Professional Documents
Culture Documents
Link Budget Parametrs
Link Budget Parametrs
2019-10-07
David Frykskog
Hjalmar Jonsson
Norrköping 2019-10-07
Upphovsrätt
Copyright
The publishers will keep this document online on the Internet - or its possible
replacement - for a considerable time from the date of publication barring
exceptional circumstances.
The online availability of the document implies a permanent permission for
anyone to read, to download, to print out single copies for your own use and to
use it unchanged for any non-commercial research and educational purpose.
Subsequent transfers of copyright cannot revoke this permission. All other uses
of the document are conditional on the consent of the copyright owner. The
publisher has taken technical and administrative measures to assure authenticity,
security and accessibility.
According to intellectual property law the author has the right to be
mentioned when his/her work is accessed as described above and to be protected
against infringement.
For additional information about the Linköping University Electronic Press
and its procedures for publication and for assurance of document integrity,
please refer to its WWW home page: http://www.ep.liu.se/
This report presents the development of a simulation platform for radio receiver sys-
tem design. The simulation platform is implemented in the AWR VSS environment. The
report goes through the underlying theory of radio receivers and the methodology of im-
plementing the receiver and receiver impairments in the simulation platform. The purpose
of the project is to evaluate the AWR VSS environment for developing a simulation plat-
form for RF budget analysis and to take advantage of built in functionality as well as its
graphical interface. The project results in two simulation platform templates for receivers
that uses different simulation types. It is demonstrated that these simulation platforms
can have much of the specified requirements implemented. Some of the listed functional-
ity requirements turns out difficult or impractical to implement however. It is concluded
that VSS can be used for developing simulation platforms with the specified requirements.
Some of the functionality that is not natively supported by the software is implemented
using calculations post simulation and VBA scripting. These methods are proposed as a
solution for adding functions to the template in future work.
Acknowledgments
This work would not be possible if not for the great support and guidance of our supervisors
at Ericsson. With the help and expertise of our main supervisor, Peter Pääkkönen, our work
was made substantially easier and the quality of the results would not be the same without
him. We would also give out thanks to the head of our team at the analog design depart-
ment, Jörgen Johansson, for pitching the idea of the master thesis and for helping us getting
started with our work. We have been fortunate to have great support from many people at
Ericsson, who has lent their time and knowledge to make sure we got all the help we needed
to complete the work.
We have also had great support from our supervisors at Linköping University who has
managed the administrative tasks and helped the work flow smoothly as well as made sure
the project did not get out of hand.
Finally we would like to extend our gratitude to our families who has supported our
efforts in all the ways possible. Thanks to their unrelenting support, we have managed to
finish our electrical engineering education and this master thesis project.
ii
Contents
Abstract i
Acknowledgments ii
Contents iii
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Delimitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 State of the art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Theory 5
2.1 Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Non linearity in RF devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Selectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Degradation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.5 Analog circuits in RF receivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.6 Process variations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.7 Modulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.8 I/Q Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.9 Receiver architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.10 Digital Signal Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3 Development 20
3.1 NI AWR Design Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2 Receiver impairments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Receiver model implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.4 Simulation platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4 Results 34
4.1 Simulation platform results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2 Budget simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3 Time domain simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5 Discussion 49
5.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.3 Improvements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
6 Conclusion 53
iii
Bibliography 55
iv
Nomenclature
In order of appearance
Ericsson is a company which mainly focuses on the development of Radio Infrastructure for
cellular, wireless connectivity. Their mobile solutions are installed all over the world in all
environments and conditions. This requires high tolerances, high precision engineering and
therefor high precision tools in the design phase. The push towards a more mobile society
with 5G connectivity on the horizon calls for these tools to be more complex, precise and
comprehensive than ever before.
1.1 Motivation
Current simulation platforms and tools used for transceiver link budget simulations in RF
design at Ericsson are often Matlab or Microsoft Excel based. These custom tools are ver-
satile and detailed, but complex and thus they require knowledge and competence in both
these tools, and in specific transceiver systems. In the hands of an expert with knowledge in
transceiver systems and the tools themselves, the tools are efficient. However for a person in
the field with less experience in either area, the threshold to start using these tools is high.
The companies developing software for electronic design are with every edition of re-
leases including more advanced solutions in their system simulation tools. This, along with
a relatively low entry bar and a visual interface, makes for an intuitive and efficient work
interface for system simulation. In addition, the system simulation tools can already include
built in functionality similar to that of the tools in Matlab and Microsoft Excel. One major
drawback of these software tools is that they might not include options and possibilities of
customization for specific needs, such as specific impairments and functionalities. However,
with the release of simulation tools such as Applied Wave Research (AWR) system Simu-
lation tool Visual System Simulations (VSS), these customization options have however been
improved upon.
For engineers, ease of use, versatility and improved customization makes VSS competitive
towards existing link budget simulation tools made in Matlab and Microsoft Excel.
1
1.2. Aim
1.2 Aim
The project aims to implement an RF link budget model in AWR VSS environment. Specifi-
cally the RX and TX chains of a homodyne transceiver should be modelled for budget analysis
purposes. The work includes analog and digital transceiver from antenna to the digital inter-
face, but not any further. In comparison to present link budget tools in Matlab and Microsoft
Excel, the tool should include frequency spectrum at each component (block) in VSS and RF
budget parameters such as gain, noise figure and compression point.
In addition, the work aims to clarify the drawbacks of the present tools in Matlab and Mi-
crosoft Excel regarding ease of use. The link budget tool in VSS should address these issues.
As mentioned in Section 1.1, the tool should be targeted towards users, i.e., RF engineers not
working with the present, more complex tools on a daily basis.
Analog blocks
Analog blocks include analog filters, amplifiers (LNA, PA), attenuators (including cable
losses), mixers and passive components. They have a lot of different properties and parame-
ters to be taken into account when making accurate models.
• Gain
• Noise Figure
• IP3
• P1dB
• HDx
• IQ-imbalance
• LO leakage
• DC offset
• Reciprocal mixing
• LO phase noise
ADC
• Sampling rate
• Aliasing
• Jitter
2
1.3. Research questions
• Decimation
• IM blocking
Other functionality
• Automatic Gain Control (AGC)
• Gain calibration
Optional functionality
If the scope of the project allows for further modelling, the following proposals are to be
investigated:
Statistical simulation
• Yield analysis
Digital blocks
• Channel filtering
• What functionalities and parameters can be implemented in VSS using built in func-
tionality and what have to be implemented using ad hoc means?
1.4 Delimitation
The scope of the project is limited to existing radio links. It aims to include functions already
in use in excel-based budget tools with addition of VSS functionality such as time-domain.
Contrasting the existing Excel-based link budget, the VSS tool will remain on system-level
complexity.
The tool aims to be practical in the sense that user experience should be prioritized over
complexity and customization. In a scientific sense, this feature is hard to define and to mea-
sure. An example of a so called "practical" aspect of the tool could be simulation time. This
limits the project to certain design choices based on this philosophy. If a model is considered
accurate but slow, it might be discarded.
