PREDICTION OF DISTORTION IN PIN DIODE RF SWITCHES WITH

INPUTS POSSESSING ARBITRARY SPECTRUMS

_____________

A Thesis

Presented to the

Faculty of

San Diego State University

_____________

In Partial Fulfillment

of the Requirements for the Degree

Master of Science

in

Electrical Engineering

_____________

by

Takavar Ghahri Saremi

Summer 2011
 iii

Copyright © 2011

by

Takavar Ghahri Saremi

All Rights Reserved

 iv

DEDICATION

This thesis is dedicated to my mother whom I view to be the highest

personification of what it means to be a mother; eternally forgiving to the point of
selflessness, her unwavering devotion to her children continues to readjust my grasp
of a mother's potential.
 v

ABSTRACT OF THE THESIS

Prediction of Distortion in PIN Diode RF Switches with Inputs Possessing Arbitrary

Spectrums
by
Takavar Ghahri Saremi
Master of Science in Electrical Engineering
San Diego State University, 2011

A paper presented at the 1989 IEEE MTTS by Robert H. Caverly and Gerald
Hiller showed that the distortion caused by RF switches utilizing reverse biased PIN
diodes can compromise system performance in RF communication systems. Their
paper presented equations predicting the two tone intermodulation distortion
intercept points of IP2 and IP3 for switches utilizing series connected reverse biased
PIN diodes. This thesis presents equations predicting the nonlinear response of an
RF switch utilizing a shunt connected reverse biased PIN diode to a narrowband
excitation possessing an arbitrary spectrum. The equations predict the amplitude
and phase of the spectral content generated by the switch's nonlinearity in the
neighborhood of the fundamental, as well as the second and third harmonics of the
excitation's center of frequency. These equations are also used to derive a formula
predicting the RF switch's performance under a noise power ratio (NPR)
measurement.
 vi

TABLE OF CONTENTS

PAGE

ABSTRACT ................................................................................................................ v
LIST OF FIGURES .................................................................................................. viii
CHAPTER
1 INTRODUCTION .........................................................................................................1 
1.1 The Mechanisms of Distortion Within the RF Switch Predicted
by Analysis .............................................................................................................3 
1.2 Frequency Generation ...................................................................................3 
1.2.1 Harmonic Generation ...........................................................................4 
1.2.2 Mixing Products .....................................................................................4 
1.3 The Analytical Method Used in the Nonlinear Analysis of the RF
Switch ......................................................................................................................4 
1.4 An Overview of the Analytical Method ........................................................4 
1.4.1 Analytical Methods for Weakly Nonlinear Networks ........................5 
1.4.2 The Limiting Assumptions of the Nonlinear Analysis ......................7 
2 ANALYSIS OF THE RF SWITCH'S RESPONSE ..................................................8 
2.1 Arriving at the Switch's Equivalent Network ...............................................8 
2.2 The Derivation of the Switch's Response ...................................................9 
2.3 Formulation of the Node Equation ...............................................................9 
2.3.1 Formulation of the Response as a Series Summation ..................11 
2.3.2 Truncation of the Summations ..........................................................17 
2.3.3 The Identification of the Multidimensional Transfer
Functions ........................................................................................................19 
2.3.3.1 The Derivation of the First Order Transfer Function ............19 
2.3.3.2 The Derivation of the Second Order Transfer
Function ....................................................................................................20 
2.3.3.3 The Derivation of the Third Order Transfer Function ...........21 
2.3.4 The Equation Describing the Response in the Time
Domain ............................................................................................................23 
2.3.5 Relating Volterra Series to Spectral Convolutions .........................24 
 vii

2.3.6 The Formulation of the Response in the Frequency

Domain ............................................................................................................27 
3 ANALYTICAL PREDICTIONS COMPARED WITH SIMULATED DATA .........34 
3.1 Circuit Details in MWO.................................................................................34 
3.1.1 The Nonlinear Capacitance ...............................................................36 
3.1.2 The Input Excitation in Simulation ....................................................38 
3.2 Details of Analytical Prediction Using the Closed Form
Solutions Implemented in MATLAB .................................................................40 
3.3 A Comparison of Predicted vs. Simulated Spectral Content .................42 
3.3.1 Simulation in the Neighborhood of the Fundamental ....................42 
3.3.2 Simulation in the Neighborhood of the Second Harmonic ............45 
3.3.3 Simulation in the Neighborhood of the Third Harmonic ................47 
4 FURTHER APPLICATIONS ....................................................................................51 
4.1 Noise Power Ratio of the RF Switch .........................................................51 
4.2 Prediction of the RF Switch's NPR ............................................................53 
5 CONCLUSION ...........................................................................................................55 
REFERENCES ......................................................................................................................56 
 viii

LIST OF FIGURES

PAGE

Figure 1.1. PIN diode RF switch. ..........................................................................................2 

Figure 1.2. Equivalent circuit of the RF switch with the PIN diode in reverse
bias. ...............................................................................................................................3 
Figure 1.3. Nonlinear network block. ..................................................................................5 
Figure 2.1. The RF source, load, and the RF switch consolidated into network
blocks. ...........................................................................................................................8 
Figure 2.2. Norton equivalent circuit as the arbitrary excitation. .....................................9 
Figure 2.3. Currents forming the node equation. .............................................................10 
Figure 2.4. A nonlinear network block. ..............................................................................11 
Figure 2.5. A depiction of the current source's spectrum. .............................................29 
Figure 3.1. Simulation setup in MWO. ..............................................................................35 
Figure 3.2. Simulation circuit of the MWO setup including the source and load.........35 
Figure 3.3. Nonlinear capacitor excited by a Norton equivalent. ...................................36 
Figure 3.4. Capacitance versus reverse bias voltage of a PIN Diode. .........................37 
Figure 3.5. The capacitance versus bias voltage of the PIN Diode in the
simulation. ..................................................................................................................38 
Figure 3.6. Magnitude of the excitation. ...........................................................................39 
Figure 3.7. Phase of the excitation....................................................................................39 
Figure 3.8. The response voltage swing across the capacitor. .....................................40 
Figure 3.9. Simulated voltage spectral density in the neighborhood of the
fundamental ...............................................................................................................43 
Figure 3.10. Predicted spectral density vs. the simulated spectral density in
the neighborhood of the fundamental. ...................................................................43 
Figure 3.11. Simulated spectral phase in the neighborhood of the
fundamental. ..............................................................................................................44 
Figure 3.12. Predicted spectral phase vs. the simulated spectral phase in the
neighborhood of the fundamental. ..........................................................................44 
Figure 3.13. Simulated spectral density in the neighborhood of the second
harmonic. ....................................................................................................................45 
 ix

Figure 3.14. Predicted spectral density vs. simulated spectral density in the
neighborhood of the second harmonic. .................................................................46 
Figure 3.15. Simulated spectral phase in the neighborhood of the second
harmonic. ....................................................................................................................46 
Figure 3.16. Predicted spectral phase vs. simulated spectral phase in the
neighborhood of the second harmonic. .................................................................47 
Figure 3.17. Simulated spectral density in the neighborhood of the third
harmonic. ....................................................................................................................48 
Figure 3.18. Predicted spectral density vs. the simulated spectral density in
the neighborhood of the third harmonic. ................................................................49 
Figure 3.19. Simulated spectral phase in the neighborhood of the third
harmonic. ....................................................................................................................49 
Figure 3.20. Predicted spectral phase vs. the simulated spectral phase in the
neighborhood of the third harmonic. ......................................................................50 
Figure 4.1. A representation of a network's stimulus-response characteristic
under an NPR test. ...................................................................................................51 
Figure 4.2. NPR excitation-response of the RF switch. .................................................52 
 1

