1
Electrical Modeling and Optimization
of Multilayer Via Transitions for
Fully-Integrated Systems
by
Gaudencio Hernández Sosa, MSc
A thesis submitted in partial fulfillment of the
requirements for the degree of
DOCTOR OF PHILOSOPHY
Department of Electronics
National Institute for Astrophysics,
Optics and Electronics (INAOE)
November 2011
Tonantzintla, Puebla
Supervised by:
Prof. Reydezel Torres Torres, PhD
INAOE
©INAOE 2011
All rights reserved
The author hereby grants to INAOE permission to
reproduce and to distribute paper or electronic copies
ii
Abstract
Due to the advances in electronics technology, today’s modern life is inevitably
dependent on the complex convergence of diverse technologies into products designed
to provide solutions in many applications (e.g., computing, communication, biomedical, etc.). In this way, the concept of fully-integrated “convergent electronic systems” has
emerged. As promissory approach for implementing reliable fully-integrated systems,
System-On-Package (SOP) technology proposes the integration of active and passive
components into a single package. Nonetheless, SOP presents important challenges
over traditional packaging due to the presence of multiple ICs, embedded passives and
other components connected within the package. On the other hand, high-density
substrates with narrow traces widths and thousands of vias in order to achieve vertical
integration of different ICs working at higher speed is a key enabler technology for SOP.
Unfortunately, electrical discontinuities introduced by via transitions are the main limiting
factors for the ultimate performance of SOP interconnects. As a result, the development
of interconnects capable of guiding broadband digital signals without degrading the
signal integrity plays a key role in the SOP implementation.
Thus, due to the importance of vias for fully-integrated systems based on SOP, fast
and accurate techniques to model vias in multilayered high-density packages are
developed in this thesis. Furthermore, techniques to mitigate the signal integrity
problems related to vias are proposed. Among the contributions that the reader will find
in this document are novel full-wave/circuit modeling methodologies for vias in
high-density packages, return loss and crosstalk mitigation techniques which allow extending
the useful bandwidth of chip-to-chip links. Rigorous theoretical analyses were carried
out accompanied by exhaustive simulations in the time and frequency domains to
develop and demonstrate the models and methodologies proposed in this thesis.
Moreover, several prototypes were designed and fabricated as part of this project to
serve as test vehicles that allow the verification of the mitigation techniques in an
iv
Dedicated to
My wife Bris and daughter Diana whose lives have
changed my entire world… my parents and brothers for
vi
Acknowledgments
First and foremost, I would like to express my sincerely gratitude to my advisor, Dr.
Reydezel Torres, for his guidance and patience during the course of this work. I thank
him for recognizing my potential to achieve research success in the exciting field of
electromagnetics, and especially in the signal integrity area which is a new emerging
field in Mexico. His kind guidance, patience and motivation have contributed to develop
my research skills.
I would like to thank and acknowledge my Intel mentors Dr. Gerardo Romo and Dr.
Adan Sánchez during my internships at Intel Guadalajara. Those internship positions
held over the course of my graduate studies provided me with the unique opportunity to
acquire hands-on experience and the industry point of view on signal integrity research
problems. In particular I am fortunate to have had the patience and understanding of my
manager at Intel Dr. Horacio Visairo for the valuable support to finish this doctoral
thesis.
Finally, I would like to thank INAOE and CONACyT for the human, technical and
financial support to develop and finish this thesis.
Gaudencio Hernández-Sosa
viii
Table of Contents
LIST OF TABLES x
LIST OF FIGURES xi
1.
Fully-Integrated
Electronic
Systems:
Requirements
and
Challenges
1.1 Introduction ……….……… 11.2 Approaches for Fully-Integrated Systems ……….. 3
1.2.1 System-On-Chip (SOC) ...……….………... 3
1.2.2 System-In-Package (SIP) ………..……….………. 4
1.2.3 System-On-Package (SOP) ……… 6
1.3 Why SOP instead of SOC? ...………. 7
1.3.1 SOP Technical Advantages ……...……….…… 8
1.4 Requirements for SOP ……….. 10
1.5 Electrical Challenges in SOP realization ……….. 12
1.6 Objective and Thesis Organization ………..……….. 14
2.
Full-Wave and Circuit Modeling of Multilayer Via Transitions in
High-Density Packages
2.1 Introduction ……….………. 172.2 Segmented Modeling of Multilayer Via Transitions ..……….. 19
2.3 The Interior Problem ……… 24
2.3.1 Full-Wave Approach ...………..……….. 26
2.3.1.1 Single Via with Arbitrary Power/Ground Via Distribution ………. 30
2.3.1.2 Multiple Vias with Arbitrary Power/Ground Via Distribution ……. 34
2.3.1.3 Effect of the Finite Thickness of the Metal Layer …………... 40
2.3.2 Circuit Modeling Approach ………. 45
2.3.2.1 Single Via with Arbitrary Power/Ground Via Distribution ………... 46
ix
2.4 The Exterior Problem for Single and Multiple Traces ………. 57
2.5 Segmented Full-Wave/Circuit Model for Multilayer Via Transitions ….………... 58
2.6 Comparison between Full-Wave, Segmented Full-Wave and Segmented Circuit Modeling Approaches for Multilayer Via Transitions ………..….. 59
2.7 Conclusions ……….……….……… 62
3.
Analysis of Return Loss and Crosstalk in Multilayer Package
Interconnects
3.1 Introduction .……….……… 633.2 Return Loss and Crosstalk Analysis for Multiple Vias with Arbitrary Power/Ground Via Distribution ……….….………….…… 65
3.2.1 Conditions for Mitigating Return Loss and Crosstalk Effects ………….... 66
3.2.2 Methodology for Mitigating the Return Loss and Crosstalk ………... 71
3.2.3 Example and Discussion ………..……… 77
3.3 Return Loss Mitigation for Single Multilayer Via Transitions ……….. 84
3.3.1 Conditions for Mitigating the Return Loss ……..……….. 88
3.4 Conclusions ……….………. 92
4.
Minimizing the Return Loss in a Practical Chip-to-Chip Link
4.1 Introduction ……….………. 954.2 Description of Prototypes ……… 95
4.3 Designing of Chip-to-Chip Links for Minimum Return Loss ...……….. 98
4.4 S-parameters and TDR Measurements ………. 100
4.5 Conclusions ……….……….. 105
5.
