Professional Documents
Culture Documents
ZTE COMMUNICATIONS
ZTE COMMUNICATIONS
Special Topics:
Advances in Digital Front-End and Software RF Processing: Part Ⅱ
Advances in Mobile Data Communications
VOLUME 9 NUMBER 4 DECEMBER 2011
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P1
Frost & Sullivan Recognizes ZTE as 2011
LTE Vendor of the Year
United States. ZTE is one of the world’ s most competitive
LTE solutions providers.
ZTE’ s influence in the global LTE market has steadily
grown. In first half 2011, the number of new LTE contracts
exceeded the total number of LTE contracts for the whole
of 2010. ZTE has won 28 contracts for LTE commercial
application and has deployed test networks in
conjunction with more than 90 operators worldwide. To
date, more than 100,000 ZTE LTE terminal units have
been ordered worldwide. The terminals, which are
purchased by high-end operators, contribute to the
commercialization of LTE technology.
“ZTE is honored to be recognized by Frost & Sullivan
as a leader in the LTE field,”said Mr. Richard Lihe Ye, the
ZTE Corporation announced on November 10, 2011 senior director of wireless product operation for ZTE
that it has been named LTE Vendor of the Year by leading Corporation.“The award illustrates the company’ s
consulting firm Frost & Sullivan. commitment to the development and commercialization
The Frost & Sullivan LTE Vendor of the Year award is of next-generation technology.”
given to companies that have achieved excellence in ZTE is a leader in developing LTE technologies. To
LTE. ZTE won the award for its growing number of date, ZTE has applied to ETSI for 235 essential patents
secured contracts and also for its patent applications. for LTE standards, accounting for approximately
The company’ s industry leading equipment has entered 7 per cent of the total number of EP applications globally.
high-end markets such as Europe, Japan, and the (ZTE Corporation)
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P2
ZTE Establishes Global Consumer
Brand Ambitions in 2011
ZTE Corporation became the fourth largest handset
vendor by market share in 2011 and set out its ambitions
for further growth with a series of high-profile celebrity
sponsorship deals.
According to a Q3 report by ABI research, IDC, and
Strategy Analytics, ZTE overtook Apple to become the
fourth-largest handset manufacturer in the world.
According to the report, ZTE shipped 19.1 million
terminals in Q3 2011, a 57.9 percent increase year on
year. ZTE’ s market share also increased to 4.9 percent,
which closed the gap between ZTE and 3rd place, LG
Electronics. In the first three quarters of 2011, ZTE’ s
revenue from terminal sales increased 53.4 percent year
on year.
ZTE is not resting on its laurels and is looking forward to
2012. Backed by strong financial and business
performance, ZTE will invest more in marketing initiatives
next year to bring its brand recognition in line with its SIM-free Skate terminals, a ticket give-away supported
global market share. by local media, and an integrated Facebook campaign
In April, ZTE announced its smart terminal strategy, on facebook.com/ZTEUK. Professor Green fans had the
part of which included partnering with the Singaporean opportunity to win free tickets to an exclusive London gig
singer/songwriter, Huang Yida, to create the new brand and some very lucky fans to meet the performer
song“Light Your Smart World,”which is also ZTE’ s face-to-face.
smart terminal slogan. Also in October, ZTE announced the exclusive
In May, ZTE unveiled the new Light Pro tablet at the sponsorship of Spanish artist, David Bisbal’ s, tour
Mutua Madrid Open, an ATP World Tour Masters 1000 ‘Acustico’ . The tour started on November 1 with a
event, with support from Spanish PR Agency, special concert in the Teatro Real in Madrid. This
Perception & Image. The device included an application sponsorship package is part of ZTE’ s marketing strategy
specifically designed for the tournament. The application for the launch of ZTE Skate.
provided statistics, player information, and match ZTE also secured exclusive sponsorship of all the
schedules in real time. The world’ s top tennis player, handsets in China’ s top-rated TV series,“Man’ s Gang.”
Carlos Moyà Llompart, was also invited to experience ZTE presents its full series of smartphones and tablets in
ZTE’s smart terminal products. the TV series, strengthening ZTE’ s smart terminal brand
In September, ZTE launched its flagship smartphone, with fashion-conscious youth.
