UPTEC F 15013
Examensarbete 30 hp Maj 2015
Performance evaluation of IQ-modulator ADL5375 at 5.8 GHz and its effect on transmitter
performance in a telecommunications system
Alexander Bergslilja
Teknisk- naturvetenskaplig fakultet UTH-enheten
Besöksadress:
Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0
Postadress:
Box 536 751 21 Uppsala
Telefon:
018 – 471 30 03
Telefax:
018 – 471 30 00
Hemsida:
http://www.teknat.uu.se/student
Abstract
Performance evaluation of IQ-modulator ADL5375 at 5.8 GHz and its effect on transmitter performance in a telecommunications system
Alexander Bergslilja
Because of the tough competition in the telecom business there is a constant push for higher capacity and data rates and the companies producing the telecommunications equipment need more cost effective products to stay ahead of competitors. It is therefore interesting to evaluate the
possibilities to use unlicensed
frequency bands at higher frequencies as a complement to the traditional lower frequency bands. This study is focusing on the 5.8 GHz band, which is mainly used for WLAN applications. A key component in most transmitter (TX) designs is is the quadrature
modulator, which upconverts the information signal to desired carrier frequency. In this study an attempt to evaluate the commercially available quadrature modulator ADL5375 at 5.8 GHz. An AWR Visual System Simulator (VSS) model based on measurements of key parameters of ADL5375 is
constructed. An attempt is made to see whether a TX design can pass the specifications set by 3rd Generation Partnership Project (3GPP) for the Long Term Evolution (LTE) standard. To test this an LTE signal source was also constructed. No certain conclusions can be drawn without putting the modulator in a complete (TX) design but the results indicate that it might be possible to use it in a (TX) design for the 5.8 GHz band.
Ämnesgranskare: Uwe Zimmermann Handledare: David Scafe
Popul¨arvetenskaplig sammanfattning
P˚a grund av den h˚arda konkurrensen p˚a telekom-marknaden finns det ett h˚art tryck f¨or att ¨oka capaciteten (antal anv¨andare telekomn¨aten kan hantera) och datahastigheten i dagens n¨at. Detta inneb¨ar att man beh¨over mer bandbredd. De frekvensband som idag ¨ar licensierade f¨or telekommu- nikation b¨orjar bli ¨overpopulerade, vilket g¨or det sv˚art att h¨oja prestandan p˚a n¨aten. Detta g¨or att man i branschen tittar p˚a olika s¨att att ut¨oka frekvensspektrumen som anv¨ands f¨or 3G och LTE. Ett s¨att att ut¨oka ¨ar att anv¨anda det olicensierade spektrumet p˚a 5.8 GHz och upp˚at. Men om frekvensen i systemen ¨okas s˚a sjunker prestandan. Detta st¨aller h¨ogre krav p˚a designen av dessa system och det blir sv˚arare att klara av de designspeci- fikationer som utf¨ardas av standardiseringsorganisationer som t.ex. 3GPP.
Den h¨ar rapporten siktar p˚a att testa en delkomponent i en s¨andarkrets, IQ-modulatorn. I all radiokommunikation s˚a konverteras den analoga sig- nalen som inneh˚aller informationen man vill ¨overf¨ora upp till en b¨arv˚ags- frekvens (fc). Den uppgiften utf¨ors av IQ-modulatorn. IQ-modulatorn best˚ar internt av n˚agon form av icke-linj¨ara elektriska komponenter, som dioder eller transistorer. De icke-linj¨ara egenskaperna anv¨ands f¨or att f˚a den ¨onskade uppkonverteringen till b¨arv˚agsfrekvensen. F¨or att se om en s¨andare kan byggas med kommersiellt tillg¨angliga komponenter s˚a testades i denna studie IQ-modulatorn ADL5375 fr˚an Analog Devices. Den ¨ar gjord f¨or att anv¨andas inom frekvensspannet 400 MHz – 6 GHz.
F¨or att unders¨oka hur ADL5375 presterar p˚a 5.8 GHz och kunna utv¨ardera huruvida den ¨ar l¨amplig i en s¨andadesign beh¨over den testas med bred- bandiga signaler av typen som anv¨ands n¨ar exempelvis LTE-standarden anv¨ands. F¨or att g¨ora detta m¨ojligt utan att designa en hel s¨andare ska- pades ist¨allet en simuleringsmodell av ADL5375 i simuleringsprogrammet AWR – Visual System Simulator. Simuleringsmodellen baserades p˚a m¨at- ningar av vissa nyckelparametrar som ¨ar typiska f¨or modulatorer.
Specifikationerna som s¨atts upp av 3GPP f¨or basstationss¨andare i LTE- system utg˚ar fr˚an en komplett s¨andardesign. F¨or att d˚a kunna utv¨ardera huruvida ADL5375 klarar av kraven som st¨alls beh¨over den s¨attas i en kom- plett s¨andardesign. P˚a s˚a s¨att kan dess effekt p˚a s¨andaren utv¨arderas. F¨or att m¨ojligg¨ora bredbandig signalgenerering s˚a skapades ut¨over en modell av ADL5375 ¨aven en LTE-signalk¨alla.