CHAPTER 1

INTRODUCTION

The ability to predict nonlinear phenomenon in electronic networks has

become exceedingly important in the design of wireless and microwave systems. In
wireless communications, the attempt to pack more channels in ever tightening
bandwidths has placed stringent requirements on unwanted emissions [1]. Digitally
modulated RF carrier signals designed to occupy limited bandwidths undergo an
undesired expansion of their spectral content upon passing through system power
stages where nonlinearities are most apparent [2].
Nonlinear modeling in high frequency work is often limited to establishing
performance metrics such as P1dB (gain compression point) or IP3 (third order
intermodulation intercept point) describing a device’s performance under a single
tone or a two tone excitation respectively [3]. Historically, this approach has had a
great deal of utility but lacks generality in the face of increasing complexity of
communication systems where digitally modulated signals are not well approximated
by a single or a two tone excitation [4].
It is within this context that a set of closed form solutions predicting the
nonlinear response of a reverse biased PIN diode RF switch to a narrowband
excitation possessing an arbitrary spectrum is presented. Distortion in RF switches
by reverse biased PIN diodes has thus far been addressed under a two tone
excitation in the literature [5].
The schematic of the RF switch utilizing a shunt connected reverse biased
PIN diode is presented in Figure 1.1.
The RF switch in Figure 1.1 permits the flow of RF power from the source to
the load, when the PIN diode is reverse biased. The reverse biased PIN diode
contains a nonlinear parasitic capacitance causing the RF switch's response to be
distorted [5, pp. 21-22].
 2

bias
resistor

RF
bypass RF Vbias
choke
Zo v(t)

DC DC
block block
Zo PIN Zo Zo
Vs(t)
diode

Source RF Switch Load

Figure 1.1. PIN diode RF switch.

The remainder of this chapter will present a brief discussion of the RF switch's
nonlinear characteristics, and an outline of the analysis method used to model its
nonlinear excitation-response relationship.
In Chapter 2, the analysis deriving equations (2.85), (2.86), (2.87), (2.88) and
(2.89) describing the RF switch's nonlinear excitation-response relationship is
presented. In Chapter 3, the validity of these equations is confirmed by simulations
of the RF switch's equivalent network as performed by a commercially available
simulation software known as Microwave Office (MWO). In Chapter 4, the utility of
the equations is further demonstrated by using them to derive a concise formula
predicting the performance metric of noise power ratio (NPR) for the RF switch.
The equivalent circuit of the RF switch with the PIN diode in reverse bias is
presented in Figure 1.2. In Figure 1.2, the RF switch's frequency domain voltage
response is denoted as . Equations (2.85), (2.86), (2.87), (2.88) and (2.89)
describe as the nonlinear response of the RF switch.
Before the derivation of as the nonlinear response of the RF switch is
presented, a brief description of distortion exhibited by the RF switch is discussed in
the next section.
 3

bias
resistor

RF
bypass RF Vbias
choke

Zo v(t), or
V( )
DC
DC block
block C(v) PIN
Zo Zo Zo
Vs(t) diode
reverse bias
equivalent
Source RF Switch Load

Figure 1.2. Equivalent circuit of the RF switch with

the PIN diode in reverse bias.

1.1 THE MECHANISMS OF DISTORTION WITHIN THE RF

SWITCH PREDICTED BY ANALYSIS
The technique of linearization has been paramount in studying countless
circuits and networks containing active elements like vacuum tubes, transistors, and
diodes, all of which are inherently nonlinear; linear analysis however fails to predict
many nonlinear phenomenon that are observed in practical circuits [6]. The
nonlinear phenomenon of signal distortion is exhibited by the RF switch in Figure 1.2
to a measureable degree when it is excited by a sufficiently large signal. The signal
distortion exhibited by the RF switch is predicted by the equations (2.85), (2.86),
(2.87), (2.88), and (2.89); these equations predict newly generated spectral content
at the switch's output that were not present at the switch's input. The next section
briefly outlines the mechanisms of frequency generation exhibited by the RF switch.

1.2 FREQUENCY GENERATION

A single tone excitation at the input of a nonlinear network may produce new
frequencies at the network's output through the generation of harmonics; multiple
tone excitations at the input of a nonlinear network may generate new frequencies at
the network's output through the generation of mixing products [6, pp. 8]. The
following sections discuss these mechanisms of frequency generation due to the RF
switch's nonlinearity.
 4

1.2.1 Harmonic Generation

If the excitation at the input of the RF switch contains a single tone at the
frequency then the response at its output may contain any of the frequencies
with 1, 2, 3, …, where the variable represents the order of the nonlinearity
responsible for the generation of the corresponding harmonics. The significance of
orders of nonlinearity will be discussed further in Section 1.2.1.1. The effect of
multiple tone excitations are discussed next.

1.2.2 Mixing Products

If two distinct tones at frequencies and ω are placed at the input of the
RF switch, the output may contain both tones as predicted by linear theory and
various mixing products at frequencies and with ,
0, 1, 2, … [7]. The term represents the order of the nonlinearity responsible for
the generation of the mixing products in case of a two tone excitation. The next
section outlines the analytical approach used to analyze the excitation-response
relationship of the RF switch.

1.3 THE ANALYTICAL METHOD USED IN THE NONLINEAR

ANALYSIS OF THE RF SWITCH
As stated previously, the closed form expressions (2.85), (2.86), (2.87),
(2.88), and (2.89) describe the nonlinear excitation-response relationship of the RF
switch in Figure 1.2. Section 1.2.1 outlines the analytical method used in the
derivation of these equations and its limiting assumptions.

1.4 AN OVERVIEW OF THE ANALYTICAL METHOD

Nonlinearities in high frequency electronics have historically been divided into
two categories; weak nonlinearities and strong nonlinearities [7, pp. 2-3]. The input-
output relationships of networks possessing weak nonlinearities vary gradually as
the function of the signal amplitude at the input of such networks; in contrast, strong
nonlinearities are characterized by abrupt changes within the operating regions of
their characteristics [6, pp. 1-3]. Analysis methods for strongly nonlinear circuits are
often limited to numerical methods [7, pp. 3].
 5

In the next sections, the analysis of weakly nonlinear networks will be

discussed; this is the method by which the RF switch in Figure 1.2 is analyzed.

1.4.1 Analytical Methods for Weakly Nonlinear

Networks
Analytic methods for weakly nonlinear networks may utilize a series to model
nonlinear transfer characteristics. Given the nonlinear network in Figure 1.3, the
response to an excitation may be modeled using power series as
presented in equation (1.0) if the network is memoryless.
(1.0)

In case of a network with memory, the response to an excitation may be

modeled as Volterra series as presented in equation (1.1) [7, pp. 155-78]. If a
network's transfer characteristics have a frequency dependency, then the network
possesses memory.

Nonlinear
g(t) y(t)
Network

Figure 1.3. Nonlinear network block.

(1.1)
… , ,… … …

If Volterra series are used to represent the output of a memoryless network,

they reduce to power series because power series are a special case of the Volterra
series [7, pp. 174].
In the power series model, the output is related to the input through constants
, , … as shown in equation (1.0). In the Volterra series model, the output is
related to the input through functions , ,… for 1, 2, … as shown in
equation (1.1).
 6

Any signal can be expressed in terms of its Fourier transform as shown with
in equation (1.2).

1 (1.2)
2
An alternate formulation of the Volterra series is shown in equation (1.3)
where the functions , ,… and are the Fourier transformations of
, ,… and respectively.

1
… , ,… … …
2

(1.3)

The highest order of harmonic generation discussed in Section 1.1.1.1 is

limited to the value of utilized in equations (1.0) and (1.3); similarly, the highest
order mixing product represented as in Section 1.1.1.2 is also limited to in
equations (1.0) and (1.3). The constant represents the highest order of
nonlinearity modeled and corresponds to the highest order of harmonic or mixing
product that can be predicted by either equation (1.0) or (1.3).
The analysis of the reverse biased PIN diode switch presented in Figure 1.2
utilizes the Volterra series because the switch is weakly nonlinear and the behavior
of its nonlinear element is frequency dependent. The Fourier transform of the
waveform denoted as in Figure 1.2 representing the response of the
switch to an arbitrary excitation will be approximated as a truncated Volterra series.
In the Volterra analysis of nonlinear circuits, the nonlinear characteristics of
nonlinear elements may be modeled as polynomials. In the analysis of the RF
switch, the nonlinear capacitance's capacitance-voltage relationship is modeled as
equation (1.4)
(1.4)

Equation (1.4) is defined about an operating point fixed by the DC source reverse
biasing the PIN diode in Figure 1.2.
 7

Equations (2.85), (2.86), (2.87), (2.88), (2.89) approximate the RF switch's

response under a set of assumptions. These limiting assumptions are discussed
next.