Skate, during Cibeles Madrid Fashion Week. Spanish “ZTE has been involved in a wide range of celebrity
supermodel, Jon Kortajarena, presented ZTE products at and entertainment marketing campaigns over the past 12
the event, further strengthening the company’ s brand. In months, marking 2011 as a watershed year in the
Hong Kong, the Skate launch was endorsed by local pop company’ s transition from white label ODM to fully
star, Kandy Wong with support from Edelman Hong Kong. fledged own-brand player,”said He Shiyou, head of
In October, ZTE, in conjunction with PR agency, ZTE’ s mobile-terminals business.“Next year looks set to
AxiCom, announced an exclusive deal with EMI be even bigger and brighter, as we continue on our
Music/Virgin Records to sponsor leading UK rapper, roadmap to becoming a top-three handset player by the
Professor Green, on an 18 day UK tour. The campaign year 2015.” (ZTE Corporation)
includes an ongoing Twitter competition @ZTE_UK for
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P3
Contents
www.zte.com.cn/magazine/English ISSN 1673-5188
CODEN ZCTOAK
ZTE COMMUNICATIONS
Http://www.zte.com.cn/magazine/English December 2011, Vol.9 No.4
Staff Members:
32 An Antenna Diversity Scheme for Digital Front-End with OFDM Technology
Editor-in-chief: Xie Daxiong
35 A Histogram-Based Static-Error Correction Technique for Flash ADCs
Deputy Editor-in-chief: Deng Xin
Executive Deputy
Editor-in-chief: Huang Xinming
Editor in Charge: Zhu Li
Editors: Paul Sleswick, Yang Qinyi, Xu Ye, Special Topic: Advances in Mobile Data Communications
Lu Dan 42 Guest Editorial
Producer: Yu Gang
Circulation Executive: Wang Pingping 43 Enhanced Cell-Edge Performance with Transmit Power-Shaping and
Assistant: Wang Kun Multipoint, Multiflow Techniques
Editorial Correspondence:
Add: 12/F Kaixuan Building, 49 Spatial Load Balancing in Wide-Area Wireless Networks
329 Jinzhai Road,
55 Uplink Power Control for MIMO-OFDMA Cellular Systems
HeFei 230061, P. R. China
Tel: +86-551-5533356
63 Mobile Backhaul Solutions
Fax: +86-551-5850139
Email: magazine@zte.com.cn
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P4
Guest Editorial
Jun Fang Fa-Long Luo Mikko Valkama Serioja Ovidiu Tatu Tomohisa Wada
I
n the first editorial of this two-part special issue, we special issue, which is organized into two parts and has been
pointed out that one of the biggest trends in wireless published in two consecutive issues in 2011.
broadband, radar, sonar, and broadcasting technology is The contents of part I is divided into two groups. The first
software RF processing and digital front-end [1]. This group consists of four papers, addressing different aspects of
trend encompasses signal processing algorithms and power amplification (PA) technologies. The second group,
integrated circuit design and includes digital pre-distortion comprising two papers, is devoted to another interesting
(DPD), conversions between digital and analog signals, digital topic: architecture design and test of millimeter wave (60 GHz)
up-conversion (DUC), digital down-conversion (DDC), DC wideband radio transceivers, which are mainly based on a
offset, and I/Q imbalance calibration. Other important related six-port device and related technology. For more details on
topics are peak-to-average power ratio (PAPR) reduction, this part, refer to [1].
crest-factor reduction (CFR), pulse shaping, image rejection, Six papers have been selected for publication in part II, and
digital mixing, delay/gain imbalance compensation, error these papers are organized again into two groups. The first
correction, noise shaping, numerical-controlled oscillator group consists of three papers that focus on theory, design,
(NCO), and various diversity methods. Digital techniques for and implementation of resampling, that is, up-conversion and
RF processing have many advantages over traditional down-conversion. The second group consists of three papers
techniques in terms of power efficiency, cost and area that describe non-uniform channelization of multiplexed
reduction, flexibility, programmability, and reconfigurability. frequency bands, selective subcarrier degradation-based
Digital RF processing is now regarded as key to enabling crest-factor reduction, and an eigenfiltering-based antenna
software-defined radio (SDR) that supports multistandard, diversity scheme.
multimode applications and fast time-to-market The first paper by Mehmood Awan, Yannick Le Moullec,
solutions [2], [3]. Peter Koch, and Fred Harris describes efficient processing
This special issue aims to stimulate and guide the engines in the digital radio front-end. These engines, based
development of new and improved systems for wireless on a polyphase channelizer, perform arbitrary sample-rate
communications and digital broadcasting. It is a timely and changes, frequency selection, and bandwidth control. The
high-quality forum for scientists, engineers, technologists, polyphase filter bank with five different resampling modes is
broadcasters, manufacturers, software developers, and other used as a case study for embedded resampling in digital
professionals to engage in discussion. In this special issue of front-ends. These modes are (i) maximally decimated, (ii)
ZTE Communications published in English, we wish to show under-decimated, (iii) over-decimated, and combined
how the theory is translated into practical technology for all the up- and down-sampling with (iv) single stride length and (v)
relevant standards. Readers are given ideal design multiple stride lengths. These modes can be used to obtain
methodologies to manage a rapidly increasing range of any required rational sampling rate change in an SDR
applications. front-end based on a polyphase channelizer. They can also
Step-by-step information for designing practical systems be used for translation to and from arbitrary center
is included, and theory, principles, algorithms, standards, and frequencies that are unrelated to the output sample rates. In a
implementation are all given comprehensive treatment. future issue, the authors will report on their work analyzing
The call-for-papers for this special issue attracted a good hardware architecture of polyphase channelization.