3
1.5. State of the art
While VSS includes functionality to implement many of the listed requirements of the
tool, it might not include others in the sense that they are either considered "impractical" or
do not exist. An example of an impracticality is long simulation times. The limitations will
be considered at the end of the project for evaluating VSS as a software for RF system level
simulations and for implementing budget-like tools.
4
2 Theory
This chapter covers the theoretical background considered related to the functionality of the
link budget tool in VSS. What we mean by functionality is the set of parameters listed in
Section 1.2.
We will cover the baseline components of an RF front end (RFFE) receiver. These include
amplifiers, filters and mixers. Other components and functionality covered in this section
include Analog-Digital-Converter (ADC), Automatic Gain Control (AGC) and channel filter-
ing. There are many other aspects of RF system modelling such as antenna characterization,
DSP and the transmitting part of the system. But for the sake of narrowing the scope of the
project, we will focus mainly on the analog front end, and theory related to the link budget
tool.
We will also cover non-ideal aspects of RF system design. When designing and evaluating
RF system performance and characteristics, we must take into consideration these aspects.
These include noise and distortion, process variations and nonlinear behavior of RF compo-
nents. It is worth noting that all non-ideal phenomena mentioned above are not necessarily
independent. This will be clarified.
2.1 Noise
Noise can be described as a random process which is present in RF systems. Noise present in
RF systems is introduced by the antenna from the external environment to the system, and is
also generated internally by active components within the system itself.
One type of noise important to radio systems is thermal noise. It is caused by random
motion of charges and is generated by anything lossy in radio systems. It can also be gen-
erated outside the system e.g., in the atmosphere from thermally excited charges. Thermal
noise can be characterized as frequency independent, noise, i.e., white Noise. When measuring
noise power in the frequency domain, we can view the thermal noise as a constant power
spectrum. across all measured frequencies due to this characteristic [11].
When measuring a signal in the time domain, we will observe the noise as a random
process on top of the signal. If the noise power in the system is stronger than the power of
the desired signal, we will not be able to observe the desired signal. Therefore, it is necessary
to minimize the amount of noise generated by any RF system.
5
2.2. Non linearity in RF devices
Noise figure
In RF systems, the increase of noise can be characterized by noise figure, which is the mea-
surement of degradation in signal-to-noise ratio (SNR) from a point to another in a system (for
example input and output). Noise figure is defined as [11]
S1 /N1
F= (2.3)
S2 /N2
where terms with notation 1 refers to port 1 and notation 2 to port 2. Single components often
have noise figure specified in their data sheets in the units of decibels as F (dB) = 10log10 ( F ).
As systems consists of many components, independently produced and tested, in order
to characterize a system by its total noise figure, Friis formula for noise can be used: [11]
F2 ´ 1 F ´1
F = F1 + + 3 + ..., (2.4)
G1 G1 G2
In order to minimize noise figure, we can derive from (2.4) that components with a large
gain (G) such as amplifiers should be put close to the input.
1 3 1 1
= ( a0 + a2 V02 ) + ( a1 V0 + a3 V03 ) cos(ω0 t) + a2 V02 cos(2ω0 t) + a3 V03 cos(3ω0 t) + ..., (2.6)
2 4 2 4
From (2.6), it can be seen that harmonic components are present at the output of a nonlin-
ear system, [11].
6
2.2. Non linearity in RF devices
As signals in RF systems are rarely a single tone, it might be of interest to investigate the
case when the input signal consists of many tones. Lets say si = V0 (cos(ω1 t) + cos(ω2 t)),
then the output becomes
so = a0 + a1 V0 (cos(ω1 t) + cos(ω2 t)) + a2 V02 (cos(ω1 t) + cos(ω2 t))2 + a3 V03 (cos(ω1 t) + cos(ω2 t))3 + ...,
(2.7)
The output spectrum will consist of harmonics given by
Gain compression
Looking back at the output signal of an nonlinear system (2.6), we can derive the voltage gain
of the fundamental frequency as:
a1 V0 + 43 a3 V03 3
Gω0 = (so /si )ω0 = = a1 + a3 V02 (2.9)
V0 4
Since a3 is negative for amplifiers, it can be seen that the voltage gain decreases as the
amplitude of the input signal V0 increases. The gain becomes saturated as signal strength
increases at the fundamental frequency, hence the name gain compression. For nonlinear RF
components, the gain compression is quantified as P1 , usually in decibels. P1 is the point
where the gain has decreased 1 decibel with respect to the linear case, either referred to the
input or the output [11]. Fig. 2.1 shows a typical nonlinear amplifier with output compression
point at 15 dB.
Figure 2.1: Input versus output signal strength of a typical nonlinear amplifier simulated in
VSS
7
2.3. Selectivity
2a31
P3 = (2.10)
3a3
where P3 is referred to the output. This characteristic is of interest in RF systems since
these intermodulation products can cause distortion in the received signal This can in turn
affect performance.
Fig. 2.2, shows the signal strength of fundamental tone (in blue) and the signal strength
of the third order intermodulation tone (in pink). The intersection of the two dotted lines is
referred to as P3 . In contrast to the dotted lines, the solid lines take into account the nonlinear
contribution in both the fundamental tone (in blue) and the third order intermodulation tone
(in pink).
Figure 2.2: VSS simulations of input versus output signal strength of a typical nonlinear
amplifier. Fundamental tone (in blue) and third order intermodulation tone (in pink).
2.3 Selectivity
An RF receiver must provide certain requirements for desired performance. RF receivers
have to be able to receive a signal in a specific channel within the band of operation, while
rejecting signals and distortions from adjacent channels. This specific requirement is referred
to as selectivity [11].
8
2.4. Degradation
2.4 Degradation
Degradation is a measurement of how well a radio channel performs with interference. The
degradation can be calculated using (2.11).
RN
F = Fmin + |Y ´ Yopt |2 (2.12)
GS S
Where YS : Source admittance, Yopt : source admittance that results in the lowest possi-
ble noise figure, Fmin : Minimum noise Figure of transistor, R N : Equivalent noise figure of
transistor and GS : The real part of the source admittance [11].
Filters
Analog passive filters are used to attenuate undesired frequency content of an incoming sig-
nal, while retaining desired frequencies. There are four basic types of passive filters; Low-
pass filters (LPF), bandpass filters (BPF), bandstop filters (BSF) and high-pass filters (HPF).
In RF applications, these types of filters can be implemented using a variety of techniques.
These include lumped component filters, microstrip filters and cavity resonators.
Lumped component filters are implemented with lumped elements such as capacitors,
inductors and resistors. Microstrip filters are implemented using microstrip transmission
lines utilizing stubs and coupled lines as filtering elements [10].
Cavity filters are filters constructed from a cavity resonator i.e. a wave guide with its in-
ternal dimensions tuned to create a standing wave for certain frequencies while frequencies
outside the passband get attenuated by not being able to propagate through the cavity. Be-
cause the filter is a cavity, it attenuates unwanted signals hard, which is why it is commonly
9
2.5. Analog circuits in RF receivers
used as the bandpass filter at the antenna of RF systems to prevent interferance from reaching
the components of the receiver. [17].
