrattjocklek lika med 0,254 mm fungerar bra i relativt låga frekvenser, men vid frekvenser högre än 4,3 GHz blir tjockleken för stor som påverkar de små mikrostrip längderna i matching networks. Problemet löses genom att använda en mindre substrattjocklek.

Genom att börja frekvensdemultiplexern vid 7 GHz, kan en teoretisk aktiv frekvensregion vara upp till 10,5 GHz. Alltsedan elva 200 MHz subband används med 50 MHz brett spärrband är det tillräckligt bra.

Contents

2.2 Quarter-wavelength transmission lines ... 7

2.3 Scattering parameters ... 10

2.4 Filters ... 12

2.4.1 Coupled Line Filter ... 13

2.4.1.1 Filter order... 14

2.5 RF front-end ... 15

3 Computer Aided Design tool 16 3.1 Microwave Office... 16

3.2 Advanced Design System... 17

4 Design process 18 4.1 Design Overview ... 18

4.2 Design specification ... 21

4.3 Design and simulation ... 23

4.3.1 Guard-band evaluation ... 29

4.3.1.1 N=3... 29

4.3.1.2 N=4... 31

4.3.1.3 N=5... 33

4.3.1.4 N=6... 35

4.3.2 Higher frequency simulations ... 38

4.3.3 Implementation of Coupled Line Filter... 40

4.3.3.1 Filter design... 41

4.3.3.2 Exchange the filters ... 44

5 Results 49

6 Discussions 54

7 Conclusions 56

APPENDICES

A Calculation of OFDM parameters for the system. 59 B The length of microstrip lines using 0.254 mm substrate thickness. 60 C The length of microstrip lines using 0.0508 mm substrate thickness. 62 D The ten remaining bandpass filters designed and optimized in ADS. 64 E The length of microstrip lines using 0.0508 mm substrate thickness and

barrier-bands. 73 F Schematics of the 11 bandpass filters designed in Microwave Office. 74

List of figures

2.1: Geometry and field distribution in two-wire parallel conductor ... 3

2.2: Coaxial cable transmission line ... 4

2.3: Transmission line representation. (a) PCB section. (b) Microstrip line ... 4

2.4: Electric field leakage. (a) Teflon epoxy. (b) Alumina. ... 5

2.5: Topology of microstrip substrate ... 5

2.6: Terminated transmission line... 8

2.7: Two-port network convention used to define S-parameters ... 10

2.8: A resonant transmission line representation ... 13

2.9: Geometry of the resonant transmission line... 14

2.10: Multiple numbers of resonators merged together ... 14

2.11: RF front-end overview... 15

4.1: The principle of multiplexer and demultiplexer ... 18

4.2: Block diagram of the concept ... 19

4.3: Multiple demultiplexers in parallel... 19

4.4: (a) Sub-bands. (b) Band covering with demultiplexers ... 21

4.5: The general principle of the demultiplexer... 22

4.6: Differences between symbol and geometry in CAD-tools ... 22

4.7: With the variable tuner-tool the microstrip length can be optimized ... 24

4.8: Verifying the matching network with help of impedance check ... 25

4.9: Circuit with an ideal bandpass filter. Order N=20... 25

4.10: Filter response... 26

4.11: Simulation of S11 ... 26

4.12: The demultiplexer with 11 sub-bands... 27

4.13: Eleven 200 MHz bandpassfilters with order N equal to 20 ... 27

4.14: Simulation of S11 ... 28

4.15: S11 represented in the same graph as the filter response... 28

4.16: No guard-band ... 29 4.17: Guard-band equal to 200MHz ... 30 4.18: Guard-band equal to 100 MHz ... 30 4.19: Guard-band equal to 50 MHz ... 31 4.20: No guard-band ... 32 4.21: Guard-band equal to 50 MHz ... 32 4.22: Guard-band equal to 25 MHz ... 33 4.23: No guard-band ... 34 4.24: Guard-band equal to 50 MHz ... 34 4.25: Guard-band equal to 25 MHz ... 35 4.26: No guard-band ... 36 4.27: Guard-band equal to 50 MHz ... 36 4.28: Guard-band equal to 25 MHz ... 37 4.29: Guard-band equal to 20 MHz ... 37

4.30: S11 represented with the filter response. Order N equal to 20 ... 39

4.31: Simulation in the 4.2 to 6.4 GHz region. Order N equal to 20 ... 39

4.32: Simulation in the 6.4 to 8.6 GHz region. Order N equal to 20 ... 40

4.33: ADS DesignGuide for passive circuits ... 41 4.34: Microwave Office Filter Wizard.

4.36: Microwave Office schematic of the bandpass filter ... 42

4.37: Simulation in ADS... 43

4.38: Simulation in Microwave Office ... 43

4.39: The first sub-band with designed bandpass filter ... 44

4.40: The two first sub-bands with designed bandpass filters ... 44

4.41: The three first sub-bands with designed bandpass filters ... 45

4.42: The four first sub-bands with designed bandpass filters... 45

4.43: The impedance of the bandpass filter in region 2 to 2.2 GHz ... 46

4.44: Transformed impedance... 46

4.45: With a lowpass filter possible harmonic signals can be filtrated away .... 47