number of excellent submissions. After two-round reviews, The paper by Yue Zhang, Li-Ke Huang, Carsten Maple and
twelve papers have been selected for publication in this Qing Xuan gives an overview of the concepts, theory, and
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P5
Guest Editorial
design principles of digital down-conversion and Jacob Wikner, Sayed Masoud Sayedi and Rasoul Dehghani
up-conversion for SDR-based communication and describes a static calibration technique for flash
broadcasting systems. After an introduction to essential analog-to-digital converters (ADC). The calibration
concepts and principles, the paper discusses design issues technique is based on histogram test methods, and estimation
in down-conversion, and super-heterodyne and of the equivalent errors in the flash ADC comparators is done
direct-conversion architectures are presented. Details of the in the digital domain, without any significant changes being
SDR design for up-conversion are described in this paper, made to the ADC comparators. In the trimming process, the
and trade-offs in the design stages for filters, mixers, NCO, reference voltages are adjusted to compensate for static
DAC, and signal processing are discussed in detail. errors. Behavioral-level simulations of a moderate-resolution
In the third paper of the first group, Álvaro Palomo Navarro, 8-bit flash ADC show that, for typical errors, ADC
Rudi Villing, and Ronan Farrell deal with the practical performance is considerably improved by the proposed
implementation of two non-uniform channelization techniques technique. Implementation of the proposed technique will be
for multiplexed frequency bands. These techniques are based discussed in a future issue.
on generalized DFT filter banks (GDFT-FB), which are a major We would like to thank again all authors for their valuable
processing task in the receivers of an SDR system. Filter contributions. We also express our sincere gratitude to all the
bank-based schemes are efficient and flexible; however, the reviewers for their timely and insightful comments on all
use of FIR filters generally leads to high filter orders, and this submitted papers. It is hoped the content of this two-part
is impractical. To overcome this problem, a multistage filtering special issue is informative and useful from various
structure is applied to the GDFT-FB to reduce the number of technological and implementation perspectives.
coefficients. In this approach, the number of filter coefficients
and operations per input sample is less than that in the
References
single-stage approach. These reductions make fixed-point [1] J. Fang, Fa-Long Luo, M. Valkama, S. O. Tatu, and T. Wada,“Advances in digital
implementation using existing FPGA platforms more practical. front-end and software RF processing: Part I,”ZTE Communications, vol. 9, no.3,
In the second group of papers, the first paper by R. Neil pp.1-3, 2011.
[2] Fadhel M. Ghannouchi,“Power amplifier and transmitter architecture for software
Braithwaite deals with crest-factor reduction. Selected data defined radio systems,”IEEE Circ. Syst. Mag., vol. 10, no.4, pp.56-63, 2010.
subcarriers are modified within an OFDM signal to reduce [3] Fa-Long Luo, Digital Front-End in Wireless Communications and Broadcasting:
Circuits and Signal Processing. Cambridge: Cambridge University Press, 2011.
peaks in the time domain, and pilot and null subcarriers
remain unchanged. In this approach, a set of peaks within an
OFDM symbol interval are identified, and data subcarriers
Biographies
whose data element has a positive or negative correlation to
the peak set are selected. For a subcarrier with an outer Jun Fang graduated from Shanghai Jiao Tong University in 1982 and
element and significant positive correlation, a bit error received his Ph.D. from Ecole Nationale Supérieure des
Télécommunications de Paris (ENST Paris) in 1987. He has been an
(reversal) is intentionally introduced in order to move the data
associate professor at Shanghai Jiao Tong University since 1987. He
element to the opposite side of the constellation. Outer worked with Alcatel Space Industries from 1990 to 2001 in several R&D
elements on negatively-correlated subcarriers are increased and management positions. He was senior vice president of Linkair
in magnitude along the real or imaginary axis. Simulations Communications USA from 2001 to 2005 and then worked as digital
described in this paper show that selecting the correct design director with TechnoConcepts from 2005 to 2006. During this time,
subcarriers for bit reversals and outward enhancements he was involved in digital RF-mixed chip projects. Dr. Fang held a senior
reduces the peak-to-average power ratio of the OFDM signal wireless system position in EDA with Cadence USA during 2006. Since
2009, he has been director of the Electronics & Information Technology
to a target value. At the same time, in-band degradation
Center of the Research Institute of Tsinghua University (RITS) and has led
caused by bit error rate (BER) is limited. R&D and industrialization activities in wireless, signal processing, EDA,
In the paper by Fa-Long Luo, Ward Williams, and Bruce and digital TV multimedia. He has also been involved in wireless activities
Gladstone a new antenna diversity system is proposed that with CARITS Inc., a subsidiary of RITS in the U.S. Dr. Fang has written one
uses eigenfiltering for OFDM-based wireless communication book on information theory and coding, and he has contributed to ITU
and digital broadcasting applications. Compared with the publications on satellite systems. He has 10 patents and published many
technical papers. His research interests include SDR, digital RF chip
existing schemes, such as post-FFT, pre-FFT, and
design, and MIMO transmission systems.