There are some parameters of importance when analyzing filters. Insertion loss (IL) de-
scribes how much power that is lost between input and output. Ripple describes the flatness
of the filter frequency response in the passband. Rejection describes how the filter attenuates
frequencies in the stopband [10].
Mixers
The RF mixer is an essential component in any RF system. It is a two-input, one-output
component that combines the two input signals by adding together their two frequency com-
ponents as: ω1 + ω2 or ω1 ´ ω2 . This can be done to either up- or down-convert a signal
depending weather the mixer is used in a transmitter or receiver. Since the mixer is a non-
linear device, its characteristics can be described as a series of Taylor polynomials according
to (2.7).
Up-conversion is used in the transmission link: Intermediate frequency (IF) signal is
mixed with a signal from a Local Oscillator (LO) that acts as a carrier for the information
contained in the IF signal. The mixed signal is then sent to the RF port. Given the two input
signals
K
VRF (t) = [cos(ω LO t ´ ω IF t) + cos(ω LO t + ω IF t)] (2.15)
2
where K is the voltage conversion loss constant of the mixer.
At the receiver end, the RF and LO signals will act as input and IF the output. The output
of the Down-converting mixer can be approximated as in (2.16).
K
VIF (t) = [cos(ω RF t ´ ω LO t) + cos(ω RF t + ω LO t)] (2.16)
2
A simple mixer can be implemented with a single diode. However, simple mixers have the
general disadvantage of having a high LO- and RF-leakage. There are two types of balanced
mixers designed to handle this problem; single and double balanced mixers.
The single balanced mixer is used to suppress the level of either the RF or LO input signal
while the double balanced mixer is used to suppress both. The advantages of using a single
balanced mixer is that it does not require high LO drive levels, it is also cheaper and less
complex. The advantages of the double balanced mixer are that all ports are isolated, This
preventing leakage, increases the linearity of the mixing and has a better spurious response
since all even order products are suppressed. This is why the double balanced mixer is in
wider use today than its single counterpart.
Mixers can be used to detect phase differences in the two input signals. Mixers however
often display some non ideal functionality such as DC-offsets when used in this configuration
[11].
Phase noise
Phase noise are random fluctuations in phase that are ever present in real signals. Since the
phase noise is generated from the uncertainty of the phase of the signal, the effect it generates
"propagates" outwards from the signal in the frequency spectrum as shown in Fig. 2.3 [8].
10
2.5. Analog circuits in RF receivers
Figure 2.3: Two systems using different oscillators showing the effect of phase noise Red:
noisy oscillator, Blue: system with ideal oscillator [8].
ADC
ADC stands for Analog-to-digital converter and is the component responsible for sampling an
analog signal and converting it into digital information. The basic idea of ADC functionality
will be described and some of the implementation types will be discussed. The reader is
referred to [7] for further theory regarding the implementation types of ADC’s.
Flash ADC
Flash ADC uses parallel connected comparators with different reference signals. The signal
is sampled one symbol at a time. Each comparator outputs one bit. They are simple in their
principle of operation and fast, but suffer from a large number of reference voltages and
relatively high power consumption.
Pipeline ADC
Pipeline ADC architectures perform conversion using multiple cascaded stages of low-
resolution ADCs. Each stage outputs the quantization error of the ADC, amplifies it and
sends it to the input of the next stage. This operation is performed until the last stage which
only consists of an ADC without quantization error output. It is slower than flash ADCs but
power consumption is not as high.
Delta-Sigma ADC
Delta-Sigma ADC utilizes a feedback loop to force input and output to similar levels, while
at the same time making use of low resolution (can be as low as 1 bit) ADC quantizers. The
quantization error (noise) is added to the output signal and shaped with the help of an inte-
grator within the delta-sigma loop. The integrator loop filter has the effect of a LPF for the
input signal and HPF to the quantization noise. We can therefore filter out the quantization
noise and keep the input signal.
AGC
AGC stands for Automatic Gain Control and is a control feature which is particularly used in
RF systems. It is used to actively adjust power levels in a system. For receivers, it is often
times used as automatic attenuation for keeping linearity and to extend dynamic range of the
ADC. In real time, we cannot know what the received signal strength will be, and we have to
take that uncertainty into account when designing a system.
11
2.6. Process variations
In the case of a signal weak enough not to be detected, an AGC will not be able to do
anything to improve performance. In the case of signal strong enough to either compress
gain in the system (destroying linearity) or being outside the dynamic range of the ADC, we
can actively adjust system gain to attenuate this signal.
Aspects of stability, noise figure and linearity all have to be taken into account for in order
to design AGC systems. By attenuating the signal for example, we will inevitably increase
noise figure as derived from (2.4).
There are many ways of implementing AGC functionality for receivers, but the one archi-
tecture focused on in this project has similarities to [14].
2.7 Modulation
This section aims to introduce the reader to the basic theory of RF modulation.
Modulation of a signal is the process of varying certain properties of a periodic waveform
(carrier signal) with another signal (modulating signal). A general way of thinking about
this scheme is that the modulating signal contains information and the carrier signal controls
where the radiated spectrum is located in frequency.
There are various ways to modulate a carrier wave using properties of waves. For a gen-
eral signal in time
12
2.8. I/Q Data
schemes AM, FM and PM are Amplitude Shift Keying (ASK), Frequency Shift Keying (FSK) and
Phase Shift Keying (PSK).
Using the basic principles of modulation presented in this section, more advanced meth-
ods of modulation can be derived. For example, both amplitude and phase modulation can
simultaneously be utilized in the same modulation technique. [11].
it gives
Acos(ω0 t + φ(t)) = A[cos(φ(t))cos(ω0 t) ´ sin(φ(t))sin(ω0 t)] (2.20)
where
S I (t) = Acos(φ(t))cos(ω0 t) (2.21)
and
SQ (t) = ´Asin(φ(t))sin(ω0 t) (2.22)
are the in-phase and quadrature phase components of S(t).
For RF application this result can be applied for modulating a single waveform (2.19) with
information contained in its I and Q components. It also means that the data stored in I and
Q can be independently set and the resulting waveform is still in the form as in (2.19) [5].
In PSK, the most simple case is storing information with two states, where φ can store
either binary 0 or binary 1. For this, the advantage of having information stored in I and Q
components does not even need to be utilized if φ0 and φ1 are 0 and π respectively.
This idea can be expanded by introducing more states. QPSK (or Quadrature Phase Shift
Keying) utilizes 4 states of φ while holding amplitude is constant. Each state can store two
bits utilizing I/Q data. These states are φ11 = π4 , φ01 = 3π π π
4 , φ00 = ´ 4 and φ10 = ´ 4 [11].
Introducing amplitude as a degree of freedom to the modulation scheme, more than 2
bits per state as in QPSK can be utilized. An example of a I/Q modulation scheme using
both amplitude and phase modulation is QAM (Quadrature Amplitude Modulation). QAM
modulation has number of states on the form as in (2.23).
M = 2n (2.23)
QAM modulation schemes are referred to by their number of states on the form M-QAM.
For example, we have 4-QAM, 16-QAM, 32-QAM, 64-QAM and so on where the number of
bits per states is
n = log2 ( M ) (2.24)
By this result we can conclude that QPSK and 4-QAM are identical modulation schemes
even though the theoretical background is different [11].