4.46: Simulation with a lowpass filter connected in series ... 47

5.1: Simulation with open filters in the 2 to 4.7 GHz region... 49

5.2: Simulation of the first half region... 50

5.3: Simulation in the 7 to 9.7 GHz region ... 51

5.4: Simulation in the 7 to 9.7 GHz region ... 52

5.5: A 2-D and 3-D view of the circuit ... 52

5.6: 2-D view of the frequency demultiplexer ... 53

6.1: Comparison of frequency response... 54

D.1: Bandpass filter, N=5. Active in the 2.2 to 2.4 GHz frequency region... 64

D.2: Bandpass filter, N=5. Active in the 2.4 to 2.6 GHz frequency region... 64

D.3: Bandpass filter, N=5. Active in the 2.6 to 2.8 GHz frequency region... 64

D.4: Bandpass filter, N=5. Active in the 2.8 to 3 GHz frequency region... 65

D.5: Bandpass filter, N=5. Active in the 3 to 3.2 GHz frequency region... 65

D.6: Bandpass filter, N=5. Active in the 3.2 to 3.4 GHz frequency region... 65

D.7: Bandpass filter, N=5. Active in the 3.4 to 3.6 GHz frequency region... 66

D.8: Bandpass filter, N=5. Active in the 3.6 to 3.8 GHz frequency region... 66

D.9: Bandpass filter, N=5. Active in the 3.8 to 4 GHz frequency region... 66

D.10: Bandpass filter, N=5. Active in the 4 to 4.2 GHz frequency region... 67

D.11: Simulation of the second bandpass filter in the 2- 4.2 GHz band... 67

D.12: Simulation of the third bandpass filter in the 2- 4.2 GHz band ... 68

D.13: Simulation of the fourth bandpass filter in the 2- 4.2 GHz band ... 68

D.14: Simulation of the fifth bandpass filter in the 2- 4.2 GHz band... 69

D.15: Simulation of the sixth bandpass filter in the 2- 4.2 GHz band ... 69

D.16: Simulation of the seventh bandpass filter in the 2- 4.2 GHz band... 70

D.17: Simulation of the eighth bandpass filter in the 2- 4.2 GHz band... 70

D.18: Simulation of the ninth bandpass filter in the 2- 4.2 GHz band... 71

D.19: Simulation of the tenth bandpass filter in the 2- 4.2 GHz band... 71

F.9: Bandpass filter, N=5. Active in the 9 to 9.2 GHz region ... 78

F.10: Bandpass filter, N=5. Active in the 9.25 to 9.45 GHz region... 78

F.11: Bandpass filter, N=5. Active in the 9.5 to 9.7 GHz region... 79

F.12: Simulation of the first bandpass filter in the 7- 9.7 GHz band ... 79

F.13: Simulation of the second bandpass filter in the 7- 9.7 GHz band ... 80

F.14 Simulation of the third bandpass filter in the 7- 9.7 GHz band ... 80

F.15: Simulation of the fourth bandpass filter in the 7- 9.7 GHz band... 81

F.16: Simulation of the fifth bandpass filter in the 7- 9.7 GHz band... 81

F.17: Simulation of the sixth bandpass filter in the 7- 9.7 GHz band... 82

F.18: Simulation of the seventh bandpass filter in the 7- 9.7 GHz band ... 82

F.19: Simulation of the eight bandpass filter in the 7- 9.7 GHz band ... 83

F.20: Simulation of the ninth bandpass filter in the 7- 9.7 GHz band ... 83

F.21: Simulation of the tenth bandpass filter in the 7- 9.7 GHz band ... 84

List of tables

2.1: Microstrip substrate properties ... 5

4.1: The estimated performance... 20

4.2: Substrate Properties ... 23

4.3: The current lengths of the demultiplexer ... 23

4.4: The new substrate definition... 38

5.1: Active frequency... 49

5.2: Resonances occur approximately 1.5 times the frequency ... 50

5.3: Realizable substrate properties ... 51

B.1: Substrate definition ... 60

B.2: Microstrip lengths at different frequencies ... 60

C.1: Substrate definition ... 62

C.2: Microstrip lengths at different frequencies ... 62

E.1: Substrate definition... 73

E.2: Microstrip lengths at different frequencies... 73

Terminology

AWR Applied Wave Research ADS Advanced Design System Balun Balanced to unbalanced BW Bandwidth

CAD Computer Aided Design

EDA Electronic Design Automation

FR-4 Flame Retardant. 4 indicate woven glass reinforced epoxy resin.

IL Insertion Loss

LNA Low Noise Amplifier

PA Power Amplifier

PCB Printed Circuit Board RAM Random Access Memory

RF Radio Frequency

RL Return Loss

SF Shape Factor

TanD Loss Tangent of Dielectric UWB Ultra Wide Band

WLAN Wireless Local Area Network

in Z Input impedance Z0 Characteristic impedance L Z Load impedance 0 Γ Reflection coefficient β Wave number

1 Introduction

The state-of-the-art standard for short range wireless data transmission is the Ultra Wide Band radio (UWB) standard. UWB provides the maximum data rate of 480 Mbit/s. However, for some data transmission applications this is far from sufficient. Since there already are a lot of research and development activities in wireless technology, the demands on even higher data rates is considered as necessary.

1.1 Background

The research group of Communication Electronics at ITN has a new concept to design a radio frequency front-end for wireless and high-speed data transmission using a multi-band technique. With help of this technique wideband signals can be transferred simultaneously and then be divided into multiple of sub-bands for down conversion. Compare to other technologies such as Bluetooth and WLAN, only a single narrow band is used, which gives much lower data rate.

1.2 Goal

The main goal for this thesis is to design and verify the new concept of a multi-band radio frequency front-end between 2-12 GHz. The work is done by using two different types of CAD tools for RF electronics, Microwave Office by Applied Wave Research Inc. and Advanced Design Systems (ADS) by Agilent Technologies Inc. Another goal for this project was to learn and evaluate Microwave Office for the research group of Communication Electronics at ITN.

1.3 Method

The CAD-tools are ADS by Agilent Technologies Inc. and Microwave Office by Applied Wave Research Inc. ADS is used at ITN for both research and educational purposes. Applied Wave Research Inc. has recently donated 20 new licenses of Microwave Office to ITN for educational purpose.

1.4 OUTLINE CHAPTER 1. INTRODUCTION

1.4 Outline

This thesis is divided in the following chapters:

• Chapter 2 covers most of the theory. This involves substrate characteristics, general issues when designing RF components with help of microstrip lines and the related RF theory.

• Chapter 3 contains a brief introduction of the used CAD-tools.

• Chapter 4 describes the entire design process in this thesis work. This includes overview, specification, research and computer simulations.

• Chapter 5 contains the result of the work and the physical layout of the circuit. • Chapter 6 consists of a discussion regarding the results of this thesis work. • Chapter 7 contains the conclusions drawn in this work.

2 Theory

This chapter covers most of the theory used in this thesis report. A basic knowledge of electrical engineering is considered as an advantage. For further information, please follow the reference list.

Dealing with RF frequencies, ordinary electronics design methodology using lumped elements such as inductors and capacitors has difficulties to achieve good design because these components are highly frequency dependent. The fact that capacitors have parasitic inductance in series and inductors commonly have parasitic capacitance in parallel is drawbacks which cannot be neglected.

2.1 Transmission Line [1] [6]

Transmission lines are ordinary conductors which are capable of transporting electric energy from one location to another at high frequencies. An example is the two-wire transmission line, where two conductors are separated over a fixed distance, as shown in Figure 2.1.

2.1 TRANSMISSION LINE CHAPTER 2. THEORY

Another example of a transmission line is the coaxial cable, and is more commonly used for data transmission. As shown in figure 2.2 a typical coaxial line consists of an inner cylindrical conductor, an outer conductor and a dielectric medium layered in between where the outer conductor is for ground.

Figure 2.2: Coaxial cable transmission line.

2.1.1 Microstrip [6]

For printed circuit board (PCBs) designs microstrip conductor lines is one of the most common. Dealing with RF circuits, a consideration of the high-frequency behavior is needed. The conducting strips etched on the PCBs, is depicted in figure 2.3. The ground plane below the current carrying conductor traces helps prevent excessive field leakage and thus reduces radiation loss.

(a) PCB section (b) Microstrip line Figure 2.3: Transmission line representation.

2.1 TRANSMISSION LINE CHAPTER 2. THEORY

A disadvantage of single layered PCBs is that they have relatively high radiation loss despite the ground plane with comparison to more advanced transmission lines. This lead to phenomena “crosstalk” which is interference between neighboring conductor traces. The field leakage depends on the relative dielectric constants in the mircostrip. As shown in figure 2.4 a greater dielectric constant leads to smaller field leakage and cross coupling effects.

(a) Teflon epoxy (εr = 2.55) (b) Alumina (εr = 10.0)

Figure 2.4: Electric field leakage.

The mircostrip topology and the parameter definition are shown in figure 2.5 and in table 2.1, respectively.

Figure 2.5: Topology of microstrip substrate. Table 2.1: Microstrip substrate properties.