1.4.2 The Limiting Assumptions of the Nonlinear

Analysis
The analysis resulting in equations (2.85), (2.86), (2.87), (2.88) and (2.89) is
valid under the following conditions:
1. The PIN diode within the RF switch is reverse biased and the excitation is not
large enough to forward bias the diode, cause diode breakdown, or otherwise
cause a discontinuous operation.
2. The voltage swing across the terminals of the parasitic capacitance does not
exceed the voltage interval for which the voltage-capacitance relationship of
the capacitance in equation (1.4) is valid.
3. The excitation is a narrowband modulated RF carrier signal. The bandwidth
of the excitation is assumed to be narrow. The spectrum of the excitation is
sufficiently narrow if the parasitic elements present within the circuit do not
exhibit an appreciable sensitivity to frequency changes across the excitation's
bandwidth.
4. The total power of spectral contents produced by nonlinear responses of
orders higher than the third is negligible compared to that of the nonlinear
responses equal to or lower than the third order.
In the next chapter, the derivation of the equations representing the RF
switch's nonlinear response is presented in detail.
 8

CHAPTER 2

ANALYSIS OF THE RF SWITCH'S RESPONSE

In Chapter 1, it was stated that the nonlinear response of the RF switch in

Figure 1.2 to a narrowband excitation possessing an arbitrary spectrum is presented
as a set of closed form expressions presented in equations (2.85), (2.86), (2.87),
(2.88), and (2.89). In this chapter, the derivation of these closed form expressions is
presented in detail.
The derivation of the switch's response denoted as in Figure 1.2 begins
with a network manipulation presented in the next section.

2.1 ARRIVING AT THE SWITCH'S EQUIVALENT NETWORK

When the PIN diode within the RF switch is reverse biased, the switch’s high
frequency equivalent circuit appears as the circuit in Figure 1.2. Seeking a solution
to the response prompts a simplification of the networks in parallel with the
nonlinear capacitance. We consolidate the circuit elements to the left and right of
the nonlinear capacitor as one port networks, denoting them "Network L" and
"Network R" as shown in Figure 2.1.

Network L Network R
bias YinL YinR
resistor RF
bypass RF
choke
Zo v(t)

DC DC
block C(v) block
Zo PIN Zo
Vs (t) diode Zo
reverse bias
equivalent

Figure 2.1. The RF source, load, and the RF switch

consolidated into network blocks.
 9

Network L and network R can be consolidated to form a Norton equivalent

circuit in parallel with the nonlinear capacitor as shown in Figure 2.2.

, or
V( )

C(v)
Is( ) Yso( )

Figure 2.2. Norton equivalent circuit as the

arbitrary excitation.

Figure 2.2 represents the final result of the network transformation carried out
on the RF switch in Figure 1.2. The subsequent analysis will be performed on this
circuit in order to arrive at the solution for the time domain voltage waveform
and its frequency domain equivalent . The next section presents this analysis.

2.2 THE DERIVATION OF THE SWITCH'S RESPONSE

The derivation of the closed form solutions for begins by formulating a
current node equation involving the quantity . The quantity in the node
equation will be expressed as a Volterra series shown in equation (1.3). The
subsequent mathematical manipulations will aim to identify transfer functions of the
series in order to approximate as a function of the arbitrary excitation .
Once the solution for is obtained, a Fourier transformation is performed on
to obtain assuming that the excitation possesses a narrow spectrum with
the constant representing its center of frequency.

2.3 FORMULATION OF THE NODE EQUATION

Equation (2.1) sums up all the currents entering and leaving the circuit node
labeled shown in Figure 2.3.
 10

i s(t) id(t) v(t)

i yso(t)
C(v)
Is( ) Yso( )

Figure 2.3. Currents forming the node equation.

(2.1)

The source current in equation (2.1) is described in the time domain and
it is given as the inverse Fourier transform of the function in equation (2.2).

1 (2.2)
2
The arbitrary admittance in Figure 2.3 has an impulse response given
by the inverse Fourier transform in equation (2.3).

1 (2.3)
2
The current flowing through the admittance is dependent on
the voltage . We shall formulate it as a convolution in the time domain as shown
in equation (2.4).

1 (2.4)
2
The current through the nonlinear capacitance denoted as is formulated
by multiplying equation (1.4) by the time derivative of the voltage waveform as
shown below to arrive at equation (2.5).
 11

(2.5)

Substituting equations (2.4) and (2.5) back into the node equation in (2.1)
produces the equation in (2.6).

1 (2.6)
2
In the next section, the node equation (2.6) will be expressed as a Volterra
series.

2.3.1 Formulation of the Response as a Series

Summation
In the continuation of the analysis, we seek to express as a function of
using the Volterra series in equation (1.1) rewritten as equation (2.7). Doing so
effectively models the response as a function of the excitation through the
network transfer functions , ,… depicted in Figure 2.4. The variable in
equation (2.7) representing the highest order of nonlinearity discussed in section
1.1.1.1 shall remain arbitrary at this time.

Nonlinear
Network
i s(t) h(t1 ,t2 ,...tn ) v(t)

Figure 2.4. A nonlinear network block.

 12

(2.7)
… , ,… … …

Equation (2.2) describes in terms of its frequency domain equivalent

. It shall be necessary to represent the nonlinear network transfer functions
, ,… as frequency domain equivalents , ,… . Therefore,
equation (2.7) shall be manipulated to formulate as a function of and
, ,… in continuation of the analysis.
Using equation (2.2), is formulated as equation (2.8).

1
(2.8)
2

Therefore, the product … can be expressed as shown

in equation (2.9).
…
1 1
…
2 2
1 (2.9)
·
2
The right hand side of equation (2.9) can be rearranged as a multiple integral
resulting in equation (2.10).
…
1
… … …
2

(2.10)
We substitute the right hand side of equation (2.10) in equation (2.7) to obtain the
expression in equation (2.11).

… , ,…

1
… … …
2
· … (2.11)
 13

The order of integration in (2.11) is changed to yield the expression in equation

(2.12).

1
… …
2

· … , ,… … …

(2.12)

The right hand side of equation (2.13) is the component inside the parenthesis in
equation (2.12) and is equivalent to , ,… which is the multi-dimensional
Fourier transform of … .

, ,… … , ,… …

(2.13)

Replacing the content inside the parenthesis in equation (2.12) with , ,…

allows to be described as the equation (2.14).

1
… , ,… … …
2

(2.14)

To facilitate a more concise notation of , we shall rewrite equation (2.14) as

equations (2.15) and (2.16).

(2.15)

where,

1
… , ,… … …
2

(2.16)
 14

The right hand side of equation (2.15) shall be substituted in every instance of
in equation (2.6) to arrive at equation (2.17). In cases where instances of
are multiplied by other instances of or their derivatives, the corresponding
summations using equation (2.15) shall be assigned unique index variables.
Equation (2.17) therefore contains new index variables n , n and n in summations
being multiplied by other summations or their derivatives.

1
2

(2.17)

We substitute the left hand side of the equality in equation (2.17) with equation (2.2),
and on the right hand side, we expand the square of the summation into a double
summation to arrive at equation (2.18).

1 1
2 2

(2.18)

It is necessary to manipulate (2.18) in a manner that eliminates the

convolution operation on its right hand side. Equation (2.19) presents a relationship
that allows the convolution operation in equation (2.18) to be absorbed.

1
2

1
… . , ,… …
2

(2.19)

The right hand side of equation (2.19) is derived in the following manner:
 15

Using equation (2.16), the left hand side of equation (2.19) is expanded as
shown in equation (2.20).

1
2

1
… , ,… … …
2

1
(2.20)
2

Recalling that the time domain convolution of signals and is given by

equation (2.21),

(2.21)

we express (2.20) alternatively as equation (2.22).

1
2

1
2

1
… , ,… …
2

1
· …
2

(2.22)

Moving the orders of integration on the right hand side of equation (2.22), we arrive
at equation (2.23).