A graphical representation of the states encoded in I/Q data is the constellation diagram. A
constellation diagram of QPSK states and 64-QAM states is presented in Fig. 2.4. X-axis data
is the in phase component I, and the y-axis data the quadrature phase component Q.
13
2.9. Receiver architectures
Superheterodyne Receiver
The main functionality which identifies superheterodyne receivers is the use of an interme-
diate frequency (IF). Superheterodyne receivers might use more than one down-conversion
to overcome problems due to LO stability at high frequency RF. After each down-conversion,
image frequencies created by the mixing process and ideally intermodulation products are
filtered out [7, 11]. A general schematic of a so called dual superheterodyne receiver, which uses
two down-conversion stages is presented in Fig. 2.5 and Fig. 2.6
Talking about superheterodyne receivers, one issue that appears is so called image rejection.
The problem is characterized by a trade off between image rejection and adjacent channel
suppression. The problem of image rejection appears since the mixer outputs two dominant
signals of frequencies |ω1 ´ ω2 | and |ω1 + ω2 |. Since we have two cases where we can mix to
a desired IF, this means RF frequencies of distance 2ω IF in frequency will mix to the same IF
[12].
14
2.9. Receiver architectures
Homodyne Receiver
Homodyne, Direct-conversion or Zero-IF receivers directly convert the RF signal into baseband
without any IF stage. Down-conversion is achieved with a quadrature down-converter split-
ting up in-phase (I) and quadrature-phase (Q) components of a signal, which is then filtered
and sampled. The reasons for quadrature down-conversion of the signal in I and Q com-
ponents is to avoid so called "folding" from the negative frequencies and in case of phase
modulated signals. In case of IQ modulated signals, the different sidebands of the RF signal
contains different information and splitting up these into two different phases at demodula-
tion avoids folding since the different sidebands fall on each side of baseband [12].
The main difference here from a super heterodyne receiver is that channel selection is
achieved with a low pass filter instead of a bandpass filter since the signal is already down-
converted to baseband. The LO signal is tuned to the incoming RF frequency for achieving
zero-IF [7].
Homodyne receivers come with some technical challenges. LO leakage from the mixers
result in so called self mixing which has the effect of adding a DC offset to the output of the
mixer. IQ imbalance is also an issue when talking about homodyne receivers, although this
is not necessarily unique to this type of receiver.
Two types of homodyne receivers are presented in Figures 2.7 and 2.8.
15
2.9. Receiver architectures
Discrete-Time Receivers
In certain applications, the use of direct RF sampling can be used to advantage. One sim-
ple case is to enable multi standard radio on one single front end architecture. This leaves
more design space for software defined radios (SDR), which have the possibility to increase
versatility and reduce bill of materials since a lot of the functionality lies in integrated digital
circuits. However, RF sampling of various kinds comes with a new set of challenges. Remov-
ing analog RF devices from the design puts high performance requirements on sampling and
signal processing in order to compete with established and well understood technologies.
A few different RF sampling architectures which have been reported on over the years are
presented here.
The so called analog processing receiver, uses analog filters and decimation instead of a
mixer after the LNA in order for the ADC to sample the signal at a reasonable rate. Interfering
signals and blocker signals are therefore not mixed down with LO harmonics as IM products,
but instead aliased when sampled unless filtered. To tackle this problem, anti-aliasing filters
can be implemented or the sample rate of the ADC must be increased. The first reported
implemented receiver of this type was presented in [15].
Direct digitization receivers apply ADC functionality directly after LNA without adding any
extra analog hardware. There are various types of these receivers depending on sampling
method and of ADC [7].
16
2.10. Digital Signal Processing
Another discrete-time receiver well worth mentioning but not discussed any further are
Hybrid-filter bank receivers [16] which use the principle of decomposing the signal in the fre-
quency domain using a bank of analog filters. There are various other kinds of discrete-time
receivers and it is an active topic of research.
Sampling
Sampling is the act of measuring a signal value at certain points in time. How often these
measurements are done is called the "sampling rate" or "sampling frequency" [ Fs ]. How this
sampling frequency relates to the signal frequency being measured dictates if the information
of the original signal will be preserved or not. Sampling frequency is derived through the
time between measurements, i.e, samples. It is the measurement of how many samples per
second that are performed.
The Nyqvist theorem states that any signal needs to be sampled for at least twice its fre-
quency or simply put:
Fs
Fsignal,max ă (2.25)
2
A signal that has a frequency higher than the Nyqvist frequency will naturally form a
mirror signal on the opposite side of the Nyqvist frequency due to having the samples fall on
the same places on the signal. The mirror signal will fall on Fs ´ FSignal as shown in Fig 2.9.
This is also known as aliasing [9].
Figure 2.9: Signal mirroring (aliasing) around the Nyquist frequency [9]
Spectrum analysis
Spectrum analysis is a tool used when conducting measurements and analyzing signals.
Fourier transformation of a signal decomposes it into its spectral components. In the case of
signal in time, the transformed signal will be a function of frequency.
The discrete version of the Fourier Transform is called Discrete Fourier Transform (DFT)
and is widely used to analyze sampled signals. The more samples analyzed, the higher the
frequency resolution will be on the Fourier Transformed signal.
As signals are sampled, some things need to be taken into account for when performing
spectrum analysis using DFT. A finite frequency means aliasing will occur when measuring
frequencies above the Nyquist frequency.
17
2.10. Digital Signal Processing
Another problem that arises with analyzing finite data is that we cannot assume first and
last sample continuously connect with perfect periodicity. This gives rise to uncertainty in the
frequency spectrum as the signals will appear to contain information far out from the center.
One way to tackle this issue is to apply so called Window functions to the spectrum to
mitigate some of these effects. Windowing does this by applying different functions in time
domain that minimizes the amplitude of the discontinuos area of the signal that would other-
wise cause high frequency spikes in the frequency domain. In Fig. 2.10 the difference between
a spectrum with applied window and one without is presented where the difference can be
seen clearly.
Figure 2.10: VSS simulation of a QPSK signal spectrum with Hanning window (Blue) and
without window function (pink)
Digital filtering
In order to minimize the risk of aliasing, the signal cannot contain frequencies above the
Nyquist frequency. This can be achieved by applying a low pass filter with a cut off frequency
at the Nyquist limit i.e., an anti-aliasing filter. However, because there is no filter with an
instant transition band, the cut off frequency needs to be slightly below the Nyquist limit in
order to ensure that the stop band attenuates the higher frequencies enough for the aliasing
to be negligible [2].
A filter that is especially well suited for this is the Finite Impulse Response or FIR-filter,
which is a type of digital filter. The FIR filter is finite because it settles to zero in a finite time.
An IIR (Infinite Impulse Response) filter however can decay towards zero for an infinite time.
FIR filters are based on a N-number of parallel "taps" as seen in Fig. 2.11.
18
2.10. Digital Signal Processing
The mathematical expression of the output signal y given an input signal x through a FIR
filter with filter coefficients h is given by (2.26).
