Heat Sink Antennas

mm-Wave Heatsink Antenna
Array Design for Low-Sidelobe

and Low-Temperature
Ashwath Balasubramanian
mm-Wave Heatsink Antenna Array
Design for Low-Sidelobe and
Low-Temperature
by
Ashwath Balasubramanian
to obtain the degree of Master of Science in Electrical Engineering
at the Delft University of Technology,
to be defended publicly on 08/10/2021 at 9:30 AM.
Student number: 5142679

Project duration: Jan 18, 2021 – Oct 8, 2021
Thesis committee: Prof. Dsc. A. Yarovoy, Professor, Chairman, MS3-EWI
Dr. ir. M. Mastrangeli, Assistant Professor, ECTM-EWI
Dr. Henk Van Zeijl, Project Coordinator, ECTM-EWI
Dr. ir. Y. Aslan, Assistant Professor, MS3-EWI
An electronic version of this thesis is available at http://repository.tudelft.nl/.

Abstract
The development of fifth generation (5G) technology is the beginning of a rapid tran-
sition in the world of wireless communications. Gbps data rates, minimal latency, and
good connectivity are the ultimate aims of 5G. To achieve them, 5G systems employ
the mm-Wave frequency band which has a frequency range above 24GHz and so al-
lows for higher bandwidth and gigabit wireless services. Because the size of the an-
tenna elements and their spacings are so small in mm-Wave, massive antenna arrays
in the base station may fit into a smaller area while yet providing high gain. How-
ever, the problem with mm-Wave integrated antenna array system is the excessive heat
generated per unit volume as there is not enough surface area to dissipate heat. Ther-
mal management of antenna system is very important as it affects the reliability and
life time of the electronic components in the system. Both active and passive cool-
ing strategies have been employed with passive cooling being the cost-effective and
energy-efficient solution. Heatsink antennas can enhance the cooling capacity by pro-
viding a dual functionality in terms of both thermal and electromagnetics. Tradition-
ally, most of the works on heatsink antennas are focused at lower frequencies and a few
at the mm-Wave frequency range. However, a proper mm-wave thermal modeling in
active integrated antennas is missing and there aren’t any research on the performance
of heatsink antennas in array designs.
This thesis work aims in designing and optimizing a heatsink antenna operating at
28 GHz to achieve a dual functionality. The second aim of the thesis is to develop
an appropriate thermal model for the designed antenna. Following the conduction
based simulations depending on assumed heat transfer coefficients, proper thermal
modeling with appropriate beamformer chip characteristics and a CFD-based natu-
ral convective simulation setup has been developed without the assumption of a heat
transfer coefficient. Optimal heatsink antenna dimensions are chosen based on the
electro-thermal performance. Then, the selected antenna has been used in 1D and 2D
arrays. Finally, a comparison study has been made with that of the conventional patch
antenna. The results obtained have shown that both 1D (1x8) and 2D (4x4) heatsink
antenna array can achieve a better heat dissipation by lowering the junction tempera-
ture of about 10−20◦C (for the investigated cases) with higher realized gain and similar
side lobe level compared to the respective patch antenna arrays. Furthermore, am-
plitude tapering of the heatsink antenna array achieved lower side lobe levels which
makes this heatsink antenna a low-sidelobe and low-temperature alternative for the
patch antenna array.
iii
Acknowledgements
I would want to use this opportunity to offer my heartfelt gratitude to everyone who
has supported and assisted me in the completion of my thesis work.
First and foremost, I would want to express my gratitude to Prof. DSc. Alexander
Yarovoy, my supervisor, for providing me with this fantastic opportunity to work on
my thesis at the Microwave Sensing, Signals, and Systems group. I enjoyed all of your
lectures, especially those on Antenna systems and Applied Electromagnetics. I thank
you for your support and guidance.
I would sincerely like to thank Dr. Yanki Aslan, my daily supervisor. I cannot express
my gratitude enough for the time you spent training me and clearing all my doubts.
Thank you for patiently listening to all of my queries and assisting me whenever I had
a problem. I appreciate all of your time and effort in supervising me over the last year.
In addition, I’d want to express my gratitude to all of my Microwave Sensing, Signals,

and Systems Group colleagues for some intelligent conversations on radars and anten-
nas.
Above all, I’d like to express my heartfelt gratitude to my Mother and friends, who have
always been there to listen to me and keep me encouraged during difficult times. With-
out you all, I would not have been able to accomplish this! I dedicate this thesis to my
father. I hope you’re having a lovely time in the afterlife. I love you!
v
Contents
Abstract iii
Acknowledgements v
List of Figures ix
List of Tables xiii
1 Introduction 1
1.1 Research Problem Description And Objectives . . . . . . . . . . . . . 1
1.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Research Approach and Novelty . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2 Heatsink Antenna Design - EM Aspects 17

2.1 mmWave Patch Antenna Design With Heatsink On Top . . . . . . . . . 17
2.1.1 Comparison With The Regular Patch Antenna . . . . . . . . . . 19
2.1.2 Analysis Of X-pol Directivity With Heatsink Dimension
21
2.1.3 Effect Of Removing Heatsink Base . . . . . . . . . . . . . . . . . 22
2.2 Effects Of Geometry And Orientation Of The Heatfins . . . . . . . 23
2.3 Pattern Directivity With Fin Dimensions . . . . . . . . . . . . . . . . 26
2.4 Infinite Array Analysis Using Periodic Boundary Condition . . 27
2.5 Conclusions - EM Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3 Heatsink Antenna Design - Thermal Aspects 35

3.1 Thermal Model Considerations. . . . . . . . . . . . . . . . . . . . . . . . . 35
3.1.1 Two-resistor Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.1.2 Properties Of The Electronic Chip Package . . . . . . . . . . 36
3.1.3 Thermal Model Of The Antenna . . . . . . . . . . . . . . . . . . . . 37
3.2 Simulation Approaches And Boundary Conditions . . . . . . . . . . . 38
3.2.1 Conduction Based Model . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.2 Airflow Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3 Results And Analysis-Conduction Based Model . . . . . . . . . . . . . 41
3.3.1 Analysis With Different h.t.c. . . . . . . . . . . . . . . . . . . . 42
3.3.2 Effects Of Chip’s Parameters On Junction Temperature . . 43
3.3.3 Radiative Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . 46
3.4 Results And Analysis-Airflow Model . . . . . . . . . . . . . . . . . . . . 47
3.5 Conclusions-Thermal Aspects . . . . . . . . . . . . . . . . . . . . . . . . . 48
vii
Contents viii
4 Heatsink Antenna - Electro-Thermal Co-Design 51

4.1 Parametric Study With Fin Dimensions . . . . . . . . . . . . . . . . . . 51
4.2 Finite Array Implementation . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.1 Linear Array Implementation . . . . . . . . . . . . . . . . . . . . . 55
4.2.2 Planar Array Implementation . . . . . . . . . . . . . . . . . . . . . 60
4.3 Conclusion-Electro-Thermal Aspects . . . . . . . . . . . . . . . . . . . . 62
5 Conclusions And Future Work 65
5.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
A Extended Heatsink Geometry 71
B Calculations For The Thermal Model 72
C Extending Ground plane For Heat Transfer 75
D Additional Parametric Study 76
E Amplitude And Phase Shift Calculations 77
List of Figures
1.1 Fabricated arrays with the Mugen MAX cooling CPU heatsink module
(right - single pol., left - dual pol. array). (from [5]) . . . . . . . . . . . . . 2
1.2 Fabricated FR4 square patch antenna with heatsink. (from [6]) . . . . . 3
1.3 Fabricated heatsink antennas with fins parallel to the nonradiating edges
(left) and fins parallel to radiating edges (right) (from [7]) . . . . . . . . . 3
1.4 Topology of the heat-sink antenna, cross section view. 1) Heatsink an-
tenna. 2) Heat spreader. 3) Short circuit. 4) Heat exchange surface and
radiating surface. 5) Ground plane. 6) Power amplifier circuit-ground
plane. 7) Transistor device. 8) Printed circuit board. 9) Surrounding
medium. (from [9]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 (Top) Al fractal,n=3 before mounting on substrate. (Bottom left) square
patch. (Bottom right) Al fractal, n=3,mounted on FR4. (from [12]) . . . . 6
1.6 Cross-sectional view of notch antenna array with radome and printed cir-
cuit board. (from [14]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.7 Massive element antenna array with FSS structure (from [15]) . . . . . . 7
1.8 Thermal deisgn parameters in a unit cell, a)Top view b) bottom view.
(from [16]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.9 Patch Dimension effect on Tmax with t g nd = 2mm for varying L max (from
[16]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.10 Illustrations of the proposed stamped metal AiP: (a) The cross-sectional
view of the entire package. (b) The exploded view of the radiating ele-
ment. (from [18]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.11 Temperature distribution of the 8x8 rectangular array a)Case C-1, b) Case
C-2, C)case C-3. (from [2]) . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.12 Multilayer heatsink structure (Left). Individual antenna element (Right).
(from [20]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.13 Triangular fin heat sink with 120◦ obtuse angle (Top). Triangluar fin heatsink
having alternating layout with 120◦ obtuse angle. (from [ [21]]) . . . . . 12
1.14 (a) Top view of the vivaldi antenna array with two fins, (b) bottom view .
(from [22]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1 CST Model of Coaxial fed patch antenna . . . . . . . . . . . . . . . . . . . 18

2.2 CST model of patch antenna with heatsink on top . . . . . . . . . . . . . . 18
2.3 Return loss S11 (dB),(a)Patch antenna,(b)Heatsink antenna . . . . . . . . 19
2.4 Co-pol pattern directivity (dBi), (a) Patch Antenna, (b) Heatsink Antenna. 20
2.5 Cross-pol pattern directivity (dBi), (a) Patch Antenna, (b) Heatsink An-
tenna. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.6 Maximum cross pol Directivity (dBi) vs Frequency . . . . . . . . . . . . . 21
2.7 Maximum Cross-Pol Directivity (dBi) vs Frequency . . . . . . . . . . . . . 22
2.8 CST Model of Coaxial fed patch antenna with heat fins on top . . . . . . 22
ix
List of Figures x
2.9 Return loss S11(dB) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.10 Pattern Directivity for the patch with heat fins (dBi) (a) co-pol, (b) cross-
pol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.11 Different heatsink antenna geometry and orientations and its Radiation
pattern directivity (dBi), (a,b)Rectangular fins parallel to non-radiating
edges (PNRE), (c,d) Rectangular fins parallel to radiating edge (PRE), (e,f )
Radial pin fins all around the top of the patch, (g,h) Radial pin fins (Edges),
(i,j) Radial pin fins (near the center) . . . . . . . . . . . . . . . . . . . . . . 25
2.12 Pattern Directivity vs fin height . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.13 Pattern Directivity vs Number of fins . . . . . . . . . . . . . . . . . . . . . . 27
2.14 Pattern Directivity vs Fin thickness . . . . . . . . . . . . . . . . . . . . . . . 27
2.15 CST Model of heatfins PNRE array with periodic boundary condition along
x axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.16 Return loss S11 (dB) of infinitely extended heatfins PNRE array,(a)along
x,(b)along y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.17 Coaxial feed with a ring inside the air region . . . . . . . . . . . . . . . . . 29
2.18 Return loss (dB) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.19 Return loss S11 (dB) of infinitely extended heatfins PNRE array with scan
angle,(a)15◦ ,(b) 30◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.20 Embedded element pattern of the heat fins (PNRE) antenna with 15◦ scan
angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.21 Embedded element pattern of the modified heat fins (PNRE) antenna
(shrunk patch) with 15◦ scan angle . . . . . . . . . . . . . . . . . . . . . . . 31
2.22 CST Model of Radial fins (centre) antenna with Periodic Boundary con-
dition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.23 Embedded element pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1 Two resistor model and its 3D space representation taken from [cite] . . 36
3.2 4-channel mm-Wave NXP analog beamformer chip from [25] . . . . . . 37
3.3 Thermal model of 4-element Linear patch antenna . . . . . . . . . . . . . 37
3.4 Airflow model of 4-element Linear heatsink antenna . . . . . . . . . . . . 40
3.5 Temperature plots for 4-element linear array (a) Patch ,(b)fins (PNRE)
heatsink antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.6 Temperature plots for 4-element 2D array (a) Patch ,(b)Radial fins (cen-
ter) heatsink antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.7 4-element linear array,(a)T J vs h.t.c,(b)T A vs h.t.c . . . . . . . . . . . . . . 42
3.8 4-element 2D array,(a)T J vs h.t.c,(b)T A vs h.t.c . . . . . . . . . . . . . . . . 42
3.9 Junction Temperature vs h.t.c at the antenna side (4-element linear array) 43
3.11 Temperature plots for 4-element 2D array (a) Patch ,(b) Radial fins (cen-
ter) heatsink antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.13 Effects of thermal resistance, (a) θ JC vs T J ,(b)θ J B vs T J . . . . . . . . . . . 45
3.14 Temperature plot with radiative heat transfer,(a)Patch,(b)Heatsink . . . . 46
List of Figures xi
3.15 Temperature plot with radiative heat transfer using Anodized Aluminium 47
3.16 Natural convection of 4-element Linear patch antenna . . . . . . . . . . . 47
3.17 Natural convection of 4-element Linear heatfins (PNRE) antenna . . . . 47
4.1 Conduction Based model-Fin height vs (a)Co-pol,(b)Cross-pol . . . . . . 52

4.2 Air flow Model - Fin height vs (a)Co-pol,(b)Cross-pol . . . . . . . . . . . . 52
4.3 Conduction Based model-Fin thickness vs (a)Co-pol,(b)Cross-pol . . . . 53
4.4 Airflow model-Fin thickness vs (a)Co-pol,(b)Cross-pol . . . . . . . . . . . 53
4.5 Conduction Based model-Number of fins vs (a)Co-pol,(b)Cross-pol . . . 54
4.6 Airflow model-Number of fins vs (a)Co-pol,(b)Cross-pol . . . . . . . . . . 54
4.7 Array Topology, (a) 1 x 8 Patch, (b) 1 x 8 heatsink antenna . . . . . . . . . 56
4.8 Return loss S11 (dB) of 1 x 8 array,(a)Patch,(b)Heatsink antenna . . . . . 56
4.9 Realized gain (dB), (a) Patch, (b) heatsink antenna . . . . . . . . . . . . . 56
4.10 Farfield realized gain (φ = 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.11 Farfield realized gain (φ = 0) for different scan angles . . . . . . . . . . . . 57
4.13 Return loss S11 (dB) of 1 x 8 heterogenous heatsink antenna array . . . . 58
4.14 Realized gain (dB), (a) Patch, (b) heatsink antenna . . . . . . . . . . . . . 58
4.15 Farfield realized gain (φ = 90) . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.16 Farfield realized gain (φ = 90) for different scan angles . . . . . . . . . . . 59
4.17 Array topology, (a) 4 x 4 patch, (b) 4 x 4 heatsink . . . . . . . . . . . . . . . 60
4.19 Realized Gain (dB), (a) Patch, (b) heatsink antenna . . . . . . . . . . . . . 60
4.20 Farfield realized gain (φ = 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.21 Farfield realized gain (φ = 90) . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.22 Conduction based model - Temperature plots, 2D array (a)Patch, (b) Heatsink 62
4.23 2D Heatsink antenna array with scan angle (θs , φs ) = 30◦ , 45◦ . . . . . . . 62
A.1 Extended heatsink geometry (along y) . . . . . . . . . . . . . . . . . . . . . 71

A.2 3D Pattern Directivity (dBi) . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
B.1 Cross-section of the heatsink antenna array . . . . . . . . . . . . . . . . . 72

B.2 Velocity slice - heatsink antenna array . . . . . . . . . . . . . . . . . . . . . 73
C.1 Extended groundplane (a) Patch, (b) Heatsink antenna . . . . . . . . . . . 75

List of Tables
1.1 Comparison of different heatsink antenna topologies . . . . . . . . . . . . 13
2.1 Comparison of EM performance . . . . . . . . . . . . . . . . . . . . . . . . 20

2.2 Comparison of EM performance for different heatsink geometries and
orientations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3 EM performance comparison of the selected heatsink orientations . . . 26
3.1 Chip parameters from [5] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.2 Geometry properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.3 Comparison of the two heat transfer simulation models . . . . . . . . . . 48
4.1 Comparison of joint EM-thermal performance . . . . . . . . . . . . . . . 57