Resultaten av projektet ¨ar en simuleringsmodell av ADL5375. Ut¨over det skapades ¨aven en LTE-signalk¨alla att anv¨anda f¨or kompletta s¨andar- designsimuleringar. Resultaten av projektet ger inga entydiga svar p˚a huru- vida ADL5375 l¨ampar sig f¨or 5.8GHz-bandet, men under vissa antaganden kan det vara m¨ojligt.
F¨or att kunna dra n˚agra vidare slutsatser beh¨over en mer komplett simuleringsmodell utvecklas som tar h¨ansyn till alla steg i en s¨andardesign.
Foreword
I would like to thank my supervisor, David Scafe, for his support and his never- fading smile. A big thank you goes to the Ericsson TX design department and especially Wojciech Mudyna for giving me the opportunity to work with this project. An extra thank you goes to Erik Vedin, Theodor Berg, Jimmy Andersson and Mathias Augustsson for their patience and willingness to answer all my questions.
I would also like to thank Andreas Tenggren on Keysight for the oppurtunity to borrow measurement equipment.
Finally I would like to thank my wife Eva for her endless support and for always picking me up when I’m down.
Contents
Foreword . . . . ii
1 Introduction 1 1.1 Background . . . . 1
1.2 Purpose . . . . 2
2 Theory 3 2.1 Modelling Transmit Signals . . . . 3
2.2 General Transmitter Design . . . . 4
2.3 Upconversion . . . . 5
2.3.1 Mixer . . . . 5
2.3.2 Upconverting Using a Mixer . . . . 6
2.4 Quadrature Modulator . . . . 7
2.5 Modulator Characteristic parameters . . . . 8
2.5.1 LO Leakage . . . . 8
2.5.2 Sideband Suppression . . . . 9
2.5.3 Conversion Gain . . . . 9
2.5.4 1 dB Compression Point . . . . 9
2.5.5 Second and Third Order Intercept Points . . . . 9
3 Hardware 12 3.1 Evaluation Board . . . . 12
4 Simulation 13 4.1 AWR’s Visual System Simulator . . . . 13
4.1.1 Simulation of ADL5375 in VSS . . . . 13
4.1.2 Wideband Signal Source . . . . 15
4.2 Simulate Wideband Behaviour . . . . 16
5 Measurements 18
5.1 Necessary Measurements . . . . 18
5.2 Measurement equipment . . . . 19
5.2.1 DAC as stimulus . . . . 19
5.2.2 81160A and N5242A PNA-X . . . . 23
5.2.3 Calibration . . . . 25
6 Results and Conclusions 27 6.1 Measurement Results . . . . 27
6.1.1 CW Tone Measurements. . . . 27
6.1.2 IM Measurements . . . . 29
6.1.3 Simulation Model . . . . 30
6.1.4 ACLR Measurements . . . . 31
6.2 Discussion . . . . 34
6.3 Summary of Results . . . . 36
6.4 Future Work . . . . 36
A Appendix 38 A.1 Plots . . . . 38
Glossary
IP2 Second Order Intercept Point. 12, 13, 25, 26, 31, 36 IP3 Third Order Intercept Point. 11–13, 25, 26, 31, 36 3GPP 3rd Generation Partnership Project. 2, 17, 36
ACLR Adjacent Channel Rejection Ratio. 17, 18, 32–34, 36 BB baseband. 5, 6, 8, 19, 20, 22–25, 28, 31, 32, 35, 37 BPF bandpass filter. 5
BS Base Station. 18
CE complex envelope. 4, 14 CW continuous wave. 19, 22, 32
DAC Digital-to-Analog Converter. 5, 6, 20, 21, 37 DDS Direct Digital Synthesis. 8
DPD Digital Pre-distortion. 36
E-UTRA Evolved Universal Terrestrial Radio Access. 17 EVM Error Vector Magnitude. 18
GPIB General Purpose Interface Bus. 23, 25 IF intermediate frequency. 5, 6, 8, 17, 32 IM intermodulation. 12, 26
LO local oscillator. 6–9, 13–16, 19, 24, 26, 27, 35, 36 LPF lowpass filter. 5
LSB lower sideband. 6, 19, 25
LTE Long Term Evolution. 2, 3, 17, 18, 32, 34, 36 NA network analyser. 20, 24
P1dB 1 dB Compression Point. 10, 13, 19, 20, 31 PA power amplifier. 5, 36, 37
PCB Printed Circuit Board. 3, 22
RF radio frequency. 2, 5, 6, 8–10, 13–15, 17, 20, 26
SBS sideband suppression. 9, 28, 29, 31, 32, 38 SSB Single Sideband. 8
TX transmitter. 2, 3, 5, 6, 17, 18, 20, 36, 37
UMTS Universal Mobile Telecommunications Systems. 2 USB upper sideband. 6, 19, 25
VI Virtual Instrument. 23, 24 VSA vector signal analyser. 20
VSS Visual System Simulator. 14–17, 19, 20, 37
WCDMA Wideband Code Division Multiple Access. 3, 17
1 INTRODUCTION
1 Introduction
1.1 Background
In the telecom business there is tough competition. The business constantly pushes for higher capacity and data rates and the companies producing the telecommunications equipment need more cost effective products to stay ahead of competitors. Both Ericsson (Ericsson, 2014) and Qualcomm (QUALCOMM, 2013) project a substantial increase in data traffic over the coming years. At the same time, the commercial spectrum used for Universal Mobile Telecommunica- tions Systems (UMTS) and Long Term Evolution (LTE) ranging between 700 MHz to 2.6 GHz is getting overcrowded as can be seen in Figure 1.