1
2

1
… , ,… …
2
 16

1
· …
2

(2.23)

Focusing on the term inside the inner parenthesis of equation (2.23), we recognize it
as a Fourier transform with respect to the frequency thereby
allowing another manipulation to take place as shown below:

2 (2.24)

The content within the inner parenthesis on the right hand side of equation (2.23) is
replaced with the right hand side of equation (2.24) to yield equation (2.25).

1
2

1
… , ,… …
2

1
· 2 …
2

1
… , ,… …
2

· …
(2.25)

The right hand side of equation (2.25) is rearranged to arrive at equation (2.26)
confirming the result in equation (2.19).

1
2

1
… , ,… …
2

…
 17

(2.26)

Going back to equation (2.18) to proceed with the analysis, we expand the
summations on its right hand side to arrive at (2.27). A truncation of the terms on
the right hand side of (2.27) will be necessary. The next section describes this
process of truncation.

1
2
1
2

(2.27)

2.3.2 Truncation of the Summations

There can be an arbitrary number of terms specified by the choice of on the
right hand side of equation (2.27); however we shall truncate its right hand side to
account only for the first, second, and third order terms to arrive at the approximation
in equation (2.28).

1
2
1
2

(2.28)

Equation (2.28) is an approximation that accounts for nonlinear effects only up to the
third order. The right hand side of equation (2.28) contains voltage components
 18

denoted as , and , where their subscripts signify their order; these

voltages (or their derivatives) can also be multiplied together to form various order
components. For example, the product of the functions produces a third
order component because the sum of their subscripts is equal to three.
To proceed with the analysis, we shall collect the terms that account for the
first order components on the right hand side of equation (2.28) and shall do the
same with the left hand side of equation (2.28). Knowing that the sum of all the first
order terms on the right hand side must equal to the sum of the first order terms on
the left hand side, the equality in equation (2.29) is formed.

1 1 (2.29)
2 2
Equaling the sum of all of the second order terms on the right hand side of
equation (2.28) to all of the second order terms on the left hand side of equation
(2.28) (in this case equaling to zero), allows the equality in equation (2.30) to be
formed.

1 (2.30)
0
2
The last group of components is the group of the third order components; the
collection of the third order terms on both sides of equation (2.28) allows the equality
in equation (2.31) to be formed.

1
0
2

(2.31)

Each instance of v t in equations (2.29), (2.30), and (2.31) is described by

equation (2.16) and therefore contains an unknown transfer function
H ω , ω , … ω , for 1,2,3. The next section describes the method by which
these transfer functions are identified.
 19

2.3.3 The Identification of the Multidimensional

Transfer Functions
The first order, the second order and the third order transfer functions
denoted H ω , ω , … ω for 1,2,3 must be identified in order to form a Volterra
series solution to . The next section identifies these transfer functions using
equations (2.29), (2.30) and (2.31).

2.3.3.1 THE DERIVATION OF THE FIRST

ORDER TRANSFER FUNCTION
The first order transfer function is derived by first considering the equality of
the first order terms in (2.29) rearranged as shown below:

1 1 (2.32)
0
2 2
Recalling from equation (2.16) that is alternatively expressed as an
integral whose integrand is composed of the first order transfer function and
the current source spectral density function , we formulate the equality in (2.32)
alternatively as equation (2.33).

1 1 1
2 2 2
1
0 (2.33)
2
The relationship in equation (2.19) allows the terms being convolved together
in equation (2.33) to be expressed alternatively as a single integral. The derivative
in equation (2.33) can also be moved inside the integral to differentiate the
exponential function with respect to time. The result of these manipulations yields
the equation (2.34).

1 1 1
2 2 2

0 (2.34)

The first order transfer function is identified by forming an equality with

respect to the integrands in equation (2.34) by performing a Fourier transform on
 20

both sides of the equation; dividing both sides by I ω next and rearranging the
terms yields equation (2.35).
(2.35)
1
Equation (2.35) is used to identify the first order transfer function presented in
equation (2.36).

1 (2.36)

2.3.3.2 THE DERIVATION OF THE SECOND

ORDER TRANSFER FUNCTION
The second order transfer function is derived in the same manner as the first
order transfer function by manipulating equation (2.30) rearranged slightly as shown
in equation (2.37).

1 (2.37)
0
2
The relationship,

1
2

is used to express equation (2.37) alternatively as equation (2.38).

1 1 (2.38)
0
2 2
Using equation (2.16) to replace and with the corresponding
integrals in equation (2.38) yields equation (2.39).

1 1
,
2 2
1
,
2
1 1 (2.39)
0
2 2
 21

Using the relationship in equation (2.19), the convolution operation on the left
hand side of (2.39) can be absorbed; the derivatives are also moved inside the
integrals to yield equation (2.40).

1
,
2
1
,
2
1 1
0
2 2
(2.40)

Equation (2.41) is formed with respect to the integrands within equation

(2.40).

1
, ,
2
(2.41)

Using equation (2.41), the second order transfer function is identified and
presented in equation (2.42).

1 (2.42)
,
2

2.3.3.3 THE DERIVATION OF THE THIRD

ORDER TRANSFER FUNCTION
The third order transfer function will be identified in a similar manner. The
third order transfer function is derived by first taking the third order equality in (2.31)
and rearranging it as shown in equation (2.43).

1
2

0 (2.43)

The relationship,

1
3
 22

is used to express equation (2.43) alternatively as equation (2.44).

1 1
0
2 3
(2.44)

Using equation (2.16) to replace all the instances of v t , v t , and v t in

equation (2.44) with the corresponding integrals yields equation (2.45).

1
, ,
2
1
2
1
, ,
2
1
,
2
1 1
2 3

0 (2.45)

Equation (2.45) is further simplified using the relationship (2.19) to absorb the
convolution operation. The derivatives are also moved inside the integrals to yield
the equality in equation (2.46).

1
, ,
2
1
, ,
2
1
,
2
·
1 1
2 3
·

0 (2.46)
 23

As with the first and the second order transfer functions, equality (2.47) is
formed with respect to the integrands in equation (2.46).

, , , ,
1
,
3
(2.47)

Equation (2.47) is used to identify the third order transfer function presented
in equation (2.48).

, , ,

1 (2.48)
3
Equations (2.36), (2.42), and (2.48) are the first order, the second order and
the third order transfer functions respectively. The next section uses these transfer
functions to arrive at a series solution to .

2.3.4 The Equation Describing the Response in

the Time Domain
The solution for the response voltage waveform as a function of the
current source and network transfer functions H ω , ω , … ω for 1,2,3 is
presented in equation (2.49), (2.50), (2.51), (2.52), and (2.53).

(2.49)

given,

1
… , ,… … …
2

(2.50)

where,

1 (2.51)
 24

1 (2.52)
,
2

, , ,

1 (2.53)
3
The above equations represent in the time domain. The frequency
domain equivalent of denoted as will be derived in Section 2.2.1.6 by
performing a Fourier transformation on . Before proceeding however, a
mathematical relationship between Volterra series and spectral convolutions must
be formulated in order to perform such transformation. This mathematical
relationship is outlined in Section 2.2.1.5.

2.3.5 Relating Volterra Series to Spectral

Convolutions
Given the time domain functions and , their product . can be
expressed in terms of their frequency domain equivalents as shown below,

1 1
.
2 2
1
2
1
. (2.54)
2

where,

1 1 (2.55)
2 2
It is seen that the double integral in equation (2.54) has the same form as the
component of the Volterra series in equation (2.16) with 2. In the derivation
below, it is confirmed that a Fourier transformation of the right hand side of equation
(2.54) is equivalent to the convolution of the spectral densities and as
shown in equation (2.56).
 25

1
.
2
1
2
1
2
2
1
2
1
2
1 (2.56)
2
Therefore, the product of two signals in the time domain can be expressed as a
double integral presented in equation (2.57) or it can be represented as an inverse
Fourier transform of the corresponding signals' spectral convolution shown in
equation (2.58).

1 (2.57)
.
2

1 (2.58)
.
2
The relationships in (2.57) and (2.58) can be extended to account for multiple
signals. If we define the signal t through t as shown below,

(2.59)

where,

1 (2.60)
2

1 (2.61)
2
 26

then the functions ω through ω can be deduced using equations (2.59),

(2.60), and (2.61) as shown in equation (2.62).