Name Description Unit Type

εr Relative dielectric constant None

H Substrate thickness Length

T Conductor thickness Length

Rho Metal bulk resistivity normalized to gold None TanD Loss tangent of dielectric None

2.1 TRANSMISSION LINE CHAPTER 2. THEORY

The relative permittivity of the substrate materialε_r, is relative to the permittivity of

free space 12 2

0 8.854 10 F/m −

× =

ε . Rho, ρ is simply the bulk resistivity of conductor metal normalized to gold, which is _2.44×10−8 _{m/S. TanD is the loss tangent of}

dielectric. Conductivity is simply the inverse ofρ.

Increasing the substrate dielectric constant will lead to smaller dimensions of the microstrip circuit, but the loss usually increase due to a greater dielectric constant. Materials with high relative dielectric constant usually have higher loss tangents, TanD. It is also the same for the characteristic impedance, where reduced width of the conductor line gives higher ohm losses. This is a typical conflicting situation between the necessary requirements for small dimensions and low loss. Generally lower dielectric constants are preferred since losses are reduced.

2.2 QUARTER-WAVELENGTH TRANSMISSION LINES CHAPTER 2. THEORY

2.2 Quarter-wavelength transmission lines [6]

The most important parameters for RF electronics design methodology are the length and width of the conductor. Generally speaking changing the width of the conductor directly affects the impedance and changing the length controls the voltage and current waves. The key of getting a proper design in RF electronics demands knowledge in this behavior.

The characteristic line impedance for a lossless transmission line is defined in equation 2.1 where Z0 is frequency independent, L is the inductance and C is the capacitance for

the transmission lines.

C L

Z₀ = / (2.1)

When dealing with high-frequency electric circuits, an important variable is the reflection coefficient, .Γ ₀ 0 0 0 Z Z Z Z L L + − = Γ (2.2)

Where ZL is the load impedance for the transmission line. With this representation the

conclusions that can be made is for an open line, Z_L is infinite and the reflection coefficient becomes 1. That means the reflected wave returns with the same polarity as the incident voltage wave. But for a short circuit, Z_L is zero and the reflected voltage wave returns with inverted amplitudes, resulting in Γ₀ =−1. In the case where the line impedance matches the load impedance Γ will be zero. ₀

2.2 QUARTER-WAVELENGTH TRANSMISSION LINES CHAPTER 2. THEORY

The input impedance Z is defined as: _in

) 1 ( ) 1 ( ) ( ) ( ) ( ₂ 0 2 0 0 j d j d d j d j in e e V e e V Z d I d V d Z _{β β} β β − + − + Γ − Γ + = = (2.3)

Where d is distance away from load and V and + V is traveling voltage waves −

aligned along the theoretical z -axis, see figure 2.6.

Figure 2.6: Terminated transmission line.

The +

V means that the traveling wave propagates in the +z-direction and the V −

propagates in the z− -direction.

Together with equation 2.2, Zin(d) can be expressed as:

0 0 0 0 0 ) ( Z e Z Z Z Z e e Z Z Z Z e d Z d j L L d j d j L L d j in β β β β − − ⎟⎟ ⎠ ⎞ ⎜⎜ ⎝ ⎛ + − − ⎟⎟ ⎠ ⎞ ⎜⎜ ⎝ ⎛ + − + = (2.4) 0 0 0 ) ( ) ( ) ( ) ( Z e e Z e e Z e e Z e e Z d j d j d j d j L d j d j d j d j L β β β β β β β β − − − − + + − − + + = (2.5) 0 0 0 ) sin( ) cos( ) sin( ) cos( Z d jZ d Z d jZ d Z L L β β β β + + = (2.6)

2.2 QUARTER-WAVELENGTH TRANSMISSION LINES CHAPTER 2. THEORY

Division by the cosine term gives the final form of the input impedance for the terminated transmission line:

) tan( ) tan( ) ( 0 0 0 d jZ Z d jZ Z Z d Z L L in _β β + + = (2.7)

This result helps when predicting the load impedance, ZL along the transmission line

of the characteristic impedance, Z₀ and length d. It also takes the wave number β into

account. β can be expressed in terms of frequency and phase velocity:

p v f ) 2 ( π β = , or wavelength, λ π β = (2 ). (2.8)

If length d is exactly a half wavelength long, the input impedance is equal to the load impedance, regardless of the characteristic line impedance. If the length d gets reduced to λ /4 the input impedance gets equal to:

L L L in Z Z jZ Z jZ Z Z d Z 2 0 0 0 0 4 2 tan 4 2 tan ) 4 / ( = ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ _⋅ + ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ _⋅ + = = λ λ π λ λ π λ (2.9)

The quarter-wave transformer allows the matching of real load impedance to real input impedance by choosing a transmission line segment whose characteristic impedance is equal to:

in LZ

Z

Z₀ = (2.10)

The theory of impedance matching is very important when designing RF components. In this thesis work, the main focus is on frequency dividing using quarter-wave principle. More details will be covered in chapter 4, Design process.

2.3 SCATTERING PARAMETERS CHAPTER 2. THEORY

2.3 Scattering parameters [6]

S-parameters are power wave descriptors defined as input-output relations of a network in terms of normalized incident and reflected power waves. With help of the S-parameters, RF devices can be measured without causing any undesired inductance or capacitance effects to the device or circuit. This is especially important in higher frequency where small magnitude of disturbance gets substantially larger. Reference to figure 2.7 the defined incident normalized power wave, an and reflected normalized

power wave, bn gives:

) ( 2 1 0 0 n n n V Z I Z a = + (2.11a) ) ( 2 1 0 0 n n n V Z I Z b = − (2.11b)

The index n refers either to port number 1 or 2. The characteristic impedance Z₀ is

over the connecting lines on the input and output side of the network. For simplicity the line impedance on the input and the output is considered as equal. In reality the impedance can vary a little bit.

2.3 SCATTERING PARAMETERS CHAPTER 2. THEORY

By using equation 2.11 together with the two-port network the definition of the S-parameters can be expressed as:

1 1 0 1 1 11

2 incident powerwaveat port

port at wave power reflected a b S a ≡ = = (2.12a) 1 2 0 1 2 21

2 incident powerwaveat port

port at wave power d transmitte a b S a ≡ = = (2.12b) 2 1 0 2 1 12

1 incident powerwaveat port

port at wave power d transmitte a b S a ≡ = = (2.12c) 2 2 0 2 2 22

1 incident powerwaveat port

port at wave power reflected a b S a ≡ = = (2.12d)

The conditions a₂ =0 and a₁ =0means that no power waves are returned to the network at either port 1 or 2. Moreover these conditions only can be ensured when the connecting transmission lines are terminated into their characteristic impedances.

2.4 FILTER CHAPTER 2. THEORY

2.4 Filter [6]

It is of particular interest in any analog circuit design to manipulate high-frequency signals in such a way as to enhance or attenuate certain frequency ranges or bands. This section will go through the theory of filtering analog RF signals.