Conclusions and Opportunities for Future Research
5.1 General Conclusions …….……….……….. 1075.2 Opportunities for Future Research .………..……… 109
Appendix: List of Publications ………. 111
x LIST OF TABLES
1. Comparative features between SOP and SOC ………..……….. 9
LIST OF FIGURES
1.1 Trend of electronic products toward the fully-integrated convergent systems ….… 2 1.2 SOC approach for fully-integrated systems …………...……….... 4
1.3 SIP approach integrating memory, memory controller, and a microprocessor into a single package/module using wirebonds ………...……… 5 1.4 Generic SOP approach ………..………….. 6
1.5 Fundamental basis of SOP with two parts: the digital CMOS IC regime, and system regime ……….……….. 10
2.1 Multilayer via transition interconnecting a surface trace and an inner trace in a high-density package/board ……….. 19
2.2 Multilayer via transitions interconnecting M coupled surface traces and M coupled inner traces in a high-density package/board ……….… 20
2.3 Decomposition of a multilayer via transition into exterior and interior problems using the equivalence principle (power/ground vias are not shown) ……….……. 22
2.4 a) Multiple vias inside a parallel-plate cavity and the corresponding dimensions and
b) model for N vias in Fig. 2.4a when the equivalent principle is applied ……….. 25
2.5 Signal via with Q power/ground vias disposed in an arbitrary spatial distribution
showing the corresponding dimensions: a) perspective view, and b) side view ... 31
2.6 a) 3D model of a single via with four P/G, and b) model implemented and simulated in HFSS ………....……….. 33
2.7 Real and imaginary parts of 1/Y11 for a single via with four P/G vias. Full-wave
versus Foldy-Lax (n=0) confrontation ………..……… 34
2.8 Real and imaginary parts of 1/Y12 for a single via with four P/G vias. Full-wave
versus Foldy-Lax (n=0) confrontation ………..……… 34
2.9 N signal vias with Q power/ground vias disposed in an arbitrary distribution showing the corresponding dimensions ……….. 35
xi
2.10 3D model of two signal vias with three P/G and the port corresponding assignation ………..……….. 38 2.11 Re(1/Y11) and Im(1/Y11) for two signal vias with three P/G vias ……….. 38 2.12 Re(1/Y12) and Im(1/Y12) for two signal vias with three P/G vias …………..…….... 39 2.13 Re(1/Y13) and Im(1/Y13) for two signal vias with three P/G vias ………..…... 39 2.14 Re(1/Y14) and Im(1/Y14) for two signal vias with three P/G vias ……….. 39
2.15 Hybrid model for including the pad and the antipad holes effects due to the finite thickness of the metal planes ……… 40
2.16 Magnitude of the S-parameters for a single via with four P/G vias considering effects due to the metal thickness ……….... 43
2.17 Phase of the S-parameters for a single via with four P/G vias considering effects due to the metal thickness ………. 43
2.18 TDR for single via with four P/G vias considering effects due to the metal thickness ………. 43
2.19 Magnitude of the S-parameters for two vias with three P/G vias considering the effects due to the finite metal thickness ………... 44
2.20 Phase of the S-parameters for two vias with three P/G vias considering effects due to the finite metal thickness ……… 44
2.21 TDR (at port 1) for two vias with three P/G vias considering effects due to the metal thickness ………... 44
2.22 TDR (at port 3 and 4) for two vias with three P/G vias considering effects due to the metal thickness ……….… 45
2.23 Equivalent circuit model for one signal via including ground power/ground vias arbitrarily disposed inside a parallel-plate cavity ……… 47
2.24 Inductive effects due to a signal via and Q power/ground vias ……… 48
2.25 a) Circuit model for one signal via sharing Q power/ground vias, and b)
combination of Q power/ground vias and one signal via into an equivalent
inductance ………... 49
2.26 Equivalent circuit model for multiple signal vias with an arbitrary distribution of ground power/ground vias inside a parallel-plate cavity ………... 51
xii 2.28 a) Circuit model for two signal vias sharing Q power/ground vias, and b)
combination of Q power/ground vias and two signal vias into equivalent
self-inductances and and mutual inductance ………... 53
2.29 a) Six signal vias with two power/ground vias, and b) the corresponding port assignation at the antipad holes ……… 56
2.30 Circuit model for six signal vias with two power/ground vias (equivalent mutual inductances are shown only for via 1 with other vias) ………... 56
2.31 Comparison between S-parameters predicted by HFSS and those predicted by the circuit model of Fig. 2.30 ……… 57
2.32 Comparison between S-parameters predicted by HFSS and those predicted by the circuit model of Fig. 2.30 ……… 57
2.33 a) Segmented full-wave model, and b) segmented circuit model for representing multilayer via transitions ………. 58
2.34 a) Segmented full-wave model for two coupled multilayer via transitions, and b) the corresponding segmented circuit model ……….. 60
2.35 Comparison of return loss (S11) in magnitude and phase predicted by the full model, the segmented full-wave model (hybrid model), and the segmented circuit model ………. 60
2.36 Comparison of insertion loss (S21) in magnitude and phase predicted by the full model, the segmented full-wave model (hybrid model) and the segmented circuit model ………. 61
2.37 Comparison of near-end crosstalk (S31) in magnitude and phase predicted by full model, the segmented full-wave model (hybrid model) and the segmented circuit model ………. 61
2.38 Comparison of far-end crosstalk (S41) in magnitude and phase predicted by full model, the segmented full-wave model (hybrid model) and the segmented circuit model ………. 61
3.1 Circuit model for an array of multiple signal vias and power/ground vias inside a parallel-plate cavity ………. 65
xiii
3.2 Circuit model for calculating the input impedance | at port 2p –1 showing
the corresponding simplification at the right side ………... 66
3.3 Structures used to investigate the impact of the number of power/ground vias, and the corresponding spatial distribution on : a) two P/G vias, b) three P/G vias, and c) four P/G vias ……… 73
3.4 versus and R for the test structures shown in: a) Fig. 3.3a, and b) in Fig. 3.3b, 3.3c ……….. 74
3.5 Structures used to investigate the impact of the number of power/ground vias and the corresponding spatial distribution in the value of : a) one P/G via, and b) two P/G vias ………. 75
3.6 versus for the test structures in Fig. 3.5a and 3.5b ………... 76
3.7 Flow chart for simultaneously minimizing the return-loss and crosstalk in multiple vias (antipad radius b and the number of the power/ground vias W are the varying parameters) ………..… 78
3.8 Six signal vias disposed in a 2×3 array with a) two GND vias, and b) seven GND vias ……… 79
3.9 and corresponding to the signal vias depicted in Fig. 3.8a ……… 80
3.10 Equivalent impedance for signal vias depicted in Fig. 3.8a ……….…... 80
3.11 and corresponding to signal vias depicted in Fig. 3.8b ………... 81
3.12 Equivalent impedance for signal vias depicted in Fig. 3.8b ………...…. 81
3.13 Magnitude of the return loss corresponding to the signal vias at the edge corners (vias 1, 2, 5 and 6) ……….. 82
3.14 Magnitude of the return loss corresponding to the signal vias at the center of the array (vias 3 and 4) ……….… 82
3.15 Magnitude of NEXT between vias 3 and 1 ……….. 83
3.16 TDR waveforms at signal via 1 for the arrays with two and with seven GND vias when no impedance matching is achieved ( ), and for impedance matching ( ) ………...……….…….… 83