Figure 1: Spectrum allocation between different network operators in Sweden (PTS, 2014).
Interest to use unlicensed spectrum with carrier aggregation as a supplement to increase capacity and data rates can be seen by Qualcomm and T-Mobile among others (Tammy Parker, 2013; Monica Alleven, 2014). 3rd Generation Partnership Project (3GPP) has started work to make specifications on how unlicensed spectrum can be used in LTE. Dino Flore (2015) states that focus will be put on implementing LTE in the 5 GHz band. Specifically, the frequency range above 5.8 GHz is of interest due to the fact that this is in many regions (the US and China) unlicensed spectrum.
Working with higher frequencies puts higher demand on the hardware in transceiver circuitry. With higher frequencies one must overcome performance degradation such as reduced range and linearity. It also makes the radio frequency (RF) design more complex due to shorter wavelengths and transmission line effects.
To cope with these issues one needs to carefully evaluate the performance degra- dation of all components in the communications system chain. One of these components is the mixer or modulator. Most telecommunications systems use complex modulation schemes and must therefore use a quadrature modulator.
Therefore it is of interest for Ericsson’s transmitter (TX) design department to test the modulator stage for transmission at higher frequencies. To make a design commercially usable it needs not only to perform well but also be cost ef- fective. This project will investigate the performance of a commercially available IQ-modulator, ADL5375 from Analog Devices (Analog Devices Inc., 2014). To
1.2 Purpose 1 INTRODUCTION
use said modulator at the 5.8 GHz band would be to push it to its upper limit.
At the same time it is interesting to see if it is possible to make a TX design with ADL5375 to examine the possibility to keep it inside performance requirements that a given communications protocol require. This would require testing of the components in a TX design and fine-tuning link budget parameters. Many aspects of this process are very time consuming, such as going from schematic to a physical Printed Circuit Board (PCB) design, redesign for different scenarios, generating different types of signals etc. An alternative approach that is less time consuming is to make a sufficiently adequate simulation model describing the ADL5375 and insert it into a customizable TX chain in a simulation soft- ware. Analog Devices provide documentation for ADL5375 of key performance parameters up to 5.8 GHz. This documentation and complementary reference measurements gives a good base for constructing a simulation model which in turn can be inserted into a TX chain. This can be used to evaluate performance issues on a system level. It also makes it easy to trim different system parameters and apply scenarios with different broadband signals, such as LTE or Wideband Code Division Multiple Access (WCDMA).
1.2 Purpose
The purpose of this project is to :
make detailed measurements of performance parameters on ADL5375,
design a simulation model for ADL5375,
use the simulation model to evaluate performance of wideband communi- cations system on the 5.8 GHz band and
give suggestions for link budget optimizations for high frequency applica- tions with ADL5375.
2 THEORY
2 Theory
In this section all relevant theory for the project will be presented. Starting in section 2.1 with how signals with complex modulation is modelled and analysed, moving on to describing general transmitter design in section 2.2 and covering the process of upconversion with mixers and modulators in section 2.3.
2.1 Modelling Transmit Signals
As stated by Goldsmith (2005) all physical signals that are transmitted are real signals. This is because all modulators and mixers are built using oscillators and circuitry that generate real sinusoidal signals. When modelling wireless channels mathematically using a complex frequency response, it is done purely for analytical simplicity. The wireless channel is only introducing a phase lag and amplitude change at each frequency, so that the received signal is also a real signal. Modulated and demodulated signals are often represented as the real part of a complex signal. The purpose of this is to make the analysis of signals in channels with complex properties possible. The transmitted signal is modelled as a bandpass signal being modulated by a carrier frequency fc
s(t) = <n
u(t)ej2πfcto
= vI(t)cos(2πfct) − vQ(t)sin(2πfct) (2.1.1) where
u(t) = vI(t) + jvQ(t) (2.1.2) is a complex baseband signal, also called the signals complex envelope (CE).The in-phase component is
< {s(t)} = vI(t)cos(2πfct) − vQ(t)sin(2πfct) (2.1.3) and the quadrature component is
= {s(t)} = vI(t)sin(2πfct) + vQ(t)cos(2πfct). (2.1.4) Equation 2.1.1 is a standard representation for bandpass signals such as the one seen in Figure 2.
Figure 2: Bandpass signal S(f ).
This means that any high frequency bandpass signal s(t) has an equivalent low- pass signal u(t). As a consequence it is easier to work with bandpass signals because it is possible to work with the equivalent lowpass signal instead. This greatly simplifies the process of applying signal processing algorithms because of the much decreased needed sampling rates and thus decreased data rates (Proakis and Salehi, 2008).
The process of going from passband to baseband as presented in equation 2.1.1 is further described in Proakis and Salehi (2008). It can be implemented by a system called a modulator and will be thoroughly explained in section 2.4.
2.2 General Transmitter Design 2 THEORY
2.2 General Transmitter Design
The TX part of a radio communications system is a device that is fed with a signal containing information which it modulates onto a relatively high carrier frequency and then amplified to be broadcast from an antenna. In modern systems, the data is created digitally and then converted to an analog signal called a baseband (BB) signal. BB signals are low frequency signals centred around 0 Hz with bandwidths depending on the type of communications system (Leon W. Couch, 2013). As stated above, no communication is conducted at BB frequencies. To avoid interfering with other communication, the signal is upconverted to a carrier frequency, fc. The upconversion is made by a frequency converting device such as a mixer or a modulator (described in depth in section 2.3). Before the signal is broadcast by an antenna it is generally amplified to a power level that is required by the communication protocol that the TX is designed for.