4.2 Comparison of joint EM-Thermal performance . . . . . . . . . . . . . . . 59
4.3 Comparison of joint EM-Thermal performance (4 x 4) array . . . . . . . 61
B.1 Thermophysical properties of Air . . . . . . . . . . . . . . . . . . . . . . . . 73
D.1 Parametric study data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
xiii
1
Introduction
This chapter deals with the background on the research project and the general out-
line of the thesis work. Motivation to do research, problem description, and research
objectives will be discussed in the first section. Next to this, a literature survey on ex-
isting state-of-art techniques for thermal management in antennas will be discussed.
In the third section, the chosen research approach and novelty of the thesis will be
elucidated. Finally, in the end, the thesis outline will be provided.
1.1. Research Problem Description And Objectives

The rapid transformation in the world of wireless communications starts with the de-
velopment of fifth generation (5G) technology. The ultimate goals of 5G are to achieve
Gbps data rates, low latency, and excellent connectivity. To achieve this, 5G systems
use mmWave frequency band with frequency range beyond 24GHz as this provides
larger bandwidth and gigabit wireless services. The advantage of mm-Wave is that the
size of the antenna elements and their spacings are so small that large antenna arrays
in the base station can fit into a smaller area providing high gain [1].
However, the small aperture size of antenna array will not provide sufficient surface
area for the heat to dissipate. In mm-Wave, the excessive heat generation per unit
volume within the integrated antenna array system is a serious problem as it nega-
tively affects the reliability and stability of the electronic components in the RF cir-
cuitry. For example, at mm-Wave the power generated is about 2W in just 4cm 2 area
of the chip.[2]
It can be stated through the Arrhenius equation that for every 10◦C rise in Integrated
Circuit (IC) junction temperature, the average lifetime of the chip gets reduced by 50%
[3]. This shows that cooling such systems is very much necessary to have a better life-
time of the electronic components. Cooling of such systems involve either active cool-
ing, passive cooling or both (hybrid). Active cooling depends on external devices such
as fans, pumped liquids to enhance the heat transfer employing forced convection
whereas passive cooling achieves a natural convective heat transfer utilizing heatsinks,
heat spreaders, and thermal interface materials. The use of external fans or liquids for
1
1.1. Research Problem Description And Objectives 2
actively cooling the electronics requires more energy and cost. On the other hand, with
passive thermal management no external energy is required and it is a cost-effective
and energy-efficient solution.
The cooling issues at the Ka frequency band where the current 5G mmWave operates
has been dealt by employing active cooling techniques using fans or liquid pumps [4].
For 5G base station antennas, reducing the cost, maintenance requirements, and en-
ergy consumption while increasing the reliability is crucial. Therefore, fully passive
cooling solutions are desirable. Currently, to achieve this, large-sized metallic heat
sinks are attached to the beamformer chips at the back of the antenna. An example
of such heatsink is shown in Figure 1.1. The disadvantage of such large heatsink is that
they are bulky and occupy a lot of space.
Figure 1.1: Fabricated arrays with the Mugen MAX cooling CPU heatsink module (right - single pol., left
- dual pol. array). (from [5])
However, without attaching a large heatsink at the bottom surface of the chip, heat will
not dissipate and forced convection techniques have to be used for the cooling of such
structures. Therefore, low-profile antenna passive thermal management is a major re-
search problem. To achieve this, heatsink antennas can be used which are 3D radiating
structures that create additional heat flow paths from the beamformer chips to the sur-
rounding air. They have the potential of providing dual-functionality in terms of both
thermal and electromagnetics. This way, the requirements on the size and weight of
the bulky metallic heatsinks attached to the chips can be relaxed.
The main objectives of the thesis are as follows.
1. To design and optimize such heatsink antenna centered at 28 GHz operating fre-
quency for dual functionality and to analyze its performance both in terms of electro-
magnetics and thermal aspects.
2. To design a proper thermal model with the beamformer chip and evaluate the ther-
mal performance of the designed mmWave antenna.
3. To realize such elements in 1D/2D arrays and investigate their scanning range abili-
ties and mutual coupling effects as well as the thermal performance of the arrays.
1.2. Literature Review 3
4. To compare the electromagnetic and thermal performance of the designed heatsink

antenna with the conventional microstrip patch antenna and with the relevant alter-
natives proposed in the literature.
1.2. Literature Review

A comprehensive list of articles on heatsink antennas has been presented here where
some of the literature use a regular antenna like patch and a heatsink attached on it for
dissipating the heat, while other use different approaches so that the antenna (or its
layout in an array arrangement) itself acts as a heatsink.
Antenna+Heatsink:
L.Covert and J.Lin (2006) in [6] suggest a new type of antenna which considers the
dual functionality aspects both electromagnetic and thermal. A simple edge-fed patch
antenna centered around 2.4 GHz has been considered for the design and the effect of
placing a heatsink on the patch was analysed. The input reflection coefficient results
show that there is a decrease in the resonant frequency to 2.26 GHz(Measured), 2.23
GHz(Simulated) due to the increase in effective length of the patch. The radiation ef-
ficiency of the antenna has been greatly improved due to the less confinement of the
electric fields in the substrate. This resulted in a higher peak realized gain for the patch
with heatsink.
Figure 1.2: Fabricated FR4 square patch antenna with heatsink. (from [6])
L.Covert et al.(2007) in [7] proposed the use of extruded pin heatsinks having two dif-
ferent orthogonal orientations. Two 5.8 GHz heatsink antennas were designed, one
with the extruded fins parallel to the non-radiating structure (Fins PNRE) and the other
is with extruded fins parallel (Fins PRE) to radiating edges.
Figure 1.3: Fabricated heatsink antennas with fins parallel to the nonradiating edges (left) and fins
parallel to radiating edges (right) (from [7])
A silver conductive epoxy layer is used between the patch and heatsink to improve
the thermal conductivity. Tuning stub is used only in the heatsink antenna case for
impedance matching, where its length and placement are determined experimentally.
Results show that there is increase in radiation efficiencies for both the cases (Fins PRE)
and (Fins PNRE) compared to the normal patch. The return loss (S11) dB shows a huge
improvement in the 10 dB bandwidth from 3.1% (basic patch) to 17.6% (PNRE) and
11.9% (PRE). However a better comparison is made measuring the Q of the antenna
with the same matching network for all cases. The basic patch had higher Q value
compared to the heatsink cases. Comparing the peak directivity and gain for the three
cases, it was found that (Fins PNRE) case had better directivity than the patch since
the fins are parallel to the current flow on the patch whereas in Fins PNRE it was the
opposite case (perpendicular to current flow) which resulted in a lower directivity than
the patch for the Fins PRE case.
Later, L.Covert et al. in [8] demonstrate the performance of the heatsink antenna by
integrating the above heatsink antenna with Power Amplifier with a thermal via be-
tween power amplifier and the heatsink. It is stated that a high thermal conductivity
material can be used for the via to dissipate heat better. This has an impact on the
resonant frequency, which can be easily compensated by adjusting the patch length.
As the via is at the center of the patch, there is no negative impact on the shape of the
radiation pattern. The performance of RF devices is strongly influenced by the temper-
ature increase due to the heat dissipation. Therefore, an active heatsink antenna for RF
transmitter has been proposed by A.Alnukari et al. in [9]. Higher thermal conductivity
and bigger surface area helps for better heat dissipation in the transmitter.
Figure 1.4: Topology of the heat-sink antenna, cross section view. 1) Heatsink antenna. 2) Heat
spreader. 3) Short circuit. 4) Heat exchange surface and radiating surface. 5) Ground plane. 6) Power
amplifier circuit-ground plane. 7) Transistor device. 8) Printed circuit board. 9) Surrounding medium.
(from [9])
Figure 1.4 gives the topology of the proposed heatsink antenna where FR4 epoxy sub-
strate with 1.58mm thickness and copper with metal thickness of 35µm has been used.
Wire fed Patch antenna designed for 2 GHz with diameter 1mm is used to feed the
patch with 67mm x 54mm dimensions. The conducting cylinder connecting the power
amplifier circuit to the patch antenna is the heat spreader which is effective in creat-
ing a thermal path for heat dissipation. About 10 mm diameter is used for the heat
spreader as increasing the heat spreader diameter larger than the wavelength might
affect the electromagnetic performance. Thermal simulations using Finite element

method results matched with the Infrared thermography measurement. Relation be-
tween the copper thickness and thermal resistance was shown where as the copper
thickness increases, the thermal resistance drops down.
A.Alnukari et al. extended their work in [10] and [11] where [10] present an ap-
proach of cooling the antenna using a distilled water channel and [11] use a sapphire
layer for cooling. Results show that the distilled water which is about 20◦C removed
about 0.73W of thermal power/cm 2 of the antenna whereas it was 0.28W for the case
without distilled water. Experimental verification was done by infrared imaging which
proved the simulation correct. With the sapphire layer having permittivity of 8.9, the
heat removed was about 1W /cm 2 of the antenna while the thermal resistance was
about 5.5◦C /W . Temperature distribution results show that the antenna maximum
temperature reduced from 140◦C to 80◦C with the effect of sapphire layer.
In [12], J.Casonova et al.(2008) suggest an increase in surface area of the heatsink to