1
2
1
2

1 (2.62)
2
Substituting all instances of into for 3 within the set of
equations in (2.62), results in equation (2.63).

1 (2.63)
…
2
Substituting all instances of into for 3 within the set of equations
in (2.61) results in equation (2.64),
(2.64)
…
A Fourier transformation of both sides of equation (2.64) yields equation (2.65).

(2.65)
…

Equating H ω in equation (2.65) to H ω in equation (2.63) produces the

equation shown below:

1
… …
2

Rewriting the above more concisely we have equation (2.66).

1
… … (2.66)
2π

An inverse Fourier transformation on both sides of equation (2.66) yields equation

(2.67) shown below:

1 1
… …
2 2
 27

1 (2.67)
…
2
The left hand side of equation (2.67) can be expressed alternatively as shown
below:

…
1 1 1
2 2 2
1 (2.68)
… … …
2
Observing equations (2.66), (2.67), and (2.68), we summarize the relationship
between Volterra series and spectral convolutions as equations (2.69) and (2.70).

…
1
… … …
2
1
… (2.69)
2

and,

1
… … …  
2
1 (2.70)
…
2
The relationships presented in equations (2.69) and (2.70) will be used in the
next section to evaluate the Fourier transformation of in equation (2.49).

2.3.6 The Formulation of the Response in the

Frequency Domain
Equation (2.49) describes the circuit response in the time domain. In
order to derive its frequency domain equivalent , we perform a Fourier
transformation on as shown below:
 28

(2.71)

Equations (2.50), (2.51), (2.52), (2.53) are substituted into equation (2.71), to arrive
at equation (2.72).

1
2
1 1
2 2

1
,
2

1 1
2 3

(2.72)

The evaluation of the Fourier integrals in equation (2.72) is possible under a

simplifying assumption. We shall consider a case where is non-zero over a
small interval of in neighborhood of for a positive range of as formalized
below:

0 0 ∞
2 2

0 0 ∞
2 2
(2.73)

The function is depicted in Figure 2.5 on a representative basis.

 29

Amplitude Is( )

0
BW
Figure 2.5. A depiction of the current source's spectrum.

Considering in equation (2.73) over a negative range of also implies the

following:

0 ∞ 0
2 2

0 ∞ 0
2 2
(2.74)

The evaluation of the Fourier integrals in equation (2.72) must be carried out
over the ranges for which both and are defined in order to evaluate
the effect of excitations properly.
Because is non-zero over a small interval of in the small
neighborhood of , the integral that characterizes the spectral content of the
second order response in (2.72) is non-zero only on a small interval of in the
neighborhood of . This term is shown below for convenience:

1 1
2 2
·
(2.75)

The left hand side of equation (2.76) varies little as and vary in the
neighborhood of and therefore can be treated as a constant in the neighborhood
of

2
,
2 2
(2.76)
 30

Consequently, the left hand side of (2.76) can be pulled out of the integral in
equation (2.75) simplifying the integral to:

1 2 1
2 2 2 2
·
(2.77)

Utilizing the relationship in equation (2.70), we can evaluate the Fourier integral in
equation (2.77) as a spectral convolution presented in (2.78), characterizing the
response's spectral content in the neighborhood of the second harmonic of the
excitation's center of frequency .

1 1 2 (2.78)
2 2 2 2
This treatment can be carried out to formulate the response's spectral content
in the neighborhood of the 3rd harmonic of the excitation's center of frequency .
Equation (2.79) recalls the Fourier integral of the third order response within
equation (2.72).

1
,
2

1 1
2 3

(2.79)

The integral in (2.79) must be evaluated for all of the possible combinations of
, , in the neighborhood of . This gives rise to the combinations of
, , equaling to , , , , , and , , as well as
, , . The combination of , , containing all positive values of
allows the integral in equation (2.73) to be simplified as shown in (2.80).

1 3 2 1 3
.
2 3 3 2 2 3 3 3
 31

1
2
(2.80)

Using the relationship in equation (2.70), the Fourier integral of the third order
response in equation (2.80) can be expressed as equation (2.81) representing the
spectral content in the neighborhood of the third harmonic of the excitation's center
frequency ω .

1 1 3 2 1 3
2 2 3 3 2 2 3 3 3

· (2.81)

For all of the combinations of , , containing instances of , the

Fourier integral of the third order response in equation (2.72) can be simplified as
shown in equation (2.82).

1 0 1
2 0 0 3
1
·
2

1 0 1
2 0 0 3
1
·
2

1 2 1
2 2 2 3
1
·
2

(2.82)
 32

Using the relationship in equation (2.70), the Fourier integral of the third order
response in equation (2.82) can be reduced and summed up to the final form
presented in equation (2.83). Equation (2.83) represents the third order response in
the neighborhood of the excitation's center of frequency ω .

1 1 2
2 2 2 2

· (2.83)

The Fourier integral of the linear response in equation (2.72) is evaluated and
presented in equation (2.84).

1 (2.84)
2
Equation (2.83) together with equation (2.84) represent the response's spectral
content in the neighborhood of the excitation's center of frequency ω .
Equations (2.78), (2.81), and (2.83) are added to the right hand side of
equation (2.84) to formulate the Fourier equivalent of the response voltage
presented as equations (2.85), (2.86), (2.87), (2.88), and (2.89).
(2.85)

1 1 2
2 2 2 2

· (2.86)

 
1 1 2 (2.87)
2 2 2 2

1 1 3 2 1 3
2 2 3 3 2 2 3 3 3

· (2.88)

Where ω is the frequency of the RF carrier signal as defined in equation (2.73) and,
 33

1 (2.89)

The closed form expression (2.85), (2.86), (2.87), (2.88) and (2.89) represent
the spectral density of the response voltage denoted as to a current
source possessing the spectrum in parallel with an arbitrary admittance
depicted in Figure 2.3. The circuit in Figure 2.3 is the equivalent circuit of the RF
switch in Figure 1.2, therefore these equations will predict in Figure 1.2
provided that the limiting assumptions in Section 1.2.1.2 are not violated.
The next chapter compares the predictions of equations (2.85), (2.86), (2.87),
(2.88) and (2.89) to numerical simulations as performed by MWO in order to verify
the validity of these equations.
 34

CHAPTER 3

ANALYTICAL PREDICTIONS COMPARED WITH

SIMULATED DATA

In Chapter 2, closed form expressions (2.85), (2.86), (2.87), (2.88), and (2.89)
were derived representing the nonlinear response of the circuit in Figure 2.2. In this
chapter, these closed form expressions are validated by comparing their predictions
evaluated in MATLAB against simulations performed by MWO.
In the subsequent sections, a description of the simulation in MWO is
detailed; details of analytical predictions performed in MATLAB are also presented.
Numerical results produced by simulation in MWO are compared to numerical
results generated by MATLAB.

3.1 CIRCUIT DETAILS IN MWO

Figure 3.1 presents the simulation setup in MWO. It is composed of a
nonlinear capacitor connected to a source and a load through two resistors of
negligible resistance.
The simulation circuit in Figure 3.1 was constructed to have the same
equivalent circuit as the one presented in Figure 2.2. To make this assertion clear,
the circuit in Figure 3.1 is presented as the schematic companion to the MWO
simulation setup in Figure 3.1.
In Figure 3.1, the graphical symbol connected to the left side of the electric
circuit represents the excitation and has an output impedance of 50 ohms. In Figure
3.2, this source is represented schematically and identified as the "source." In
Figure 3.1, the graphical symbol connected to the right side of the simulation circuit
represents the load and it has an impedance of 50 ohms. In Figure 3.2, this load is
represented schematically and identified as "load."
 35

SOURCE LOAD

PORTMOD_F
P=1
Z=50 Ohm
Pwr=16 dBm
NET="sqwr"
FRes=10 MHz
WINDOW=NONE

RES RES PORT

ID=R1 ID=R2 P=2
R=1e-6 Ohm R=1e-6 Ohm Z=50 Ohm

+
VCNL
ID=CN1
AFAC=1

Figure 3.1. Simulation setup in MWO.

negligible negligible
v(t),
resistance V( )or resistance

C(v)
Is( ) 50 50
ohms ohms

source simulation load

circuit
Figure 3.2. Simulation circuit of the MWO setup
including the source and load.