(a)
(b)
Figure 3: (a) Schematic representation of a a complex intermediate frequency (IF) TX design. After the Digital-to-Analog Converter (DAC) there is a bandpass fil- ter (BPF). Before the final power amplifier (PA) stage there is another BPF. (b) Schematic representation of a zero IF design. After the DAC stage there is a lowpass filter (LPF) stage. After upconversion to the RF there is a BPF and a PA.
Two common TX designs is depicted in Figure 3. These two designs are the ones
2.3 Upconversion 2 THEORY
that are intended to be simulated in this project. In most of today’s systems, the data is produced digitally in the form of quadrature BB signals and converted in a DAC before it can be upconverted to a carrier frequency. There are different types of TX designs, which employ different techniques to get from digital BB signals to a transmitted signal on a carrier frequency. In the design in Figure 3(a) the BB signal is upconverted to the carrier frequency in two steps. First it is upconverted digitally to an IF. After that it is converted to an analog signal by the DAC after which it is upconverted to an RF. In Figure 3(b) the BB signal is converted by the DAC and then directly upconverted to an RF. The difference between the two designs is that in the Zero IF design you can use a lower sampling speed, but you will get the local oscillator (LO) leakage in the middle of your signals bandwidth. If using the Complex IF design you will need a higher sampling frequency but will separate the LO from your signal which can be beneficial (Razavi, 2011).
2.3 Upconversion
In section 2.3.1 the process of upconverting a signal to a carrier frequency by using a mixer will be presented. The mixers role in frequency conversion when inside of a modulator will be explained in section 2.4.
2.3.1 Mixer
As stated by Pozar (2012) a mixer is a three port device that uses a non-linear device for either up or down conversion of a signals frequency. Figure 4 shows a functional diagram of the upconverting process.
Figure 4: Upconversion with a mixer
If the two input signals are
v1(t) = sinω1t and v2(t) = sinω2t (2.3.1) and the mixer is an ideal, lossless mixer, the output will be
y(t) = sin(ω1t)sin(ω2t) = 1
2[sin(ω1− ω2)t + sin(ω1+ ω2)t]. (2.3.2) These two frequencies, ω1− ω2 and ω1 + ω2, are called upper sideband (USB) and lower sideband (LSB) (Pozar, 2012). One of these sidebands are used for communications and the other is filtered out.
In practice, a mixer uses the non-linear properties of either a diode or a transistor to produce the desired mixing products (Pozar, 2012). Therefore the mixer
2.3.2 Upconverting Using a Mixer 2 THEORY
output can be modelled using the junction diode equation
I(V ) = Is(eqV /nkT − 1) (2.3.3) where q is the electron charge, k is the Boltzmann’s constant, T is temperature, n is the ideality factor and Is is the saturation current. These parameters are dependant on the type of diode that is considered.
The output of a physical mixer deviates from that of the ideal mixer due to the non-linearities that are introduced in 2.3.3. If the input voltage is expressed as
V = V0+ v (2.3.4)
where Vo is a DC bias and v is a small AC signal voltage. According to Pozar (2012) 2.3.3 can be expanded in a Taylor series about V0 as
I(V ) = I0+ vdI dV V0
+1 2v2d2I
dV2 V0
+ .... (2.3.5)
This derives to
I(V ) = I0+ vGd+v2
2 G0d+ ... (2.3.6) where Gd is the dynamic conductance of the diode. This is called the Small signal approximation of the diode and is useful when analysing the diode mixer.
2.3.2 Upconverting Using a Mixer Given the voltage signals
vIF = VIF cos ωIFt (2.3.7) and
vLO = VLOcos ωLOt (2.3.8)
as the inputs of a diode mixer and using the small signal approximation in equation 2.3.6, the total mixer current is
i(t) = I0+ Gd(vLO+ vIF) +G0d
2 (vLO+ vIF)2+ .... (2.3.9) The DC term I0is easily blocked and the second term is a replication of the input signals where the vIF term can easily be filtered out. vLO is an in-band signal and is difficult to filter out. It is one of the contributions to the modulator’s LO leakage and will be further explained in section 2.5. The third term can be rewritten using trigonometrical identities to
i(t)quadratic = G0d
2 (VIF cos ωIFt + VLOcos ωLOt)2
= G0d
4 [VIF2 (1 + cos 2ωIFt) + 2VLOVIFcosωLOcosωIF + VLO2 (1 + cos2ωLOt)
= 1
4[2VIFVLOcos(ωLO− ωIF)t + 2VIFVLOcos(ωLO+ ωIF)t]
+ VLO2 (1 + cos 2ωLO) + VIF2 (1 + cos 2ωIF)]. (2.3.10)
2.4 Quadrature Modulator 2 THEORY
This yields several signal components that are unwanted. The DC terms are again blocked and the 2ωIF and 2ωLO are filtered out (given that they are sufficiently far away in frequency to ωLO). This leaves the last two terms,
vRF1(t) = 1
2VIFVLOcos(ωLO− ωIF)t (2.3.11) and
vRF2(t) = 1
2VIFVLOcos(ωLO+ ωIF)t (2.3.12) which are consistent with (2.3.2). One of these two frequencies are used as the carrier frequency, fC. **
2.4 Quadrature Modulator
Not only is the process of modulating the BB signals made in the digital domain in high-end communications systems, but also the act of upconverting it to an IF fIF, by using Direct Digital Synthesis (DDS) (Cushing, 2000). The final upconversion to an RF is done in the analog domain. It is then possible to use a similar method as when implementing a Single Sideband (SSB) mixer (Pozar, 2012) to suppress one of the sidebands. this enables more spectrum effective transmissions. By using a quadrature modulator, such as the one schematically represented in Figure 5 it is possible to obtain similar suppression. In the same way as when implementing a SSB mixer, both the two IF signals and the LO signals fed to their respective mixer core need to be 90◦out of phase to each other if maximum suppression is to occur. When the signal is digitally upconverted to an IF and then converted to an analog signal, the two signals are on the form of 2.1.3 and 2.1.4 and thus 90◦ out of phase. The two signals can be fed into a quadrature modulator as the one in Figure 5.