polyphase-based filter-banks, the proposed scheme
performs optimally and has very low computational Fa-Long Luo is chief scientist at two leading international companies
complexity. The adaptive algorithm for updating the headquartered in Silicon Valley, CA, that deal with SDR and wireless
eigenfiltering coefficients has the same complexity as the LMS multimedia. He has been the editor-in-chief of the International Journal of
algorithm. The proposed scheme offers a better compromise Digital Multimedia Broadcasting since 2007. Dr. Luo is currently chairman
between performance, power consumption, and complexity in of the IEEE Industry DSP Standing Committee and technical board
member of the IEEE Signal Processing Society. He has 28 years of
the real-time implementation of receivers in broadband
research and industrial experience in multimedia, communication and
communication and digital broadcasting systems. broadcasting with real-time implementation, applications, and
The last paper in this special issue deals with standardization. He has received worldwide recognition. Dr. Luo has
analog-to-digital conversion and describes a histogram authored and edited four books, more than 100 technical papers, and 18
approach to error correction. The paper by Armin Jalili, J. patents on these and closely related fields. ➡To P.14
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P6
Polyphase Filter Banks for Embedded Sample Rate Changes in Digital Radio Front-Ends Special Topic
Mehmood Awan, Yannick Le Moullec, Peter Koch, and Fred Harris
Abstract: This paper presents efficient processing engines for software-defined radio (SDR) front-ends. These engines, based on a
polyphase channelizer, perform arbitrary sample-rate changes, frequency selection, and bandwidth control. This paper presents an
M-path polyphase filter bank based on a modified N-path polyphase filter. Such a system allows resampling by arbitrary ratios while
performing baseband aliasing from center frequencies at Nyquist zones that are not multiples of the output sample rate. This resampling
technique is based on sliding cyclic data load interacting with cyclic-shifted coefficients. A non-maximally-decimated polyphase filter
bank (where the number of data loads is not equal to the number of M subfilters) processes M subfilters in a time period that is less than
or greater than the M data loads. A polyphase filter bank with five different resampling modes is used as a case study for embedded
resampling in SDR front-ends. These modes are (i) maximally decimated, (ii) under-decimated, (iii) over-decimated, and combined
up- and down-sampling with (iv) single stride length, and (v) multiple stride lengths. These modes can be used to obtain any required
rational sampling rate change in an SDR front-end based on a polyphase channelizer. They can also be used for translation to and from
arbitrary center frequencies that are unrelated to the output sample rates.
(IF). In the second stage, the IF filter image must be lower than the desired
1 Motivation output is again down-converted to spectral term, and this is difficult to
T
here are several generations of baseband by a matched-quadrature sustain over time and at varying
architectures for digital radio mixer and matched baseband filters temperatures.
transceivers. A base station in a that perform final bandwidth control. The need for extreme I/Q balance
cellular mobile communication Next, the signal passes into the digital gave rise to the next generation of
system is an example of a multichannel domain where the output of the radios where second-stage (IF)
radio receiver that simultaneously analog-to-digital converter (ADC) is down-conversion and, consequently,
down-converts and demodulates processed by digital signal processing the channelization process are
narrowband radio frequency (RF) (DSP) engines. These engines perform digitized, as shown in Fig. 1(b). Digital
channels [1], [2]. Traditional the required baseband processing, that conversion at IF provides greater
heterodyne architecture, considered is, synchronization, equalization, control over the imbalance by
the first generation of digital radio demodulation, detection, and manipulating the number of bits
architecture, is shown in Fig. 1(a) for an decoding. The problem with this type of involved in the arithmetic operation. The
N-channel receiver. Each subreceiver architecture is that amplitude and precision of the coefficients used in the
consists of a dual-stage phase are imbalanced. This results in filtering process sets an upper limit for
down-converter, and only the cross-talk between the narrowband spectral artifacts at -5 dB/bit. This
baseband processing is done in the channels because of aging (time, means that a 12-bit ADC can achieve
digital domain [2]. In the first stage, the temperature) of the analog components image levels below -60 dB [2].
RF signal is down-converted to of the quadrature down-converter. DSP-based complex
bandlimited intermediate frequency Each imbalance-related spectral down-conversion, however, has two
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P7
Special Topic Polyphase Filter Banks for Embedded Sample Rate Changes in Digital Radio Front-Ends
Mehmood Awan, Yannick Le Moullec, Peter Koch, and Fred Harris
Singel Channel
Receiver LO-1_B BB Digital BB LO-1_B Digital BB
IF ADC
RF Base Base
Band Digital Band
Proc IF Proc
ADC
Splitter Network
…
LO-N_B LO-N_B
IF LO-A
RF ADC Base Base
Band Band
ADC Proc Proc
(a) (b)
ADC: analog-to-digital converter BB: baseband IF: internediate frequency LO: local oscillator RF: radio frequency
▲Figure 1. (a) First generation of digital radio architecture for an N-channel receiver, and (b) Second generation of digital radio architecture
for an N-channel receiver.
advantages: the spectral images are multirate signal processing specifies concludes the paper.