3.17 Exterior problem and the corresponding circuit model ……….…. 85
xiv 3.19 Four-layer via transition and the corresponding dimensions and circuit model … 88
3.20 Comparison of the return-loss (S11) predicted by the circuit model and full-wave simulations for the via transition depicted in Fig. 3.19 ………..…… 88
3.21 Simplified circuit model partitioned into several T-networks for a multilayer via transition embedded between two traces ……… 89
3.22 Circuit model to calculate the partial input impedance at the k-th T-network …… 89
4.1 Cross section of a symmetrical chip-to-chip link including multilayer via transitions ……….… 96
4.2 A four-layer via transition and the corresponding dimensions and circuit model .. 97 4.3 A six-layer via transition and the corresponding dimensions and circuit model … 97
4.4 Chip-to-chip links including four-layer via transitions with probing pads for measurement purposes ………..… 97
4.5 Chip-to-chip links including six-layer via transitions with probing pads for measurement purposes ………. 98
4.6 and , versus de corresponding to the four-layer via transition ……...…. 99
4.7 , , and , , and versus de corresponding to the six-layer via
transition ………..………… 99
4.8 Equivalent impedance Ze versus de corresponding to four-layer and six-layer via
transitions ………...…… 100
4.9 Photograph of the fabricated chip-to-chip links using four-layer via transitions and the probes used to collect the measurements ……….………. 101
4.10 Photograph of the fabricated chip-to-chip links using six-layer via transitions and the probes used to collect the measurements ………..………… 101 4.11 De-embedded return-loss for the fabricated chip-to-chip links ……….. 102
4.12 Measured TDR profile for the three chip-to-chip links using four-layer via transitions ……….. 103
4.13 Operation regions (dominantly capacitive, inductive and matched) for chip-to-chip links using four-layer via transitions ………...………… 103
4.14 Measured TDR profile for the two chip-to-chip links using six-layer via transition ……….. 104
xv
4.15 Eye diagram at 20 Gbps corresponding to chip-to-chip links using a) non-matched six-layer via transitions and b) matched six-layer via transitions …………..…… 105
1
CHAPTER I
Fully-Integrated Electronic Systems:
Requirements and Challenges
1.1 Introduction
During the last four decades, electronics technology based on semiconductor
devices has revolutionized every aspect of human life. Thus, electronic products
have emerged in several fields such as automotive, computing,
telecommunications, aerospace, military, and medical care. In this regard, the
shrinking of transistor size has yield the integration of hundreds of millions of
transistors on a single chip, allowing a considerably high computing capacity. In
this way, a wide variety of electronic systems with powerful computing capabilities
has been developed.
Thus, due to the advances in electronics technology, today’s modern life is
inevitably dependent on the complex convergence of technologies into products
designed to provide solutions in many applications. In fact, there is an emerging trend within the electronics industry which is referred to as “convergent electronic systems” [1]. This trend is characterized by the convergence of computer,
communications, consumer, and biomedical functions (among others) into one
product. As shown in Fig. 1.1, the development of these systems is leading to
miniaturized and portable multi-function systems. Several examples of electronic
systems in the past and in the near future are depicted in Fig. 1.1. In fact,
fully-integrated systems are expected to be about a cubic centimeter in size with not
only computing and communication capabilities, but also with devices to sense and
digitize signals, to monitor, to control, and to communicate anyone anywhere.
Thus, when developing these systems, several technical aspects have been
2
Fig. 1.1 Trend of electronic products toward the fully-integrated convergent
systems [2].
On the other hand, the projected market for fully-integrated systems will
demand increased performance, smaller size, lower power and lower cost than that of today’s systems. Unfortunately, these demands cannot be fulfilled with
conventional packaging and interconnect technologies because of the
corresponding limitations in interconnect density, thermal dissipation, power
integrity, and signal integrity [3]. In consequence, advanced packaging
technologies are starting to be developed to allow three-dimensional (3D) integration of digital, analog, and mixed-signal chip technologies to build reliable systems [4].
In this chapter, an overview about fully-integrated electronic systems is
presented including the aspects related to the corresponding implementation.
Thus, by recognizing the requirements of these systems, the related challenges
are exposed. Visualizing these challenges allows identifying the opportunities for
research and development of packaging technologies, in particular of
system-on-package (SOP). As it will be shown later, vias are key structures in the
development of a reliable wiring technology for SOP and therefore; accurate
modeling of these structures at microwave frequencies is a vital part of the design
cycle. Furthermore, mitigation of the undesirable effects due to vias such as
3
1.2 Approaches for Fully-Integrated Systems
Before reviewing the system integration approaches for fully-integrated convergent
systems, it is important to mention that the ultimate goal of an entire electronic
system is to capture input signals, carrying out a processing, and providing output
signals. In general, the input and output signals are provided/received by the user
in response of a need. So, the necessary basic building blocks of an electronic
system are: power sources, integrated circuits (ICs), passive components,
interconnections, packages/substrates, heat removal structures, sensors, and
actuators. On the other hand, traditionally, the electronics industry has been guided
by the following system aspects: high-reliability, high-performance, low power
consumption, small size, and fast time to market. Thus, it is not a surprise that
fully-integrated convergent systems are driven by the same aspects. In this way,
the full integration of all building blocks of an electronic system is the goal of the
integration approaches. For this reason, in the following sections, the most
important concepts associated with system integration paradigms are presented
and discussed.
1.2.1 System-On-Chip (SOC)
SOC is a system integration approach that integrates a large number of transistors
as well as various mixed-signal active and passive components into a single planar
chip in order to build fully-integrated systems. Furthermore, SOC can be defined as the realization of an entire system’s functionality in a single, large IC with various
semiconductor technologies. Thus, SOC ICs must integrate digital, RF, analog,
and other functions in order to perform all the operations demanded to the
electronic system. Fig. 1.2 highlights the SOC integration approach. As expected,
SOC is guided by the technology roadmap for semiconductors, which is in turn based on Moore’s Law. If fully-integrated systems based on the SOC approach can
be cost-effectively designed and fabricated while offering high performance, high
4
Fig. 1.2 SOC approach for fully-integrated systems [2].