Nÿ
´1
y[n] = h [i ] x [ n ´ i ] (2.26)
i =0
In order to reduce the processing required if the sample rate is a lot higher than the signal
frequency, downsampling reduces the amount of samples that has to be processed. By ap-
plying Half-band FIR filter and then downsampling (decimating) the signal, we can keep the
information within the boundaries of the Nyquist frequency while minimizing effects from
aliasing.
19
3 Development
This chapter presents the detailed approach to create a simulation platform with focus on
the budget analysis simulation tool. After some consideration about the complexity of a
communication system in terms in RF circuit design when non-idealities of the real-world
components, process variations etc. must be considered, the receiver model is introduced
along with the added functionality as required by project description. The way temperature-
and frequency-dependency are implemented in the model is detailed. Models for simulat-
ing the effect of LO phase noise, mixer M x N products and IQ-imbalance are described as
well as models for gain calibration and automatic gain control. The receiver model includes
also aspects of digital signal processing (DSP) for complete characterization of the receiver
performance. Finally, the simulation platform and the developed templates for link budget
simulations and time-domain simulation are presented in detail.
20
3.2. Receiver impairments
• Process variations affecting all components in their electrical parameters, matching, etc.
• Receiver modelling as a chain of components (filters, LNA, mixers, etc) according to the
actual receiver architecture
• Compare the simulation results to the required specifications of the system (receiver)
21
3.3. Receiver model implementation
From this pseudo flow-chart the results of how important top-level simulations are in the
design of complex, wireless systems. Since they comprise of a multitude of circuits operat-
ing under different circumstances, e.g., signal levels, frequency etc., and performing signal
processing of high complexity, these system level simulations are necessary to make sure the
system is working as intended.
Homodyne RX
The basic receiver lineup is based on a general homodyne receiver model presented in Fig.
2.8. A system diagram of the receiver lineup is presented in Fig. 3.1 with a few differences.
Since VSS supports the use of complex signals on the form S(t) = I (t) + jQ(t) a single mixer
acts as an IQ demodulator by itself with the right settings. Another difference is that there is
no Variable Gain Amplifier (VGA) in Fig. 3.1. The reason for this design choice was that this
feature was not assessed as "basic". This feature will be presented in the next lineup used for
the project.
The receiver lineup implementation for this project however has a few more features in-
cluded for a more detailed model. This lineup is constructed based on some of the specifica-
tions of an existing 5G radio receiver and the implementation is presented in Fig. 3.2. It uses
features such as temperature dependent parameters, losses and VGA’s.
22
3.3. Receiver model implementation
Temperature dependency
In order to approximate real system performance, the temperature dependency of certain
component parameters needs to be taken into account and subsequently modelled. In a lot of
cases, when temperature is static, component properties such as gain/loss behaves the same
independent of temperature and can be modelled easily. However, when there is a system
that will operate in a wide range of temperatures and the system itself is prone to self heating,
the physical structure of the components causes the performance to drift.
In order to ensure correct performance and that the system meets regulatory and/or in-
ternally set requirements, the temperature dependency needs to be taken into account.
Temperature variation is implemented with respect to a nominal temperature. The nomi-
nal temperature is defined as the arithmetic average of the maximum and minimum temper-
ature given in 3.1.
Tmin + Tmax
Tnom = (3.1)
2
where temperatures are defined as either Celsius or Kelvin. For a temperature varying
parameter Q, the value is modelled as a function of T
T ´ Tnom
Q( T ) = Qnom + Qtemp (3.2)
Tmax ´ Tnom
where Qtemp is the user specified parameter for temperature variation. To give an example
of how this works, lets say we have an amplifier with nominal gain Gnom = 10 dB and a
temperature varying parameter Qtemp = ´1.2. At T = Tmax , G = 8.8 dB and at T = Tmin ,
G = 11.2 dB.
The gain/loss variation due to temperature can be compensated for by using a variable
gain amplifier that can attenuate/amplify a signal depending on what temperature is in the
system. This behaviour can be modelled in AWR by knowing the target gain for two different
temperatures and interpolate the gain linearly between these two temperature values. The
drift in gain/loss can thereby be compensated for in the model for static (time independent)
simulations.
For time domain simulations, it is convenient to use a look up table of pre-measured gains
at different temperatures while sweeping temperature values for the system.
23
3.3. Receiver model implementation
Frequency dependency
As well as in the case above with temperature, components tend to vary in performance
depending on frequency. This is can be modelled in various ways in AWR depending on what
simulations are being done and what components are being used. Frequency dependency
can be modelled linearly, like temperature dependency is modelled in this project, but since
temperature dependent data for the different components may be difficult to obtain, it is
instead made part of the AGC system by taking the frequency dependent attenuation of the
system blocks in S-parameter files. A number of points are placed on the S-parameter curves
and then an interpolation is made between these. The power measurement for the AGC is
made with regard to S21.
Frequency dependency could be implemented in this way for the entire lineup, but it
would require a complete redesign of the lineup as well as lose a lot of what makes VSS
simple to perform simulations in due to that the component properties would be complex to
change and would have to be specified in non-intuitive ways.
The various filters implemented in the model are also a way of representing the frequency
dependency of the system. Frequencies are attenuated differently depending on where on the
spectrum they fall while still being propagated through the system.
Process variation
There are two types of process variations implemented. One for each of the simulation plat-
forms respectively. This was done because of how the component parameters for the analog
lineup are specified in the different platforms. The budget document can perform yield anal-
ysis as part of VSS built in functionality, where the user can specify the type of distribution
and how much the chosen component parameter should vary Fig. 3.3.
Figure 3.3: Process variation for Gain, P1dB and IP3 for a component in the analog lineup in
the budget platform
When performing a yield analysis in VSS, a random number is generated in the specified
interval and used as the parameter value. This can be done multiple times as a sweep, leading
to a different approach being used in the time domain platform to reduce simulation time.
In the time domain platform, the process variation is specified alongside the other com-
ponent values in the "Global Definitions"-folder as seen in Fig. 3.4.
24
3.3. Receiver model implementation
The process variation is then multiplied with a factor K_[...] indicating the maximum, min
and normal case for the parameter. This is then added in turn to the component parameter
in the analog lineup. This is suitable for the time domain platform since there are no sweeps
involved and the most interesting cases of i.e. maximum, minimum or normal gain can be
simulated with the manual input of the K_[...] factor. This can be seen in detail in Fig. 3.5.
Figure 3.5: Factor determining max, min and nominal case for process variation.
LO phase noise
Phase noise is modelled in AWR with a text file, a phase noise mask, attached to the LO input
tone of the mixer. The text data file contains frequency offset-dBc/Hz data in a format as
Table 3.1.
The text file is then applied to a tone source as phase noise as shown in Fig. 3.6.
Mixer M x N products
The mixers inside the IQ modulator outputs a signal on the form (2.7). However, to model
the specific behavior of the mixer, the user needs to specify the output levels (dBc) of the
intermodulation products on the form (2.8), where m and n are indices in a resulting matrix.
In AWR, the user needs to input this matrix in the mixer model as a text file. The AWR
implementation of this feature is presented in Fig. 3.7. This example shows a matrix with
indices 0x0 to 4x4.