Generally it exist four types of filters: low-pass, high-pass, bandstop and bandpass. The bandpass filter allows high, in ideal infinite attenuation before and after the lower respectively upper cut off frequency points, reducing the amplitude signal to the output. At the same time the attenuation is low in the passband to increase the amplitude signal to the output signal in the active frequency region.

To choose the right type of filter in a RF circuit design an adjustment to some parameters is needed. The following parameters play key roles:

Bandwidth. For a bandpass filter, bandwidth defines the difference between upper and

lower frequencies typically recorded at the 3 dB attenuation points. The 3 dB point is the half power effect.

dB L dB u dB f f BW3 = 3 − 3 (2.13)

Shape factor. This factor describes the sharpness of the filter response by taking the

ratio between the 60 dB and the 3 dB bandwidths:

dB L dB u dB L dB u dB dB f f f f BW BW SF _{3 3} 60 60 3 60 − − = = (2.14)

Ripple. The flatness of the signal in the passband can be adjustable by specifying the

ripple or difference between maximum and minimum amplitude response. The common unit of filter ripple is in dB. A common type of filter which relatively easy can adjust the ripple is the Chebyshev filter. A benefit of adjustable ripple filters is that it has faster and steeper transition in the passband and stopband, which usually is important in high demanding electronic design. A disadvantage due, is when the ripples get too large; this will increase the power losses.

Insertion loss. Ideally, a perfect filter inserted into the RF circuit path would not give

any power loss in the passband. In other words, it would have zero insertion loss. In reality due, a certain amount of power loss associated with the filter is expected. The insertion loss quantifies how much below the zero dB line the power amplitude response drops. ) 1 log( 10 log 10 _in 2 L in P P IL= =− −Γ (2.15)

2.4 FILTER CHAPTER 2. THEORY

in

P is the input power from the source and P_L is the power delivered to the load. Γ is _in the reflection coefficient to the filter. This input coefficient is obtained by looking from the source into the transmission line of length d = l:

0 0 ) ( Z Z Z Z l d in in in + − = = Γ = Γ (2.16)

Return loss. Circuit realizations always suffer a certain degree of mismatch between

the source and the power delivered to the transmission line. This mismatch is called return loss and is the ratio of reflected power to incident power:

in in i r P P RL _⎟⎟=− Γ =− Γ ⎠ ⎞ ⎜⎜ ⎝ ⎛ −

= 10log 10log 2 20log (2.17)

2.4.1 Coupled Line Filter [6]

A Coupled Line Filter includes an input port, an output port and a number of resonant elements of order N. Together they form an electromagnetic coupling effect which gives the characteristics of the filter. In figure 2.8 the geometric arrangement for a Coupled Filter representation is depicted.

2.4 FILTER CHAPTER 2. THEORY

3-D view of the same type coupled microstrip lines as figure 2.8 is shown in figure 2.9. The whole concept consists of two lines separated over a distance S and attached to a dielectric substrate with a specially defined thickness and a dielectric constant,ε . The r

microstrip lines are wide W, and the thickness is negligible compared to the substrate thickness.

Figure 2.9: Geometry of the resonant transmission line. 2.4.1.1 Filter order [6]

In common RF circuit design the performance of the filter plays a decisive role. Often a filter with steep passband to stopband transitions is most desirable. To achieve this type of performance a single bandpass resonator is not good enough. The solution for this is to multiply the resonators and merge the components together as shown in figure 2.10 for optimal filter characteristics.

Figure 2.10: Multiple numbers of resonators merged together.

Increasing order N improves the selectivity in the passband. The disadvantages are the cost of size and higher insertion loss. To define the right filter order can be difficult to optimize, especially if the amount of tunable variables is too many. The time complexity to find the optimized solution will be much lower if the number of tunable variables is cut down. When using an electromagnetic simulator in CAD-programs it is probably even more important to not have more variables than necessary.

2.5 RF FRONT-END CHAPTER 2. THEORY

2.5 RF front-end [7]

The general purpose of a RF front-end is to transmit and receive signals between an antenna and an active radio integrated circuit component. Every wireless communication standard such as Bluetooth and UWB demands a custom designed front-end. But the general principle for a front-end looks the same. In figure 2.11 a typical RF front-end block diagram is shown, where every component has its assignments. The components in a front-end with their tasks are the following.

Switches. RF switches are devices which select a specific RF path between transceiver

and receiver mode, often controlled by a logical control.

PA. Power amplifier is required to amplify the low-level modulated signal. It is a very

high current device because it must provide significant gain. One problem issue is it demands a relatively large amount of output power.

LNA. Purpose of a Low noise amplifier is to amplify the receive signal path before

down conversion. Because of the incoming RF signal has to have a much lower noise characteristic, i.e. noise figure, compared to the down conversion components.

Filters. Used to reject harmonics, noise and spurs in the transmit path while

attenuating out of band in the receive path. This can be accomplished with lowpass, highpass or bandpass filters. A lowpass filter is more suitable at the output of the PA. So the unwelcome harmonics fully can be terminated away. Bandpass filter before the modulator output is the best way to reduce harmonics and spurs.

Balun. A balancing transformer to convert unbalanced input signal into balanced

output signal and vice versa. At the same time an impedance ratio conversion is performed. The Balun is also specified in terms of insertion loss, phase accuracy and together with the variables impedance ratio and balance. Insertion loss is a key figure for receiver noise. However it will be less important if the Balun is placed between the modulator and the PA. Phase inaccuracy results in losses and noise suppression which gives a less ideal performance.

3 Computer Aided Design tool

3.1 Microwave Office [2] [4]

Applied Wave Research Inc. (AWR) product, Microwave Office is a CAD-tool for RF and microwave circuit design. This simulation tool includes the essential technology: linear and nonlinear circuit simulators, electromagnetic (EM) analysis tools, layout-vs.-schematic checks, statistical design capability, and parametric cell libraries with built-in design rule checking.

The Microwave Office solution is designed around a single, object-oriented database that is inherently synchronized with schematic, simulation, and layout data. This integrated solution provides designers need to take an idea from concept through simulation and directly into physical implementation. On the electromagnetic front AWR has added a dynamic EM extraction that permits layout elements to be selected and targeted for simulation using any of the EM solvers supported through the EM Socket interface. This capability helps users to optimize circuit schematics while simulating with full wave EM-based accuracy. It also includes an optional time-domain engine to augment the harmonic balance and EM simulators. This time-time-domain engine is more than another SPICE simulator. AWR has integrated Synopsys HSpice simulation technology because it is very accurate for time-domain simulations.

AWR´s EMSight simulator uses a full-wave spectral-domain approach based on the method-of-moments and is multi-threaded to take advantage of multiprocessor computers. EMSight includes new block matrix solvers that enable users to simulate 5 times larger problems within the same amount of RAM. Advanced animation capabilities let users view true time and frequency domain current animations. Users can display the magnitude and direction of currents and electric fields on any layer within the structure.