Unfortunately, at this time, there are many technical, financial, and strategy
challenges that obstruct the realization of the SOCs for all applications. One of the
most important problems to be faced by technologists developing SOC is the long
design time due to the complexity of integrating diverse semiconductor
technologies (such as bipolar, CMOS, Silicon-Germanium (SiGe), and
optoelectronics) for simultaneously performing several system functionalities (e.g.
digital, analog, mixed-signal processing, RF, optoelectronics, etc.) Another problem
is the high cost of integrating different semiconductor technologies into a single
chip with multiple voltage levels and dozens of mask steps. Thus, as can be
concluded, SOC presents major challenges for the realization of cost effectively
fully-integrated systems. Fortunately, it is not the only available alternative.
1.2.2 System-In-Package (SIP)
Due to the problems related to SOC, alternative system integration approaches
have been proposed. SIP is among these approaches, and can be defined as 3D
or vertical stacking of ICs into a single package or module, which contrasts the
planar nature of SOC. In fact, vertical stacking in SIP can be achieved using
5 (TSV), flip-chip, etc. [1]. Thus, SIP overcomes some of the above SOC limitations,
such as cost, design complexity and time to market.
SIP is often alternatively called 3D ICs or 3D stacked die/package. This
approach allows designers to stack multiple ICs at a low cost and into a SSF
package. In this case, the vertical die-to-die via pitch in a 3D stacked package is
very small; so, designers can arrange digital functional modules across multiple
dies. Moreover, stacked digital ICs have shorter wires than conventional
packaging, which can be translated into less wire delay and less power
consumption. As can be noticed, there are major benefits offered by the SIP
approach: simpler design and operation verification, a process with minimal mask
steps, minimal time-to-market, and minimal intellectual property (IP) issues. Thus,
SiP provides major opportunities in both miniaturization and integration for
advanced and portable electronic products.
Fig. 1.3 SIP approach integrating memory, memory controller, and a
microprocessor into a single package/module using wirebonds [2].
There are, however, some problems associated with the SIP approach, which
are in a certain way similar to SOC. An SIP, defined above as vertical stacking of
ICs, includes only the IC integration and hence addresses only about 10 to 20
percent of the system. Additionally, the integration of discrete passive elements
such as resistors, inductors and capacitors is not covered in current SIP
6
SIP approaches only integrate ICs, this may be considered as a “pseudo full-integration approach” because the ultimate goal of a fully-integrated system is not
achieved. In fact, some of the roadblocks for full system integration remain
unsolved such as in SOC. Even though limited benefits are offered by SIP,
currently, about fifty IC and packaging companies are designing and fabricating
products based on SIP modules. For example, using SIP, a 3G cell phone with
multiple semiconductor technology ICs, discrete passive components, and RF
radios can be developed in a few months. Integration of this functionality into an
SOC would be cost prohibitive, technically impossible, or beyond a viable market
window.
1.2.3 System-On-Package (SOP)
SOP is a system integration approach that seeks to integrate multiple system functions into one compact, low-cost, and high-performance packaged system [1] –
[4]. To implement SOP, a high-density substrate with embedded passives, active
devices, and integrated optics are required. A generic SOP is illustrated in Fig. 1.4, showing the similarity with SOC in the fact that it is a “single component” system.
Fig. 1.4 Generic SOP approach [1].
However, in contrast to SOC, SOP is a system-level package containing multiple
large ICs that integrate all the system functions. Furthermore, SOP can be seen as
7 single package, while maintaining a low-cost profile and SFF supporting mixed IC
technologies. SOP accomplishes this goal by achieving ultra-high wiring density,
using multiple layers, and applying a variety of embedded ultrathin-film component
integrations [5]. For this reason, a very important point to be mentioned here is the
fact that microvias are a key feature enabling SOP realization. The Association
Connecting Electronic Industries defines any via that is equal or less than 150 µm (6 mils) in diameter as a “microvia”. In this regard, considerable advances in
fabrication at large scale of buried and blind microvias within high-density
packages and boards have been reported to date.
Notice that the SOP paradigm proposes a unified chip-package view of the
design process. Thus, SOP is more than SIP, which contains only digital wiring to
interconnect ICs and requires two levels of packaging, including a motherboard.
Finally, it is important to clarify that when all the system components (e.g., passive
components, interconnections, and thermal structures), power sources, and
system boards are integrated into a single complete system, there is no difference
between SIP and SOP.
1.3 Why SOP instead of SOC?
As previously discussed, the SOC paradigm seeks to integrate numerous system
functions on one planar substrate, called the chip. This chip requires external
connections for power supplies and input/output signals. If this can be totally
implemented on chip, SOC offers the most compact, light-weight system that can
be massively produced. However, the electronics industry is figuring out that the
SOC paradigm has fundamental engineering and financial limits [6]. The technical
SOC problems include:
1) Difficulty of capturing the requirements and composing the
specifications of a mixed digital-analog-RF microwave system in a unified
and machine-processable form. In other words, there is a lack of experience
about integrating disparate semiconductor technologies (CMOS, SiGe,
8
deteriorate the overall performance; therefore, the SOC benefits can be
marginal.
2) Lack of design test tools for mixed-semiconductor technologies, for
trade-off evaluation, synthesis, and verification. Additionally, currently there
are limited model libraries that capture material-processing, circuit design,
and architectural information for modeling and design.
3) Lack of design tools for partitioning, floor planning, device placement,
routing, and other physical design tasks. This problem limits the integration
of heterogeneous technologies and components on the same IC.
4) Mechanical, thermal, and manufacturing-related problems. For
example, a high cost for large ICs is expected due to the high cost of wafer
fabrication and the low yield for technologies with minimum feature size of
less than 0.1-micron.
Due to these problems, fully-integrated systems based on the SOC approach
require long design and test cycles due to the large complexity of the system. In
fact, due to the long design cycle, an SOC system may face obsolescence due to a
rapid increase in product requirements in the market, and thus, it represents a
high-risk investment for the foundry and the product developer. However, SOP can
address the drawbacks of both SOC and SIP approaches, as well as those of
traditional packaging. In this way, SOP overcomes both the computing and
integration limitations of SOC, SIP, multichip modules (MCMs), and traditional
system packaging. Moreover, 3D SOP addresses the wire delay problem by
enabling the replacement of long, slow global interconnects with short, fast vertical
routes based on microvias.