25
3.3. Receiver model implementation
IQ imbalance
IQ imbalance models non ideal aspects of the IQ demodulator. It is implemented as a block,
containing parameters DC offset, amplitude imbalance and phase imbalance. Units are volts,
dB and radians.
Gain calibration
Gain calibration is used as to compensate for process variations and temperature variations
in the system. It is also vital for the system to estimate its gain for DSP operations. It uses
VGA blocks as variable attenuators. It is therefore important that the "Target gain" is not set
to high since we only compensate by attenuation.
Two different gain calibration methods were implemented in AWR. The first model calcu-
lates the sum of gain parameters from the RX lineups (3.3) and the total temperature varying
parameter (3.4).
N
ÿ
Gnom = Gn (3.3)
n =1
N
ÿ
Gtemp = Gtempn (3.4)
n =1
Process variations are modelled by adding together the statistical gain variations in each
element in the RX lineup as in (3.5).
N
ÿ
G process = G processn (3.5)
n =1
Adding together nominal gain, temperature varying gain and statistical gain variations
we can determine the total gain by (3.6).
T ´ Tnom
Gtot = Gnom + Gtemp + G process (3.6)
Tmax ´ Tnom
Gain calibration was then applied as an attenuation block attenuating the signal as in (3.7).
#
Gtot ´ GTarget , if Gtot ą GTarget
Lcal = (3.7)
0, otherwise
Implementation of this method in AWR is presented in Fig. 3.8.
26
3.3. Receiver model implementation
The second implementation method of gain calibration uses a PID (Product, Integrator,
Derivative) controlled gain calibration loop to calculates attenuation by gain compensation
before the main simulation starts. The AWR implementation method is presented in Fig. 3.9.
27
3.3. Receiver model implementation
figure. In the moment we attenuate the signal, we also generate discontinuity in the data
stream.
There are other problems that arise with control systems. Stability is one aspect that has
to be taken into account since we rely on negative feedback with added delays.
VGA placement in the receiver lineup along with power measurement placement also
matter for optimal control.
Feedback AGC
In order to model AGC functionality in the system for time-domain simulations, feedback
control can be used instead of manual control.
The receiver feedback AGC model implemented in AWR utilized comparators with hys-
teresis built in. In contrast to the regular logical comparator which outputs either True or
False depending on if input passes the condition, the comparator with hysteresis utilizes two
conditions, one condition for initial state and when the first condition was passed. The elec-
trical analog to this behavior is the well known Schmitt trigger design which is described in
[1]. The comparator with hysteresis implemented in AWR is presented in Fig. 3.11.
28
3.3. Receiver model implementation
For an AGC, we are interested in power measurement in order to control VGA’s. An AGC
controller consisting of four comparators with different threshold levels was implemented in
order to control a maximum of four VGA’s. This can be extended to any number of compara-
tors. The AGC controller implemented in AWR is presented in Fig. 3.12.
29
3.3. Receiver model implementation
Power measurement is important for implementing AGC functionality. In the case for the
AGC implementation for the project, the running average of N samples of the incoming RMS
power is utilized, where N is user specified. Other user specified parameters for the AGC are
hysteresis levels, delay between triggering VGA’s and a start delay for simulation purposes.
The power measurement AWR implementation is presented in Fig. 3.13.
30
3.3. Receiver model implementation
ADC model
The Delta-Sigma (∆-Σ) ADC implemented and tested in VSS was a 2:nd order ∆-Σ as indi-
cated by the number of integrators and/or the number of feedback loops in Fig. 3.14.
This ADC model was not implemented due to the drastically increased simulation times
when included in the receiver lineup for time domain simulations. It was also not compatible
with budget simulations and was therefore not used for that platform either.
Digital interface
The digital interface was modelled using frequency transformation, frequency decimation
along with FIR filters and a channel filter. A simple model of this is presented in Fig. 3.15.
31
3.4. Simulation platforms
32
3.4. Simulation platforms
ADC sampling and aliasing are all taken into account for since finite sampling frequency and
decimation are used.
33
4 Results
The results of the project are presented in this chapter. This includes implemented function-
ality along with simulation examples from the AWR project documents.
Functionality
The AWR project ended up containing two different simulation platforms. The core function-
ality includes:
• Temperature dependency
• Frequency dependency
• Gain calibration
• AGC
• IQ imbalance
34
4.1. Simulation platform results
The project aimed to implement PLL functionality and a proper ADC model as well, but
this was not achieved, however the other goals of the project listed in Section 1.2 were. The
functionality can also be extended to include other types of measurements with the given
simulation platforms.
The budget simulation compensate for variations in power by a PID-controlled gain cal-
ibration block as shown in Fig. 3.9. The budget simulations use the PID version of the gain
calibration since the VSS PID block simulates steady state when budget simulations are used.
This is suitable for budget simulations since they are time independent and only perform
calculations on the boundary conditions of the system. The PID block also supports time do-
main simulations, but when implemented in those types of systems, the transient time of the
controller causes the simulation time to increase drastically which is undesirable for a real
time simulation platform.
35
4.1. Simulation platform results
Figure 4.2: Time domain simulation platform - Blocking signal and noise source
Gain correction for AGC is also included for time-domain simulations. Gain correction is
a digital operation used to compensate for attenuation from AGC When a VGA attenuator
is turned on inside the analog RX block, the corresponding amplifier is turned on inside the
gain correction block. In real systems, this operation is performed digitally after the ADC in
order to an extent achieve constant system gain. The system diagram for the gain correction
operation is presented if Fig. 4.4.
Figure 4.4: Time domain simulation platform - Gain correction system diagram
The gain calibration system used in the time domain simulation platform is equation
based as seen in Fig. 3.8. This is an entirely equation based system and is used because
the PID version of the gain calibration was considered too slow when implemented in time
domain simulations.
36
4.1. Simulation platform results
ADC
The ADC model chosen is the same for both simulation platforms. It is a VSS amplifier block
with 0 gain and ADC parameters such as IP3 and NF specified. This results in a simple
general component without the ADC characteristics such as jitter.
Feedback AGC
The power of the signal and the measured power by the AGC is plotted in Fig. 4.5. The power
level of the signal is what the AGC uses to trigger the control signals of the VGAs as shown
in Fig. 4.6.
Figure 4.5: Power measurements of feedback AGC. Blue: Instant power in channel. Pink:
measured power
The power meter block used to measure the power of the channel measures the average
power and outputs it as a voltage for the AGC to use as input. Because it is an average, there
is a slight delay in the measurements, adding to the time it takes to simulate and making it so
that the response is not instant, as seen in Fig. 4.5. In the blue plot, three VGAs are triggered,
reducing the power each time.
37
4.1. Simulation platform results
The control signals are triggered and released dependent on if the measured power ex-
ceeds the user specified trigger and hysteresis levels of the individual VGAs. In the case
of Fig. 4.6, three VGAs are triggered, at the 100 ns, 1100 ns, and 2100 ns respectively. The
corresponding power level can be seen in Fig. 4.5 at the same time stamps.