AWR has created the EM Socket with open standard interface, which enables users to access a broad variety of EM simulators from leading vendors, without leaving the Microwave Office design environment. The underlying technology can interface with virtually any EM simulator and seamlessly integrates third party design tools into the Microwave Office design flow. The solution facilitates greater flexibility in the design methodology, while providing a common user interface. For an example in the 2004 Microwave Office release includes an interface to Sonnet Software’s EM product, Zeland Software’s IE3D, EM3DS from MEM Research and Analyst from Simulation Technology through the EM Socket interface. It is even possible to run the Momentum EM analysis tool in Microwave Office. [2]

3.2 ADVANCED DESIGN SYSTEM CHAPTER 3. COMPUTER AIDED DESIGN TOOL

3.2 Advanced Design System [5]

Agilent Technologies ADS is a very powerful product for general electronic design automation (EDA) software system. It is well suited for designers of cellular and portable phones, wireless networks and radar and satellite communications systems. For RF and microwave circuits ADS also have integrated schematics, synchronized layout and an extensive and accurate vendor part library as Microwave Office. Similar to Microwave Office’s Socket interface ADS's DynamicLink can interface with Cadence, Spice and other foundry partners.

Momentum is a 2.5-D planar EM simulator that enables RF and microwave designers to significantly expand the range and accuracy of their passive circuits and circuit models. Momentum works together with ADS to compute S-, Y-, and Z-parameters of general planar circuits. Microstrip, stripline, slotline, coplanar waveguide, and other circuit topologies can be analyzed quickly and accurately with Momentum. Vias that connect one layer to another can also be simulated; enabling design engineers to more fully and accurately simulate multi-layer RF/MMICs, PCBs etcetera.

4 Design Process

The multi-band front-end work has been done using Microwave Office together with ADS. The design methodology can roughly be divided into three steps. The first step is to create a schematic with ideal circuits and components for the verifying purpose. The next step is to change the ideal components to real ones. Step three is to examine the results and eventually come up with new solutions.

4.1 Design Overview

The multi-band concept is to gather or divide the number of information signals in a specific frequency spectrum. If a 1 GHz signal-band consisted of ten 100 MHz wide bands is desirable, a frequency demultiplexer can be placed in between. The same principle is valid if ten 100 MHz bands are gathered to one single wide-band equal to 1 GHz. For this purpose instead of using a demultiplexer a multiplexer is used.

4.1 DESIGN OVERVIEW CHAPTER 4. DESIGN PROCESS

This thesis work is restricted to the receiver mode, i.e., the frequency demultiplexer. This frequency demultiplexer will be placed between the receiver antenna and the LNA in the front-end. The assignment is to collect the frequency signal with help of the antenna and then divide the signal to an amount of several narrower signals, i.e., sub-bands. These sub-bands will in turn be the input RF signal to each front-end, see figure 4.2.

Figure 4.2: Block diagram of the concept.

The advantage with this technique is a multiple times of data rate speed can be achieved by placing the front-ends in parallel. If the principle works well it is possible to use several frequency demultiplexers in parallel to achieve even greater data rate, see figure 4.3.

4.1 DESIGN OVERVIEW CHAPTER 4. DESIGN PROCESS

Calculate the data rate is achieved with help of some assumptions and equations dealing with ideal components. Please see APPENDIX A for the detailed information where each sub-band has a 200 MHz bandwidth. The result from APPENDIX A gives a theoretical data rate of 584 Mbit/s, with 64-QAM modulation technique on each sub-band. Eleven sub-bands gives a total data rate speed of 6.42 GHz/s. Table 4.1 shows how the data rate increases using multiple frequency demultiplexers as in figure 4.3.

Table 4.1: The estimated performance. Number of

receivers bands Sub- bandwidth [GHz] Required Data rate [Gbit/s] Estimated power consumption[W]

1 11 2.4 6.42 3.5

3 33 7.2 19.2 10.5

5 55 12 32.1 17.5

10 110 24 64.2 35

20 220 48 128 70

The principle of the demultiplexer is based on the properties of quarter-wave transmission line at a certain frequency. When designing the demultiplexer, the quarter-wave properties are matched with help of an input matching network (IMN). Similarly, an output matching network (OMN) is required when designing a frequency multiplexer. The general purpose of matching networks is to achieve maximum power effects between the load and the source. This is achieved by using transmission lines.

4.2 DESIGN SPECIFICATION CHAPTER 4. DESIGN PROCESS

4.2 Design specification

The assignment is to create one functionally working frequency demultiplexer with 11 sub-bands where each sub-band has a bandwidth equivalent to 200 MHz. This can give a theoretical maximum data rate speed of 6.42 Gbit/s according to table 4.1. The relevant frequency spectrum is between 2 and 12 GHz. The easiest implementation of the frequency demultiplexer is first to design it in the 2-4.2 GHz region. It is also of interest to check the abilities in a higher frequency region close to 12 GHz. Figure 4.4 shows the frequency band covering several demultiplexers, where one band contains eleven sub-bands.

(a) (b)

Figure 4.4: (a) Sub-bands. (b) Band covering with demultiplexers.

When the RF signal has been divided by the demultiplexer, each sub-band signal will pass through a bandpassfilter before down conversion to get rid of distortions. The filters will be active in each sub-band region. Together with the open type convention in the IMN this is possible. As shown in figure 4.5 the IMN is designed after the quarter wave principle, where each microstrip line junction is given a specific length equal to quarter-wavelength.

4.2 DESIGN SPECIFICATION CHAPTER 4. DESIGN PROCESS

Figure 4.5: The general principle of the demultiplexer.

Using CAD-tools the characteristics of each microstrip line can be tuned to an optimized result. The T-junctions has the geometrical properties where the conductor width is equal to the length. To achieve the right microstrip length that corresponds to the quarter-wavelength, the estimated T-junction length is set equal to the width. See figure 4.6.

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

4.3 Design and simulation

The first step is to verify the concept by using the CAD-tools ADS and Microwave Office. All the components with its substrate definition are set to ideal. No possible power losses and distortions between the adjacent placed conductors exist, especially from the filter. When the parameters are set to ideal the conductors are treated as lossless, see table 4.2:

Table 4.2: Substrate properties

Dielectric thickness 0.254 [mm] Dielectric constant 3.48 [F/m]

Loss tangent of dielectric 0

Metal conductivity 5.88*10e11 [S/m]

Metal thickness 0.035 [mm]

The frequency band signal starts at 2 GHz and ends at 4.2 GHz. With the IMN the dividing of the band signal happens in each centre frequency of every sub-band. In table 4.3 the centre frequencies are shown with its quarter-wavelength respectively.

Table 4.3: The current lengths of the demultiplexer. Centre frequency [GHz] Quarter-wavelength [mm] Microstrip length [mm] 2.1 21.79 1.89 2.3 19.90 1.59 2.5 18.31 1.36 2.7 16.96 1.17 2.9 15.79 1.02 3.1 14.77 0.90 3.3 13.88 0.79 3.5 13.08 0.71 3.7 12.38 0.63 3.9 11.74 0.57 4.1 11.17 11.17

The microstrip lengths are also represented in table 4.3 following the concept in figure 4.5. The length of the microstrip can be calculated with ADS calculation tool, LinCalc or the Microwave Office, TXLine. An alternative option is to use equation 4.1:

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

chosen to the 50 Ohm. This gives a microstrip width, w, equal to 0.545 mm. Table 4.3 is only an approximate length value which must be tuned. This is done by using the ADS parameter sweep or Microwave Office tool. An example of the tuning-tool is shown in figure 4.7. The figure shows how the microstrip length can be tuned to get an optimized result at a specific frequency.