1.3.1 SOP Technical Advantages
Recent advances in manufacturing technology will lead to low cost and
high-density packages and boards. Thus, SOP can provide strategic and tactical
advantages when using high-density packaging for high-performance digital
systems but also for fully-integrated systems. The advantages that SOP offers
9
Higher Memory bandwidth.
Transfer of interconnection functionality to a high-density substrate. Improved system integration (noise isolation and feasibility of integrating
passive components).
SOP has rapidly penetrated most major market segments: consumer
electronics, mobile, automotive, computing, networking, communications, medical
electronics, etc. The SOP benefits are for different market segments but they share
some common elements. Time to market, size, power requirements and cost have
resulted in the strongest initial penetration in mobile communications. A summary
about the pros and cons related to SOP and SOC are listed in Table I.
Technical Features
SOP SOC
PROS Different front-end technologies; GaAs, InP, Si, SiGe, etc.
Better yield at maturity (this depends upon complexity)
Different device generations Greater miniaturization
Re-use of common devices Improved performance
Reduced size vs. conventional
packaging
Lower cost in volume
Active & passive devices can be embedded
CAD systems automate interconnect design
Individual components can be upgraded Higher interconnect density
Better yield for smaller chip sets Higher reliability (not true for very large die)
Individual chips can be redesigned cheaper
Simple logistics
Noise & crosstalk can be better isolated Faster time to market
CONS
More complex assembly Difficult to change
More complex procurement & logistics Single source Power density for stacked die may be
too high
Product capabilities limited by chip technology selected
Design tools may not be adequate Yield limited in very complex, large
chips
10
The SOP paradigm creates synergy between CMOS and system integration
and this synergy overcomes both the fundamental and integration limitations of
SOC and SIP, which are limited by CMOS. In this regard, while Si technology is
great and matured for transistor density improvements from year to year, it is not
the optimal platform to integrate system components such as power sources,
thermal structures, packages, boards, and passives. These are highlighted in Fig.
1.5.
Fig. 1.5 Fundamental basis of SOP with two parts: the digital CMOS IC regime,
and system regime [2].
1.4 Requirements for SOP
The general requirements for SOP are many and the relative importance varies
with the application. However, these general requirements can be summarized in
the following areas [5]:
Electrical interconnects. SOP devices require high-density substrates with
narrow trace widths and thousands of microvias in order to achieve vertical
integration of different ICs and passive components. Thus, SOP provides a unique
opportunity for translating the global IC wiring to the package for enhanced signal
integrity performance. In this regard, recent developments in advanced build-up
11 dielectrics, conductor geometries with submicron precision, and low-cost
processes for multilayer stacked via interconnects [8]. Moreover, because SOP
devices require more ICs and other components embedded into the package, the
signal routing complexity is continuously increasing and therefore vertical
interconnects (e.g., vias and microvias) have become one of the most important
structure for reliable SOP implementations. Vias are extensively used for signal
interconnections when a signal flows between layers, or when a plane is connected
to another plane providing a low impedance current path. In this way, vias
decrease the routing complexity, provide a low impedance power/ground network
and allow 3D system integration. Nevertheless, two signaling features are very important for vertical interconnects based on vias and microvias: a) high-speed, for supporting high-data rates and high-frequency analog signals, and b) high-density, to allow 3D integration of the entire systems into a compact space. However, even though vias play a key role in the implementation of high-density chip-to-chip links,
these structures also introduce impedance mismatch, crosstalk and
electromagnetic interference (EMI), causing severe signal and power integrity
problems. Consequently, fast and accurate techniques to model hundreds of
thousands of vias in multilayered high-density packages are required for a
successful SOP implementation.
Chip-to-package interface: Current approaches for lead-free soldering
techniques present major challenges in both dispensing the underfill and ensuring
fatigue resistance as solder bumps shrink in height. Thus, recent SOP research
advances concerning the chip-to-package interface include the extension of solder
bumps to stretched-solder columns and improvements in underfill technology [9].
High-quality embedded passives: Designers have begun to use multilayer
ceramic and multilayer organic structures with liquid crystal polymer (LCP)
technology to embed passives elements in an efficient way in the package,
including high-Q inductors, capacitors, matching networks, low-pass and
12
Analog/RF components: The recent development of thin-film RF materials and
processes allows designers bringing the SOP concept into the RF area to meet the
rigorous needs of wireless communication. Thus, researchers are addressing
critical problems such as board-compatible embedded antennas and switches,
low-loss and low-cost boards, low-crosstalk embedded transmission lines, and
single-mode packages, as well as design rules for vertically integrated transceivers over a
wide frequency range [11].
Optical interconnects: A high-speed optical clock and data transport simplifies
the digital architecture by requiring fewer parallel transmission lines. Moreover,
optical links present low crosstalk and they are not susceptible to electromagnetic
interference (EMI) noise. Thus, researchers have developed a low-temperature
polymer process to fabricate and integrate optoelectronic components such as
micro lens arrays, lasers, waveguides, splitters, couplers, gratings, and photo
detectors on printed wiring boards for mixed-signal SOP applications [12].
Additionally to these requirements, small form factor (SFF) packages with high
functional density, high-frequency operation, and low cost must be designed. Also,
a package for SOP must include a good thermal dissipation for high reliability of
interconnects (e.g., avoiding cracks due to thermal and mechanical stress). Although SOP takes advantage of SOC and SIP approaches under the “More than Moore” approach [6], there are many challenges that impede the ultimate
realization of optimal SOP at the moment. In the next section, vias as key enabler
of high-density interconnects is discussed.
1.5 Electrical Challenges in SOP Realization
SOP presents new design complexities and challenges over traditional packaging
due to the presence of multiple ICs, embedded passives, and other components
connected into the package. In consequence, SOP is expected to present
interactions between different components of the system originating the need of
13 example, thermal dissipation imposes the most serious bottleneck for SOP
realization due to temperature gradients in multilayer substrates with different coefficient of thermal expansion (CTE) [15] – [16]. Because high-density packages
can use different dielectric materials with different CTE, mechanical stress on
copper-based interconnects may cause catastrophic electrical failure. Moreover,
due to the high-density integration in SOP devices, the interaction between the
thermal, mechanical, and electrical properties of the materials would significantly
impact the electrical properties of interconnects. Consequently, SOP would require
multi-physics analysis, design, and optimization in order to realize reliable
fully-integrated systems.
In this work, the electrical challenges related to SOP implementations are
considered. Among these challenges, there is a lack of package/chip co-design
methodologies and signal integrity/power integrity co-analysis at high-frequencies.