Simulation times
In Tables 4.1 and 4.2, simulation time results are presented for the budget and the time do-
main platform with a few different settings. In Table 4.1, number of sweeps as well as gain
calibration method are varied. Since there is only one AGC method (Equation based) used
for the budget platform, these are the only parameters that the user can vary for different
results. We find that using equation based gain calibration will improve simulation times,
with the downside that yield analysis using the AWR built in tool cannot be performed in
this configuration.
38
4.2. Budget simulations
The time domain platform contains more parameters that can be varied unlike the budget
platform such as simulation stop time and sampling rate. For the sake of keeping things sim-
ple, sampling rate is kept constant for all simulations and only one gain calibration method is
used (equation based). Table 4.2 clearly shows that when the feedback loop AGC is included,
simulation times are negatively impacted. The simulation stop time setting is also important
for total simulation time. Having a longer simulation stop time will give a more accurate rep-
resentation of the output spectrum of the signal. The downside is a longer total simulation
time.
These results might vary depending on which computer is running the simulation. Simu-
lation times can also vary from simulation to simulation. The results presented in this section
should therefore be interpreted relative to each other. They should give an idea of how sim-
ulation configuration parameters can impact simulation times.
39
4.2. Budget simulations
Figure 4.9: Degradation simulation with budget simulation platform. Left: Blocking signal 20
MHz from channel. Right: Blocking signal 70 MHz from channel.
Gain calibration
Gain calibration in the budget document uses a PID model to control the attenuation level of
a VGA. The user can set a variable to ON/OFF to control this function. With gain calibration
turned off, the VGA gain level is constantly set to 0 dB. The cascaded gain of the receiver
lineup with gain calibration turned off is presented in Fig. 4.10. Fig. 4.11 shown the results
when gain calibration turned on. Gain calibration target level was set to 28 dB.
40
4.2. Budget simulations
AGC
AGC functionality in budget simulations uses the feed forward model presented in Section
3.3. This function can be toggled by setting a variable to either ON/OFF. When AGC is turned
off, the power sweep of system gain is presented in Fig. 4.12 and with AGC turned on in Fig.
4.13.
41
4.2. Budget simulations
Yield analysis
Yield analysis can be performed to emulate process variations in components. If yield anal-
ysis is performed, the result is presented as in Fig. 4.14. The faded blue lines represent the
various yield sweeps and the one darker blue line in each graph is the nominal result.
42
4.3. Time domain simulations
43
4.3. Time domain simulations
Figure 4.15: Time domain channel simulation results. Channels: 10 MHz, 5 MHz, 3 MHz and
the in- and out-spectrum. Blocking signal 20 MHz from channel. Channel at center frequency.
The graphs presented as the results of the degradation simulations are presented in Fig.
4.16, Fig. 4.17, Fig. 4.18 and Fig. 4.19. Degradation is a measurement of how a channel is
affected by interference with respect to the same channel without interference.
Figure 4.16: Output spectrum with blocking signal 70 MHz from channel. Channel at center
frequency.
44
4.3. Time domain simulations
Figure 4.17: Channel degradation with blocking signal 70 MHz from channel. Left: Time
domain simulation. Right: Budget simulation.
Figure 4.18: Output spectrum with blocking signal 20 MHz from channel. Channel at center
frequency, with LO-phase noise on signal and intermodulation spurios interferers.
45
4.3. Time domain simulations
Figure 4.19: Channel degradation with blocking signal 20 MHz from channel. Left: Time
domain simulation. Right: Budget simulation.
If spurious effects from the IQ demodulator land in the channel, the degradation becomes
more evident than in the cases only affected by noise increase from AGC attenuation and
phase noise. This is presented in Fig. 4.20. Here, we see a mixer product falling inside the 10
MHz channel, but not in the 3 MHz and 5 MHz channels.
Figure 4.20: Channel degradation with mixer product in channel. Left: Channel filter spec-
trum with mixer product from blocking signal. Right: Channel degradation with mixer prod-
uct from blocking signal.
Digital interface
The digital interface is the block after the gain correction block in the schematics. In the real
system, the gain correction is done digitally in the digital IC and is therefore everything post
ADC in the DUT. The digital interface consists of three channels with different channel FIR-
filters. The filter lineups are presented in Fig. 3.15.
The resulting spectrum graphs from the filter lineups are presented in Fig. 4.21, Fig. 4.22
and Fig. 4.23. These show the different filter characteristics as well as the resulting channel
filter output marked in black.
46
4.3. Time domain simulations
47
4.3. Time domain simulations
These three spectral graphs above show the narrowing down of the digital filters to the in-
dividual channels, marked in black, and the correct channel widths. each of the three figures
corresponds to a different channel with different widths and the coloured graphs represents
the filters in the filter chain of the channel. the first filter is the widest and the last in the chain
is the most narrow as the signal propagates through the chain and is decimated and filtered
in each step down to the final channel filter.
48
5 Discussion
In the discussion part of the thesis, the result, method and work as a whole is analyzed to see
what compromises and decisions were made. What could be improved or done differently
will also be discussed as well as a reflection on the subject and its wider context, applications
and where it could go from here.
5.1 Results
The results obtained from this project are not necessarily only what is shown in the graphs,
but more the presentation and simulation setups that provide the results. The result is rather
the set of functions included in the VSS project. Using the simulation platforms and models
as a template for further work, a user can test various receiver lineups in this environment
for system-level simulations in a more ”easy to understand” manner compared to previous
simulation tools.
However, not all functionalities from the project description were included in the final
AWR projects. This mainly includes functioning ADC models, which has an effect on the
reliability of the simulation results. There are a number of design choices for modelling
functionality such as gain calibration and AGC that have limitations. This will be further
discussed in Section 5.2.
Since VSS is not a perfect platform, there are some functions that cannot be implemented
that might become possible to do if National Instruments is given feedback on the platform.
For example, the component parameters (such as gain, NF, IP3 etc.) in the lineup are specified
differently in the time domain simulation platforms and in the budget one. This is because of
how yield analysis is implemented in VSS. Since yield analysis is not practically possible in
time domain it is only implemented in the budget platform.
The issue is that component parameter specified with variables rather than numbers can-
not be used for yield analysis. This was one of the reasons why the simulation platform was
split up in two project files, budget and time domain.
Since the project aimed to examine the possibilities to use VSS as a simulation tool for
RF design that could be more user friendly than previous versions, how well the results are
depends not only on the accuracy of the simulations but on the simulation times and the us-
ability of the system. In regards to the accuracy and usability, the platform performs very
well. However the simulation times varies with what version of the implemented gain cali-
49
5.2. Method
bration systems and AGC is being used. This is however the reason there are two version of
each implemented, in order to give options for the designer when performing simulations on
the receiver.
5.2 Method
Part of the process in designing this simulation platforms was the testing of different solutions
to various problems. The methodology was in many regards learning by doing what worked
best in VSS. Some solutions proved more suitable than others in regards to user friendliness
and simulation times. For example, three different AGC implementation methods and two
different gain calibration methods were tested and evaluated.
50
5.3. Improvements
version contains all the functionality and behaviour needed. However, it is a complicated
system to understand, and most importantly of all, the simulation times become far too long
for any simulations to be practical for a designer as shown in Table. 4.2.