1.9 2.4 2.9 3.4 3.9 4.3 Frequency (GHz) divided -80 -60 -40 -20 0 R etu rn L o ss ( dB ) 2.1 GHz -72.36 dB DB(|S(1,1)|) divided

Figure 4.7: With the variable tuner-tool the microstrip length can be optimized. Another property beside the return loss is the impedance match in chosen frequency region. In figure 4.8 the optimal microstrip length gives impedance in port one equal to infinity. Confirmation that no incident impedance in port one occur. This shall happen simultaneously when the return loss is zero.

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS 1.9 2.4 2.9 3.4 3.9 4.3 Frequency (GHz) divided 0 10000 20000 30000 40000 50000 Im pe dance ( O hm ) 2.1 GHz 45460 |Z(1,1)| divided

Figure 4.8: Verifying the matching network with help of impedance check. After the confirmation, one Chebychev bandpass filter of order N equal to 20 is connected between the matching network and port 2, see figure 4.9. As previously mentioned, the centre frequency of filter one is equal to 2.1 GHz. The results are shown in figure 4.10 and figure 4.11, where figure 4.10 shows the filter response together with the IMN attached. Figure 4.11 shows S11 corresponding to the simulation in previous figure.

1 2 3 MTEE ID=TL1 W 1=0.545 mm W 2=0.545 mm W 3=0.545 mm NBPFC ID=NBPFC1 N=m FP1=2 GHz FP2=2.2 GHz AP=0.1 dB PORT P=2 Z=50 Ohm MLIN PORT P=1 Z=50 Ohm

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS 1.9 2.4 2.9 3.4 3.9 4.3 Frequency (GHz) divided -800 -600 -400 -200 0 F or w ar d gai n (d B ) DB(|S(2,1)|) divided

Figure 4.10: Filter response.

1.9 2.4 2.9 3.4 3.9 4.3 Frequency (GHz) divided -30 -20 -10 0 10 R eturn L os s (d B ) 2.1 GHz -16.53 dB DB(|S(1,1)|) divided Figure 4.11: Simulation of S11.

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

Same principle as in figure 4.7 and figure 4.8 is used to match the other microstrip lines at their respectively outputs. The result of the whole IMN in the frequency region: 2 to 4.2 GHz is represented in figure 4.12. This is the whole frequency demultiplexer with ideal filters.

Figure 4.12: The demultiplexer with 11 sub-bands.

Corresponding simulation of the frequency demultiplexer in the 2 to 4.2 GHz region is shown in figure 4.13 and figure 4.14.

1.9 2.4 2.9 3.4 3.9 4.3 Frequency (GHz) Demultiplexer -50 -40 -30 -20 -10 0 For w ar d g a in ( dB) DB(|S(2,1)|) Demux DB(|S(3,1)|) Demux DB(|S(4,1)|) Demux DB(|S(5,1)|) Demux DB(|S(6,1)|) Demux DB(|S(7,1)|) Demux DB(|S(8,1)|) Demux DB(|S(9,1)|) Demux DB(|S(10,1)|) Demux DB(|S(11,1)|) Demux DB(|S(12,1)|) Demux

Figure 4.13: Eleven 200 MHz bandpassfilters with N=20.

Further on in this thesis the forward gain in each sub-band will not be represented in the figures. It is presupposed S21 is in the first sub-band region and S31 covers the second and so on.

ML IN ID = T L 1 W = 0. 54 5 m m L=1 .352 m m ML IN ID = T L 2 W = 0. 5 45 m m L=1 .04 9 m m ML IN ID = T L 3 W = 0. 54 5 m m L=0 .813 6 m m ML IN ID = T L 4 W = 0. 54 5 m m L=0 .626 9 m m ML IN ID = T L5 W= 0 .5 4 5 m m L = 0. 47 62 m m ML IN ID = T L 6 W = 0 .54 5 m m L = 0. 35 3 m m ML IN ID = T L 7 W = 0 .54 5 m m L= 0. 25 08 m m ML IN ID = T L 8 W = 0. 54 5 m m L=0 .16 52 m m ML IN ID = T L 9 W = 0 .545 m m L= 0 .09 27 1 m m ML IN ID = T L 1 0 W = 0. 5 45 m m L=0 .03 091 m m 1 2 3 MT EE ID = T L1 1 W 1 = 0 .5 4 5 m m W 2 = 0 .5 4 5 m m W 3 = 0 .5 4 5 m m 1 2 3 MT E E ID =T L12 W 1 =0. 5 45 m m W 2 =0. 5 45 m m W 3 =0. 5 45 m m 1 2 3 MT E E ID =T L13 W 1 =0. 5 45 m m W 2 =0. 5 45 m m W 3 =0. 5 45 m m 1 2 3 MT E E ID =T L14 W 1 =0. 5 45 m m W 2 =0. 5 45 m m W 3 =0. 5 45 m m 1 2 3 MT E E ID =T L15 W 1 =0. 5 45 m m W 2 =0. 5 45 m m W 3 =0. 5 45 m m 1 2 3 MT E E ID =T L16 W 1 =0. 5 45 m m W 2 =0. 5 45 m m W 3 =0. 5 45 m m 1 2 3 MT E E ID =T L17 W 1 =0. 5 45 m m W 2 =0. 5 45 m m W 3 =0. 5 45 m m N BPF C ID =NBPFC1 N= 2 0 FP1=2 .2 G H z FP2=2 .4 G H z AP=1 dB N BPF C ID = N BPFC2 N= m FP1=2 .4 G H z FP2=2 .6 G H z AP=1 d B 1 2 3 MT E E ID =T L18 W 1 =0. 5 45 m m W 2 =0. 5 45 m m W 3 =0. 5 45 m m 1 2 3 MT E E ID =T L19 W 1 =0. 5 45 m m W 2 =0. 5 45 m m W 3 =0. 5 45 m m 1 2 3 MT E E ID =T L20 W 1 =0. 5 45 m m W 2 =0. 5 45 m m W 3 =0. 5 45 m m N BPF C ID =NBPFC3 N= m FP1=2 G H z FP2=2 .2 G H z AP=1 dB NBP F C ID = N BPF C 4 N=m FP1=2 .6 G H z FP2=2 .8 G H z AP=1 d B N BPF C ID =NBPFC5 N= m FP1=2 .8 G H z FP2=3 G H z AP=1 dB N BPF C ID =NBPFC6 N= m FP1=3 G H z FP2=3 .2 G H z AP=1 dB NBPFC ID =NBPFC7 N=m F P 1= 3 .2 G H z F P 2= 3 .4 G H z AP=1 dB NBPFC ID =NBPFC8 N=m F P 1= 3 .4 G H z F P 2= 3 .6 G H z AP=1 dB NBPFC ID =NBPFC9 N=m F P 1= 3 .6 G H z F P 2= 3 .8 G H z AP=1 dB N BPF C ID =NBPFC1 0 N= m FP1=3 .8 G H z FP2=4 G H z AP=1 dB ML IN ID = T L2 1 W = 0. 54 5 m m L = 1 0 .9 m m N BPF C ID =NBPFC1 1 N= m FP1=4 G H z FP2=4 .2 G H z AP=1 dB 1 2 3 MT E E ID =T L22 W 1 =0. 5 45 m m W 2 =0. 5 45 m m W 3 =0. 5 45 m m PO RT P=1 Z=50 O h m PO RT P=2 Z=50 O h m PO RT P=3 Z=50 O h m PO RT P=4 Z=50 O hm PORT P=5 Z=5 0 Ohm PORT P= 6 Z= 5 0 O h m PO RT P=7 Z=50 O h m PORT P= 8 Z= 5 0 O h m PORT P= 9 Z= 5 0 O h m POR T P=1 0 Z=5 0 Ohm PO RT P=11 Z=5 0 O h m PO RT P=12 Z=5 0 O h m