In Figure 3.1, the series resistors satisfy a requirement by the software that
the source is not directly tied to the load. The value of the resistors has been
chosen sufficiently small so as to not influence the simulation to a noticeable degree.
MWO effectively simulates the circuit in Figure 3.3 when simulating the circuit
setup in Figure 3.1. The admittance in Figure 3.3 is equivalent to the output
admittance of the source in parallel with the admittance of the load. In the
simulation, the impedance of is equal to 25 ohms.
 36

v(t), or
V( )

25 C(v)
Is( ) ohms Yso( )

Figure 3.3. Nonlinear capacitor excited by a Norton equivalent.

In MWO, the excitation source is specified to possess an arbitrary spectrum

approximated by a set of discrete tones. The source can also possess an arbitrary
"available power" [8]. The description of the nonlinear capacitance shall be
discussed next.

3.1.1 The Nonlinear Capacitance

In Figure 3.3, the function provides a voltage-capacitance description of
the nonlinear capacitance, and this model was assumed in the analysis in Chapter 2.
An alternate description of a nonlinear capacitance models the charge-voltage
relationship of nonlinear capacitors [7, pp. 36-9]; this model is used in the MWO
software and it appears as equation (3.1).
(3.1)

In order to provide MWO with the correct charge model coefficients, a

conversion was performed on the capacitance-voltage model coefficients to arrive at
the charge-voltage model coefficients.
The current through a capacitor can be expressed in terms of a time
derivative of the capacitor's instantaneous charge or the time derivative of its
terminal voltage. Equation (3.2) shows the relationship between a capacitor's
current, instantaneous voltage, instantaneous charge, and capacitance.

(3.2)
 37

To arrive at the equivalence between the charge model and the capacitance
model of the nonlinear capacitance, we differentiate equation (3.1) with respect to
the variable to arrive at equation (3.3).

(3.3)
2 3

Comparing equation (3.3) to equation (1.4) establishes a relationship between

the charge coefficients and the capacitance coefficients. This relationship is
presented in equations (3.4), (3.5) and (3.6) provided,

(3.4)

(3.5)
2
(3.6)
3
In the simulation, the coefficients of the nonlinear capacitor were chosen to
describe one of the PIN diodes studied in the paper "Distortion in Microwave
Switches by Reverse Biased PIN Diodes" by Caverly and Hiller [5, pp. 21-22]. The
capacitance-voltage plot of the reverse biased PIN diode in their paper is shown in
Figure 3.4. The corresponding coefficients are shown in equations (3.7), (3.8), and
(3.9).

Figure 3.4. Capacitance versus reverse

bias voltage of a PIN Diode.
 38

The coefficients in (3.7), (3.8), and (3.9) were determined by fitting equation
(1.4) to the data in Figure 3.4 at a bias point of -5.0 volts. These coefficients are
presented in equations (3.7), (3.8), and (3.9). Equation (3.10) was constructed to
plot this voltage-capacitance curve in MATLAB; Figure 3.5 presents this plot.
(3.7)
0.12

(3.8)
6.444 · 10

(3.9)
8.888 · 10

(3.10)
5.0

0.2

0.19

0.18

0.17

0.16

pF0.15

0.14
Bias Point
0.13

0.12

0.11

-8 -7 -6 -5 -4 -3 -2 -1
Volts

Figure 3.5. The capacitance versus bias

voltage of the PIN Diode in the simulation.

The details associated with the input excitation within the simulation will be
discussed in the next section.

3.1.2 The Input Excitation in Simulation

The excitation is composed of nine discrete tones of equal amplitude and
random phases spanning from 960 MHz to 1040 MHz. These tones are spaced at
10 MHz apart and represent a band-pass modulated tone of about 9% bandwidth.
 39

The center frequency of the excitation is 1000 MHz and is represented as in the
closed form equations describing . Figure 3.6 and Figure 3.7 represent the
magnitude and phase of the excitation's voltage spectrum respectively. The phase
angles of the individual tones within the excitation were selected at random.

Magnitude of Excitation
10
0
-25 1040 MHz
-50 -3.542425298 dB
960 MHz
-75 -3.542425298 dB
-100
-125
-150
-175
-200
-225 DB(|Vharm(PORT_2)|)[*] (dB)
portmod

-250
p1: FREQ = 1000 MHz
-275
-300
700 800 900 1000 1100 1200 1300
Frequency (MHz)

Figure 3.6. Magnitude of the excitation.

Phase of Excitation
180
160 1040 MHz
78.5 Deg
140 1010 MHz
980 MHz 57 Deg
120 30.4 Deg
100
80 970 MHz
11 Deg
60
40 960 MHz
0 Deg
20 p1
0
-20
-40
1030 MHz
-60 990 MHz -7.8 Deg
-80 2.3 Deg

-100 Ang(Vharm(PORT_2))[*] (Deg)

portmod 1000 MHz 1020 MHz
-120 -15.6 Deg -23.5 Deg
-140 p1: FREQ = 1000 MHz
-160
-180
700 800 900 1000 1100 1200 1300
Frequency (MHz)

Figure 3.7. Phase of the excitation.

 40

The excitation was specified to deliver 16.0 dBm to a matched load. This
power level was selected to maximize the voltage swing across the nonlinear
capacitance thereby simulating the highest power level allowable by the analysis.
Figure 3.8 presents a plot (generated by MWO after the simulation) of the
voltage swing across the nonlinear capacitor.

Voltage Envelope
6

Re(Venv(PORT_2,128,1))[1] (V)
portmod
4
9.904e-008 s
4.998 V
0s
4.982 V
2
p1 p1: FREQ = 1000 MHz
0

8.327e-008 s
-1.78 V
-2
0 20 40 60 80 100
Time (n sec)

Figure 3.8. The response voltage swing across the

capacitor.

The next section provides the details associated with the analytical
predictions of the nonlinear response exhibited by the simulation setup in MWO.

3.2 DETAILS OF ANALYTICAL PREDICTION USING THE

CLOSED FORM SOLUTIONS IMPLEMENTED IN MATLAB
Utilizing equations (2.85), (2.86), (2.87), (2.88) and (2.89) in MATLAB to
predict the nonlinear response of the MWO simulation setup in Figure 3.1 requires a
knowledge of the excitation's spectrum and peak amplitude.
The one sided (positive side) spectrum of the peak current excitation utilized
in MATLAB is presented in equations (3.11), (3.12) and (3.13). The peak current as
represented by i in equation (3.11) is calculated based on the available power
specified in the simulation.
 41

(3.11)

where,
(3.12)
960 970 980 990 1000 1010 1020 1030 1040 · 2 · 10 1,2,3, … 9
(3.13)
0, 11, 30.4, 2.3, 15.6, 57, 23.5, 7.8, 78.5 1,2,3, … 9
The available power integrated over the bandwidth of the excitation's
spectrum in the simulation is 16.0 dBm; this leaves the available power at each tone
as equation (3.14). The power per tone expressed in Watts is given in equation
(3.15).

1 (3.14)
16 10 6.4575
9
. (3.15)
0.001 · 10 0.0044233
The available power at each tone in a 50 ohm system allows for a calculation
of the peak current provided by the source as shown in equation (3.16).