The third and final solution to the AGC problem was the equation based system that
was described in Section 3.3. This version had good performance but lacked hysteresis and
was also relatively easy to use and fast to simulate. The issue with this version of AGC
implementation is that it assumes linear behavior over the full dynamic range of the system.
As input signal strength increases, the system goes into compression and this affects linearity.
This aspect is not taken into account. This version was finally implemented in both projects.
The implementation of the gain calibration was also iterated upon in three steps. The
simplest model implemented was a VGA with a look up table that compensated for the tem-
perature dependent gain shift. The implementation of this version was however not accurate
and flexible enough to make it to the final version.
There were two different solutions implemented after the look up table. These two are
described in Section 3.3. The PID controlled gain calibration system is used in the budget
platform since the VSS PID block instantly goes to a steady state, which is needed for budget
simulations, since they assumes state and only simulates on the boundary condition of the
system. this version is therefore not appropriate for time domain simulations. It is also not
possible to do gain calibration in steps of i.e., 0.1 dB which is desirable to model real system
behavior. For the Time domain platform, the most "real-life" accurate version of the gain
calibration is used. This version is viable because of the VSS function Round( X1 , X2 ) that
enables the gain compensation step size. Both versions work well in their respective platform.
It would be preferable to use the same version in both platform for ease of use and consequent
designing.
5.3 Improvements
Since this simulation platform is brand new, it will act as a demonstration concept for what
a VSS can do in terms of budget simulations. This platform can be iterated and built upon
to achieve new functionality not present current tools. Some of these new improvements
might even lead to new ways of looking at the radio systems and how to approach the design
process of new radio links.
Right now, the resulting simulation platforms have modeled frequency dependency only
with the help S-Parameter filter models and their effect on signals. Frequency dependency
however is an intrinsic part of each and every component, including traces and lines. This
could be modeled with the built in frequency dependency parameters in each blocks. Even
adding S-parameter blocks after the amplifiers and normalizing their gain would greatly in-
crease the accuracy of the AGC. This however could give a higher entry level for usage and
make the gain harder and more time consuming to change if necessary.
While on the subject of the AGC, a more accurate time domain model of the AGC could
be implemented if VSS started supporting that functionality natively.
As of the latest version of the VSS platform, the temperature dependency of the system
is modeled with linear interpolation between two temperature points. This however is not
necessarily accurate for all components and even so, the implementation of the temperature
dependency could be different. The linear interpolation could for example be implemented
using more than two points, to more accurately replicate temperature dependency in real
components.
An improvement to the digital part of the lineup is the addition of more signal types. It
is for example possible to add decimation filters for all LTE-channels of interest instead of
the three general channel bandwidths as seen in this report in Fig. 4.21, Fig. 4.22 and Fig.
4.23. The way filters are handled in the current version of the digital interface could also be
51
5.3. Improvements
improved with more accurate models of the filters. This would increase the accuracy of the
results even further.
In the current version of the project, the ADC at the end of the analog lineup is only
represented as a zero gain amplifier block with noise, IP3, and P1dB impairments. This is
however only the tip of the iceberg when it comes to the types of ADC impairments affecting
an RF receiver, such as jitter due to the sampling clock signal. As mentioned in chapter:
5.2 under compromises and decisions, the implemented delta-sigma ADC was too slow to use,
however with another method of implementation, a more complex ADC could perhaps be
put into place, that more accurately exhibits the true characteristics of an ADC.
52
6 Conclusion
In this project, the possibility of implementation of a simulation platform for radio receiver
development with focus on budget link evaluation was firstly investigated. The commercial
AWR Design Environment with Visual System Simulator (VSS) were chosen to implement
the simulation platform, mainly due the fact that AWR allows customization of codes accord-
ing to company requirements and specifications. The work included modeling of a receiver
lineup with necessary functionalities so that that it would be a powerful tool to design com-
plex systems such as modern receivers. In contrast to the existing simulation platform, the
new one should offer more opportunities to visualize the simulation results concerning differ-
ent receiver impairments while being able to perform complex frequency- and time domain
simulations. The main idea was to utilize as many as possible of the built-in functions in VSS
to create a more user-friendly experience for the radio designers than the current existing
simulation platforms provide, without compromises in complexity and accuracy.
It was shown that the VSS-environment offers the possibilities to create simulation plat-
forms that are easy to understand and work with. Moreover, the VSS and the developed
simulation platform provided the grade of adaptability to the receiver architectures and to re-
ceiver specifications as the company has required. Compromises because of usability turned
out to be few. The functionalities that were not natively supported by AWR VSS, were still
possible to be implemented via a VBA scripting editor that is featured in the AWR envi-
ronment. Making changes to this is not as intuitive as making changes in the VSS system
diagrams, but a user performing simulations in the developed platform it is not required to
make changes to that part of the system in order to perform simulations, once the scripts
were implemented. Some problems were encountered in how the time-domain simulations
affected the budget simulation run times when parameter sweep were performed. This lead
to the decision to split the platform into two separate files that had different strengths and
weaknesses e.g., a lack of yield analysis in time-domain and simpler input signals in the bud-
get file.
Finally, the project clearly demonstrates the strengths of using a simulation environment
such as VSS when designing radio systems. It can provide fast, reliable receiver model imple-
mentation and receiver simulations. Also, it was shown that using VSS and its customization
features, actual receiver specification in terms of receiver architectures, receiver characteris-
tics and individual block specifications can be modelled and simulations can be performed in
an practical way with short run times while simulation results can be graphically presented
53
and hence they can be easily interpreted. The developed receiver model has included the re-
ceiver front-end, from the antenna to the ADC, including the AGC, gain calibration VGA and
down-converter as well as elements after the ADC such as digital gain correction and three
of the filter branches present in the digital signal processing unit of the modeled receiver.
54
Bibliography
55
Bibliography
[15] Robert Bogdan Staszewski, Khurram Muhammad, Dirk Leipold, Chih-Ming Hung,
Yo-Chuol Ho, John L. Wallberg, Chan Fernando, Ken Maggio, Roman Staszewski,
Tom Jung, Jinseok Koh, Soji John, Irene Yuanying Deng, Vivek Sarda, Oscar Moreira-
Tamayo, Valerian Mayega, Ran Katz, Ofer Friedman, Oren Eytan Eliezer, Elida de-
Obaldia, and Poras T. Balsara. “All-Digital TX Frequency Synthesizer and Discrete-
Time Receiver for Bluetooth Radio in 130-nm CMOS”. In: IEEE Journal of Solid-State
Circuits (Volume: 39 , Issue: 12 , Dec. 2004 ) (2004).
[16] S.R. Velazquez, T.Q. Nguyen, and S.R. Broadstone. “Design of hybrid filter banks for
analog/digital conversion”. In: IEEE Transactions on Signal Processing ( Volume: 46 , Issue:
4 , Apr 1998 ) (1998).
[17] N. Omar A. R. Othman V. R. Gannapathy Z. Zakaria1 M. S. Jawad. “A Low-Loss Coax-
ial Cavity Microwave Bandpass Filter with Post-Manufacturing Tuning Capabilities”.
In: International Journal of Engineering and Technology (IJET) (2013).
56