Besides, a good package design with accurate characterization and modeling must
be part of chip-package co-design methodology. Hence, excellent performance of
packages and interconnects is a crucial factor to achieve high-speed and wider
bandwidths. This is due to the fact that poor electrical performance of packages
and interconnects may degrade the entire system performance, especially when
digital signals processed by ICs present fast edge rates. Since faster edge rates of
digital signals imply higher frequency harmonics, the accurate characterization and
modeling of packages and interconnects at microwave frequencies is a critical
design step in the SOP design flow.
Since SOP devices require high-density substrates with thousands of vias,
multilayer via transitions need to be accurately modeled and optimized at
microwave frequencies. If multilayers via transitions are not well designed
(electrically optimized), when digital signals propagate throughout them crossing
power/ground planes with different voltages, the return current paths could behave
as electrical discontinuities causing severe signal degradation. In fact, the return
current for multilayer via transitions behaves very different from return path for
14
that the displacement current at the vias is a distributed property of the parallel
power/ground planes and this fact makes very hard to model the return path in
multilayer via transitions. Different modeling strategies can be found in the
literature to address the problem of modeling vias in multilayer packages. Most of
them are based on numerical methods, whose computational requirements rapidly
grow as the size and complexity of the interconnect structure increase. In contrast,
to overcome computational limitations, circuit models for vias have been proposed
with the purpose to perform large system-level signal integrity simulations.
As can be noticed, efficient modeling of vias in high-density packages becomes
essential in the SOP design process to obtain the best tradeoff between cost and
performance. This is a challenging task because of the large number of vias that
must be considered to model a realistic SOP package. As it will be explained, vias
and microvias can excite parallel-plate modes in the power/ground planes of a
multilayered package. These parasitic modes are the dominant source of parasitic
radiation, crosstalk, and impedance mismatch, causing signal integrity problems.
Since crosstalk and impedance mismatch due to vias are exacerbated as the
frequency operation of chip-to-chip links increases, fast and accurate modeling
techniques are needed for successfully designing SOP packages. Although SOP
can address the SOC and SIP limitations, much work is required to be done,
mainly in the signal and power integrity area.
1.6 Objective and Thesis Organization
As discussed throughout this chapter, vias and microvias play a key role in the
implementation of reliable 3D high-density interconnects for SOP devices.
Nevertheless, due to the signal and power integrity problems caused by them,
electrical modeling and optimization is mandatory in order to achieve low-power,
high-bandwidth links for SOP based fully-integrated systems. From an exhaustive
review in the literature, it was found that contributions for the electrical modeling
and optimization of multilayer via transitions in high-density packages and boards
15 currently used as the main tool for analyzing vias, the number of these structures
required for real SOP implementations (tens of thousands or even more) makes
unpractical using the full-wave approach. Furthermore, since fast to time to market
are required for SOP devices, the accurate characterization and modeling of vias in
high-density packages must be faster than full-wave approaches.
For this reason, the objective of this thesis is to develop a systematic electrical
modeling for multilayer via transitions at microwave frequencies as well as the
optimization of these structures from a point of view of signal integrity. In this way,
the contributions of this work will allow exploring the design space in a systematic
and fast manner, especially when the SOP system is highly constrained (in terms
of power, area and cost) and when a fast time to market is needed.
The full-wave modeling of vias based on the Foldy-Lax theory and the
equivalent principle is reviewed in Chapter 2 in order to develop a simplified
framework to calculate the Y-parameters of vias in a parallel cavity (taking into
account the distributed property of the return path). Additionally, a circuit model is
proposed and a general methodology to calculate the corresponding electrical
parameters is developed in order to analyze the reflection loss and crosstalk in
multilayer via transitions structures.
Based on the results presented in Chapter 2, in Chapter 3 reflection loss and
crosstalk analyses are performed in order to develop the corresponding mitigation
techniques. The benefits of the proposed solutions are demonstrated in Chapter 4
by using as test vehicles several chip-to-chip prototypes. As demonstrated by
S-parameter measurements corresponding to optimized and non-optimized
chip-to-chip links, the proposed mitigation techniques provide a great insight regarding
how to minimize the reflection and crosstalk in practical chip-to-chip buses and
how this minimization is translated into a great improvement of the signal-to-noise
(SNR) ratio of signals propagating in multilayer packages. Finally, Chapter 6
summarizes the contributions of this thesis. Also research directions are discussed
17
CHAPTER II
Full-Wave and Circuit Modeling of Multilayer
Via Transitions in High-Density Packages
2.1 Introduction
Vias play a key role in the reliable implementation of high-density interconnects for
3D system integration. Because vias are extensively used in high-speed links to
interconnect signal traces and power/ground planes at different layers, these
structures have become one of the most important features for the implementation
of high-speed and high-density interconnects for SOP devices. Furthermore, vias
reduce the signal routing complexity, provide a low impedance power/ground
network and allow 3D system integration. Consequently, the characterization,
modeling, and optimization of high-density, high-speed, and low-power
interconnects based on via structures represent a fundamental part of a packaging
design cycle. On the other hand, with the ever-rising clock rate in digital systems,
the spectral content of digital signals is becoming wider and wider and
interconnects must be designed to operate at higher frequencies than before.
Unfortunately, vias cause impedance mismatch, crosstalk, mode conversion, and
other signal integrity issues in a signal link path [17]. Vias passing through a
parallel power/ground plate pair could be affected by emissions originated at the
power distribution network, resulting in degraded signal quality. Similarly,
high-speed transient currents flowing along vias could also excite the multiple
parallel-plate environments existing in packages and PCBs, causing serious voltage
fluctuations in the power distribution network and introducing electromagnetic
interference (EMI) problems. Therefore, the electromagnetic modeling of vias in
parallel plates plays a critical role when analyzing signal integrity, power integrity,
and EMI for multilayer packages and boards in early design stages.
Basically, the modeling of vias may be accomplished through: a) carrying out
18
behavior of vias by means of equivalent circuit models. As it is well-known,
full-wave analysis of complex structures is often computationally prohibitive. Moreover,
full-wave simulations require considerable time and effort on the part of the
designer; this becomes a drawback when short time-to-market is needed.
Fortunately, equivalent circuit models can be developed using full-wave
simulations. Circuit models for vias are preferred for signal integrity analysis
because these can be easily integrated in SPICE-like environments with the
purpose of performing system-level signal integrity simulations. Furthermore,
physics-based circuit models relate geometry parameters with electrical properties.
This fact allows the fast analysis and design of optimized vias because circuit
models can be scaled according to physical geometry and material properties.