8 0.0044234 (3.16)
0.026603
50

The remaining constants of the equations are as follows:

(3.17)
0.12

(3.18)
6.444 · 10

(3.19)
8.888 · 10

(3.20)
2 10 /

1 (3.21)
25
The next section will carry out a comparison of the data generated by the
analysis against the data simulated by MWO.
 42

3.3 A COMPARISON OF PREDICTED VS. SIMULATED

SPECTRAL CONTENT
The validity of equations (2.85), (2.86), (2.87), (2.88) and (2.89) is confirmed
by examining their predictions of the simulation's results in the neighborhood of the
fundamental, the second harmonic as well as the third harmonic of the excitation's
center frequency.
All magnitude plots generated by MWO represent the response's peak
voltage in frequency domain. In MATLAB, the response's voltage magnitude in
frequency domain denoted as |V| B is calculated by equation (3.22); the response's
phase denoted V is calculated using equation (3.23).

| | (3.22)
| | 20 ·

180 (3.23)
·

3.3.1 Simulation in the Neighborhood of the

Fundamental
Figure 3.9 was generated by MWO and it represents the magnitude of the
simulated voltage response in the neighborhood of the excitation's center frequency
of 1000 MHz.
In Figure 3.10, the magnitude of the voltage spectrum in the neighborhood of
the fundamental as predicted by equation (2.86) is plotted against the magnitude of
the voltage spectrum as generated by the MWO simulation.
Figure 3.11 was generated by MWO and it represents the phase of the
simulated voltage response in the neighborhood of the excitation's center frequency
of 1000 MHz.
In Figure 3.12, the phase of the voltage spectrum in the neighborhood of the
fundamental as predicted by equation (2.86) is plotted against the phase of the
voltage spectrum as generated by the MWO simulation.
 43

Magnitude about the Fundemental

10 1050 MHz
0 1000 MHz -73.91256572 dB
-25 -3.541770852 dB
-50 1120 MHz
-98.84951943 dB
-75
-100
-125
-150
-175
-200
-225 DB(|Vharm(PORT_2)|)[*] (dB)
portmod

-250
p1: FREQ = 1000 MHz
-275
-300
700 800 900 1000 1100 1200 1300
Frequency (MHz)

Figure 3.9. Simulated voltage spectral density

in the neighborhood of the fundamental

X: 1000 Predicted by Analysis

Y: -3.542 Simulated by MWO

-50

X: 1120
X: 1050 Y: -98.85
-100 Y: -73.91

dBm

-150

-200

-250
750 800 850 900 950 1000 1050 1100 1150 1200 1250
MHz

Figure 3.10. Predicted spectral

density vs. the simulated spectral
density in the neighborhood of the
fundamental.
 44

Phase about the Fundemental

180 960 MHz
160 1040 MHz
-1.064850572 Deg 77.3586926 Deg
140
120
100
80
60
40
20
0 p1
-20
-40
-60
-80
-100
Ang(Vharm(PORT_2))[*] (Deg)
portmod

-120
-140 p1: FREQ = 1000 MHz
-160
-180
700 800 900 1000 1100 1200 1300
Frequency (MHz)

Figure 3.11. Simulated spectral phase in the

neighborhood of the fundamental.

Predicted by Analysis
150
Simulated by MWO

X: 1040
100 Y: 77.36

50
X: 960
Y: -1.065
Deg
0

-50

-100

-150

750 800 850 900 950 1000 1050 1100 1150 1200 1250
MHz

Figure 3.12. Predicted spectral phase vs. the

simulated spectral phase in the neighborhood of the
fundamental.
 45

Figures 3.10 and 3.12 confirm that equation (2.86) predicts the linear and
nonlinear response in the neighborhood of the excitation's center of frequency to a
good degree for both magnitude and phase. As evident however, equation (2.86)
does not predict the fifth and the seventh order components of the distortion due to
the truncation of equation (2.28) as described in Section 2.2.1.2.
The next section will examine the response in the neighborhood of the
second harmonic of the excitation's center of frequency.

3.3.2 Simulation in the Neighborhood of the

Second Harmonic
Figure 3.13 was generated by MWO and it represents the magnitude of the
simulated voltage response in the neighborhood of the excitation's second harmonic
frequency of 2000 MHz.
In Figure 3.14, the magnitude of the voltage spectrum in the neighborhood of
the second harmonic as predicted by equation (2.87) is plotted against the
magnitude of the voltage spectrum as generated by the MWO simulation.

Magnitude about the Second Harmonic

0
1920 MHz 2000 MHz
-25 -73.35645242 dB -55.8947701 dB
-50 2080 MHz
-75 -72.74199129 dB
-100 2050 MHz
-79.61342757 dB
-125
-150
-175
-200
-225 DB(|Vharm(PORT_2)|)[*] (dB)
portmod
-250
-275 p1: FREQ = 1000 MHz
-300
1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500
Frequency (MHz)

Figure 3.13. Simulated spectral density in the

neighborhood of the second harmonic.
 46

Predicted by Analysis
X: 2000
Simulated by MWO
Y: -55.9
-50 X: 1920
Y: -73.36

X: 2080
X: 2050 Y: -72.74
-100 Y: -79.61

dBm

-150

-200

1800 1850 1900 1950 2000 2050 2100 2150 2200

MHz

Figure 3.14. Predicted spectral density vs.

simulated spectral density in the
neighborhood of the second harmonic.

Figure 3.15 was generated by MWO and it represents the phase of the
simulated voltage response in the neighborhood of the excitation's second harmonic
frequency of 2000 MHz.

Phase about the Second Harmonic

180
160 2080 MHz
62.55339146 Deg
140
120
100
80
60
40
20 p1
0
-20
-40
-60
-80
-100 Ang(Vharm(PORT_2))[*] (Deg)
portmod
-120
-140 p1: FREQ = 1000 MHz
1920 MHz
-94.86839875 Deg
-160
-180
1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500
Frequency (MHz)

Figure 3.15. Simulated spectral phase in the

neighborhood of the second harmonic.
 47

In Figure 3.16, the phase of the voltage spectrum in the neighborhood of the
second harmonic as predicted by equation (2.87) is plotted against the phase of the
voltage spectrum as generated by the MWO simulation.

Predicted by Analysis
150
Simulated by MWO

100 X: 2080
Y: 62.55

50

Deg
0

-50
X: 1920
Y: -94.87

-100

-150

1800 1850 1900 1950 2000 2050 2100 2150 2200

MHz

Figure 3.16. Predicted spectral phase vs.

simulated spectral phase in the
neighborhood of the second harmonic.

Figures 3.14 and 3.16 confirm that equation (2.87) predicts the nonlinear
response in the neighborhood of the excitation's second harmonic frequency to a
good degree for both magnitude and phase. As evident however, equation (2.87)
does not predict the fourth and the sixth order components of the distortion due to
the truncation of equation (2.28) as described in Section 2.2.1.2.
The next section will examine the response in the neighborhood of the third
harmonic of the excitation's center of frequency.

3.3.3 Simulation in the Neighborhood of the

Third Harmonic
Figure 3.17 was generated by MWO and it represents the magnitude of the
simulated voltage response in the neighborhood of the excitation's third harmonic
frequency of 3000 MHz.
 48

Magnitude about the Thrid Harmonic

0 2880 MHz 3000 MHz
-25 -100.1044635 dB -67.50557761 dB
-50 3100 MHz
-85.74430322 dB
-75
-100 3120 MHz
-99.56119618 dB
-125
-150
-175
-200
-225
DB(|Vharm(PORT_2)|)[*] (dB)
-250 portmod

-275 p1: FREQ = 1000 MHz

-300
2500 2600 2700 2800 2900 3000 3100 3200 3300 3400 3500
Frequency (MHz)

Figure 3.17. Simulated spectral density in the

neighborhood of the third harmonic.

In Figure 3.18, the magnitude of the voltage spectrum in the neighborhood of

the third harmonic as predicted by equation (2.88) is plotted against the magnitude
of the voltage spectrum as generated by the MWO simulation.
Figure 3.19 was generated by MWO and it represents the phase of the
simulated voltage response in the neighborhood of the excitation's third harmonic
frequency of 3000 MHz.
In Figure 3.20, the phase of the voltage spectrum in the neighborhood of the
third harmonic as predicted by equation (2.88) is plotted against the phase of the
voltage spectrum as generated by the MWO simulation.
Figures 3.18 and 3.20 confirm that equation (2.88) predicts the nonlinear
response in the neighborhood of the third harmonic of the excitation's center of
frequency to a good degree for both magnitude and phase. As evident however,
equation (2.88) does not predict the fifth and the seventh order components of the
distortion due to the truncation of equation (2.28) as described in Section 2.2.1.2.
This concludes the validation of equations (2.85), (2.86), (2.87), (2.88), and
(2.89) by simulation.
 49

Predicted by Analysis
Simulated by MWO
X: 3000
-50 Y: -67.51

X: 2880 X: 3120
Y: -100.1 Y: -99.56
-100 X: 3090
Y: -88.79
dBm

-150

-200

-250

2750 2800 2850 2900 2950 3000 3050 3100 3150 3200 3250
MHz

Figure 3.18. Predicted spectral density vs. the

simulated spectral density in the
neighborhood of the third harmonic.