improve the heat dissipation thereby preventing burnout. A 3D fractal structure has
been proposed with a cube of side w and for every iteration there is an addition of
cubes of side w/2n with centre on each corner and these fractals are placed on top of
the patch. From n=0 to n=4, the fractals were designed and a comparison is made in
terms of thermal and electromagnetic properties. The design from [7] (Fins PNRE and
Fins PRE) were also considered for comparison. With the operating frequency set to
5.8 GHz, electromagnetic properties were compared and it is seen that all the heatsink
antennas had increased radiation efficiency compared to the patch. The resonant fe-
quency shift is observed for each cases and with n=1 it increases up to 6.75 GHz due
to the decrease in effective length. Similar radiation patterns were observed in all the
cases but the beamwidth in E-plane varies. The orientation of the main beam showed
variation for different cases. Also, a change in linear polarization (patch) to the ellipti-
cal polarization for fractals was observed.
At steady state, thermal resistance was calculated by simulation under natural or forced
convection scenario. This was done to characterize the heatsink performance. It is ob-
served that except for n=0, all the other fractal heatsinks showed less thermal resistance
than the extruded pin-fin heatsinks. This is due to the fact that as n increases, the sur-
face area of the heatsink increases resulting in better heat conduction. However there
is also some increase in cross-sectional area of the antenna as n increases. So, with a
constant volumetric flow rate, the air speed increases resulting in better convection.
Figure 1.5: (Top) Al fractal,n=3 before mounting on substrate. (Bottom left) square patch. (Bottom
right) Al fractal, n=3,mounted on FR4. (from [12])
Antenna as a Heatsink / Layout optimization as a heatsink:
A Solder based self assembly (SBSA) technique has been advocated in [13] as a ro-
bust and novel technique for integration of 3D antennas. Both at low (2.4 GHz) fre-
quency and high (23.8 GHz) frequency the two 3D antennas, 3D cube antenna and 3D
TSP Antenna were compared with the conventional patch antenna with heatsink on
top. The three-dimensional constructs were made up of an array of three-dimensional
cubes.On a scale of 300 µm in three dimensions, a 3x3 array was developed on the
patch substrate. The patch is 2.87mm X 2.87mm in size. On the substrate, truncated
square pyramids with a 150m square base and a 75 m height were built in a 3 X 3 array.
In self-assembly experiments, truncated structures yielded the highest yield, so they
were used to model the 3D antenna. Results show that the 3D cube antenna and the
3D truncated structured antenna had better gain and radiation efficiency.
A new approach of using stepped notch antenna array as heat fins is presented in [14].
A radome is placed above the notch antenna array for trapping the air such that it re-
duces the thermal resistance. A 26 x 11 array for Ku band has been designed where the
actual radiating elements are 24 x 8. Thermal analysis show that the array is competent
of giving up to 0.4◦C /W thermal resistance.
Figure 1.6: Cross-sectional view of notch antenna array with radome and printed circuit board. (from
[14])
In recent times, different approaches for better heat transfer enhancement in the an-
tennas at the mm-Wave frequency range have been introduced.
Massive MIMO antennas are required to achieve large capacity communications for
5G. This leads to large amount of heat dissipation for which the antennas would re-
quire the placement of heatsink to the beamformer chips at the back of the antenna.
NEC [15] suggest an alternative approach for dissipating heat efficiently which re-
duces the size of heatsink and thus reducing the entire antenna array size. Two new
features of this technology were to exploit the antenna itself to dissipate heat along
with the heatsink and providing a thermal pathway for better dissipation of heat. For
the dissipation of heat from the antenna element, the surface area has been expanded
and Frequency Selective Structure (FSS) has been placed to act as additional heat dis-
sipation fins. FSS structure has been selected as it allows only specific frequencies to
pass through thereby making them electromagnetically transparent. For better heat
transfer, a split ring structure is considered reducing the total size of the array and
vertical orientation of the antennas with checker plate style pattern arrangement for
enabling good air flow. The structure is shown in Figure 1.7.
Figure 1.7: Massive element antenna array with FSS structure (from [15])
By varying the thickness of the ground plane and the aperture size of the antenna ar-
ray, Y. Aslan et al. in [16] studied the thermal and electromagnetic properties of the
antenna array at mm-Wave frequency. Figure 1.8 gives the thermal model design. For
the EM model, Patch antenna is used which is fed by a 50 ohm microstrip feeeding line
on the chip substrate and it is connected with the patch through a vertical feeding pin.
For the thermal model, two -resistor model [17] has been used where it consists of
three nodes (Junction, Case and Board) connected by two thermal resistors (junction-
to-case and junction-to-board). Here the heat flow is either through the Case or Board
node and not through the sides.
Figure 1.8: Thermal deisgn parameters in a unit cell, a)Top view b) bottom view. (from [16])
An 8x8 array has been designed and the thermal simulation results show that as the
ground plane thickness is varied from 0.05mm to 2mm, the maximum junction tem-
perature (Tmax ) decreases. It is observed that the effective thermal conductivities in
lateral and vertical directions increased with the increase in ground plane thickness.
Effect of varying the aperture size has also been studied. However, for thinner ground
plane the variation of the aperture size (L max ) doesn’t significantly reduce the temper-
ature. EM performance has been computed for the best and worst cooling scenario,
where it is found that patch dimensions need to be changed to compensate the effect
of mutual coupling. Figure 1.9 gives the table showing the effect of patch dimension
on the junction temperature.
Figure 1.9: Patch Dimension effect on Tmax with t g nd = 2mm for varying L max (from [16])
A new approach for better heat dissipation for phased array applications has been
studied by Junho park et al. in [18], where metal stamping technology is used for
the Antenna in package (AiP) concept. This approach proposes the use of RFIC and
antennas to be mounted on the same side. Planar Inverted-F antenna (PIFA) is consid-
ered for their large aperture size, wide feeding structure and easy realization through
the stamping process.
Figure 1.10: Illustrations of the proposed stamped metal AiP: (a) The cross-sectional view of the entire
package. (b) The exploded view of the radiating element. (from [18])
1 x 8 patch antenna has been designed through stamped metal AiP at the centre fre-
quency of 28 GHz. Conduction based analysis with 10W /m 2 K heat transfer coefficient
show that maximum temperature was 11◦C less than that of the maximum tempera-
ture in a conventional patch array topology.
The problem of thermal control in 5G base station antenna arrays at mm-waves has
been solved by Y.Aslan et al. in [2]. For the first time, it has been proposed to use
passively cooled planar AESAs with fanless CPU coolers as an energy-efficient and rel-
atively low-cost solution with a significant beamforming flexibility. Sparsity has been
implemented into antenna layouts by increasing inter-element spacing in order to im-
prove cooling efficiency. A number of sparse array topologies have been proposed and
their thermal and electromagnetic efficiency evaluated.
With the 8x8 normal square array, a parametric analysis of the maximum junction tem-
perature was conducted, varying the heat dissipated P h (W ) , h.t.c of the passive cooler
h hs (W /m 2 K ), and the spacing between the elements (d e ). Figure 1.11 shows the tem-
perature distribution of the regular 8x8 square array layout for a varying inter-element
spacing. Here, Case C-1,C-2,C-3 has spacing 0.5λ, 1λ, 2λ respectively with P h = 2W
and h hs = 1000W /m 2 K
Figure 1.11: Temperature distribution of the 8x8 rectangular array a)Case C-1, b) Case C-2, C)case C-3.
(from [2])
However, with the increase in spacing, undesirable grating lobes occur which is a trade-
off. So, alternative array layouts like sparse array, spiral array, thinned array, sparse cir-
cular ring array and Heatsink antenna array topologies were considered and its thermal
performance has been simulated.
Comparisons of the junction temperature with the alternating layout has been made
where it is found that these alternating layouts achieved better Electromagnetic per-
formance while maintaining its cooling. The average distance between the elements
was found to be the most important factor in lowering the temperature for the various
sparse topologies. The sunflower topology was found to be the best in this regard, as it
allows for the largest inter-element spacing of all investigated topologies while main-
taining low side lobe levels.
[19] suggest a large-scale (256-element) aperiodic array topology for future 5G/6G
communications. The addition of more elements resulted in an increase in gain and
the reduction of side lobe level. This increase in elements also decreased the heat pro-
duced per element for the same EIRP performance. Through conduction-based ther-
mal simulations, it was shown that large sparse arrays can help in reducing the heat
generated per unit volume thereby getting rid of the heatsink attached at the back of
the array.
[20] present a 60 GHz Heatsink antenna array fabricated by 3D printing where the
fin heatsink is soldered on a low-temperature cofired ceramic (LTCC) substrate with a
chip mounted at the bottom. An open-ended waveguide and two vertical metallic fins
make up the antenna part, which is shaped like a horn. The proposed heatsink antenna
array, according to simulation and calculation, not only performs well in terms of radi-
ation but also has a strong heat dissipation capability. Figure 1.12 gives the multilayer
heatsink antenna array structure and the single antenna element
Figure 1.12: Multilayer heatsink structure (Left). Individual antenna element (Right). (from [20])
The (4x4) antenna array’s feeding network is made up of two-layer substrate integrated
waveguides (SIWs) with various metallic vias, and the 3-D printed heatsink is installed
on an LTCC substrate. Metal layers are denoted by M1–M11 in the LTCC substrate,
while ceramic layers are denoted by D1–D10 from the bottom up. A step-profiled trans-
verse electromagnetic mode (TEM) horn structure is constructed as the antenna por-
tion by combining the fins with an open ended waveguide, as shown in Figure 1.12.
Parametric studies of pattern gain with fin width, fin height and spacing between the
fins has been conducted to determine the best suited dimensions. Furthermore Elec-
trothermal analysis has been made now with the parametric study of the hotspot tem-
perature. The dimensions of the fins are chosen based on better EM and thermal per-
formance. The hotspot temperature dropped as the fin height and width increased,
indicating that the surface area expanded, resulting in better heat dissipation.
The chip’s dissipated power is 0.2W . Since the heatsink structure is exposed to the air,
the convection boundary in thermal simulation is set to h = 10W /(m 2 K ) at 300K am-
bient temperature. The entire structure’s hotspot temperatures are shown to be 313.1
and 353.2K , respectively. The fin-shaped heatsink on the LTCC package achieves a
temperature reduction of 40.1 K. The thick metallic vias of SIWs in the LTCC substrate
act as thermal spreaders by default, providing an efficient means of transferring the
chip’s dissipated thermal energy to the heatsink.
The structure of the heatsink plays an important role in thermal management where
[21] compare different layout of the heatsink with its thermal performance. About 5
different heatsink layouts were compared, Rectangular fins, Triangular fins with right
angle , obtuse angle and alternating triangular fins with right angle, obtuse angle. It is
seen that the effective Heat Transfer Coefficient (HTC) has been improved for all the
cases relative to the rectangular fin case. The alternating triangular layout with obtuse
angle had the maximum increase in HTC about 12.1% relative to the rectangular fin
case. CFD analysis show that in the alternating trainagular fin case, increase in the
angle θ reduced the temperarture at the base and the fins. Optimal angle was cho-
sen to be 150◦ since a further increase in the angle θ after 150◦ resulted in reduced fin
efficiency.
Figure 1.13: Triangular fin heat sink with 120◦ obtuse angle (Top). Triangluar fin heatsink having
alternating layout with 120◦ obtuse angle. (from [ [21]])
In [22], a vivaldi antenna array has been designed to show better heat dissipation and
electromagnetic radiation capabilities. Here, the vivaldi antenna is manufactured with
the help of 3D printing technology where Substrate integrated Waveguide (SIW) with
a longitudinal slot is used as the feed for the antenna. A 1x4 Vivaldi antenna array de-
signed achieved a bandwith of 28.72% (21.63-28.81 GHz) and a gain range of 10.5-12.5
dBi. In addition to this a fin shaped structure is added in the array as shown in Figure
1.14 where the metallic fins are parallel to the E-field of the pattern which according to
image theorem can enhance the directivity. Also this fin shaped structure can enhance
heat dissipation as it provides more surface area.
Figure 1.14: (a) Top view of the vivaldi antenna array with two fins, (b) bottom view . (from [22])
The thermal validation is done using COMSOL where a silicon chip with 0.5 W dis-
sipated power is condiered as the heat source and a heat transfer coefficient of h =
10W /(m 2 .K ) at 300K ambient temperarure is considered. About 18K reduction is
hotspot temperature is observed.
Most research papers use a conduction-based simulation with 10W /(m 2 K ) h.t.c in the
solid-air area, whereas a few papers include experimental results and others simply ex-
plain the theory.
-10dB
Heat dissipating
Antenna Frequency Number of Peak Gain Impedance
Type Ref structure /
Topology (GHz) Elements (dB) Bandwidth
technique
(GHz)
Heatsink-loading
0.04 Fin-shaped
[6] microstrip 2.4 1 4.12
(1.76%) heatsink
patch
Heatsink-loading
1.02 Fin-shaped
[7] microstrip 5.8 1 6.25
(17.6%) heatsink
patch
Air-dielectric
0.07 Patch with
Antenna+Heatsink [9] microstrip 2 1 5
(3.5%) heat spreader
patch
Patch with heat
Air-dielectric
0.03 spreader &
[10] microstrip 2 1 3.85
(1.5%) distilled water
patch
channel.
Air-dielectric Patch with heat
0.03
[11] microstrip 2 1 4.06 spreader &
(1.5%)
patch sapphire layer
3-D fractal heatsink
0.04 3-D fractal
[12] -loading microstrip 1.43 1 7.15
(2.80%) heatsink
patch
3D cube 1 SBSA 3-D
[13] 23.78 1 4.89
antenna (4.2%) structures
Stepped notch Stepped notch
5
Antenna as a heatsink [14] antenna 13.2 24 x 11 - array with
(37.88%)
array radome
Planar inverted 0.7
[18] 28.5 1x8 13.78 Stamped metal AiP
-F antenna (2.45%)
Fin-shaped
Open-ended
7 heatsink with
[20] waveguides in 60 4x4 18.8
(11.7%) numerous
heatsink
thermal vias
Vivaldi
3-D Antenna
7.2
[22] Vivaldi 25 1x4 15.5 and
(28.8%)
Antenna Fin-shaped
heatsink
Thick ground plane
Layout optimization Microstrip patch 1.7
[16] 28 8x8 22.68 and
as a heatsink array (6%)
large aperture size
Aperiodic arrays
[2] Sunflower array 28.5 64 24.6 - (Larger inter
element spacing)
Table 1.1: Comparison of different heatsink antenna topologies
Based on the comparison of different heatsink antennas made in [20],[22] a compar-

ison table of different antenna topologies and the heat dissipating technique is shown
in Table 1.1.
Conclusions:
To summarize, research works on heatsink antennas fell into three categories, antenna
as a heatsink, heatsink+antenna and layout optimization as a heatsink. Both active
and passive cooling approaches for thermal management in antennas has been used
in various studies. Previous research on heatsink antennas has mostly focused on low
frequencies, with only a few studies at mm-Wave frequencies examining the effects of
heatsinks on standard patch antennas and other approaches to antenna thermal man-
1.3. Research Approach and Novelty 14
agement. Return loss, pattern gain, directivity, and radiation efficiency are all shown
as EM results for these heatsink antennas. A conduction based simulation with the
assumption of heat transfer coefficient is used for thermal simulations in most of the
research works. However, there are few things that hasn’t been explored in the litera-
ture.
• The proper mm-wave thermal modeling in active integrated antennas is missing
• The usefulness and performance trade-offs of heatsink+antenna concepts in mm-

wave bands are not known.
• The optimality of the proposed heatsink antennas in terms of joint electromagnetic

and thermal performance is not clear
• There is lack of studies in performance investigation of heatsink antennas in array

configurations
• The effects of heatsink antennas on cross-polarization and scanning abilities of the

heatsink antenna array has not been investigated
• At mm-Wave frequencies, the interference problem due to high sidelobes in heatsink

antenna array has not yet been addressed.
• In the thermal simulations of the preceding studies involving passive cooling, a conduction-
based heat transfer or natural convection heat transfer with an h.t.c value assump-
tion was used. Nonetheless, thermal simulations must have the accurate determi-
nation of h.t.c value for natural convection and also involve heat transfer through
radiation which would give close to accurate results with the natural scenario
1.3. Research Approach and Novelty

The antenna+heatsink strategy is chosen from among the three approaches for ther-
mal management in antennas due to its straightforward applicability to the existing
prototypes. All of the research conducted under this approach is focused at low fre-
quencies. However, at high mm-wave frequencies, overheating becomes a severe issue
and need to be addressed.
• For the first time, a novel design of mm-Wave heatsink antenna is investigated in this
thesis project. With the designed heatsink antenna, the effects of heatsink dimen-
sions, different orientations on both the EM and Thermal performance are explored.
• A novel thermal model considering appropriate beamformer chip parameters is de-

veloped. The developed antenna thermal model is then simulated using an unique
CFD-based natural convection scenario setup that does not make any assumptions
about heat transfer coefficient values.
1.4. Thesis Outline 15
• The influence of h.t.c. and thermal resistance on junction and antenna temperatures
are researched. The patch antenna is used as a benchmark, and a novel comparison
is done between patch and heatsink antennas in terms of EM and thermal perfor-
mance.
• Based on the results of the parametric study, optimum fin dimensions are selected for
the array design. Another novel aspect includes the examination of scanning abili-
ties, mutual coupling, sidelobe level, and junction temperature of such antennas in
finite and infinite arrays.
1.4. Thesis Outline

The outline of the thesis is the following:
Chapter 2 presents the design of mmWave patch antenna and heatsink antenna de-
signs. The electromagnetic performance between the two designed antennas has been
compared. Focus is given at examining the maximum Co-polarization and Cross-Polarization
Directivity for different heatsink geometry, which helps in modification of the designed
heatsink antenna. Effects of different orientation of heatsink antenna has been dis-
cussed and comparison in terms of EM aspects were done. Finally, infinite array anal-
ysis of the designed heatsink antenna has been done in order to check the mutual cou-
pling effects and embedded element pattern for very large arrays.
Chapter 3 consists of the detailed thermal modelling of the designed structure con-
sidering appropriate parameters of the NXP beamformer chip which is used as the
heat source. The assumptions and boundary conditions for the conduction based
heat transfer with the assumption of heat transfer coefficient and air flow model are
stated. The models are simulated to obtain the maximum junction temperature at the
chip. The change in junction junction temperature and temperature at the antenna
has been observed for different values of heat transfer coefficient. Results from the
air flow natural convective model without the assumption of heat transfer coefficient
are also obtained for comparison with the simplified model based on the approximate
heat transfer coefficients.
Chapter 4 presents electro-thermal parametric studies with the heatsink dimensions

on co-pol, x-pol directivity, and junction temperature. Appropriate heatsink dimen-
sions are selected for the finite array design. Scanning abilities of such both linear and
planar arrays are studied. A comparitive study of the designed heatsink array and the
regular patch array are made.
Chapter 5 summarizes the thesis work and gives concluding remarks and future rec-
ommendations.
2
Heatsink Antenna Design - EM Aspects
The generic goal of this chapter is to design a mm-Wave heatsink antenna which pro-
vides the electromagnetic and thermal dual functionality. The antenna is designed
to resonate at a frequency of 28 GHz and its EM performance has been evaluated.
To achieve this, a conventional coaxial-fed patch antenna is designed first and the
heatsink is attached on top of the microstrip patch. Then, EM performance will be
compared between the regular patch and the heatsink antenna. Maximum co-polarization
and cross-polarization directivity will be analysed for different heatsink geometry. Based
on the analysis, heatsink antenna design will be modified to have the best EM perfor-
mance. The effects of different geometries and orientation of the heatsinks are studied.
Two different heatsink geometries are considered: rectangular extruded fin and cylin-
drical fin. Finally, the coupling and scanning abilities of such heatsink antenna in a
very large array is examined using infinite array analysis.
2.1. mmWave Patch Antenna Design With Heatsink On Top
Firstly, a microstrip patch antenna is designed resonating at 28 GHz frequency and

then a heatsink with rectangular fins is attached on the top. Using the basic patch de-
sign equations shown below, the initial length and width were calculated and then they
were tuned so that the patch resonates at 28 GHz. The initial width and length are cal-
culated using Eq.(2.1)-(2.4)
c
Wp = q (2.1)
²r +1
2f 2
where Wp is the width of the patch, f is the frequency, c is speed of light in vacuum and
²r is the relative permittivity of the substrate. Here, Rogers RT5880 material is used as
the substrate for the patch with permittivity ²r = 2.2 and height H = 0.2718mm.
17
2.1. mmWave Patch Antenna Design With Heatsink On Top 18
Wp
∆L (²r e f f + 0.3)( H + 0.264)
= 0.412 Wp
(2.2)
H (²r e f f − 0.258)( H + 0.8)
where ²r e f f is the effective dielectric permittivity given by
²r + 1 ²r − 1 H −1/2
²r e f f = + (1 + 12 ) (2.3)
2 2 Wp
c
Lp = p − 2∆L (2.4)
2 f ²r e f f
The patch size is L p = 3.3mm, Wp = 4.77mm and it is fed by a coaxial feed beneath
the groundplane which is suitable for potential 3D integration. The groundplane and
patch are made of copper with 17µm thickness. The patch design and the heatsink
antenna design are shown in Figure 2.1 and 2.2 respectively.
Figure 2.1: CST Model of Coaxial fed patch antenna
Figure 2.2: CST model of patch antenna with heatsink on top
For the heatsink antenna, the coaxial pin inset, feed pin diameter and patch length
are tuned for it to resonate at 28 GHz. Here, the patch size is L p = 3.16mm and Wp =
4.77mm. A thermal grease material is attached between the patch and the heatsink in-
order to eliminate the air gaps and maximize heat transfer to the heatsink. The thermal
grease material is silicone heat transfer compound with dielectric permittivity 3.81.
Aluminium is used as the heatsink material. The heatsink base thickness, fin thick-
ness, fin height, and fin width are marked in Figure 2.2 as b t , f t , f h , f w respectively.
Here, the entire heatsink geometry is enclosed within the patch size and this is done to
take care of the mutual coupling effects while designing the element in an array. Also if
the heatsink geometry is extended beyond the patch length (along y), the pattern be-
comes distorted as seen in Appendix A. The fin width f w is the same as the length of
the patch L p = 3.16mm, the fin height f h is 2mm, the fin thickness f t is 0.5mm, and
the heatsink base thickness b t is 1mm. In order to make sure that the geometry of the
heatsink is bounded within the size of the patch, the spacing between the fins becomes
dependent on the fin thickness, number of fins, and the width of the patch if the fins
of the heatsink are oriented along the width of the patch like in Figure 2.2. The spacing
can be calculated as from Eq.2.5
Wp − (N ∗ f t )
S= (2.5)
N −1
where S is the spacing between the fins and N is the number of fins. L p replaces Wp
in Eq.2.5 if the fins are oriented along the patch’s length and in this case the fin width
will be the patch width. The values of the fin dimensions considered here are based on
intuition and more parametric studies on the heatsink dimensions will follow.
2.1.1. Comparison With The Regular Patch Antenna
The two antennas were designed and simulated using CST 3D electromagnetic sim-
ulation tool. The return loss shown in Figure 2.3 indicates that both the antennas
resonate at 28 GHz. The co-polarization and cross-polarization pattern directivity for
both patch and heatsink antenna are shown below in Figures 2.4, and 2.5 respectively.
The desired polarization is the co-polarization and the polarization that is undesirable
is the cross-polarization. High cross-polarization is generally undesired as it may lead
to high interference in systems using dual polarization for frequency re-use and capac-
ity enhancement.
(a) (b)
Figure 2.3: Return loss S11 (dB),(a)Patch antenna,(b)Heatsink antenna