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS 1.9 2.4 2.9 3.4 3.9 4.3 Frequency (GHz) Demultiplexer -50 -40 -30 -20 -10 0 R e tu rn lo ss (d B ) DB(|S(1,1)|) Demux Figure 4.14: Simulation of S11.

For a better overview the return loss is represented together with the filter response in figure 4.15. 1.9 2.4 2.9 3.4 3.9 4.3 Frequency (GHz) Demultiplexer -50 -40 -30 -20 -10 0 Re tu rn L oss & For w ar d g a in ( dB )

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

The conclusion which can be drawn from here is the general principle of the multi-band front-end is working from a theoretical point of view. The IMN gives a satisfying result and works well together with the filters. A problem, despite the relative high order, N=20, is the return loss. Between each sub-band the S11 signal does not manage to recover from the filter effects. This happens when the filters are too close and affect each other in the guard-band. The guard-band is the band where every sub-band rejects out-bands signals, as return loss for an example.

4.3.1. Guard-band evaluation

The next step will be to examine the width of guard-bands with respect to different types of filter orders. The filters are still ideal and it is done with help of the already designed demultiplexer in previous work. The filter order variation is restricted to3≤ N ≤6. If the order N is higher than 6 the actual size of every filter will be physically too large.

4.3.1.1 N=3

With no guard-band between each filter the intersection point is equal to -5dB approximately, see figure 4.16. The simulation result shown in figure 4.17 with a guard width equal to 200 MHz gives a satisfying result. When S11 (red line) is approximately equal to 0 dB the adjacent filters crosses each other in -22 dB.

2.4

2.9

1.9

3.3

-50

0

Frequency (GHz)

R

et

urn

L

os

s & F

orw

ar

d g

ai

n (

dB

)

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS 2.4 2.9 1.9 3.3 -50 0 Frequency (GHz) R et urn L os s & F orw ar d ga in ( dB )

Figure 4.17: Guard-band equal to 200MHz.

2.4 2.9 1.9 3.3 -50 0 Frequency (GHz) R et urn L os s & F orw ar d ga in (dB )

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS 2.4 2.9 1.9 3.3 -50 0 Frequency (GHz) Ret urn L os s & F orw ar d ga in (d B )

m1

m1

freq=

dB(S(1,1))=-1.334

2.201GHz

Figure 4.19: Guard-band equal to 50 MHz.

As shown in figure 4.18 the filter crosses each other in approximately -22 dB. In figure 4.19 the guard-band is set to 50 MHz. The intersection point is around -10 dB, which is a result coherent with S11. The performance of S11 is directly dependent on the filters intersection point. The simulation with filter order N equal to three stops here. 4.3.1.2 N =4

As shown in figure 4.20 the performance of the demultiplexer is not satisfying. The passband width of each filter in the -50 dB region is relatively too large which affects the S11 negatively. The intersection point of the filters happens already in the -10 dB. An optimized guard-band for order N = 4 filter can be accomplished, see figure 4.21 and figure 4.22, respectively.

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

Figure 4.20: No guard-band.

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

Figure 4.22: Guard-band equal to 25 MHz.

According to figure 4.21 the filters intersection points is approximately between -12 to -10 dB. The variances are dependent on the filters insertion loss, which naturally intends to increase with the frequency.

As shown in figure 4.22 the tangential point in the guard-band has too small value, approximately -8 dB. Looking at the -50 dB limit the influence of each filter is too wide.

4.3.1.3 N =5

As shown in figure 4.23 S11 is still not close to 0 dB in the guard-band region, due to the low intersection point. Figure 4.24 shows a satisfying result where the intersection point is approximately -20 dB.

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

Figure 4.23: No guard-band.

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

Figure 4.25: Guard-band equal to 25 MHz.

As shown in figure 4.25 a smaller guard-band than 25 MHz is not recommended when the filter is of order five, since the intersection point is relatively too small.

4.3.1.4 N =6

The -50 dB bandwidth gets smaller with increasingly filter order. Verification can be seen comparing the figure 4.26 together with figure 4.23. Figure 4.27 shows a low intersection point, almost -30 dB. The value of the filter order N affects the whole performance of the frequency demultiplexer.

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

Figure 4.26: No guard-band.

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

4.3.2 Higher frequency simulations

Previous in this thesis report, simulations of a frequency demultiplexer have been done in the 2 to 4.2 GHz region. Further work is to test and verify the concept in a higher frequency region. Referring to figure 4.3, the next step is to check the band between 4.2-6.4 GHz. In similar way the next frequency region will be 6.4 to 8.6 GHz and so on, up to 12 GHz.

This can not be done using the current substrate, see table 4.2, because of the relatively large substrate height. The width of each T-junction is approximately 0.545 mm, with impedance matched to 50 Ohm. As illustrated before in figure 4.6, the width of the T-junction is equal to the length. With increased frequency the wavelengths turn to be smaller, accordingly the matching networks also. At a certain frequency the quarter-wave matching network length will be equal to the T-junctions width, i.e., 0.545 mm. This results in a negative microstrip length in the matching network which makes it impossible to divide the signal as desired. This happens already in the 4.3 GHz region, see APPENDIX B.

The solution to this problem is to exchange the substrate with a smaller substrate height. This will drastically decrease the conductor width to 0.104±0.002 mm. The properties of the current substrate are shown in table 4.4.

Table 4.4: The new substrate definition. Dielectric thickness 0.0508 [mm] Dielectric constant 3.4 [F/m]

Loss tangent of dielectric 0

Metal conductivity 5.88*10e7 [S/m]

Metal thickness 0.02 [mm]

Simulation with the new substrate is shown in figure 4.30. The result is almost identical compare to the simulation in figure 4.15 with a substrate height equal to 0.254 mm. The result from the guard-band investigation in the previous chapter (4.3.1) is still useable thanks to this result.

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS 1.9 2.4 2.9 3.4 3.9 4.3 Frequency (GHz) Demultiplexer -50 -40 -30 -20 -10 0 R et ur n Loss & Fo rw ar d gai n ( dB )

Figure 4.30: S11 represented with the filter response, N=20.