Figure 5: Schematic representation of quadrature modulator (Agilent Technologies, 2007).
If similar assumptions are made about the low frequency mixing products as in section 2.3 it is sufficient to calculate the quadratic term in equation 2.3.9.
When rewritten using the same trigonometric identities for the two inputs gives
2.5 Modulator Characteristic parameters 2 THEORY
in-phase component iin−phase(t) = G0d
2 h<n
u(t)ej2πfIFto
+ vLO(t)i2
= G0d
2 − 2VLOVQsin(ωIF − ωLO)t − 2VLOVQsin(ωIF + ωLO)t + 2VLOVIcos(ωIF − ωLO)t + 2VLOVIcos(ωIF + ωLO)t
− VQ2cos 2ωIFt − 2VQVIsin 2ωIFt + VI2cos 2ωIFt + VLO2 cos 2ωLOt + VLO2 + VQ2+ VI2
(2.4.1) and quadrature component
iquadrature(t) = G0d 2
h
=n
u(t)ej2πfIFto
+ vLO(t)i2
= G0d
2 2VLOVQsin(ωIF − ωLO)t − 2VLOVQsin(ωIF + ωLO)t
− 2VLOVIcos(ωIF − ωLO)t + 2VLOVIcos(ωIF + ωLO)t
+ VQ2cos 2ωIFt + 2VQVIsin 2ωIFt − VI2cos 2ωIFt − VLO2 cos 2ωLOt + VLO2 + VQ2+ VI2
(2.4.2) as quadrature component. If the two are summed together as in Figure 5 the output of the modulator is
isum(t) = G0d2VLOVIcos(ωIF + ωLO)t − 2VLOVQsin(ωIF + ωLO)t
+ VLO2 + VQ2+ VI2. (2.4.3)
As can be seen, only one of the two sidebands is left in the output signal, ωIF + ωLO, but all information is preserved.
Described above is the theoretical operation of an ideal quadrature modulator.
There is however a number of circumstances that degrade the performance of the modulator. To make the performance of the modulator measurable there are a number of important parameters that characterizes it. These will be presented in section 2.5.
2.5 Modulator Characteristic parameters
Contrary to the situation in above sections, where a schematic modulator is considered, there are imperfections internally that create asymmetries and non- idealities. This will affect the output of the modulator (Nash, 2009).
2.5.1 LO Leakage
LO leakage, as described by Razavi (2011), refers to the situation when the LO signal is somehow leaking to the output of the mixer. It can be caused by the non-linear properties of the mixer, as seen in equation 2.3.6, where a replication of the input signal is produced. It can also be caused by device capacitances between the LO and RF port or by emissions from the substrate to the output pad. Both of the latter causes what is referred to as self-mixing (Nash, 2009). It can be caused by the finite isolation either between the LO port and one or both of the I and Q ports of the modulator or between the LO port and the actual RF output. It can also occur when a DC offset is present on the input of the mixer.
2.5.2 Sideband Suppression 2 THEORY
2.5.2 Sideband Suppression
The sideband suppression (SBS) of a quadrature modulator is defined as SBS = PDSB
PSSB (2.5.1)
which is the ratio between the desired sideband, PDSB, and the suppressed sideband, PSSB. Consider the event where the gain in the I channel is greater then that of the Q channel. This could be caused by any number of reasons such as input mismatch, internal I/Q gain imbalance etc., but the effect would be the same. It would cause degradation of the sideband suppression of the modulator.
2.5.3 Conversion Gain
The conversion gain specifies the power transfer of the modulator. A modulator is generally a lossy component, but can have a small gain in certain designs. The most general definition of gain is
Gconv= Pout/Pin (2.5.2)
where Pout is the output signal power and Pin is the input signal power. There is no convention for the definition of Pout when talking about modulators. Most logical is to define it as the power confined inside the desired sideband of the output signal, since this is the only part of the output that is interesting. Simi- larly there is no convention for Pin either. You could either define it as the sum of the I and Q input powers, or just one of the two channels power.