(a) (b)
Figure 2.4: Co-pol pattern directivity (dBi), (a) Patch Antenna, (b) Heatsink Antenna.
(a) (b)
Figure 2.5: Cross-pol pattern directivity (dBi), (a) Patch Antenna, (b) Heatsink Antenna.
Parameters Patch Antenna Patch antenna with heatsink

Resonant Frequency
28.03 27.9
(GHz)
Peak Directivity (dBi) 8.273 8.307
Radiation efficiency 0.94 0.974
Peak Gain (dBi) 8.024 8.2
Realized Gain (dBi) 7.961 8.031
S11 (dB) (f=28 GHz) -35.2 -14.44
Peak Cross-Pol
-7.5 0.165
Directivity (dBi)
-10dB Impedance 0.987 2.668
Bandwidth (GHZ) (3.52%) (9.56%)
Table 2.1: Comparison of EM performance
Table 2.1 shows the simulated EM performance comparison between patch antenna
and the heatsink antenna. From Figure 2.4 and Table 2.1, it is seen that there is a slight
increase in directivity with the heatsink antenna case. The pattern directivity at broad-
side is similar for both the antennas but with higher back lobe for the heatsink antenna.
The radiation efficieny is high for both the antennas as the substrate permittivity and
the loss tangent are low. A higher ²r results in tightly contained fringing fields resulting
in less radiation whereas a smaller ²r results in high radiation due to loosely bound
fringing fields [23]. The radiation efficiency in the case of heatsink antenna is slightly
improved due to the heatsink structure. However, a large improvement on radiation
efficiency with the heatsink structure is discussed in Chapter 4 with the FR-4 substrate
that has high loss tangent (δ) and electrical permittivity (²r ). A significant enhance-
ment in bandwidth is observed for the heatsink antenna and this is because of more
efficient use of available volume in a sphere by the 3D heatsink antenna [7].
The peak cross-polarization directivity (x-pol) is higher in case of the heatsink an-
tenna which is undesirable, whereas the patch antenna has relatively low cross polar-
ization level. The peak cross-polarization directivity was found to be around -7.5 dBi
at (φ = 307◦ , θ = 65◦ ) for the patch and 0.165 dBi at (φ = 305◦ , θ = 62◦ ) for the heatsink
antenna. The peak co-polarization directivity is found to be about 8.273 dBi and 8.31
dBi at θ = 0◦ , φ = 0◦ to 360◦ for the patch and heatsink antenna respectively. As it is
important to have less cross-polarization level, the effects of heatsink dimensions on
both co-polarization and cross-polarization directivity will be studied in the following
sections.
2.1.2. Analysis Of X-pol Directivity With Heatsink Dimension
The maximum cross-polarization level was found to be around 0.165 dBi for the heatsink
antenna which is much higher than the regular patch which was -7.5 dBi. A paramet-
ric study has been made to check if the change in heatsink dimensions has impact on
this maximum cross polarization level. The heatsink base thickness and fin height of
the heatsink antenna are varied to see the effects on cross-pol directivity. Figure 2.6
and 2.7 shows the peak cross-polarization directivity with different fin height ( f h ) and
heatsink base thickness (h b ). respectively
Figure 2.6: Maximum cross pol Directivity (dBi) vs Frequency

Figure 2.7: Maximum Cross-Pol Directivity (dBi) vs Frequency
The heatsink parameters f t = 0.5mm, h b = 1mm, f w = 3.16mm were kept constant and
height of the fin was varied from 1mm to 3mm. As the fin height is increased, the cross-
polarization level increased to small degree. Peak cross-pol level at 28 GHz was found
to be around 0.023 dBi for f h = 1mm. For the analysis with the heatsink base thickness,
the fin height f h = 2mm, f t = 0.5mm, f w = 3.16mm were kept constant and the thick-
ness of the heatsink base was varied from 0.1mm to 1mm. The behaviour of cross-
polarization level was alike as in the case with fin height. The impact of base thick-
ness on cross-pol level was slightly more than that of the fin height’s impact. With less
heatsink base thickness of 0.1 mm, peak cross pol directivity was found to be around
-0.8 dBi.
2.1.3. Effect Of Removing Heatsink Base
The dimensions of the heatsink can be chosen such that antenna has a lower cross
polarization level. Reducing the heatsink dimension has shown some improvement
in lowering the cross-polarization directivity. Contrary to this, shrinking the heatsink
dimension reduces it’s thermal performance. As reducing the heatsink base thickness
has more impact on lowering cross-polarization and also improving the co-polarization
directivity, the base of the heatsink has been removed. The fins are directly placed on
the top of the patch as shown in Figure 2.8 with thermal paste between them.
Figure 2.8: CST Model of Coaxial fed patch antenna with heat fins on top
2.2. Effects Of Geometry And Orientation Of The Heatfins 23
This modified heatsink antenna design showed improvement in cross-polarization and

co-polarization directivity. The peak cross-polarization directivity is found to be around
-1.39 dBi at φ = 305◦ , θ = 70◦ and the peak co-polarization directivity is found to be im-
proved from 8.3 to 9.1 dB at θ = 0◦ , φ = 0◦ to 360◦ and the max cross polarization level
is -10.49 dB lower than the co-polarization level. The direction of maximum cross-
polarization is important as it causes interference with the desired polarization direc-
tion. Low cross polarization levels close to the regular patch antenna can be achieved
with change in substrate material and permittivity which will be discussed in Chap-
ter 4. The return loss and pattern directivity for the modified heatsink antenna are
shown in Figures 2.9 and 2.10 respectively. The EM performance comparison will be
discussed in the next subsection.
Figure 2.9: Return loss S11(dB)
(a) (b)
Figure 2.10: Pattern Directivity for the patch with heat fins (dBi) (a) co-pol, (b) cross-pol
2.2. Effects Of Geometry And Orientation Of The Heatfins

Two different heatsink geometries are considered here: 1) Extruded rectangular fin
structure 2) Cylindrical/Radial pin fin structure. Two different orientations for the rect-
angular fins and three different orientations for the radial pin fin case are designed and
their corresponding EM performance are studied. Figure 2.11 shows the different ge-
ometry and orientation of heatsink antenna and its 3D Radiation pattern respectively.
The different orientations considered in the rectangular fin geometry are (a). Rectan-
gular fins with fins parallel to the non-radiating edges of the patch (PNRE), (c). Rectan-
gular fins with fins parallel to the radiating edges of the patch (PRE). The rectangular
fins are aligned along the patch’s length in the case of PNRE and it is aligned along the
patch’s width in the case of PRE.
Three different orientations are considered in the radial pin fin geometry, (e). Radial
pin fins all around the top of the patch, (g). Radial pin fins across the edges of the
patch, and (i). Radial pin fins near the centre of the patch. Peak directivity, peak gain
and radiation efficiency of the respective heatsink antenna orientation has been tabu-
lated in Table 2.2
Peak Directivity Peak Gain Radiation

Heatsink orientation
(dBi) (dBi) Efficiency
Rectangular fins parallel
to non-radiating edges 9.096 8.9 0.956
(PNRE)
Rectangular fins parallel
to radiating edges 6.234 6.16 0.983
(PRE)
Radial pin fins all around
4.918 4.73 0.962
the top of the patch
Radial pin fins (Edges) 5.863 5.637 0.95
Radial pin fins (near the
9.074 8.772 0.933
center)
Table 2.2: Comparison of EM performance for different heatsink geometries and orientations
The radiation efficiency for all the cases are above 90 % with rectangular fins (PRE)
having the highest radiation efficiency of about 98.3 % indicating that it has the lowest
conduction and dielectric loss.
With the fins parallel to non-radiating edges (PNRE) orientation, the peak gain is around
8.9 dBi which is 0.87 dB greater than that of the patch antenna as the broadside lobe is
more focused due to the heatfin structure. Whereas, with the fins parallel to radiating
edges of the patch, the peak gain is reduced to 6.16 dBi which is around 2 dB lower
than the regular patch antenna. This is due to the fact that the fins in this case lie per-
pendicular to the current flow across the patch which lets the vertical surface currents
flowing across the fins. This vertical surface currents is responsible for the radiation of
side lobes in end-fire direction [7] as seen in Figure 2.11d.
(a) (b)
(c) (d)
(e) (f)
(g) (h)
(i) (j)
Figure 2.11: Different heatsink antenna geometry and orientations and its Radiation pattern directivity
(dBi), (a,b)Rectangular fins parallel to non-radiating edges (PNRE), (c,d) Rectangular fins parallel to
radiating edge (PRE), (e,f) Radial pin fins all around the top of the patch, (g,h) Radial pin fins (Edges),
(i,j) Radial pin fins (near the center)
With the Radial pin fin case, the fin thickness is the diameter of the fin which is around
0.3mm. The Radial pin fins all around the top of the patch and along the edges has
lower directivity compared to the regular patch antenna. These two types have simi-
lar radiation pattern like that of the fins PRE case. But when the radial fins are placed
around the center and not along the edges, the pattern has focused broadside lobe
similar to that of the fins PNRE case.
2.3. Pattern Directivity With Fin Dimensions 26
Since, the rectangular fins (PNRE) and the radial fins (center) has higher directivity
(focused broadside beam) than that of the regular patch antenna. These two heatfin
orientations are considered for futher EM and thermal analysis. The EM performance
comparison of these two orientations with the regular patch antenna is shown in Table
2.3.
Patch Antenna with Patch Antenna with Patch Antenna with

Parameters Patch Antenna
heatsink heat fins (PNRE) Radial fins (centre)
Resonant Freq
28.03 27.9 28.03 28.07
(GHz)
Peak Directivity
8.273 8.307 9.096 9.074
(dB)
Radiation efficiency 0.94 0.974 0.956 0.934
Peak Gain
8.024 8.2 8.899 8.772
(dB)
Realized Gain
7.961 8.031 8.868 8.568
(dB)
S11 (dB) -35.2 -14.44 -25.6 -16.1
Peak Cross-pol
-7.5 0.162 -1.39 -2.8
Directivity (dB)
-10 dB Impedance 0.987 2.668 2.56 0.694
Bandwidth (GHz) (3.52 %) (9.56 %) (9.13 %) (2.5 %)
Table 2.3: EM performance comparison of the selected heatsink orientations
2.3. Pattern Directivity With Fin Dimensions

The heatfin orientations rectangular fins PNRE and radial pin fins (center) were con-
sidered for parametric analysis with the fin parameters. Fin height, fin thickness and
number of fins were varied to see the effects on pattern directivity. Figure 2.12, 2.13,
and 2.14 shows the pattern directivity of the rectangular fin PNRE heatsink antenna
with respect to different fin height, number of fins, and fin thickness respectively. Pat-
tern directivity at constant φ = 0◦ and φ = 90◦ are shown.
Here, the initial parameters are N = 4, f h = 2mm and f t = 0.5mm. One of the pa-
rameter is varied while the other two are kept constant considering their initial values.
With the change in the fin height, directivity at constant φ = 90◦ is almost similar for
all the cases whereas with the pattern directivity at φ = 0◦ , back lobe increases as the
fin height increases. Pattern directivity at φ = 90◦ is almost similar for the other para-
metric study as well. Except for N=3 fins, the pattern directivity at φ = 0◦ looks alike.
Similar analaysis has been made for the radial pin fin (centre) case which is not shown
here. The optimum parameters are chosen based on the maximum co-pol and cross-
pol directivity which will be shown in Chapter 4.
2.4. Infinite Array Analysis Using Periodic Boundary Condition 27
Figure 2.12: Pattern Directivity vs fin height
Figure 2.13: Pattern Directivity vs Number of fins
Figure 2.14: Pattern Directivity vs Fin thickness
2.4. Infinite Array Analysis Using Periodic Boundary

Condition
In this section, the effects of mutual coupling between antenna elements in a very
large array has been analysed. For this, an infinite antenna array approximation is
used where periodic boundary condition is applied to represent it as an infinitely ex-

tended antenna array. Using periodic Boundary conditions, mutual coupling effects
can be analysed and the pattern obtained will be the embedded element pattern of the
element in the infinite array. These boundary conditions will simulate a simultaneous
excitation of all elements in the spatially periodic array, where the element-to-element
phase shifts are given. However, the phase shift property works only with Frequency
domain solver in CST, which is used in this work.
Firstly, the heatfins PNRE oriented antenna is taken and periodic boundary condition
is applied on x-axis (along the width of the patch). The y-axis (along the patch length)
is kept open which means that it creates infinite 1D (linear) array along x-axis. The
spacing between the array elements is taken as d x = 0.6λ. Similarly the infinitely ex-
tended array is created along y axis with d y = 0.6λ. Figure 2.15 shows the heatfins
PNRE antenna with the periodic boundary condition applied along x axis.
Figure 2.15: CST Model of heatfins PNRE array with periodic boundary condition along x axis
The return loss plots for both the infinitely extended linear arrays are shown in Figure
2.16.
(a) (b)
Figure 2.16: Return loss S11 (dB) of infinitely extended heatfins PNRE array,(a)along x,(b)along y
It is seen that the heatfins PNRE infinitely extended along x axis has better matching at
28 GHz fequency. With the inifinitely extended array along y-axis, the resonating fre-
quency is shifted lower to 26.7 GHz and this is due to the mutual coupling between the
antenna elements. Mutual coupling along y-oriented heatsink antenna array affects
the pattern and thus this heat fins (PNRE) geometry in the current form of the design
is not suitable for a very large 2D array. Although the mutual coupling can be reduced
by increasing the spacing between the elements, this will result in grating lobes in the
radiation pattern. The return loss plot for array infinitely extended array along y shows
almost constant value of -8 dB for a wide band from 27-31 GHz (not simulated beyond
31 GHz). The S11 value can be lowered down below -10 dB to have a good matching in
the wide band from 26-31 GHz by lowering the imaginary part of the input impedance.
To achieve this one option could be to add lumped elements or by adding a metallic
ring in the air region of the coax feed line.
A case study has been made to see the effects of inserting a metallic ring in the air
region of the coax line as shown in Figure 2.17
Figure 2.17: Coaxial feed with a ring inside the air region
The addition of this coax ring reduces the imaginary part of impedance and increases
the real part of impedance providing good matching when it is normalized to 80Ω
impedance. Figure 2.18 shows return loss properly matched having a wide bandwidth
of 17.87 GHz from 26.4-44.3 GHz.
Figure 2.18: Return loss (dB)
The impedance matching effects in finite arrays are also interesting to investigate, thus
this is being considered as a future research study.
Scanning range abilities can be observed by entering scan angle value in the phase
shift/scan angle dialog box in CST. The heatfins PNRE infinitely extended linear array
along x could achieve only ±15◦ scanning range. For 30◦ scanning and higher angles,
the matching became worse. Figure 2.19 gives the return loss plot for 15◦ and 30◦ scan
angle respectively. The matching is worse for 30◦ scan angle and this is due to the
mutual coupling. This can be improved by increasing the spacing which reduces the
coupling between elements.
(a) (b)
Figure 2.19: Return loss S11 (dB) of infinitely extended heatfins PNRE array with scan angle,(a)15◦ ,(b)
30◦
With 15◦ scan angle, parametric study is made for different spacing to check the effects
of mutual coupling. The spacing is varied from d x = 0.5λ to 1λ and the embedded pat-
tern directivity is shown in Figure 2.20. For d x = 0.5λ spacing, the pattern is distorted
and this is because of mutual coupling. The pattern is also distorted at 0.9λ and 1λ
which might be due to poor matching.
Figure 2.20: Embedded element pattern of the heat fins (PNRE) antenna with 15◦ scan angle
As coupling is due to the heatsink and making them close to each other increases cou-
pling, the patch size has been reduced. Reducing the patch length reduces the width
of the heat fins. The length of the patch is changed from 3.33mm to 2.25mm. Shrink-
ing the patch size and increasing the substrate permittivity makes the patch resonate
along the same 28 GHz frequency. The substarte used here is FR-4 with dielectric per-
mittivity ²r = 4.3. Figure 2.21 shows the embedded element pattern for the shrunk
antenna. The effects of changing the permittivity and shrinking the patch size on EM
and thermal aspects will be discussed further on Chapter 4.
2.5. Conclusions - EM Aspects 31
Figure 2.21: Embedded element pattern of the modified heat fins (PNRE) antenna (shrunk patch) with
15◦ scan angle
Similarly, the analysis is done for the radial fins (centre) orientation of the heatsink
antenna. Since the variations due to the coupling effects are less significant, with x-
extended and y-extended infinite arrays, periodic boundary conditions were applied
across both x and y axis to form a 2D infinitely extended array. It is found to achieve
about ±60◦ scanning range.
Figure 2.22: CST Model of Radial fins (centre) antenna with Periodic Boundary condition
Figure 2.23: Embedded element pattern
2.5. Conclusions - EM Aspects