4.1 4.6 5.1 5.6 6.1 6.5 Frequency (GHz) Demultiplexer -50 -40 -30 -20 -10 0 R et u rn L os s & F orw ar d ga in (dB )

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS 6.3 6.8 7.3 7.8 8.3 8.7 Frequency (GHz) Demultiplexer -50 -40 -30 -20 -10 0 R e tu rn Lo ss & Fo rw ar d g ai n ( d B)

Figure 4.32: Simulation in the 6.4 to 8.6 GHz region, N=20.

As shown in figure 4.31 and figure 4.32 the frequency demultiplexer works well even at this high stage of frequency. The length of microstrips using the substrate properties shown in table 4.4 is shown in APPENDIX C.

4.3.3 Implementation of Coupled Line Filter

The next step in the design process is to exchange the ideal filter to real ones. The bandpass filters are of type Coupled Lines, the order N is equal to five. This number is chosen because it gives suitable characteristics for the implementation without being too physically large. Using the Microwave Offices Filters Synthesis Wizard alternative ADS DesignGuide for passive circuits, desired filters can be designed with right characteristics.

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

4.3.3.1 Filter design [3]

The first filter is designed in the frequency region 2-2.2 GHz. Using ADS DesignGuide for passive circuits the bandpass filter can easily be made.

Figure 4.33: ADS DesignGuide for passive circuits.

As shown in figure 4.33 the design of a Coupled Line Filter is made by following the instructions from the DesignGuide overview tab. It is a complete guide where every design, simulation and optimizations steps are structurally displayed.

For Microwave Office Filter Synthesis Wizard the same principle is used. Here you also can specify the filter parameters such as order, bandwidth and ripple, etc. As

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

(a) Selection menu of the filter type. (b) Parameter specification menu. Figure 4.34: Microwave Office Filter Wizard.

As shown in figure 4.35 and figure 4.36 the designed bandpass filter for 2 to 2.2 GHz is represented. Port P2 Num=2 Port P1 Num=1 MCFIL CLin5 L=22.747 mm S=0.00768 mm W=0.0842 mm Subst="MSub1" MCFIL CLin4 L=22.177 mm S=0.0667 mm W=0.115 mm Subst="MSub1" MCFIL CLin3 L=22.112 mm S=0.1 mm W=0.116 mm Subst="MSub1" MCFIL CLin2 L=22.177 mm S=0.0667 mm W=0.115 mm Subst="MSub1" MCFIL CLin1 L=22.747 mm S=0.00768 mm W=0.0842 mm Subst="MSub1"

Figure 4.35: ADS schematic of the bandpass filter active in the 2-2.2 GHz region.

MSUB Er=3.4 H=0.0508 mm T=0.02 mm Rho=2.44e-7 Tand=0 ErNom=3.4 Name=SUB1 MCFIL ID=TL1 W=W0 mm S=S0 mm L=L0 mm MSTEPX$ ID=MS1 Offset=-abs(W @2-W@1)/2 mm MCFIL ID=TL2 W =W 1 mm S=S1 mm L=L1 mm MSTEPX$ ID=MS2 Offset=-abs(W @2-W @1)/2 mm MCFIL ID=TL3 W=W2 mm S=S2 mm L=L2 mm MSTEPX$ ID=MS3 Offset=-abs(W @2-W@1)/2 mm MCFIL ID=TL4 W =W 2 mm S=S2 mm L=L2 mm MSTEPX$ ID=MS4 Offset=-abs(W@2-W @1)/2 mm MCFIL ID=TL5 W=W1 mm S=S1 mm L=L1 mm MSTEPX$ ID=MS5 Offset=-abs(W @2-W@1)/2 mm MCFIL ID=TL6 W =W 0 mm S=S0 mm L=L0 mm PORT P=1 Z=50 Ohm PORT P=2 Z=50 Ohm L0=22.51 L1=22.24 L2=22.21 S0=0.01336 S1=0.05234 S2=0.06132 W 0=0.0767 W 1=0.09237 W 2=0.09334

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

The simulation result from ADS and Microwave Office is respectively represented in figure 4.37 and figure 4.38. The schematics in ADS have been optimized but not the schematics in Microwave Office. Optimizing every filter is complex and time consuming. Optimizing the filters will only be done when the demultiplexer is working correctly in a specific frequency region from now.

2.0 2.2 2.4 2.6 2.8 1.8 3.0 -20 -10 -30 0

Frequency (GHz)

dB

(S(1

,1

))

dB

(S(2

,1

))

m1

m2

m3

_m1

freq=

dB(S(2,1))=-3.928

2.000GHz

m2

freq=

dB(S(2,1))=-1.016

2.100GHz

m3

freq=

dB(S(2,1))=-4.082

2.200GHz

Figure 4.37: Simulation in ADS.

Filter Response -50 -40 -30 -20 -10 0 2.2008 GHz -3.541 dB 1.9947 GHz -3.525 dB 2.1 GHz -0.5095 dB DB(|S(1,1)|) Filter 1 DB(|S(2,1)|) Filter 1

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS

The schematics and simulations for the other ten filters listed in table 4.3 are shown in APPENDIX D.

4.3.3.2 Exchange the filters

Each designed filter will now replace the ideal ones, one by one in the demultiplexer. As shown in figure 4.39 the simulation of the demultiplexer is done first by simulating the first sub-band, i.e., 2 to 2.2 GHz. Simulation number two is done with the next designed filter, i.e., 2.2 to 2.4 GHz. The result can be seen in figure 4.40 to figure 4.42.

2.0 2.5 3.0 3.5 4.0 1.5 4.5 -40 -30 -20 -10 -50 0

Frequency (GHz)

dB

(S

(1

,1

))

dB

(S

(2

,1

))

Figure 4.39: The first sub-band with designed bandpass filter.

2.0 2.5 3.0 3.5 4.0 1.5 4.5 -60 -40 -20 -80 0 Frequency (GHz) d B (S (1 ,1 )) d B (S (2 ,1 )) d B (S (3 ,1 ))

4.3 DESIGN AND SIMULATION CHAPTER 4. DESIGN PROCESS 2.0 2.5 3.0 3.5 4.0 1.5 4.5 -80 -60 -40 -20 -100 0 Frequency (GHz) dB (S (1, 1)) dB (S (2, 1)) dB (S (3, 1)) dB (S (4, 1))

Figure 4.41: The three first sub-bands with designed bandpass filters.

2.0 2.5 3.0 3.5 4.0 1.5 4.5 -100 -80 -60 -40 -20 -120 0 Frequency (GHz) d B (S (1, 1 )) d B (S (2, 1 )) d B (S (3, 1 )) d B (S (4, 1 )) d B (S (5, 1 ))

Figure 4.42: The four first sub-bands with designed bandpass filters.

The characteristic from the simulation with one filter attached to the demultiplexer (figure 4.39) gives a satisfying result. The filter response twice the frequency is normal dealing with bandpass filter. This happens when the wavelength is equal to the desired passband divided with an even factor, in this case by 2. Problem occurs with the simulation shown in figure 4.40. Undesirable resonances happen in the 2.7 to 2.8 GHz region, approximately 1.5 times the centre frequency. As shown in figure 4.41 and figure 4.42 this also happens when filter number three and four get attached to the circuit, although in more wide frequency region. The cause of the poor result is