Phase about the Thrid Harmonic

180
160
p1
140
120 3120 MHz
100 138.4043515 Deg
80
60
40
20
0
-20
-40
-60
-80
-100 Ang(Vharm(PORT_2))[*] (Deg)

-120
portmod

-140
-160 p1: FREQ = 1000 MHz 2880 MHz
-99.75231268 Deg
-180
2500 2600 2700 2800 2900 3000 3100 3200 3300 3400 3500
Frequency (MHz)

Figure 3.19. Simulated spectral phase in the

neighborhood of the third harmonic.
 50

Predicted by Analysis
150
Simulated by MWO
X: 3120
Y: 138.4
100

50

Deg
0

-50

-100
X: 2880
Y: -99.75
-150

2700 2800 2900 3000 3100 3200 3300

MHz

Figure 3.20. Predicted spectral phase vs. the

simulated spectral phase in the neighborhood
of the third harmonic.
 51

CHAPTER 4

FURTHER APPLICATIONS

In Chapter 2, it was shown that the closed form equations (2.85), (2.86),
(2.87), (2.88) and (2.89) predict the nonlinear response of the reverse biased PIN
diode RF switch in Figure 1.2 to a narrowband excitation possessing an arbitrary
spectrum in a general manner. This chapter offers a discussion on one possible
application of these closed form equations. The next section discusses a
performance metric used in high frequency work known as the Noise Power Ratio or
NPR and presents a method by which the above equations can be used to predict
the RF switch's NPR.

4.1 NOISE POWER RATIO OF THE RF SWITCH

NPR is a distortion metric that helps to determine a network's maximum
spurious free dynamic range [9]. Figure 4.1 presents the stimulus-response
relationship of this test as it relates to a nonlinear network. In this test, the excitation
consists of a spectrum composed of equally spaced tones at uniform amplitudes
centered at frequency . The excitation's spectral density contains an absence of
spectral power in the vicinity of referred to as a "notch" as shown in Figure 4.1.

NPR Stimulus

notch
Nonlinear
Amplitude
Network

NPR Response
0
distortion
NPR level
Amplitude

Figure 4.1. A representation of a network's stimulus-

response characteristic under an NPR test.
 52

The response contains newly generated spectral content inside the notch due
to the network's distortion characteristics. NPR is the measure of the linear
response's spectral level (outside of the notch) relative to the nonlinear response's
spectral level in the neighborhood of .
Figure 4.2 presents the excitation-response plot of the RF switch' equivalent
circuit in Figure 3.3 to an NPR excitation described in equations (4.1), (4.2), (4.3),
and (4.4). The parameters of the RF switch's equivalent circuit in Chapter 3 were
preserved for this simulation.

Noise Power Ratio

10
0
-25
-50
-75
-100 p1
-125
-150 1038 MHz
962 MHz
-175 -17.64935975 dB -17.64960996 dB
-200 1000 MHz
-81.87270348 dB
-225 DB(|Vharm(PORT_2)|)[*] (dB)
portmod
-250
p1: FREQ = 1000 MHz
-275
-300
900 1000 1100
Frequency (MHz)

Figure 4.2. NPR excitation-response of the RF switch.

(4.1)
∆ 0

Where,
(4.2)
0 , 0.0052
(4.3)
39
(4.4)
2 · 10 / ∆ 2 · 2 · 10 /
The simulated NPR is calculated by subtracting the magnitude of the third
order response from the magnitude of the linear response as shown in equation
(4.5).
 53

(4.5)
17.65 81.87 64.22
The next section presents a method by which NPR of the RF switch can be
predicted using equation (2.86).

4.2 PREDICTION OF THE RF SWITCH'S NPR

Since NPR is measured in the neighborhood of the fundamental frequency of
the excitation, equation (2.86) is used to predict the NPR of the RF switch'
equivalent circuit in Figure 3.3. NPR is calculated by calculating the ratio of the
magnitude of the linear response in the non-zero neighborhood of to the
magnitude of the nonlinear response at .
An examination of equation (2.86) reveals that the linear component of the
voltage response is given as equation (4.6), and the nonlinear 3rd order component
of the voltage response is given as equation (4.7).
(4.6)

1 1 2
2 2 2 2

· (4.7)

The magnitude of the ratio of equation (4.6) and equation (4.7) evaluated at
ω ∆ω and ω respectively expressed in dB equates to NPR as shown in
equation (4.8).

| ∆ (4.8)
20 ·
|
In equations (4.6) and (4.7), if we assume that the linear transfer function
does not vary appreciably with frequency in the neighborhood of , we can
treat it as a constant by pulling all three instances of it out of the convolution
in equation (4.7). Doing so, allows one instance of in the denominator within
equation (4.8) to cancel out with one instance of in the numerator within
equation (4.8); further, one instance of is multiplied with one instance of
in the denominator within equation (4.8).
The simplification above allows equation (4.7) to be described as equations
(4.9), (4.10), (4.11), (4.12), and (4.13).
 54

(4.9)
| | | | | | | |
Where,

1 1 2
2 2 2 2
(4.10)
1 (4.11)

(4.12)
|
(4.13)
0
Equation (4.9) was evaluated in MATLAB with the constants shown below:
(4.14)
0.12

(4.15)
6.444 · 10

(4.16)
8.888 · 10

(4.17)
2 10 /

1 (4.18)
25

0.0052 (4.19)

NPR was calculated by MATLAB as shown in equation (4.20) showing a very good
agreement with the simulated NPR in equation (4.5).
(4.20)
109.03 55.91 46.74 35.66 64.2
 55

CHAPTER 5

CONCLUSION

In this paper, the analysis of distortion in a reverse biased PIN diode switch
was presented relating its nonlinear response to an arbitrarily defined narrowband
excitation at its input.
The expressions presented in this paper formulating the input-output
relationship of such switch predict the amplitudes and phases of the output's spectral
content due to distortion in the neighborhood of the fundamental, the second
harmonic, as well as the third harmonic of the excitation's center of frequency.
The ability of the closed form expressions to predict the switch's distortion at
the output was verified by simulation. The comparison of the simulations against
predictions of the switch's nonlinear response presented a good agreement.
The closed form expressions describing the RF switch's excitation-response
relationship were used to derive a concise formula predicting the switch's NPR
performance. The formula's ability to predict NPR was confirmed by simulation.
 56

REFERENCES

[1] Q. Gu. RF System Design of Transceivers for Wireless Communications.

New York, NY: Springer, 2005, pp. 332-40.

[2] G. T. Zhou. (2002, May). Predicting spectral regrowth of nonlinear power

amplifiers. IEEE Trans. on Commun.. [Print], 50(5).

[3] T. Lee. The Design of CMOS Radio-Frequency Integrated Circuits. New York,
NY: Cambridge University Press, 2004, pp. 390-96

[4] J. C. Pedro. (1999, December). On the use of multitone techniques for

assessing RF components' intermodulation distortion. IEEE Trans. on
Microwave Theory and Techniques [Print]. 47(12).

[5] R. H. Caverly and G. Hiller. (1989). “Distortion in microwave switches by

reverse biased PIN diodes,” presented at IEEE MTT International Microwave
Symposium, Long Beach, CA 1989.

[6] D. D. Weiner and J. F. Spina. Sinusoidal Analysis and Modeling of Weakly

Nonlinear Circuits. New York, NY: Van Nostrand Reinhold, 1980, pp. 1-4

[7] S. A. Mass. Nonlinear Microwave Circuits. New York, NY: IEEE Press, 1997.
pp. 14.

[8] D. M. Pozar. Microwave Engineering. New York, NY: John Wiley & Sons Inc.,
1998, pp. 606.

[9] "Noise power ratio (NPR) measurement using the HP E2507B/ E2508A multi-
format communications signal simulator," in HP Product Note, 1997.