Thus, due to the massive number of vias in high density packages and boards,
physics-based circuit models for vias are preferred over full-wave-only approaches.
Notice, however, that although circuit models provide fast analysis and design,
full-wave analysis is often required in order to capture the details that are ignored by
the circuit modeling approach. This means that both types of analysis are
complimentary and can be of great help when properly applied.
In order to provide a clear and solid understanding of the electrical properties of
vias in compact space environments such as SOP, the full-wave as well as the
equivalent circuit modeling of vias in multilayered structures is presented in this
chapter. For this purpose, a review of the state-of-the-art regarding full-wave
numerical methods and equivalent circuit models for analyzing vias is carried out.
Additionally, in this chapter, a physics-based equivalent circuit model for coupled
vias in a parallel-plate environment is developed as well as the corresponding
closed-form equations for calculating the model parameters. Thus, there are two
main contributions presented in this chapter: 1) a new and systematic approach for
full-wave modeling of multilayer via transitions based on a simplification of the
Foldy-Lax Theory, and 2) the development of a physics-based circuit model which
allows representing multilayer via transitions in compact spaces when the distance
19
2.2 Segmented Modeling of Multilayer Via Transitions
Vias are mainly used to interconnect traces and devices located at different layers
within a package; therefore, these structures present a multilayer nature. Fig. 2.1
shows a multilayer via transition interconnecting two single traces, one located at
the surface of the package/board and another located at an inner layer level.
Similarly, Fig. 2.2 shows multilayer via transitions interconnecting M traces located
at the surface of the package/board and M inner traces. In both figures, the metal
layers are labeled as , where p = 1, 2,…, P, which is used to number the layer
level from the top of the structure. In this case, the metal planes can be either used
as power (PWR) or ground (GND) planes for the power delivery network (PDN).
Fig. 2.1 Multilayer via transition interconnecting a surface trace and an inner trace
in a high-density package/board.
Nowadays, a common practice for modeling multilayer via transitions similar to
Fig. 2.1 and 2.2 is using full-wave methods. Unfortunately, this approach becomes
unsuitable for analyses including many multilayer via transitions because of the
required computational resources and simulation time, which may be prohibitive at
early design stages. Furthermore, simultaneously minimizing undesired effects
such as return loss and crosstalk due to via transitions is very hard to achieve
20
dimensions of these structures with the corresponding electrical properties.
Nevertheless, in spite of these inconveniences, using full-wave solvers as the main
tool for analyzing effects due to multilayer via transitions in modern packaging is
still a common practice both in the academy and industry. In order to overcome the
drawbacks associated to model multilayer via transitions using full-wave
approaches, a segmented modeling methodology has been developed here. In this
section, this methodology is explained in detail.
Fig. 2.2 Multilayer via transitions interconnecting M coupled surface traces and M
coupled inner traces in a high-density package/board.
When a signal is transmitted from port 1 to port 2 through the multilayer via
transition depicted in Fig. 2.1, the signal firstly propagates in a quasi-transversal
electromagnetic (TEM) wave mode supported by the surface trace that presents a
characteristic impedance Zc. Moreover, if the metal layer M1 is included in the
layout of the package/board, the surface trace can be designed for operating in
coplanar waveguide-like (CPW) mode configuration. Then, this quasi-TEM wave
excites a vertical current in the signal via 1 and the quasi-TEM wave is converted
into a coaxial TEM wave at the antipad hole due to the finite thickness of the metal
layer M2. In most practical designs, the coaxial TEM mode assumption at the
antipad holes is valid even at very high frequencies. This assumption is based on
the fact that the cutoff wavelength of the higher-order modes for coaxial
21
n a b
c
2
(2.1)
where n is an integer number indicating the higher coaxial mode index, b is the
antipad radius and a is either the pad radius or the via radius when no pad is used.
For many practical PCBs, the term (b–a)is typically less than 0.5 mm. This implies
that the cutoff wavelength of the first high-order coaxial mode is about 1 mm, i.e.,
the corresponding cutoff frequency is about 300 GHz (beyond the frequency range
of interest in this project). Due to the fact that the dimensions in high-density
packages and boards are smaller than PCB dimensions, the cutoff frequency in
this case is even higher than 300 GHz. Therefore, all the higher order coaxial
modes at the antipad holes are neglected from the analysis presented in this work.
Other approaches have reported that the TEM approximation for a coaxial-type
structure provides accurate results, presenting additional justification of this assumption [19] – [20]. Continuing with the explanation about propagation through
the multilayer structure, once the signal reaches the via, part of the coaxial energy
traveling in TEM mode is converted into evanescent/propagating cylindrical waves
at the parallel-plate cavity formed by the metal layers M2 and M3. Since the edges
of the substrate between the metal layers is mostly terminated as open circuits,
propagating cylindrical waves reflect back and forth from these edges, leading to
the formation of standing waves at certain frequencies. At/near the resonance
frequencies of the planes, these waves cause significant coupling, resulting in EMI
within the cavity as well as in radiation from the edges of the substrate. The return
path for the current due to these cylindrical waves is provided by the ground/power
(PWR/GND) vias. Afterwards, the cylindrical waves in the cavity is converted again
into a coaxial TEM mode at the antipad hole in the metal layer M3 and similar
conversions occur at the different layers until the signal reaches the inner trace
located at the metal layer MP-1. Similarly to the quasi-TEM wave supported by the
surface trace, the inner trace supports a TEM wave if this trace is designed for
22
On the other hand, when many signals are transmitted throughout the
multilayer via transitions depicted in Fig. 2.2, the signals propagate as quasi-TEM
waves supported by the surface traces. Similarly to the single trace case, each
trace in Fig. 2.2 presents a characteristic impedance Zc. However, it is important to
point out that the characteristic impedance associated to a single trace having
neighbor traces is different from the characteristic impedance of a single isolated
trace. Thus, in this case the quasi-TEM waves are converted into coaxial TEM
waves at the antipad holes and then these are converted into cylindrical waves.
Again, these conversions occur until the signal reaches the inner traces. As can be
noticed, the propagation of signals throughout multilayer via transitions is a
consecutive conversion of coaxial TEM waves into cylindrical waves. Fortunately,
the presence of coaxial TEM waves at the antipad holes allows using the
equivalence principle in order to simplify the analysis of multilayer structures. The
equivalence principle has been widely used to model multilayer via transitions in
packages and boards [21] – [35]. Basically, this principle allows assuming that a
coaxial TEM wave at the antipad holes can be replaced with an equivalent
magnetic frill current that excites the cylindrical waves at the parallel-plane cavity.