The definition that is used in this report for all conversion gain calculations is that Pin is the power of one of the two input channels and Pout is the power of the desired sideband. It has been established that this is the same definition used in the datasheet for ADL5375 by reverse engineering the values for conversion gain and input voltage levels presented in Analog Devices Inc. (2014).
2.5.4 1 dB Compression Point
Many RF circuits have non-linear properties and thus exhibits a compressive behaviour when they are driven to hard. The power level of the output signal is compressed which can create unwanted distortion. It is therefore of interest to quantify how severe the compression of a component is. As described by Pozar (2012) and Razavi (2011) it is common to present the point at which the output power of a device (e.g. amplifier or mixer) has compressed to a point 1 dB below the ideal linear behaviour, the 1 dB Compression Point (P1dB). It can be specified either as the input power (IP1dB) or the output power (OP1dB), usually the one giving the highest value. It is calculated by extrapolating the linear response as depicted in Figure 6.
2.5.5 Second and Third Order Intercept Points
As discussed previously, the modulator is a device that takes advantage of non- linear properties of either a diode or a transistor to produce certain frequency shifted output signals. It is clear in section 2.3 that there are several unwanted
2.5.5 Second and Third Order Intercept Points 2 THEORY
signal components at other frequencies than the wanted. To be able to quantize the effects of said non-linearities the concepts of second and third order inter- cept points are widely used (Razavi, 2011; Proakis and Salehi, 2008). If using equations (2.3.6) – (2.3.9) you can measure the intercept points. To do this the inputs needs to be changed to two signals with equal amplitude according to
v1(t) = V cos ω1t (2.5.3)
and
v2(t) = V cos ω2t (2.5.4)
with frequencies ω1and ω2 close to each other. Simplifying the non-linear model in equation 2.3.6 to
i(t) = I0+ α1v + α2v2+ α3v3. (2.5.5) If v is substituted to v(t) + v2(t) we get
i(t) = I0+ α1(V cos ω1t + V cos ω2t) + α2(V cos ω1t + V cos ω2t)2
+ α3(V cos ω1t + V cos ω2t)3. (2.5.6)
Discarding everything else but the third order terms we get, after simplification i(t) = α3V3 3
4cos ω1t + 1
4cos 3ω1t
+ α3V3 3
4cos ω2t + 1
4cos 3ω2t
+ α3V3 3
2cos ω2t + 3
4cos(2ω1− ω2)t +3
4cos(2ω1+ ω2)t
+ α3V3 3
2cos ω1t + 3
4cos(2ω2− ω1)t +3
4cos(2ω2+ ω1)t
(2.5.7) The interesting terms are 2ω1− ω2 and 2ω2− ω1. If ω1 and ω2 are close to each other then 2ω1− ω2 and 2ω2− ω1 will be in the vicinity of these frequencies as well. This causes a distortion of the signal of interest in the situation of a broadband signal that is impossible to filter out without distorting the signal of interest. It is easily seen, when looking at equations 2.5.6 and 2.5.7, that as the input voltage V increases the third order terms increase as V3. This means that for small voltages the third order terms will be very small but will increase rapidly but will increase rapidly as the input voltage increases. In Figure 6 we can see that at one point the power of the linear terms will intersect the power of the third order terms. This is called the Third Order Intercept Point (IP3).
It is also apparent from Figure 6 that this point is a strictly theoretical point, because it occurs well above the 1 dB compression point.
2.5.5 Second and Third Order Intercept Points 2 THEORY
Figure 6: A diagram of the third order intercept point and the 1 dB compression point in log-log scale (Pozar, 2012).
Looking at the second order terms in equation 2.4.1 it is clear that one can define a Second Order Intercept Point (IP2) in much the same manner. The increase of amplitude of the second order terms is as the square of the input voltage V . To calculate these points extrapolation of a fundamental tone and both second and third order intermodulation tones is needed. The point where the extrapolated plot of the intermodulation (IM) products intersects the extrapolated plot of the fundamental tone is the IP2 and IP3. The process of extrapolating demands that long measurement series are performed, which can be very time-consuming or not possible for other measurement technical reasons. It is then possible to estimate the IIP3. Starting with the scenario that the input power is at the level of PIIP3. If the input power then is lowered to an arbitrary level Pin1, it will then have lowered with 10logPIIP3 − 10logPin. On a log-log scale the IM3 product power will fall with a slope of 3 and the fundamental will fall with a slope of 1.
This means that the difference between the two plots increase with a slope of 2.
This can be concluded by
10logPf − 10logPIM3 = 2(10logPIIP3 − 10logPin) = ∆P (2.5.8) where PIM3 is the power of the strongest IM3 product and Pf is the power of the fundamental tone. This gives
10logPIIP3 = ∆P
2 + 10logPin (2.5.9)
and more generally
10logPIIPX = ∆P
x − 1+ 10logPin. (2.5.10) Since there can be dynamic non-linearities, this is only an estimate and not as accurate as the method of extrapolation, but can be used as an approximation (Razavi, 2011).
3 HARDWARE
3 Hardware
3.1 Evaluation Board
The center of this project is the ADL5375-05 (Analog Devices Inc., 2014). It is a broadband quadrature modulator, designed to work in the frequency range of 400 MHz to 6 GHz. The I and Q baseband input ports are both differential and the LO input port is single-ended. The output RF port is single-ended and matched to 50 Ω. The baseband input needs a 500 mV biasing for optimal performance. Some of the key parameters of the modulator can be seen in table 1. All tests and measurements were made on the evaluation board (rev. B Table 1: Performance Parameters for ADL5375@5.8 GHz (Analog Devices Inc., 2014).