controlled so that they are below the new ways of performing DSP tasks, and
quantization noise floor of the ADC these ways are not normally available in
involved in the conversion process, and traditional DSP designs. A multirate 2 Introduction
the digital filters before and after the polyphase filter can perform the tasks A multirate polyphase filter can
mixers are designed to have linear of a multichannel receiver. In such a perform the tasks of a multichannel
phase characteristics [2]. The second receiver, an input signal is receiver. These tasks are equivalent to
generation of radio, with digital composed of many equal-bandwidth, down-conversion, filtering, and
front-end, is a realizable version of equally spaced resampling of multiple narrowband
SDR. The range of applications for frequency-division-multiplexed (FDM) signals [5]. The step-by-step
second-generation architecture, channels. These channels are digitally conversion of a standard
shown in Fig. 1(b), is restricted to those down-converted to baseband single-channel demodulator into a
with IF center frequencies of a couple of (bandwidth is constrained by digital multichannel polyphase channelizer is
hundred megahertz. This is due to the filters) and subjected to a sample rate described in [2] and [3]. A brief
limited dynamic range of high-speed reduction commensurate with the introduction is given here. In the
ADCs. The dynamic range is often bandwidth reduction. This significantly standard single-channel demodulation
extended by using a hybrid scheme in reduces the number of system process, shown in Fig. 2(a), the
which the initial complex resources required to perform carrier-centered spectrum is translated
down-conversion is performed by multichannel processing and, to baseband (where a filter reduces the
analog I/Q mixers, and channelization is consequently, reduces costs [2], [3]. bandwidth), and a resampler reduces
performed digitally after the ADC. DSP The remainder of this paper is the sample rate in proportion to the
techniques are applied to the digitized organized as follows: In section 2, we bandwidth reduction. Standard
I/Q data to balance the gain and phase briefly introduce a polyphase single-channel demodulation is
offsets in the analog ADC [3]. channelizer and describe how it is described by
A digital front-end with a standard formulated from a conventional -jθkn
design that includes frequency channelizer. In section 3, we categorize y (n, k) = [x (n )e ]×h (n) (1)
N-1
selection, bandwidth reduction, and the embedded resampling cases y (n, k) = Σx (n-r )e
-jθk (n-r )
h (r ) (2)
sample rate reduction is one of the most described in [5] into five different r =0
power- and time-critical functionalities resampling cases: maximally where x (n ) is the carrier-centered
of an SDR terminal. This is due to the decimated, under-decimated, input signal, θk is carrier angular
large bandwidth and high dynamic over-decimated, and combined frequency for k th channel, h (n) is the
range of the signal to be processed. up- and down-sampled with single baseband filter, and y (n, k) is the output
Consequently, the digital signals may and multiple commutator stride lengths. baseband signal for k th channel.
have high sample rates and large word In section 4, we perform MatLab According to the Equivalency
lengths. High sample rates not only simulations and the simulation results Theorem [3], down-conversion
increase power consumption but also demonstrate the performance of followed by baseband filtering can be
make the use of time-shared hardware polyphase channelizers that deliver the reordered so that filtering at the carrier
infeasible [4]. On the other hand, targeted output sample rates. Section 5 occurs first, followed by
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P8
Polyphase Filter Banks for Embedded Sample Rate Changes in Digital Radio Front-Ends Special Topic
Mehmood Awan, Yannick Le Moullec, Peter Koch, and Fred Harris
…
condition, the down-sampled e
j (M-1)k (2π/Μ)
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P9
Special Topic Polyphase Filter Banks for Embedded Sample Rate Changes in Digital Radio Front-Ends
Mehmood Awan, Yannick Le Moullec, Peter Koch, and Fred Harris
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P10
Polyphase Filter Banks for Embedded Sample Rate Changes in Digital Radio Front-Ends Special Topic
Mehmood Awan, Yannick Le Moullec, Peter Koch, and Fred Harris
2nd Data Load 1st Data Load 2nd Data Load 1st Data Load Case 3: Over-Decimated
R0 n+9 n+4 R0 n+4 R0 n+2 R0 n+2 Mode
R1 n+8 n+3 R1 n+3 R1 n+1 R1 n+1 In a over-decimated
R2 n+7 n+2 R2 n+2 3 R2 n+5 n+0
3 R2
n+0 system, data is loaded in
5 R4
R3 n+6 n+1
5 R4
R3 n+1 R3 n+4 R3 stride lengths of 6 mod 5,
R4 n+3
n+5 n+0 n+0 R4 and a computed output
(a) (b) sample has a 6-to-1
2 Data Load
nd
1 Data Load
st
2 Data Load
nd
1 Data Load
st down-sampling within a
R0 n+10 n+5 n+0 R0 n+5 n+0 R0 n+2 R0 n+2 5-stage polyphase filter.