Additionally, the metal layers (including the antipad hole) are replaced with a
perfect electric conductor (PEC) boundary. Thus, according to the equivalence
principle, the signal propagation through the multilayer via transitions depicted in
Fig. 2.1 and 2.2 can be decomposed into interior and exterior problems as shown
in Fig. 2.3.
Fig. 2.3 Decomposition of a multilayer via transition into exterior and interior
23 The two problems can be separately solved and combined in order to obtain the
response of the multilayer via transitions. Thus, multilayer via transitions can be
segmented into many smaller blocks, which is computationally more efficient than
solving the whole multilayer via transitions at once. It is important point out that
when the equivalence principle is applied, the metal planes are replaced by PEC
boundaries; therefore, the effects due to the finite thickness of the metal planes
must be separately included in the cascaded model in order to capture the
corresponding effects to preserve accuracy.
The exterior problem consists of traces that are bent into vias, whereas the
interior problem consists of magnetic sources between metal planes and vias
modeled as conducting cylindrical scattering surfaces. Extensive research effort
has been devoted to study the via-plate interactions in the literature. Thus, for vias
crossing a single plate (similar to the exterior problem), either full-wave methods or
quasi-static approaches are effective due to the localized field distribution near the vias [36] – [43]. Nevertheless, for vias crossing a parallel-plate cavity (i.e., for the
interior problem) formed by power/ground planes, the quasi-static approximation
and simple lumped circuit modeling are not suitable anymore since the
electromagnetic fields are not localized near vias. On the other hand, although
higher-order cylindrical waves are localized in the proximity of the vias as energy
stored in the electric and magnetic fields, there may be cylindrical modes
propagating over the entire parallel-plate cavity. Furthermore, since the separation
between vias is comparable to the wavelength corresponding to the highest
frequency of interest in high-speed digital circuits, lumped circuits fail to capture the
propagation nature of cylindrical waves into the cavity. This makes more difficult to
model vias crossing a plate pair than vias crossing a single plate. Nevertheless,
due to the small dimensions in high-density packages in SOP, the separation
between vias can be treated as electrically short even at frequencies of tens of
gigahertz and therefore can be modeled using circuit models without a significant
loss of accuracy. Hence, equivalent circuit models can be more efficient than
24
mentioned before, circuit models for vias are preferred for signal integrity analysis
because these can be easily implemented in SPICE-like circuit simulators.
A review of the state-of-the-art in full-wave numerical methods as well as
equivalent circuit models for modeling interior and exterior problems related to vias
are presented in order to provide a clear and solid understanding of the
corresponding electrical properties. This understanding will be fundamental in
Chapter III where techniques to minimize the reflection losses and crosstalk in
high-speed chip-to-chip links are developed. Also, the explanations presented in
this chapter will show the feasibility and accuracy of using circuit models instead of
full-wave 3D simulations to simulate vertical interconnect containing via structures.
2.3 The Interior Problem
Generally speaking, the interior problem consists of many vias inside a
parallel-plate pair formed by two metal planes as depicted in Fig. 2.4. This figure shows the
via radius a, the radius of the antipad on the power/ground planes b, the radius of
the pads c, and the relative dielectric constant and thickness of the layer between
the two P/G planes and h, respectively. The distance between vias p and q is
spq, whereas the metal layers present a thickness tc.
At the antipad hole of each via, there is a coaxial TEM mode that excites the via
and transversal magnetic (TM) cylindrical waves inside the cavity. However, the
coaxial TEM modes can be replaced by magnetic frill currents and (where i
= 1,2,…, N) located at the bottom and top of the antipads holes, which excite each
via in the cavity. In addition, the metal planes are replaced by PEC boundaries as
shown Fig. 2.4b. Notice that in the resulting model, the effect of the finite thickness
of the metal planes is not considered at this point, but it can be taken into account
after solving the model shown in Fig. 2.4b. Analytical methods for modeling vias
inside a parallel-plate cavity were firstly reported in [21] – [22]. In both cases, the
segmentation of the entire multilayer via transition was carried out based on the
equivalence principle. Thus, magnetic frill currents were assumed in the antipad
25 Unfortunately, methods in [21] – [22] were developed to model only two vias.
Moreover, those methods are restricted to an infinitely large plate pair, and
possible reflection of the propagating cylindrical waves due to the edges was not
considered in model (due to the analytical nature of the formulation).
a)
b)
Fig. 2.4 a) Multiple vias inside a parallel-plate cavity and the corresponding
dimensions and b) model for N vias in Fig. 2.4a when the equivalent principle is
applied.
Recently, however, an algorithm referred to as the Foldy–Lax multiple
scattering approach has been proposed to extend the analytical method to multiple vias in a parallel-plate cavity [23] – [29]. In fact, methods based on the Foldy–Lax
multiple scattering equations in vector spherical/cylindrical wave functions have been developed [33] – [44]. The Foldy–Lax method was originally used for multiple
26
each other’s far-field region [45]. On the other hand, a unified method has been
reported in [34], which takes into account the coupling of multiple vias with and
without pads in a much simpler and more comprehensive manner. As shown later
here, the preferred tool for analyzing vias in a parallel-plate cavity is the Foldy-Lax
approach which is faster than full-wave modeling using three-dimensional meshing
and solving the Maxwell equations for each node of the model mesh. Thus, the
multiple scattering methods can be regarded as efficient semi-analytical methods.
Although semi-analytical methods based on Foldy-Lax theory are more efficient
in time and computational resources than commercial full-wave solvers (such as
HFSS, CST, Sonnet, etc.), the required time to evaluate the semi-analytical
expressions for thousands of vias may become a drawback if high-order effects are
included in the model. Fortunately, analytical equations can be derived from the
Foldy-Lax approach bearing in mind that via structures in high-density packages
for SOP present smaller dimensions than those found on PCBs.
2.3.1 Full-Wave Approach
The vias in Fig. 2.4b can be associated to a multiport microwave network whose
admittance matrix can be obtained from the current distributions in the ports and
the voltage at the antipad holes. The detailed process to calculate the current
distribution inside vias and therefore the corresponding admittance parameters is
described in [25] – [26]. To simplify the analysis, every via in Fig. 2.4b can be
considered as a two-port network, where the upper ports ui can be grouped into a
single upper port u and the bottom ports bi into a single bottom port b. Thus, the
admittance matrix corresponding to the via array is given by:
bu bb
ub uu
Y Y
Y Y
Y (2.2)
where u and b stand for upper and bottom ports, and Y presents a size of 2N × 2N
matrix. Then, the Foldy–Lax multiple scattering method is applied to compute the