Parameter Value Unit
Modulator Voltage Gain -5.3 dB
Output P1dB 4.9 dBm
Output IP2 39.1 dBm
Output IP3 14.6 dBm
Noise Floor -153.0 dBm/Hz
LO Leakage -19.5 dBm
Sideband Suppression -3.2 dBc
(Analog Devices Inc., 2014)) supplied from the manufacturer of ADL5375. The board has two possible RF output paths; one path with a RF drive amplifier (ADL5320 (Analog Devices Inc., 2013)) and another with a direct path from the modulator RF output. All I/O connections are of the type SMA 3.5 mm.
4 SIMULATION
4 Simulation
To be able to make a more detailed performance analysis of the modulator and see how it performs when using a broadband communication standard in a trans- mitter design it is convenient to develop a simulation model of the modulator.
This gives the possibility to evaluate the board in a complex and more in-depth manner, without the need of the time consuming process of designing hardware, verifying the design etc. It can be very useful in the pre-development stage when a new product for the intended spectrum is to be investigated. In the next sec- tions the simulation software suite used will be presented, along with how the modulator simulation model was constructed.
4.1 AWR’s Visual System Simulator
AWR’s Visual Systems Simulator is a software suite which is directed towards system design of todays modern communications systems (National Instruments, 2015a). It is suitable for the purpose of this project, because it comes with a comprehensive library of RF component models (system blocks) which ranges from analog devices such as mixers and amplifiers to logical operators and signal processing blocks. A complete and more detailed descriptionis given in the Visual System Simulator (VSS) Getting Started Guide (National Instruments, 2014b).
VSS uses the CE seen in equation 2.1.2 as representation of a signal whenever it is possible. The reason for this is that even though a signal is modulated around a high carrier frequency, the actual data is represented by the lowpass CE. This is sufficiently modelled with a sampling frequency orders of magnitude lower than is needed for proper sampling of the upconverted signal. Therefore VSS propagates a center frequency through the simulation which the CE is centred around. This reduces the sampling frequency and thus the simulation time considerably (National Instruments, 2014b).
4.1.1 Simulation of ADL5375 in VSS
Properties that a simulation model should adequately simulate are:
Non-linearities
Conversion gain
Gain compression
I/Q imbalance
These properties should preferably be simulated over a broad frequency range.
To model the internal structure of ADL5375 properly the block diagram depicted in Figure 7(a) was used as a starting point. It consists of drivers for the two baseband inputs, a quadrature phase splitter for the LO input, two mixer cores for upconversion and an RF combiner to sum the I and Q component. To model the mixer cores the system block MIXER B2 was used. It implements a behavioural model of a non-linear double-balanced mixer with an equivalent circuit as in Figure 7(a).
4.1.1 Simulation of ADL5375 in VSS 4 SIMULATION
(a) (b)
Figure 7: (a) Block diagram of ADL5375 internal design (Analog Devices Inc., 2014).
(b) VSS system block MIXER B2 equivalent diagram (National Instruments, 2015b).
MIXER B2 uses a set of user specified parameters to model a mixer. This makes it suitable for creating a model based on measurement data. It is also possible to specify a vector of frequency values to model a wideband frequency dependant behaviour. The frequency dependency is valid for the baseband frequency range.
Certain key parameters were specifically chosen to characterize the modulator (table 2), all of which can use the frequency dependency setting, except for the Noise Figure parameter.
Table 2: Parameters used to model ADL5375 in MIXER B2.
Name Description
GCONV Conversion gain of the mixer.
P1DB The 1 dB compression point of the mixer.
IP3 3rd order intercept point of the mixer.
IP2 2nd order intercept point.
LO2OUT The LO leakage between the LO input and the RF output.
NF Noise Figure of the mixer.
FREQS Vector for frequency dependent settings.
Apart from the parameters in Table 2 it is also possible to model temperature dependencies, impedance mismatch on input and output, etc. To conserve time these aspects were chosen not to be implemented in this study. The non-linear amplifier in Figure 7(b) is similiar to another system block, AMP B2 (National Instruments, 2015b). Since measurements of the mixer cores is not done individ- ually it makes more sense to implement the non-linearities on the whole mod- ulator by using AMP B2. The LO generation along with the quadrature phase splitter was realized with the combination of system blocks TONE, SPLITTER and PHASE in the manner seen in Figure 8.
4.1.2 Wideband Signal Source 4 SIMULATION
Figure 8: LO signal generation and quadrature phase splitter implementation in VSS.
TONE generates a sinusoidal tone at a specified frequency. SPLITTER splits the incoming signal power according to
Vout= Vin
√2. (4.1.1)
After that the PHASE block shifts the phase in one of the LO paths with 90◦. Here it is also possible to insert a phase error to model a sideband that is not perfectly suppressed. The upconverted I and Q signals are then summed together with an ADD block, which sums together the inputs and outputs the result. The whole implementation can be seen in Figure 9(a).
(a) (b)
Figure 9: (a) VSS realization of ADL5375. (b) ADL5375 implemented as a sub- circuit element.
For ease of use the modulator model was implemented as a sub-circuit element as in Figure 9(b) to make the model more compact.