R1 n+9 n+4 6 R1 n +4 2 R1 n +4 R1 The LCM of 6 and 5 is 30,
R2 n+8 n+3 R2 n+3 R2 n+1 3 R2 n+1 which means that the state
R3 n+3
6 R4
R3 n+7 n+2 R3 n+2 R3 engine cycles in 30 inputs,
R4 n+1 R4 n+0 R4 n+0
n+11 n+6 n+1 and because 6-point data
(c) (d) is delivered at a time, there
2 Data Load
nd
1 Data Load
st must be 5 distinct states in
R0 n+12 n+7 n+2 R0 n+7 n+2 the state machine. Fig. 4(c)
R1 n+14 n+9 n+4 R1 n+4 shows the data-loading
7
R2 n+11 n+6 n+1 R2 n+6 n+1 processes for the first two
R3 n+13 n+8 n+3 8 R3 n+3 states.
R4 n+10 n+5 n+0 R4 n+5 n+0
In the first data load,
(e) loading starts from subfilters
R0, R4 up to R0, and the
▲Figure 4. Data loading processes for a 5-path polyphase filter performing (a) 5-to-1 down-sampling
(case 1), (b) 3-to-1 down-sampling (case 2), (c) 6-to-1 down-sampling (case 3), (d) 5/2 down-sampling second load starts from R4
(case 4), and (e) 15/2 down-sampling (case 5). up to R0 and R4 again and
so on for the 5 distinct
▼Table 1. Register loading sequence and states. states. Consequently, there is a residue
corresponding subfilters coefficients for a In the first state, data loading starts of 4 for each data-loading operation.
5-path maximally decimated system from subfilter R2 up to R0; and in the To align the data time-origin with the
State Loading Sequence Filter Coefficients second state, data loading starts from FFT time-origin, the subfilter
R4 to R2 and so on for the five distinct coefficients are cyclic-shifted by the
0 R4 R3 R2 R1 R0 C0 C1 C2 C3 C4
states. Consequently, there is a residue residue address of the data time-origin
of 2 for each data-loading operation. mod 5. Table 3 shows the state
▼Table 2. Register loading sequence and To align the data time-origin with the machine for register-loading and the
the corresponding subfilter coefficients for FFT time-origin, the subfilter corresponding coefficients for
3-to-1 sample rate change in a 5-path coefficients are cyclic-shifted by the performing 6-to-1 down-sampling in a
polyphase filter residue address of the data time-origin 5-stage polyphase filter.
State Loading Sequence Filter Coefficients mod 5. The time-origin that is being Case 4: Combined Up- and
cyclically shifted is also periodic in the Down-Sampling Mode (Single Stride)
0 R2 R1 R0 C0 C1 C2 C3 C4
LCM of 3 and 5. Thus, the cyclic shift of In the previous three cases,
1 R4 R3 R2 C3 C4 C0 C1 C2 the polyphase subfilter coefficients has down-sampling was performed by
2 R1 R0 R4 C1 C2 C3 C4 C0 the same period as the data register different factors. The polyphase filter is
load and is controlled by the same state also capable of embedding the
3 R3 R2 R1 C4 C0 C1 C2 C3
machine. The data-loading sequence up-sampling factor with the
4 R0 R4 R3 C2 C3 C4 C0 C1 is always to the next 3 registers that down-sampling so that the sample rate
have indexing of mod 5, which means change is rational. In this case, we
(under-decimated) system, data is that the next register to accept data
▼Table 3. Register loading sequence and the
loaded in stride lengths of 3 mod 5, and when moving from state 0 to state 1 is corresponding subfilter coefficients for
a computed output sample has a (R0)-1, which is actually R4. Similarly, 3-to-1 sample rate change in a 5-path
3-to-1 down-sampling within a the filter coefficients assigned to polyphase filter
5-stage polyphase filter. The least perform the inner products with the State Loading Sequence Filter Coefficients
common multiple (LCM) of 3 and 5 is registers are always offset 3 mod 5
0 R0 R4 R3 R2 R1 R0 C0 C1 C2 C3 C4
15, which means that the state engine relative to the previous filter set. Table 2
cycles in 15 inputs, and because shows the state machine for 1 R4 R3 R2 R1 R0 R4 C1 C2 C3 C4 C0
3-point data is delivered at a time, register-loading and coefficients of 2 R3 R2 R1 R0 R4 R3 C2 C3 C4 C0 C1
there must be 5 distinct states in the each corresponding subfilter for
3 R2 R1 R0 R4 R3 R2 C3 C4 C0 C1 C2
state machine. Fig. 4(b) shows the performing 3-to-1 down-sampling in a
4 R1 R0 R4 R3 R2 R1 C4 C0 C1 C2 C3
data-loading processes for the first two 5-stage polyphase filter.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P11
Special Topic Polyphase Filter Banks for Embedded Sample Rate Changes in Digital Radio Front-Ends
Mehmood Awan, Yannick Le Moullec, Peter Koch, and Fred Harris
▼Table 4. Register loading sequence and successive filter index increments by is 30 MHz, and each channel has a 2.5
the corresponding subfilter coefficients for 6 mod 10; and between the states, the MHz symbol rate shaped by a square
5/2 sample rate change in a 5-path filter index increments by 5 mod 10. The root Nyquist filter with 20% excess
polyphase filter integer 6 is the offset between two data bandwidth. Three of the five channels,
No Loading Filter samples in the zero-packed load in two which are occupied by 3 MHz
State
of Inputs Sequence Coefficients adjacent rows. Because of bandwidth signals, are centered at 0, 6,
0 3 R4 R2 R0 C0 C6 C2 C8 C4 up-sampling by a factor of 2, the and 12 MHz. The remaining two
prototype filter has to be designed to channels, centered at -12 and -6 MHz,
1 2 R3 R1 C5 C1 C7 C3 C9 operate at 2 × fs , that is, 60 MHz. are intentionally kept empty. The input
Consequently, the filter becomes twice signal spectrum comprising 5 channels
up-sample by a factor of 2 and then as long as the standard design. at 30 MHz is shown in Fig. 5(a).