In this chapter, a mm-Wave heatsink antenna element has been designed to resonate
at 28 GHz frequency. The designed antenna has been modified to improve the co-pol
and x-pol directivity. Among the different geometries and orientations of the heatsink
antenna, rectangular fins PNRE and radial fins (center) showed preferable EM perfor-
mance. The EM performance like co-pol, cross-pol directivity, realized gain, and radi-
ation efficiencies of these two antennas are compared with that of the patch antenna.
2.5. Conclusions - EM Aspects 32
Using periodic boundary conditions, both of these antennas are infinitely extended to
see the coupling effects of such antennas in very large arrays. The x-oriented heatfins
PNRE linear array is found to have a scan range of ±15◦ while the radial fins (center)
2D array showed a good scan range ±60◦ .
Heatsink Antenna Design - Thermal
3
Aspects
This chapter deals with the thermal modelling and simulation of the designed mm-
wave heatsink antenna integrated with an active integrated beamformer integrated
circuit unit. Simulation is done with the help of two thermal models, one is a con-
duction based model with the assumption of convective heat transfer coefficient and
the other is the air flow natural convection model without any heat transfer coefficient
assumptions. Assumptions and boundary conditions for both the models are given.
Junction temperature at the heat source is computed and the simulated results were
compared between both the models. Effects of h.t.c, thermal resistance and radiative
heat transfer on the temperature has been analysed.
3.1. Thermal Model Considerations

To model the heat flow in electronics, a standard thermal model for electronics has
been used. There are two types of thermal modelling in electronic design. One is the
detailed thermal model (DTMs) and the other is the compact thermal model (CTMs)
[24]. Though DTMs models the IC packages more accurately, they are computation-
ally complex. In contrast, the CTMs don’t accurately model the IC packages, but they
are computationally simple in predicting the junction temperature. This makes CTM
more suitable for analysing heat flow problem in an efficient way [5]. Here, Two-
resistor CTM is used to represent the thermal model for the designed heatsink antenna.
3.1.1. Two-resistor Model

The two resistor model as described by the JESD15-3 standard, consists of two thermal
resistors connecting three nodes as shown in Figure 3.1. The three nodes are the case
node, board node and the junction node. The two resistors are junction-to-case θ JC
connecting the Junction and case nodes, junction-to-board resistor θB connecting the
junction and board nodes. The junction node is where the source of the heat comes.
In this case, the heat source is the beamformer chip placed on the opposite side of the
antenna. The case node comes at the top of the package and the board node is the
Printed Circuit Board (PCB) which is below the package’s footprint.
35
3.1. Thermal Model Considerations 36
Figure 3.1: Two resistor model and its 3D space representation taken from [cite]
For simulation in 3D space, the two resistor model is represented as shown in Figure
3.1 using block and plate method where nodes are represented by block and resistors
by plate. The junction block is a volume heat source and a highly conductive material.
The heat flow is considered only in the normal direction. The sides are insulated to
prevent leakage. The block’s size is the actual size of the package and the plate’s thick-
ness and thermal conductivity can be chosen such that it obtains the actual thermal
resistance value [17].
t
θ= (3.1)
K .A
where
θ JC or θB - thermal resistance [K /W ]
K - thermal conductivity [W /m.K ]
t - thickness of the plate [m]
A - surface area of the plate [m 2 ]
The accuracy of the two resistor model is low, but it is higher in case where most of
the heat flows either through heatsink or the board [5]. An h.t.c, heat transfer coeffi-
cient value is applied at the solid-air interface to apply convective heat transfer. Heat
transfer coefficient is the ratio of the heatflux to the temperature difference between
the solid and surrounding fluid. At the solid-fluid regions, the thermal resistance is
calculated by
1
R= (3.2)
hA
where,
R - thermal resistance at solid-fluid interface [K /W ]

h - heat transfer coefficient [W /m 2 .K ]
A - surface area on which h.t.c acts [m 2 ]
3.1.2. Properties Of The Electronic Chip Package

3.1. Thermal Model Considerations 37
The heat source here is the NXP beamformer chip attached at the back of the an-
tenna. The 4-channel mm-Wave analog beamforming chip shown in Figure 3.2. Am-
plitude and phase control in channel can be done by interfacing the chip to a computer
through I2C or SPI interfaces [25]. The chip dissipates heat of about 2.3 W in Tx mode.
Table 3.1 shows the parameters required for modelling the package using two-resistor
model.
Heat produced by the

2.3W
chip in Tx mode
4.3 x 3.5 x 0.5 mm
Chip dimensions
(W x L x H)
junction-to-case
10K /W
resistance (θC )
junction-to-board
14K /W
resistance (θB )
h.t.c at the solid-air
10W /m 2 K
interface
Ambient Temperature 20◦C
Figure 3.2: 4-channel mm-Wave NXP analog
beamformer chip from [25] Table 3.1: Chip parameters from [5]
3.1.3. Thermal Model Of The Antenna

Based on the NXP beamformer chip parameters, the thermal model of the antenna has
been desiged using COMSOL Multiphyics simulation software. Figure 3.3 shows the
thermal model of the patch antenna.
Figure 3.3: Thermal model of 4-element Linear patch antenna
As per the two resistor model representation in 3D space, the electronic chip is a highly
conductive Aluminium cuboid block with the appropriate chip dimensions. The plates
above and below are the junction-to-board and junction-to-case thermal resistances.
From Figure 3.2, we can see that this 4-channel analog beamformer chip is capable of
3.2. Simulation Approaches And Boundary Conditions 38
serving 4 antennas. Here, a 4-element linear patch array for single beamformer chip is
designed. It can also be a 4-element 2D array depending on the heatsink geometry. The
heatfins PNRE is ideal for a linear array, but the radial fins (centre) instance is suitable
for a 2D array, as previously stated. For the heatsink antenna case, a thermal grease
and the heat fins will be added above the patch in Figure 3.3. Above the electronic chip
are the chip substrate, ground plane, the antenna substrate and the 4 patch elements.
Rogers RT 5880 material with 0.254mm thickness is used for both the antenna and
chip substrate. Table 3.2 shows the material and thermal conductivity used for the
geometries.
Thickness Thermal conductivity

Geometry Materials
(mm) (W /m.K )
Antenna & chip substrate Rogers RT 5870 0.254 0.2
Patch & Ground plane Copper 0.017 400
Chip & heatsink fins Aluminium 0.5 238
Thermal grease/
Thermal grease 0.05 3-8
thermal paste
Table 3.2: Geometry properties
For the solid-air interfaces, heat transfer coefficient value is assigned to enable convec-
tive heat flow. Above the patch, the fluid surrounding is air with ambient temperature
of about 20◦C . An h.t.c value of h ai r = 10W /m 2 K is assumed from 2−30W /m 2 K which
is the approximate heat transfer coefficient value for natural convection cooling regime
[17]. At the bottom surface of the chip, an h.t.c value is assumed at the case-heatsink
interface and this value is large from about 1000 − 3000W /m 2 K . The h.t.c value is sup-
posed to be high in this case to indicate a larger heatsink at the bottom that can remove
more heat via conduction or a forced convection scenario. This is done to investigate
the impact of the heat transfer coefficient at the chip’s bottom surface on the maxi-
mum junction temperature T J (◦C ). Later, instead of assuming an h.t.c value, a larger
heatsink is attached at the bottom of the chip.
Now that the model is designed according to the geometry properties, the heat transfer
simulation environment has to be set up. For the heat transfer simulations, COMSOL
simulation environment is used.
3.2. Simulation Approaches And Boundary Conditions

Approaches used for heat transfer simulations:
• Conduction based model
• Air flow model
The assumptions and boundary conditions for these two heat transfer simulation ap-
proaches will be discussed in subsections 3.2.1 and 3.2.2. One of the main differences
between the two simulation approaches is that the conduction based model assumes
a heat transfer coefficient for a natural convective scenario, which makes the simula-
tion easier, whereas the air flow model is based on CFD without any h.t.c assumption,
making it computationally complex.
3.2.1. Conduction Based Model
Conduction is one of the three primary methods through which heat energy is trans-
ferred. It takes place by the transfer of energy within a body through small particle col-
lisions and electron mobility. This model accounts for conductive heat transfer along
with the convective cooling boundary condition, where the convective heat transfer is
given by the following equation.
q = h.A.(Text − T ) (3.3)
where,
q - heat transferred per unit time [W ]

h - Convective heat transfer coefficient [W /m 2 K ]
A - Area of heat transfer [m 2 ]
Text - Surface temperature of the object [K ]
T - Fluid’s temperature (air) [K ]
Assumptions And Boundary Conditions
• The heat rate P = 2.3W dissipated by the chip is applied to the heat source
where the volumetric heat rate is calculated as Q 0 = P 0 /V
• Convective heat flux is applied for all the solid regions for conductive and
convective heat transfer where, the heat transfer coefficient is assumed to be
10W /m 2 K .
• The Ambient temperature is set to 293.15K (20◦C ).
• Physics controlled meshing is selected with Finer meshing after checking con-
vergence.
• Stationary study solver is used for computing the temperature field at thermal
equilibrium.
3.2.2. Airflow Model
This model is based on CFD techniques, where a large cuboid box as shown in Figure
3.4 is designed enclosing the thermal model setup of the antenna in order to simulate
heat transfer through natural convection. Gravity is the primary factor behind natural
convection where it pulls denser fluid creating convection. The material of this cuboid
box is a fluid which is air in this case.
Air flow inlet and outlet are specified in the box. The dimensions of this air box is such
that it is 10 hydraulic diameter downstream (antenna to outlet), 5 hydraulic diameter
upstream (antenna to inlet) and 3-6 hydraulic diameter for the sides. In the case of
fluid flow through a non-circular geometry, the hydraulic diameter is calculated by
4A
D= (3.4)
P
where,
A - Cross sectional Area [m 2 ]

P - Wetted Perimeter of the cross-section [W /m 2 K ]
For more details on calculation of hydraulic diameter and the dimensions of the air
box, see Appendix B.
Figure 3.4: Airflow model of 4-element Linear heatsink antenna
Assumptions and Boundary conditions
• Similar to the previous model, a volumetric heat rate of 2.3 W is applied as the
heat source
• For the fluid flow, a cuboid box is designed enclosing the antenna setup where
the inlet and outlet domains are selected.
• The fluid is air where the thermophysical properties of air such as dynamic
viscosity, thermal conductivity, density and specific heat at 20◦C are consid-
ered.
• For the walls of the air box (except inlet and outlet boundaries), No-slip bound-
ary condition is applied where the fluid velocity relative to the wall is zero.
• For the fluid flow, laminar flow regime is taken where a parabolic velocity pro-
file defines the inlet velocity. The normal stress equals the outlet pressure at
the outlet, and the tangential stress is cancelled.
• Gravity is enabled by the include gravity check box where it acts along +y axis
(the flow direction, see Figure 3.4) and it is equal to 9.8m/s 2
3.3. Results And Analysis-Conduction Based Model 41
• A non-isothermal multiphyiscs coupling is activated which couples the fluid

flow and heat transfer. Incompressible air flow is assumed, which means the
fluid density remains constant during the flow. The Boussinesq approxima-
tion is selected for using non-isothermal flow with incompressible air flow as-
sumption.
• Finer physics controlled mesh and Stationary solver are used as in the previ-
ous case
For more details on the calculations of estimate of air flow velocity, selection of fluid
flow regime and boussinesq approximations, see Appendix B
3.3. Results And Analysis-Conduction Based Model

Simulations are carried out using the assumptions and boundary conditions men-
tioned above to calculate junction temperature T J . This subsection deals with the re-
sults and analysis with the conduction based model. Both the traditional patch an-
tenna and the heatsink antenna are compared based on the thermal aspects. As previ-
ously described, a four-element linear antenna for the heatfins PNRE oriented heatsink
antenna and a four-element two-dimensional array for the radial pin fin oriented at the
center are constructed, simulated, and compared to the patch antenna. At the solid-
air interface, the heat transfer coefficient is 10W /m 2 K . Figure 3.5 and 3.6 shows the
Temperature plots for both the heatsink antennas compared with the patch antenna.
(a) (b)
Figure 3.5: Temperature plots for 4-element linear array (a) Patch ,(b)fins (PNRE) heatsink antenna
(a) (b)
Figure 3.6: Temperature plots for 4-element 2D array (a) Patch ,(b)Radial fins (center) heatsink antenna
In all the circumstances, maximum temperature is found at the heat source and the
temperature is extremely high. For the linear patch array the maximum junction tem-
perature is found to be 909◦C and with the heatfins PNRE linear array, the maximum
junction temperature is found to be 671◦C . This temperature decrease is due to the
heatfins present above the patch which gives additional surface area for the heat to dis-
sipate via conduction. Even though there is a large reduction in temperature, the com-
ponents cannot withstand this high temperature. The maximum junction temperature
should be maintained below 125◦C for safe and reliable operation of the components
[26]. As a result, a larger surface area is needed to remove more heat. A heatsink can
be attached to the bottom of the chip to accomplish this.
3.3.1. Analysis With Different h.t.c

Rather than using a heatsink at the bottom of the chip, a higher h.t.c value is used at
the bottom surface of the chip to examine how it affects the junction temperature (T J )
and antenna temperature (T A ). The temperature in the middle of the central antenna
element’s microstrip patch is used to calculate the antenna temperature. It is element
2 or 3 in a linear array, but any antenna element’s center can be used for measurement
in a 2D array since all the elements are equidistant from the chip. The h.t.c value at
the bottom of the chip is varied from 1000 − 3000W /m 2 K . The junction and antenna
temperature plots with varying h.t.c is shown in Figure 3.7 and 3.8
(a) (b)
Figure 3.7: 4-element linear array,(a)T J vs h.t.c,(b)T A vs h.t.c
(a) (b)
Figure 3.8: 4-element 2D array,(a)T J vs h.t.c,(b)T A vs h.t.c