4.1.2 Wideband Signal Source
To make a performance analysis of ADL5375 and put it in a communications system perspective one needs to feed it with a wideband signal that corresponds
4.2 Simulate Wideband Behaviour 4 SIMULATION
to a communications standard used in telecommunications system, such as LTE or WCDMA. Then it is possible to make measurements that is comparable with the limits that is specified in standard specifications set by 3GPP (European Telecommunications Standards Institute, 2015). For this to be possible, it is necessary to be able to feed the modulator model with a wideband LTE signal.
there is a system block in VSS, LTE DL TSIG, which is a LTE downlink signal source (National Instruments, 2014a). It is constructed to comply with Euro- pean Telecommunications Standards Institute (2008), and is therefore of interest to use to create test signals for measurements. To create a test signal that cor- responds to the signal components presented in equation 2.1.3 and 2.1.4 in the digital domain and enable for upconversion to an IF, one needs to perform the same multiplication with a complex exponential. This is done by first splitting the complex baseband signal into two real data streams and then producing the signal components by using ideal mixers to convert the signals to an IF as in Figure10(a).
(a) (b)
Figure 10: (a) Creation of quadrature and in-phase component in AWR. (b) Inter- nals of LTE signal generation sub-circuit.
This was then inserted as a sub-circuit block inside the LTE sub-circuit block (Figure 10(b)). Using this method to create a baseband envelope at an IF requires that you sample at a higher sampling speed, to make sure that the Nyquist Criteria is met(Goldsmith, 2005).
4.2 Simulate Wideband Behaviour
To evaluate the impact of using ADL5375 in the 5.8 GHz band it is necessary to put it in a TX design perspective. When design of a TX is considered you need to adhere to the limits specified for the region that is intended for the product.
All specification limits in this report will be taken from European Telecommu- nications Standards Institute (2015), where all minimum RF characteristics and minimum performance requirements of Evolved Universal Terrestrial Radio Ac- cess (E-UTRA) base station in Europe are specified, which is the air interface for the LTE standard. Some of the parameters for base station transmitters that is regulated in European Telecommunications Standards Institute (2015) are presented in Table 3.
4.2 Simulate Wideband Behaviour 4 SIMULATION
Base Station (BS)
Property Config Limit
Rated Output Power
Wide Area BS Medium Range BS Local Area BS
Home BS (One TX Antenna) -
< +38 dBm
< +24 dBm
< +20 dBm Error Vector Magnitude
(EVM) Requirements
QPSK 16QAM 64QAM 256QAM
17.5 % 12.5 % 8.0 % 3.5 % Table 3: Some of the requirements for LTE Downlink base station found in European Telecommunications Standards Institute (2015).
There are several TX specific limits specified by 3GPP. The ones that are the most interesting when examining the modulator is the Adjacent Channel Rejec- tion Ratio (ACLR) and Rated Output Power. The ACLR is the ratio between the transmitted power in one channel to the power in an adjacent channel. This adjacent channel is usually defined as an identical channel with equal bandwidth as the source channel, but can be specified differently. In European Telecommu- nications Standards Institute (2015) there are specifications for many different source and adjacent channel combinations depending on what type of spectrum that is measured on (unpaired/paired, contiguous/non-contiguous). An example of this is presented in Figure 11.
Figure 11: Base station ACLR in paired spectrum (European Telecommunications Standards Institute, 2015).
5 MEASUREMENTS
5 Measurements
In order to characterize the performance of ADL5375 a number of key parameters needed to be measured. The parameters measured for characterization were chosen to conform to the measurement diagrams that are available in Analog Devices Inc. (2014) to make a comparison possible. In the coming sections the equipment used for measuring and the measurements that were performed will be presented.
5.1 Necessary Measurements
The measurements were chosen to comply to the simulation model and to be comparable to the ADL5375 datasheet (Analog Devices Inc., 2014). Measure- ments were performed with two types of stimulus; continuous wave (CW) and two-tone signals. The former type is used when measuring the intermodulation properties of the modulator, while CW stimulus is used for all other measure- ments. In Table 4 the parameters that can be extracted when using CW tones as stimulus are shown.
Swept Parameter
Measured Parameter
Performance Parameter
VSS Parameter
BB Frequency vs.
USB Power LSB Power LO leakage LO Gain
Sideband Suppression Conversion Gain LO leakage
GCONV LO2OUT FREQS
BB Power vs.
USB Power LSB Power LO leakage LO Gain
Sideband Suppression Conversion Gain LO leakage
P1DB(scalar)
LO Frequency vs.
USB Power LSB Power LO leakage LO Gain
Sideband Suppression Conversion Gain LO leakage
LO Power vs.
USB Power LSB Power LO leakage LO Gain
Sideband Suppression Conversion Gain LO leakage
BB Phase vs.
USB Power LSB Power LO leakage LO Gain
Sideband Suppression Conversion Gain LO leakage
Table 4: Acquired VSS specific parameters and general parameters when measuring using CW as stimulus.
The measurement data gathered when sweeping BB frequency with a CW input gives frequency dependent values to the parameters GCONV, LO2OUT and FREQS and can easily be imported into VSS and used in the simulation model.
Sweeping BB input power gives a value of P1dB by using the method presented in section 2.5.4. This only generates a scalar value for P1dB and is thus only valid