down-sample by a factor of 5 to obtain However, because only half of it is used In a system operating in maximally
a 5/2 sample rate change. The per processing cycle, there is no decimated mode, the input data is
up-sampling, performed by processing penalty [5]. channelized and down-sampled
zero-packing the input data, is actually Case 5: Combined Up- and 5-to-1 for an output rate of 6 MHz.
achieved by data-load addressing, in Down-Sampling Mode (Multiple Each of the 5 polyphase filter stages
which one address is skipped so that Strides) are 6 taps long and are anchored to the
1-to-2 up-sampling can be realized. This case is similar to case 4 but 5 input registers being fed by the
Down-sampling is performed by down-sampled by a factor of 15 to periodic input commutator. Fig. 5(b)
cyclic-loading the data (zero-packed) have a 15/2 sample rate change. shows the spectra of the 5 output
through the filter in stride of length 5. Up-sampling is performed by channels with an output rate of 6 MHz.
The two data-loading cycles for a 5/2 data-load addressing, which skips the In a system operating in
sample rate change in a 5-path next address, and down-sampling is under-decimated mode, the same
polyphase filter are shown in Fig. 4(d). performed by cyclic-loading the data input data is channelized and
In the first data load, 3 data samples through the filter in stride lengths of 15. down-sampled 3-to-1 for an output
are delivered to the 5 register The two states of the loading cycle for rate of 10 MHz. Fig. 5(c) shows the
addresses. In the second load, 2 data 15/2 sample rate change in a 5-path spectra of the 5 output channels with
samples are delivered to the 5 register polyphase filter are shown in Fig. 4(e). 10 MHz output rate. Similarly, in a
addresses. The data-loading process In the first data load, 8 data samples system operating in over-decimated
is periodic in 2 load cycles, and 2 are delivered to the 5 register mode, the same input data is
states are needed to control the addresses. In the second load, 7 data channelized and down-sampled
process. The data-loading process for samples are delivered to the 5 register 6-to-1 for an output rate of 5 MHz.
the 2 states and the corresponding addresses. The data-loading process Figure 5(d) shows the spectra of the 5
coefficient sets are listed in Table 4. In is periodic in 2 load cycles, and 2 output channels with 5 MHz output rate.
the 2 states, 5 inputs are delivered, and states are needed to control the In a system operating in combined
2 outputs from the polyphase engine process. Table 5 lists the data-loading up- and down-sampling mode, the
are taken to realize the desired 5/2 process for the 2 states and the input spectrum is channelized,
embedded resampling. The loading corresponding coefficient sets. In the up-sampled by a factor of 2, and
scheme has a constant offset of process, 15-to-2 down-sampling is down-sampled by 5-to-1 and
-2 mod 5 within a sequence and also in performed in a 5-path polyphase filter. 15-to-1 for output rates of 12 MHz and
the transition between sequences. The filter down-converts the spectral 4 MHz, respectively. Because of
The -2 offset is a result of the 1-to-2 regions from multiples of fs /5 (or up-sampling by a factor of 2, there are
up-sampling, represented by the zero 30/5 = 6 MHz) and maintains a sample 10 polyphase filter coefficient stages
packing. rate of fs (2/15) (or 60/15 = 4 MHz). each with 6 taps. The filters’
There are normally 5 subfilters in the coefficients are periodically rotated
polyphase partition of a 5-stage through the 5 input registers (which
polyphase filter. Because of the 1-to-2 4 Simulations have a periodic sliding input
up-sampling implemented by the The MatLab simulations show the commutator) according to the state
zero-packing, only half the coefficients embedded sample rate changes in a machine described in Table 4 and
in each stage actually contribute to the 5-path polyphase filter and
subfilter output [5]. Thus, each stage is DFT operating as a ▼Table 5. Register loading sequence and the corresponding
subfilter coefficients for15/2 sample rate change in a
further partitioned into 2 subsets of 5-channel channelizer. The 5-path polyphase filter
coefficients, which results in 10 subfilter FDM input signal has 5
coefficient sets. These sets are denoted channels that are each State No of Inputs Loading Sequence Filter Coefficients
C0, C1,..., C9 where the integer is the 16-QAM modulated and 0 8 R4 R2 R0 R3 R1 R4 R2 R0 C0 C6 C2 C8 C4
starting index from the original separated by 6 MHz center
1 7 R3 R1 R4 R2 R0 R3 R1 C5 C1 C7 C3 C9
non-partitioned prototype filter. The frequencies. The sample rate
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P12
Polyphase Filter Banks for Embedded Sample Rate Changes in Digital Radio Front-Ends Special Topic
Mehmood Awan, Yannick Le Moullec, Peter Koch, and Fred Harris
10 10 10 10 10 10
cellular mobile systems,”in Software Radio
0 0 0 0 0 0 Technologies and Services, E. Del Re, Ed., Berlin,
-10 -10 -10 -10 -10 -10 Germany: Springer-Verlag, 2001.