Increasing h.t.c increases heat transfer, which is evident from these plots. The patch
antenna is represented by the blue line, while the heatsink antenna is represented by
the orange line. The junction temperature for the linear patch and 2D patch array is
determined to be 169◦C at 1000W /m 2 K h.t.c, whereas it is roughly 155◦C for the heat
fins PNRE antenna and 164.4◦C for the heatsink radial fins (center) antenna. The heat
fins PNRE antenna has better heat dissipation compared with the radial fins center an-
tenna. As the h.t.c is increased further to 3000, the blue line and orange line converges
for the radial pin (centre) case. This means that the heat dissipation through the ra-
dial fins is negligible when a very large heatsink structure is attached below the chip.
Whereas for the rectangular fins PNRE case, the orange and blue line doesn’t converge
at higher h.t.c as there is always a significant difference in temperature compared to
that of the patch antenna. The reason for this is due to the fact that fins PNRE heatsink
antenna provides more surface area than that of the radial fins (center) heatsink an-
tenna. There is a significant drop in antenna temperature for both the heatsink ge-
ometries when compared to the patch.
With 1000W /m 2 K h.t.c at the bottom of the chip, the h.t.c at the antenna side is changed
from 10W /m 2 K to higher values in order to investigate the junction temperature re-
sults with a forced convection scenario (forced air from a fan). The h.t.c at the antenna
side is varied from 100 − 300W /m 2 K to have a forced air cooling. Figure 3.9 shows the
junction temperature plot with respect to different h.t.c at the antenna side.
Figure 3.9: Junction Temperature vs h.t.c at the antenna side (4-element linear array)
The junction temperature of the patch reduced from 169◦C to 154◦C after applying
100W /m 2 K h.t.c at the antenna side, while the temperature of the PNRE fins dropped
from 155◦C to 136◦C , increasing the temperature difference between patch and heatsink
antenna. The increase in h.t.c can be attributed to the increase in velocity of the forced
air which reduces the temperature even more.
3.3.2. Effects Of Chip’s Parameters On Junction Temperature
To keep the junction temperature below 125 degrees, an aluminum heatsink is mounted
to the bottom of the chip. The heatsink’s base dimension is 4.3 x 3.5 x 0.5 mm, thick-
ness of the fin is 0.5 mm and the fin height is about 15 mm. The simulated temperature
results are seen in Figures 3.10 and 3.11.
(a) (b)
(a) (b)
Figure 3.11: Temperature plots for 4-element 2D array (a) Patch ,(b) Radial fins (center) heatsink
antenna
The patch, PNRE fins, and radial fins(center) have maximum junction temperatures of
120, 112.7, and 118 degrees Celsius, respectively. Because most of the heat is transmit-
ted via the aluminum heatsink mounted at the bottom, the temperature difference is
smaller. The chip substrate and antenna substrate have a very poor thermal conduc-
tivity, which resists the flow of heat above the chip. As a result, it’s critical to look at the
effects of the materials used in the antenna and chip. The effects of the chip substrate
and the chip’s thermal resistance on the junction temperature have been researched
since changing the material used for the antenna substrate will influence the electro-
magnetic results. From now on, only the fins PNRE case will be further investigated, as
though the radial pins (center) case has good EM performance, it is not sufficient for
better heat dissipation.
Material Of Chip Substrate
The chip substrate is made of Rogers RT 5880, which has a low thermal conductivity of
0.2W /mK and thus inhibits heat flow above the chip. As a result, the chip substrate can
be made of a highly conductive substance that can conduct more heat to the copper
ground plane above the chip’s substrate. Aluminium Nitride (AlN), a ceramic substrate
with excellent thermal properties, is used as the new substrate material. The thermal
conductivity of AlN is about 180W /mK . Figure 3.12 shows the patch array and the
heatfins PNRE array with AlN chip substrate.
(a) (b)
About 4◦C temperature drop is seen with the case of regular patch and 7◦C tempera-
ture drop is seen with the case of heatsink antenna. This enhancement is due to the
AlN substrate. Further heat enhancement is possible by increasing the surface area
above the antenna or by extending the copper ground plane as shown in Appendix C.
But this has additional complexity for integration.
Effects Of Thermal Resistance
The effects of junction-to-board and junction-to-case thermal resistances on the junc-

tion temperature is studied. Figure 3.13 shows the Junction temperature varying with
both the thermal resistances.
(a) (b)
Figure 3.13: Effects of thermal resistance, (a) θ JC vs T J ,(b)θ J B vs T J
The thermal resistance values can be altered by altering the thermal conductivity and
plate thickness. When the junction-to-case resistance is reduced, the temperature re-
duction is greater than when the junction-to-board resistance is reduced. This is due
to the fact that the aluminum heatsink beneath the chip provides higher surface area
for heat to dissipate.
3.3.3. Radiative Heat Transfer

Apart from conduction and convection, heat is also dissipated by thermal radiation,
which is carried out in the form of electromagnetic waves. Unlike black bodies, which
emit all thermal radiation, other surfaces that are not perfect black bodies have an
emissivity factor that determines how much heat they radiate. The thermal radiation
emitted by a Gray body is given by
P = ²σAT 4 (3.5)
where,
P - heat transferred per unit time [W ]
² - surface emissivity
σ - Stefan-Boltzman constant = 6.67x10−8 [W m −2 K −4 ]
A - radiating area [m 2 ]
T - temperature of the body[K ]
Surface-to-Surface radiation physics is added to the existing model in COMSOL where
the radiating surfaces and direction are selected. Gray body radiation node is selected
and surface emissivity (²) is taken from the material properties. Figure 3.14 shows the
patch and heatsink antenna after radiative heat transfer has been added. In both cir-
cumstances, a temperature reduction of about 2◦C is seen.
(a) (b)
Figure 3.14: Temperature plot with radiative heat transfer,(a)Patch,(b)Heatsink
Thermal radiation is directly proportional to temperature and emissivity. The materi-

als used here have a low emissivity, particularly the polished aluminum, which has a
surface emissivity of 0.05. Increasing the emissivity of the materials used is one tech-
nique to improve heat transmission by radiation. The emissivity of the aluminum
heatsink can be increased up to 0.85 by anodizing it [27]. Anodizing, on the other
hand, has an effect on the electrical characteristics of aluminum. Figure 3.15 shows
the temperature plot of the heatsink antenna with anodized aluminium for the an-
tenna and the heatsink placed at the bottom of the chip. The temperature reduced
from 104◦C to 94.2◦C after using anodized aluminium.
3.4. Results And Analysis-Airflow Model 47
Figure 3.15: Temperature plot with radiative heat transfer using Anodized Aluminium
3.4. Results And Analysis-Airflow Model
Based on the simulation settings discussed in section 3.2.2, temperature and velocity
plots are obtained after simulation. Figure 3.16 and 3.17 shows temperature and ve-
locity distribution for the linear patch antenna array and the heatsink antenna array
respectively.
Figure 3.16: Natural convection of 4-element Linear patch antenna
Figure 3.17: Natural convection of 4-element Linear heatfins (PNRE) antenna

3.5. Conclusions-Thermal Aspects 48
4-element patch 4-element heatsink PNRE

Thermal Model
antenna array antenna array
Conduction Based 114.19◦C 104.05◦C
Air flow 126.82◦C 105.23◦C
Table 3.3: Comparison of the two heat transfer simulation models
The temperature plots show that using a heatsink antenna array lowers the junction
temperature by 21.6◦C compared to the patch antenna. Table 3.3 shows the simulated
junction temperature values based on both the models. While the junction tempera-
ture obtained is almost the same for both the models in the case of the heatsink an-
tenna array, it is not the same in the patch antenna case. This is due to the fact that
the air flow model does not assume any heat transfer coefficient because the h.t.c is
unknown in most cases. Many factors influence the heat transfer coefficient, includ-
ing geometry, fluid temperature, and other thermophysical parameters like as viscosity
and thermal conductivity. In the case of patch antenna geometry the h.t.c will be lower
when you compare to the heatsink antenna geometry and this can be also explained
with the change in velocity. For heatsink antenna the typical velocity was about 0.455
m/s and for the patch antenna it came about 0.429 m/s. This velocity here is calcu-
lated at specific points selected in (x,y,z) plane. For more details on the estimate of air
velocity and fluid flow see Appendix B.
3.5. Conclusions-Thermal Aspects

To summarize, this chapter introduces a two-resistor thermal model for electronics
that includes the proper beamformer chip parameters. Two simulation methods are
considered: 1) conduction with convective cooling boundary condition and 2) natu-
ral convective laminar air flow model. The heatsink antenna’s simulated temperature
plots are compared to the patch antenna’s. Heat fins PNRE antenna dissipated heat
more effectively than radial fins (center) and patch antennas.The effects of thermal
resistance and the chip’s material on the junction temperature is studied where low-
ering the junction-to-case resistance resulted in a substantial temperature drop while
compared to lowering the junction-to-board resistance. The temperature dropped by
roughly 4 degrees when the chip substrate was changed from a low thermally conduc-
tive Rogers RT5880 to a highly thermally conductive Aluminium Nitride. In the case of
the heatsink antenna array, the junction temperature achieved is about the same for
both models, but not in the case of the patch antenna. This is owing to the fact that
the h.t.c for patch antenna geometry might be smaller than the assumed 10W /m 2 K
h.t.c value, which can also be explained by the velocity change. The heatsink antenna’s
junction temperature is reduced by 10◦C in the conduction model, while the air flow
model shows a 21.6◦C degree reduction in comparison to the patch antenna.
Heatsink Antenna - Electro-Thermal
4
Co-Design
This chapter begins with parametric analyses of antenna dimensions and design se-
lection based on performance in both EM and thermal aspects. The sample array im-
plementation is then shown where scanning abilities, side lobe level and junction tem-
perature of both patch and heatsink antenna arrays are compared. Finally, a compar-
ison of electro-thermal performance between patch array and heatsink antenna array
is made where, heatsink antenna provides a better joint electro-thermal performance
trade-off.
4.1. Parametric Study With Fin Dimensions

4-element fins PNRE heatsink antenna array is considered with the heatsink attached
at the bottom surface of the chip as discussed in section 3.3.2. Keeping all the other di-
mensions and materials constant as discussed earlier, the fin’s dimensions, such as fin
height, thickness, and number of fins are modified to examine how they affect thermal
and electromagnetic aspects. It’s crucial to compare and contrast the EM and thermal
results because this will aid in the design of a heatsink antenna with the best fin dimen-
sions. The effects of fin dimensions on pattern shape were already examined in section
2.3 of Chapter 2. The maximal co-pol and cross-pol directivity in relation to various fin
dimensions is now being researched. As the surface area above the antenna varies due
to the change in fin dimensions, the junction temperature changes. The surface area
is calculated as the fin area above the patch antenna multiplied by the number of fins.
The fin area here is the area of the cuboid above the patch and it is given by
S = 2 ∗ ( fh ∗ fw + fh ∗ ft ) + fw ∗ ft (4.1)
S t ot = N ∗ S (4.2)
where,
S - Fin area [mm 2 ]

f h - Fin height [mm]
51
4.1. Parametric Study With Fin Dimensions 52
f t -Fin thickness [mm]

f w - Fin width [mm]
S t ot - Total surface area above the patch [mm 2 ]
N - Number of fins
For various heatsink dimensions, co-pol directivity, cross-pol directivity, and junction
temperature are investigated. The EM parameters are calculated after the designed an-
tennas have been properly matched. These results will help in the selection of the best
fin dimensions for the design.
Fin Height
The fin height is varied from 1mm to 4mm while the number of fins is kept at 4 and
the fin thickness is kept at 0.5 mm. Both the conduction-based and the air flow models
are simulated to evaluate the junction temperature. Figures 4.1, 4.2 gives the junction
temperature values based on conduction model and air flow model respectively.
(a) (b)
Figure 4.1: Conduction Based model-Fin height vs (a)Co-pol,(b)Cross-pol
(a) (b)
Figure 4.2: Air flow Model - Fin height vs (a)Co-pol,(b)Cross-pol

When the fin height was varied from 1 to 1.5 mm, the co-pol directivity increased at
first, but then declined as the fin height was increased further. However, the cross-pol
directivity increased as the fin height is increased. For better performance, high co-pol
and low cross-pol directivity are desirable. Both can be achieved by having reduced
fin height for the heatfins, but the junction temperature will be high. Increase in fin
height increases the surface area thereby reducing the surface temperature. This leads
to a trade-off between EM and thermal performance. The junction temperature drops
even more with the increase in fin height in the air flow model, which could be related
to the geometry-induced increase in h.t.c. Based on the foregoing findings, a fin height
of 2mm would be ideal, since it has a good balance of EM and thermal performance.
Fin Thickness
Here, the fin thickness is changed from 0.2 mm to 1 mm , keeping f h = 2mm and N = 4.
Figures 4.3, 4.4 gives the junction temperature values based on conduction model and
air flow model respectively.
(a) (b)
Figure 4.3: Conduction Based model-Fin thickness vs (a)Co-pol,(b)Cross-pol
(a) (b)
Figure 4.4: Airflow model-Fin thickness vs (a)Co-pol,(b)Cross-pol

Figure 4.3 follows the same trend as with the fin height. However, as shown in Figure
4.4, the findings of the air flow model showed an increase in junction temperature as
fin thickness increased. The junction temperature rises despite the fact that the surface
area increases slightly. This is because the h.t.c is also determined by the fin spacing.
The greater the spacing, the greater the air flow, which results in an increase in con-
vective heat transfer [28]. The spacing between the fins reduces as the fin thickness
or number of fins increases, according to Eq.(2.5). As a result, increased fin thickness
leads to reduced fin spacing, resulting in a lower h.t.c.
Number of fins
Figures 4.5 and 4.6 shows the results based on Conduction and Air flow model respec-
tively.
(a) (b)
Figure 4.5: Conduction Based model-Number of fins vs (a)Co-pol,(b)Cross-pol
(a) (b)
Figure 4.6: Airflow model-Number of fins vs (a)Co-pol,(b)Cross-pol
The number of fins varies from 3 to 8, while the fin height and thickness remain stable
at 2mm and 0.5mm, respectively. With N=4 fins, the co-pol directivity is highest and
4.2. Finite Array Implementation 55
the cross-pol directivity is lowest. The surface area increases as the number of fins
increases, but the spacing reduces, reducing h.t.c. As a result, ideal number of fins
would be 4. Additional parametric studies with different fin dimensions can be seen
in Table D.1 in Appendix D, where one can observe that the conduction based model
shows a reduce in junction temperature for increase in surface area (increase in fin
dimensions), whereas the airflow model takes in to account the spacing and its effects
on h.t.c as previously stated.
4.2. Finite Array Implementation
In this part, a finite array heatsink antenna is developed and compared to the patch
antenna, which serves as the benchmark, in terms of realized gain, junction temper-
ature, and side lobe level. The scanning abilities of the heatsink antenna have been
investigated for both linear and planar arrays. The array with 0.5λ spacing produced
pattern distortion because the heatsink antenna elements are relatively close to each
other, as explained in Chapter 2. In order to construct the array with standard spacing
of 0.5λ, the patch size is reduced and FR-4 substrate with ²r = 4.3 is used. The thermal
conductivity of the FR-4 and Rogers RT5880 are comparable.
Since the permittivity of the FR-4 substrate and the loss tangent are high, fringing
fields are tightly contained resulting in low radiation efficiency . For the patch antenna
the efficiency is about 66.2% and for the heatsink antenna, the efficiency is improved
to 80.6% because the electric field is less confined in the substrate, the surrounding
medium has a greater influence on the effective dielectric constant [7]. Also, the
change in the substrate had significant effect on the co-pol and cross-pol directivity.
The co-pol directivity of the heatsink antenna is reduced from 9.096 to 7.4 dBi and the
x-pol directivity is reduced from -1.39 to -6.06 dBi. The fin dimensions are chosen as
N = 4, f h = 2mm, and f t = 0.2mm based on the results obtained in the previous section
assuming that a similar trend holds for this case with the change in substrate.
4.2.1. Linear Array Implementation
Both x-oriented and y-oriented 8-element linear heatsink antenna array with 0.5λ spac-
ing are constructed. As previously stated, the patch antenna is used as a bench mark.
Simultaneous excitation is used where all the ports/sources are excited simultaneously
according to the excitation list settings.
X-oriented Linear Array
The array design, return loss and the 3D pattern realized gain plots for both the patch
array and heatsink array are shown in Figures 4.7, 4.8, and 4.9 respectively.
(a) (b)
Figure 4.7: Array Topology, (a) 1 x 8 Patch, (b) 1 x 8 heatsink antenna
(a) (b)
Figure 4.8: Return loss S11 (dB) of 1 x 8 array,(a)Patch,(b)Heatsink antenna
(a) (b)
Figure 4.9: Realized gain (dB), (a) Patch, (b) heatsink antenna
In comparison to the patch, the heatsink antenna array provides a higher realized gain
and more surface area for heat removal. The thermal model will have two heat sources
with the 8-element array, and the temperature results will be comparable to those pro-
duced with the 4-element linear array.
Both the arrays have a side lobe level (SLL) of roughly -13.5 dB. Using the Taylor or
Chebyshev window, amplitude tapering can be used to further minimize the side lobe.
The amplitude values for side lobe suppression for 20 and 30 dB are obtained using
Chebyshev windowing. The amplitude values can be seen in Appendix E. Figure 4.10
shows the farfield realized gain (φ = 0◦ ) cut for the designed array. As the amplitude
tapering is introduced, we can see the reduction in Side Lobe Levels but the main
lobe becomes wider making the power spread over larger angle i.e. beamwidth be-
comes greater which results in less directivity. Table 4.1 shows the performance of the
x-oriented array in terms of realized gain, SLL and junction temperature.
Figure 4.10: Farfield realized gain (φ = 0)
X-oriented Realized Gain SLL Junction Temperature

antenna array (dB) (dB) (◦ C )
Patch array 13.12 -13.5 124.37
Heatsink array 14.14 -13.4 104.21
Tapered heatsink array
13.8 -21.1 104.21
(20dB suppression)
13.2 -33.0 104.21
(30dB suppression)
Table 4.1: Comparison of joint EM-thermal performance
The heatsink array’s scanning abilities were investigated by applying appropriate phase
shifts between the antenna elements to direct the beam at specified angles. For phase
shift calculations, see Appendix E. Figure 4.11 shows the realized gain for different scan
angles. The x-oriented heatsink array is capable of scanning up to 45 degrees eleva-
tion. At 60◦ elevation angle, the matching is worse and the realized gain goes down to
8.2 dB. The realized gain decreases and the side lobe level increases as the scan angle
increases.
Figure 4.11: Farfield realized gain (φ = 0) for different scan angles

Y-oriented Linear Array
The return loss plots and the realized gain for the y-oriented patch and heatsink array
are shown in Figures 4.12 and 4.14
(a) (b)
The middle elements 2-7 resonate at a higher frequency, while 1 and 8 resonate at 28
GHz, as seen in the heatsink array’s return loss plot. The patch length for elements 2
to 7 is raised from 2.26 mm to 2.35 mm to shift the resonant frequency, making it a
hetreogenous array.
Figure 4.13: Return loss S11 (dB) of 1 x 8 heterogenous heatsink antenna array
(a) (b)
Figure 4.14: Realized gain (dB), (a) Patch, (b) heatsink antenna
The y-oriented heatsink array achieves a realized gain of 13.2 dB, while the patch array
achieves 11.9 dB. The side lobe level for both the arrays are about -13.3dB. Amplitude
tapering is applied to reduce this sidelobe further.
Y-oriented Realized Gain SLL Junction Temperature

antenna array (dB) (dB) (◦ C )
Patch array 11.9 -13.3 124.37
Heatsink array 13.2 -13.3 104.09
13 -21.0 104.09
(20dB suppression)
12.4 -27.9 104.09
(30dB suppression)
Table 4.2: Comparison of joint EM-Thermal performance
By applying appropriate phase shifts between the antenna elements to direct the beam
at specified angles, the scanning abilities of the heatsink array were investigated. The
y-oriented heatsink array is capable of scanning up to 30 degrees elevation. At 45◦ ele-
vation angle, the matching is worse and the realized gain is around 6.5 dB. The realized
gain decreases as the scan angle increases.
Figure 4.16: Farfield realized gain (φ = 90) for different scan angles
4.2.2. Planar Array Implementation
In this section a 4 x 4, 2D array has been implemented. Figures 4.17, 4.18, and 4.19
show the array topology, Return loss and the 3D Realized Gain of the designed patch
and heatsink 2D arrays.
(a) (b)
Figure 4.17: Array topology, (a) 4 x 4 patch, (b) 4 x 4 heatsink
(a) (b)
(a) (b)
Figure 4.19: Realized Gain (dB), (a) Patch, (b) heatsink antenna
2D heatsink array designed here is a heterogeneous array with middle elements 5-12
having a patch length of 2.45mm and the other elements with a patch length of 2.26
mm. This is done inorder to shift the resonant frequency near to 28 GHz. The 2D
heatsink array provides slighly higher realized gain of 15.3 dB with a comparable side
lobe level when compared to the 2D patch array. Amplitude tapering is applied here
with 0.5 for the 4 center elements and 1 for the other elements to reduce the side lobe
level. The comparison is given in Table 4.3. The farfield realized gain for both the cut
plane φ = 0◦ and φ = 90◦ are shown in Figures 4.20 and 4.21 and the conduction-based
temperature plots for the array are shown in Figure 4.22. The junction temperature
for the tapered heatsink array is the same as that of the non-tapered heatsink antenna
array as this is computed under the assumption of no heat power change with modified
power amplifier output (which is needed for amplitude tapering).
Realized Gain SLL

2D (4 x 4) Junction temperature
(dB) (dB)
antenna array (◦ C )
φ = 0◦ φ = 90◦ φ = 0◦ φ = 90◦
Patch array 15 15 -14.2 -13.2 123.2
Heatsink array 15.3 15.3 -13.7 -11.6 108.21
Tapered heatsink
14.7 14.7 -19.0 -17.2 108.21
array
Table 4.3: Comparison of joint EM-Thermal performance (4 x 4) array

4.3. Conclusion-Electro-Thermal Aspects 62
(a) (b)
Figure 4.22: Conduction based model - Temperature plots, 2D array (a)Patch, (b) Heatsink
As a study case, progressive phase shifts are applied for the array to steer the beam at
(θs , φs ) = 30◦ , 45◦ . The realized gain for the scanning is shown in Figure 4.23. The Real-
ized Gain is reduced to 13.9 dB. In case of the array, high cross pol might be amplified
by the array gain if the steering angle is in line with the cross pol angle. Otherwise it
may be suppressed.
Figure 4.23: 2D Heatsink antenna array with scan angle (θs , φs ) = 30◦ , 45◦
4.3. Conclusion-Electro-Thermal Aspects

The joint EM and thermal issues are discussed in this chapter. Optimal fin dimensions
for array design are chosen based on EM and thermal performance. The results of
the conduction-based model show that increasing surface area enhances heat transfer,
however in the air flow model, even though the surface area rises, heat transmission
does not, and this is due to the fact that the h.t.c also varies in Eq.3.3. N = 4, f h =
2mm, and f t = 0.2mm is chosen as it achieves a good balance between EM and thermal
4.3. Conclusion-Electro-Thermal Aspects 63
performance. Both linear (1 x 8) and planar (4 x 4) arrays are implemented. The x-

oriented (1 x 8), Y-oriented (1 x 8) and planar(4 x 4) heatsink antenna arrays provided a
greater realized gain and heat dissipation with a similar side lobe level when compared
with that of the linear patch array. Even for the tapered heatsink antenna case, the
realized gain is slightly higher than that of the patch for the linear arrays giving the
advantage of low side lobe and low temperature.
5
Conclusions And Future Work
5.1. Conclusion
The goal of this thesis was to design a low-temperature, low-side-lobe mm-Wave heatsink
antenna element and its array. The thesis was broken down into four sections. The first
section contains a comprehensive literature review of existing antenna thermal man-
agement technologies. The heatsink antenna design phase was then completed, with
its electromagnetic performance being assessed. The thermal model for the antenna
was then included, and the simulated temperature values were compared to those of a
standard patch antenna. Finally, suitable heatsink dimensions are chosen for the con-
struction of finite arrays in order to improve their EM and thermal performance.
Thermal management in antennas was divided into three sections: antenna+heatsink,

antenna as heatsink, and antenna layout optimization, based on a thorough review
of the literature. The majority of antenna thermal management work is done at low
frequencies. Due to the growing popularity of mm-Wave design and the importance
of heating effects at high frequencies, a few research studies have recently been pub-
lished. However, with a few exceptions, none of the works provide a proper thermal
model that takes into account the relevant chip parameters. Furthermore, the thermal
simulations reported in all these research papers are primarily conduction-based or
conduction-based with h.t.c assumptions that are typically unknown. In this thesis,
for the first time, a novel mm-Wave heatsink antenna design was developed, com-
plete with thermal modeling and natural convective CFD simulation setup.
An mm-Wave patch antenna was initially created and a heatsink was placed on top of
it to see its effects. Bandwidth, radiation efficiency, and co-pol directivity have all im-
proved. However, in this case, the cross-pol directivity was high. The heatsink base was
removed, resulting in a significant increase in co-pol directivity and a modest decrease
in cross-pol directivity. Then, different orientations of the rectangular fin and radial
pin fin heatsinks was designed and the EM performance was tabulated. The rectan-
gular fins with fins parallel to the patch’s non-radiating edges (PNRE) and the Radial
pin fins at the patch’s center both performed well. As a result, these configurations are
65
5.1. Conclusion 66
taken into account for additional EM and thermal analysis. Infinite array approxima-
tion was used to investigate the impact of these antennas on a very large array. Fins
PNRE design was capable of realization in a very large x-oriented linear array with a
scan range of ±15◦ , while the radial fins (center) design was capable of realization in
a 2D array with ±60◦ , according to infinite array approximation employing periodic
boundary conditions.
The antenna’s thermal model was developed after it was analyzed in terms of EM char-
acteristics. Two-resistor model was investigated, having the necessary beamformer
chip features such as heating power, junction-to-board, and junction-to-case resis-
tance values. Because a single beamformer chip can feed four antenna elements, ther-
mal simulations for a four-element (linear or two-dimensional array) were performed.
There were two simulation methodologies evaluated. One was a conduction-based
model with a convective cooling boundary condition where an h.t.c value is assumed
for a natural convective scenario. The other was a natural convective laminar air flow
model, in which heat is transferred via gravity. For this simulation, fluid flow and heat
transfer physics were coupled, and the h.t.c value was evaluated rather than assumed.
Several case studies were conducted using the conduction-based model to examine
the effects on junction temperature. The first case study involved increasing h.t.c at
the heat source’s bottom. The junction temperature was found to be lower when h.t.c
was increased. When compared to patch and Radial fins (center) heatsink antennas,
the fins PNRE antenna showed greater heat dissipation. The purpose of the second
case study was to determine whether the chip’s attributes have an impact on thermal
performance. The temperature was dropped by 4◦C when the chip’s substrate material
was changed from RogersRT5870 to Aluminium Nitride. Heat dissipation was found to
be improved by lowering the junction-to-board and junction-to-case resistances. Fi-
nally, a case study based on radiative heat transfer was done where the effects of ther-
mal radiation has been observed. In the end, a comparison was made between the two
simulation models where it was found that the temperature results were different for
the patch antenna. This was due to the difference in heat transfer coefficient values.
In the conduction model, the junction temperature of the heatsink antenna array was
decreased by 10◦C , while the air flow model revealed a 21.6◦C degree drop when com-
pared to the patch antenna array.
The major challenge of this thesis was about the trade off between EM and thermal
performance of the designed heatsink antenna. As a result, it was critical to choose
the right fin dimensions so that the designed antenna has a suitable EM and thermal
balance. Parametric study with fin dimensions was made to choose the proper dimen-
sions. The conduction based model results showed decrease in temperature whenever
the fin dimension was increased as this increases the surface area. However, the air
flow model accounts for the effects on spacing on heat transfer coefficient. Based on
the air flow model, the dimensions were chosen for the array design. Both linear and
planar arrays were investigated. For the x-oriented (1 x 8), y-oriented (1 x 8) linear ar-
rays, and (4 x 4) planar array the heatsink antenna produced greater realized gain and
heat dissipation while maintaining the same side lobe level. This side lobe level was
further reduced by amplitude tapering making the heatsink antenna, a low-sidelobe
5.2. Future work 67
and low-temperature alternative to the patch antenna.
5.2. Future work

The benefits of designed heatsink antenna in dissipating heat while maintaining good
EM characteristics have been emphasized in this thesis. The following are potential
future projects that could be pursued as a result of this thesis:
• Heatsink Antenna Optimization: Optimum heatsink dimensions are selected only

based on the parametric sweep that is done in this thesis. However, this may not
be the best as the parametric sweep for one parameter considered here is only a
small range while keeping the other parameters constant. Because manual param-
eter sweeping has limitations, an optimization process must be used to determine
near-optimal values for the design parameters. To generate the results, numerous
parameters are varied at the same time.
For the optimization, the critical parameters of the design has to be selected. The
values and range of the critical parameters must be given. In our case, the crit-
ical parameters would be the fin dimensions and the patch dimensions. Then,
geometrical and performance constraints have to be specified. Finally, an opti-
mization goal function has to be selected. The goal function in this situation could
be to increase the surface area while keeping the constraints in place. One of the
major constraints here would be spacing between the fins as it affects the h.t.c. For
this work, we are collaborating with Dr.Bo Liu, Senior Lecturer (Systems Power En-
ergy) at the University of Glasgow, and Dr.Mobayode Akinsolu, Assistant Professor
at the Wrexham Glyndwr University who has developed the optimizer. The work
has already started where goal function, constraints and critical parameters were
selected. We are hoping to finish the optimization to obtain the best parameters
for the design that would maximize the surface area and then simulate the design
with the conduction and air flow model developed in this thesis to see its thermal
performance.
• Wideband Array: When a ring is inserted in the air region of the coax, we observed
a wide band property for the Y-oriented infinite heatsink array. Impedance match-
ing may be examined further by creating very large finite arrays instead of using
Periodic Boundary conditions.
• Heat Transfer Enhancement: Due to the antenna substrate’s resistance to the ma-
jority of heat flow to the antenna side, metallic vias or other thermal enhancement
techniques can be used to increase heat flow to the antenna side. For the designed
antenna, forced convection with liquid cooling or cooling via external fans can also
be tested.
• Realization Of Aperiodic Heatsink Array: Aperiodic heatsink arrays can be cre-

ated at optimized positions instead of amplitude tapering to minimize the side
5.2. Future work 68
lobe level. Because the antenna effective aperture is large, this decreases side lobe
while maintaining a higher directivity. This larger effective aperture can dissipate
more heat than a regular array.
• Dual-polarized Antenna/Feeding Techniques: Only single-polarization elements

are considered in this thesis, so antenna with dual-polarization can be implemented.
Also, different antenna feeding mechanisms can be investigated.
• Different Heatsink Geometry: Alternating layout such as the triangular heat sink
which can provide better thermal performance can be designed above an antenna
and EM performance can be investigated.
A
Extended Heatsink Geometry
Here, the heat fins are extended beyond the patch length (along y) where L p = 3mm
and the fin width (along y) f w = 4.5mm as shown in Figure A.1. The 3D pattern direc-
tivity is shown in Figure A.2. where several nulls are observed in the distorted pattern
with lower directivity compared to when the heatsink geometry is bounded within the
patch size.
Figure A.1: Extended heatsink geometry (along y)
Figure A.2: 3D Pattern Directivity (dBi)
The heatsink geometry extended along x i.e, 5 fins instead of 4 fins exceeding the width
of the patch (along x) did not have much impact on the pattern directivity but when
the antenna element is reliazed in array the heatfins will be very close to each other
causing mutual coupling effects.
71
B
Calculations For The Thermal Model
The calculations for the assumptions and conditions for the air flow model are pro-
vided in this section.
The hydraulic diameter is calculated using Eq.3.4 for designing the air box and esti-
mating fluid velocity. In contrast to circular pipes, where the hydraulic diameter is
simply twice the circle’s radius, the hydraulic diameter here is computed using the
cross-sectional area and the wetted perimeter. Wetted perimeter is the portion of the
perimeter touching the fluid. Figure B.1 shows the dimensions for the cross-sectional
area.
Figure B.1: Cross-section of the heatsink antenna array
Here the cross-section can be taken as a rectangle box and the hydraulic diameter can
be calculated as
4A 4∗lb 4 ∗ 0.0234 ∗ 0.0212
D= = = = 0.0222m (B.1)
P 2 ∗ (l + b) 2 ∗ 0.0446
Based on this, the air box is constructed where the cuboid dimensions are 40 x 114.75
x 40 mm. The fluid velocity and the flow regime is computed. The thermophysical
properties of air at 20◦C is shown in Table B.1
72
73
Thermal Thermal Dynamic Kinematic Prandtl

Temp Density Specific heat
Conductivity Diffusivity Viscosity Viscosity Number
(◦ C ) ρ (kg /m 3 ) c ρ (J /K g .K )
k(W /m.K ) α(m 2 /s) µ(K g /m.s) v (m 2 /s) Pr
20 1.204 1007 0.02514 2.074 x 10−5 1.825x10−5 1.516x10−5 0.7309
Table B.1: Thermophysical properties of Air
According to [29], typical velocity due to buoyancy force is calculated as

q
U= g αp ∆T L (B.2)
where,
g - Acceleration due to gravity = 9.81 [m.s −2 ]

αp - Coefficient of thermal expansion = 3.43 x 10−3 [K −1 ]
L - Hydraulic diameter = 0.0222 [m]
∆T - Typical temperature difference = 343.15 [K ]
The velocity estimate is found to be 0.5m/s and the simulated velocity slice for the
heatsink array shows a closer value of 0.48m/s. The velocity slice plot is shown in Fig-
ure B.2
Figure B.2: Velocity slice - heatsink antenna array
The Reynolds number and the Grashof number are used to determine whether the
flow is laminar, transitional, or turbulent. The Reynolds number is the ratio of inertial
to viscous forces, and it is calculated as follows
ρV D
R= (B.3)
µ
where,
R - Reynolds Number
ρ - Fluid density = 1.204 [K g /m 3 ]
D - Hydraulic diameter or distance = 0.0222 [m]
µ - Dynamic Fluid viscosity = 1.825 x 10−5 [K g /m.s]
74
Reynolds number is found to be 697 which means the flow is laminar where the fluid
follow smooth path in layers and has a parabolic velocity profile. For R values less than
2000, the flow is laminar and the transition regime is between 2000 and 104 [30].
The Grashof number, which is equal to the ratio of buoyant to viscous forces, is given
by
αp g ∆T L 3
Gr = (B.4)
ν2
where,
Gr - Grashof number
αp - Coefficient of thermal expansion = 3.43 x 10−3 [K −1 ]
L - Hydraulic diameter or distance = 0.0222 [m]
∆T - Typical temperature difference = 343.15 [K ]
ν - Kinematic Viscosity = 1.516 x 10−5 m 2 /s
For natural convection from vertical flat plates, the transition to turbulent flow oc-
curs in the range 108 < Gr < 109 . The boundary layer is turbulent at higher Grashof
numbers and at lower Grashof numbers, the boundary layer is laminar in the range
103 < Gr < 106 [31]. The Grashof number is found to be 5.497 x 105 which indicates
the flow is laminar.
Having known the flow regime, laminar fluid flow physics is selected in COMSOL and
the gravity is activated. To determine the compressibility, Mach number is calculated
which is a dimensionless number. It is the ratio of fluid velocity to the velocity of sound.
u
M= (B.5)
c
where,
M - Mach number
u - Fluid velocity = 0.5 [m/s]
c - Speed of sound in the medium = 343 (20◦C ) [m/s]
Here, M < 0.2−0.3 which indicates incompressible fluid flow where material density is
constant within the fluid and incompressible Navier-Stokes equation can be used.
Extending Ground plane For Heat
C
Transfer
The major part of the heat is transferred through the bottom heatsink attached directly
to heat source. Above the chip, the chip substrate and ground plane are highly con-
ductive and then comes the antenna substrate which has lower thermal conductivity
of about 0.2W /m.K . So, one way to dissipate the heat could be to extend the ground
plane and attach heatsink at the bottom of the ground plane as shown in Figure C.1
(a) (b)
Figure C.1: Extended groundplane (a) Patch, (b) Heatsink antenna
Extending the ground plane and attaching heatsink at the bottom of the ground plane
has reduced the maximum junction temperature from 116◦C to 99.6◦C for the patch
and from 106◦C to 91.5◦C for the heatsink antenna array. The difference in junction
temperature between patch and heatsink antenna has been reduced from 10◦C to 8◦C .
This difference will further reduce if you increase the heatsink dimensions at the bot-
tom of the chip or the ground plane as most of the heat flows through it.
75
D
Additional Parametric Study
Spacing Surface area Conduction Air flow Co-Pol Cross-Pol

Fin thickness Number of
between fins above the patch Model Model Dir Dir
(mm) Fins
(mm) (mm^2) (c) (c) (dBi) (dBi)
0.5 9 0.033 145.98 101.017 107.520 8.873 0.29
9 0.259 133.09 101.95 107.327 9.025 -0.1916
10 0.197 147.88 101.33 107.454 8.991 -0.0072
0.3
11 0.147 162.67 100.75 107.393 8.969 0.1116
12 0.106 177.46 100.18 107.521 9.024 -0.2574
12 0.325 160.27 101.38 107.217 9.074 -1.636
0.1 16 0.21 213.7 99.51 107.654 8.923 -1.182
20 0.146 267.12 97.8 107.745 8.944 -0.7938
Table D.1: Parametric study data
76
E
Amplitude And Phase Shift Calculations
Using Chebwin function in MATLAB, the amplitude values for tapering is obtained.
chebwi n(N , SLL) is the function where N is the array size and SLL is the suppression
factor. The amplitude values are entered in the excitation list in CST. For 20 dB suppres-
sion, the amplitude values are [0.5799, 0.6603, 0.8751, 1.0, 1.0, 0.8751, 0.6603, 0.5799] and
for 30 dB suppression, the values are [0.2622, 0.5187, 0.8120, 1.0, 1.0, 0.8120, 0.5187, 0.2622].
Progressive phase shifts has to be entered in order to steer the beam to a specific angle.
For linear arrays, the progressive phase shift is calculated by Eq.E.1
β = −kd si nθs (E.1)
where,
β - Progressive phase shift between elements
k - Wavenumber = 2π −1
λ [m ]
d - spacing between elements = 0.5λ [m]
θs - Scan angle ◦
β for scan angles 15, 30, 45, and 60 degrees are calculated as −46.58, −90, −127.28 and
−155.88 degrees respectively.
Because the steering is in 2D for Planar arrays, the progressive phase shift along x and
y is calculated individually using Eq.E.2 and E.3
βx = −kd x si nθs cosφs (E.2)
β y = −kd y si nθs si nφs (E.3)
where,
βx , β y - Progressive phase shift between elements along x
and y
k - Wavenumber = 2π λ
[m −1 ]
d x = d y - spacing between elements along x and y = 0.5λ [m]
θs , φs - Scan angle ◦
For θs = 30◦ elevation and φs = 45◦ azimuth, βx = β y = −63.64◦
77
Bibliography
[1] 5. Americas, “Understanding mmwave spectrum for 5g networks,” White paper,

2020.
[2] Y. Aslan, J. Puskely, J. Janssen, M. Geurts, A. Roederer, and A. Yarovoy, “Thermal-
aware synthesis of 5g base station antenna arrays: An overview and a sparsity-
based approach,” IEEE Access, vol. 6, pp. 58 868–58 882, 2018.
[3] V. Lakshminarayanan and N. Sriraam, “The effect of temperature on the reliabil-
ity of electronic components,” in 2014 IEEE international conference on electron-
ics, computing and communication technologies (CONECCT), IEEE, 2014, pp. 1–
6.
[4] B. Döring, “Cooling system for a ka band transmit antenna array,” Dec. 2005.
[5] Y. Aslan, C. E. Kiper, A. J. van den Biggelaar, U. Johannsen, and A. Yarovoy, “Pas-
sive cooling of mm-wave active integrated 5g base station antennas using cpu
heatsinks,” in 2019 16th European Radar Conference (EuRAD), IEEE, 2019, pp. 121–
124.
[6] L. Covert and J. Lin, “Simulation and measurement of a heatsink antenna: A
dual-function structure,” IEEE transactions on Antennas and Propagation, vol. 54,
no. 4, pp. 1342–1349, 2006.
[7] L. Covert, J. Lin, D. Janning, and T. Dalrymple, “5.8 ghz orientation-specific extruded-
fin heatsink antennas for 3d rf system integration,” Microwave and Optical Tech-
nology Letters, vol. 50, no. 7, pp. 1826–1831, 2008.
[8] ——, “Dual-function 3-d heatsink antenna for high-density 3-d integration,” in
2007 IEEE International Workshop on Radio-Frequency Integration Technology,
IEEE, 2007, pp. 26–29.
[9] A. Alnukari, P. Guillemet, Y. Scudeller, and S. Toutain, “Active heatsink antenna
for radio-frequency transmitter,” IEEE transactions on advanced packaging, vol. 33,
no. 1, pp. 139–146, 2009.
[10] A. Alnukari, S. Toutain, Y. Scudeller, and P. Guillemet, “Microstrip heatsink an-
tenna for integrated rf transmitters,” in Proceedings of the Fourth European Con-
ference on Antennas and Propagation, IEEE, 2010, pp. 1–5.
[11] A. Alnukari, Y. Mahe, S. Toutain, and Y. Scudeller, “Microstrip heatsink antenna
cooled with sapphire layer for integrated rf transmitters,” in 2012 6th European
Conference on Antennas and Propagation (EUCAP), IEEE, 2012, pp. 1263–1266.
[12] J. J. Casanova, J. A. Taylor, and J. Lin, “Design of a 3-d fractal heatsink antenna,”
IEEE Antennas and Wireless Propagation Letters, vol. 9, pp. 1061–1064, 2010.
[13] M. Rao, N. Oraon, and P. Sravani, “Self assembly based high frequency 3d heatsink
antenna,” in 2014 IEEE REGION 10 SYMPOSIUM, IEEE, 2014, pp. 10–14.
78
Bibliography 79
[14] S. M. Duffy, G. A. Brigham, K. B. Newman, and J. S. Herd, “Stepped notch antenna

array used as a low thermal resistance heat sink,” in 2013 IEEE Antennas and
Propagation Society International Symposium (APSURSI), IEEE, 2013, pp. 622–
623.
[15] NEC. (2018). “Antenna heat dissipation technology that enables miniaturization
of wireless units for 5g,” [Online]. Available: https : / / www . nec . com / en /
global/rd/technologies/201803/index.html (visited on 01/22/2021).
[16] Y. Aslan, J. Puskely, A. Roederer, and A. Yarovoy, “Heat transfer enhancement in
passively cooled 5g base station antennas using thick ground planes,” in 2019
13th European Conference on Antennas and Propagation (EuCAP), IEEE, 2019,
pp. 1–5.
[17] J. Standard, “Two-resistor compact thermal model guideline,” JESD15-3, JEDEC
Solid State Technology Association, 2008.
[18] J. Park, D. Choi, and W. Hong, “Millimeter-wave phased-array antenna-in-package
(aip) using stamped metal process for enhanced heat dissipation,” IEEE Anten-
nas and Wireless Propagation Letters, vol. 18, no. 11, pp. 2355–2359, 2019.
[19] Y. Aslan, A. Roederer, and A. Yarovoy, “System advantages of using large-scale
aperiodic array topologies in future mm-wave 5g/6g base stations: An interdis-
ciplinary look,” IEEE Systems Journal, 2021.
[20] J. Qian, M. Tang, Y.-P. Zhang, and J. Mao, “Heatsink antenna array for millimeter-
wave applications,” IEEE Transactions on Antennas and Propagation, vol. 68, no. 11,
pp. 7664–7669, 2020.
[21] W.-X. Chu, Y.-C. Lin, C.-Y. Chen, and C.-C. Wang, “Experimental and numerical
study on the performance of passive heat sink having alternating layout,” Inter-
national Journal of Heat and Mass Transfer, vol. 135, pp. 822–836, 2019.
[22] L. Zhou, M. Tang, J. Qian, Y.-P. Zhang, and J. Mao, “Vivaldi antenna array with
heat dissipation enhancement for millimeter-wave applications,” IEEE Transac-
tions on Antennas and Propagation, 2021.
[23] D. Pozar, “Considerations for millimeter wave printed antennas,” IEEE Transac-
tions on antennas and propagation, vol. 31, no. 5, pp. 740–747, 1983.
[24] S. Shidore, “Compact thermal modeling in electronics design,” Electronics cool-
ing, vol. 13, no. 2, p. 22, 2007.
[25] Y. Aslan, “Antenna array synthesis and beamforming for 5g applications: An in-
terdisciplinary approach,” 2020.
[26] D. Nicholls, System reliability toolkit. Riac, 2005.
[27] Gabrian. (2016). “Anodized aluminum heatsinks: What you need to know,” [On-
line]. Available: https://www.gabrian.com/anodized-aluminum-heatsinks-
what-you-need-to-know (visited on 05/22/2021).
[28] M. Srikanth, “Optimization of vertical heat sink fin spacing for natural convec-
tion,” in AIP Conference Proceedings, AIP Publishing LLC, vol. 2200, 2019, p. 020 054.
Bibliography 80
[29] C. M. 5.6. (2020). “Buoyancy flow in water,” [Online]. Available: https://www.

comsol.nl/model/download/819061/models.heat.buoyancy_water.pdf
(visited on 06/22/2021).
[30] C. Kothandaraman, Fundamentals of heat and mass transfer. New Age Interna-
tional, 2006.
[31] A. Bejan and A. D. Kraus, Heat transfer handbook. John Wiley & Sons, 2003, vol. 1.

Heat Sink Antennas

Uploaded by

Copyright:

Available Formats

Heat Sink Antennas

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Heat Sink Antennas

Uploaded by

Copyright:

Available Formats

mm-Wave Heatsink Antenna

Array Design for Low-Sidelobe

Student number: 5142679

An electronic version of this thesis is available at http://repository.tudelft.nl/.

In addition, I’d want to express my gratitude to all of my Microwave Sensing, Signals,

2 Heatsink Antenna Design - EM Aspects 17

3 Heatsink Antenna Design - Thermal Aspects 35

4 Heatsink Antenna - Electro-Thermal Co-Design 51

2.1 CST Model of Coaxial fed patch antenna . . . . . . . . . . . . . . . . . . . 18

2.9 Return loss S11(dB) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

4.1 Conduction Based model-Fin height vs (a)Co-pol,(b)Cross-pol . . . . . . 52

A.1 Extended heatsink geometry (along y) . . . . . . . . . . . . . . . . . . . . . 71

B.1 Cross-section of the heatsink antenna array . . . . . . . . . . . . . . . . . 72

C.1 Extended groundplane (a) Patch, (b) Heatsink antenna . . . . . . . . . . . 75

1.1 Comparison of different heatsink antenna topologies . . . . . . . . . . . . 13

2.1 Comparison of EM performance . . . . . . . . . . . . . . . . . . . . . . . . 20

3.1 Chip parameters from [5] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.1 Comparison of joint EM-thermal performance . . . . . . . . . . . . . . . 57

B.1 Thermophysical properties of Air . . . . . . . . . . . . . . . . . . . . . . . . 73

D.1 Parametric study data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

1.1. Research Problem Description And Objectives

The main objectives of the thesis are as follows.

4. To compare the electromagnetic and thermal performance of the designed heatsink

1.2. Literature Review

affect the electromagnetic performance. Thermal simulations using Finite element

In [12], J.Casonova et al.(2008) suggest an increase in surface area of the heatsink to

Antenna as a Heatsink / Layout optimization as a heatsink:

Table 1.1: Comparison of different heatsink antenna topologies

Based on the comparison of different heatsink antennas made in [20],[22] a compar-

• The proper mm-wave thermal modeling in active integrated antennas is missing

• The usefulness and performance trade-offs of heatsink+antenna concepts in mm-

• The optimality of the proposed heatsink antennas in terms of joint electromagnetic

• There is lack of studies in performance investigation of heatsink antennas in array

• The effects of heatsink antennas on cross-polarization and scanning abilities of the

• At mm-Wave frequencies, the interference problem due to high sidelobes in heatsink

1.3. Research Approach and Novelty

• A novel thermal model considering appropriate beamformer chip parameters is de-

1.4. Thesis Outline

Chapter 4 presents electro-thermal parametric studies with the heatsink dimensions

2.1. mmWave Patch Antenna Design With Heatsink On Top

Firstly, a microstrip patch antenna is designed resonating at 28 GHz frequency and

Figure 2.1: CST Model of Coaxial fed patch antenna

Figure 2.2: CST model of patch antenna with heatsink on top

2.1.1. Comparison With The Regular Patch Antenna

Figure 2.3: Return loss S11 (dB),(a)Patch antenna,(b)Heatsink antenna

Parameters Patch Antenna Patch antenna with heatsink

Table 2.1: Comparison of EM performance

2.1.2. Analysis Of X-pol Directivity With Heatsink Dimension

Figure 2.6: Maximum cross pol Directivity (dBi) vs Frequency

Figure 2.7: Maximum Cross-Pol Directivity (dBi) vs Frequency

2.1.3. Effect Of Removing Heatsink Base

This modified heatsink antenna design showed improvement in cross-polarization and

Figure 2.9: Return loss S11(dB)

2.2. Effects Of Geometry And Orientation Of The Heatfins

Peak Directivity Peak Gain Radiation

Patch Antenna with Patch Antenna with Patch Antenna with

Table 2.3: EM performance comparison of the selected heatsink orientations

2.3. Pattern Directivity With Fin Dimensions

Figure 2.12: Pattern Directivity vs fin height

Figure 2.13: Pattern Directivity vs Number of fins