-20 -20 -20 -20 -20 -20 [2] f. harris, C. Dick, M. Rice,“Digital receivers and
-30 -30 -30 -30 -30 -30 transmitters using polyphase filter banks for
dB
(e) (f)
Biographies
▲Figure 5. (a) Input signal spectrum with 5 channels of 3 MHz bandwidth and 6 MHz channel
spacing at 30 MHz sample rate. The spectra of the 5 output channels are (b) 6 MHz for 5-to-1 Mehmood Awan (mura@es.aau.dk) received his
MSc degree in electronic engineering with
down-sampling, (c)10 MHz for 3-to-1 down-sampling, (d) 5 MHz for 6-to-1 down-sampling, specialization in applied signal processing and
(e) 12 MHz for 5/2 resampling, and (f) 4 MHz for 15/2 resampling. implementation from Aalborg University in 2007. He
was a research assistant for one year and started
Table 5. The spectral locations of the performs embedded resampling that is his Ph.D. in resource-optimal SDR front-ends in
2008. His research interests include multirate signal
channels are reordered as a result of uncoupled from frequency selection processing, SDR, hardware architectures, and
processing the up-sampled data in the and bandwidth control. Five embedded embedded systems.
polyphase filter [3], [5]. The 5-point resampling modes in polyphase filter Yannick Le Moullec (ylm@es.aau.dk) received
FFT processes the polyphase data banks have been presented, namely, his Ph.D. degree in electrical engineering from
Université de Bretagne Sud, Lorient, France, in
output frequencies in the order maximally decimated,
2003. From 2003 to 2005, he was a post-doctoral
[0, 2, 4, 1, 3], which is seen to be under-decimated, over-decimated, fellow at the Center for Embedded Software
indexing stride of 2 mod 5. These are and combined up- and Systems, Aalborg University, Denmark. From 2005
to 2008, he was an assistant professor at the
reordered back to their natural order. down-sampling. These correspond to Department of Electronic Systems, Aalborg
Figs. 5(e) and (f) show the spectra of single, short, long, and multiple University, where he is now an associate professor.
His research interests include methods and tools for
the 5 output channels at 12 MHz and 4 commutator stride lengths. For various HW/SW co-design, embedded systems, and
MHz, which correspond to 5/2 and 15/2 applications, these modes can be used reconfigurable computing.
sample rate changes, respectively. for any required rational sampling-rate
Peter Koch (pk@es.aau.dk) received his M.Sc.
The simulations show that embedded change in an SDR front-end using a and Ph.D. degrees in Electrical Engineering from
sample rate changes can be polyphase channelizer. The suggested Aalborg University, Denmark, in 1989 and 1996.
Since 1997, he has been an associate professor at
successfully implemented in a modes are highly useful for designing the Department of Electronic Systems, Aalborg
polyphase channelizer. All the output flexible and resource-optimal University, working in the interdisciplinary field
channels have 60 dB of spectral architectures for advanced software between DSP and resource-optimal real-time
architectures. From 2006 to 2010, he headed the
side-lobe attenuation, selected by the radios. In a subsequent paper Center for Software Defined Radio, Aalborg
prototype low-pass filter. The “Hardware Architecture Analysis of University. His research interests include
optimization between DSP algorithms and
processing engines used in all the 5 Polyphase Filter Banks Performing architectures, and low-energy HW/SW design.
cases are identical except that each Embedded Resampling for Software
has different state machines, register Defined Radio Front-Ends”[6], we Fred Harris (fred.harris@sdsu.edu) holds the
Signal Processing Chair of the Communication
loading schemes, and subfilter analyze FPGA based hardware Systems and Signal Processing Institute at San
coefficient sets. architecture of these resampling Diego State University where he teaches DSP and
communication systems. He holds 20 patents for
engines in terms of area, time, and digital receivers, and he lectures around the world
power tradeoffs. on DSP applications. He is an adjunct of the
5 Conclusion Princeton IDA-CCR Center for Communications
Research and is the author of“Multirate Signal
References
In this paper, we have shown the [1] A. M. Badda and M. Donati,“The software defined Processing for Communication systems.”
versatility of a polyphase engine that radio technique applied to the RF front-end for
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P13
Special Topic Design of Software-Defined Down-Conversion and Up-Conversion: An Overview
Yue Zhang, Li-Ke Huang, Carsten Maple, and Qing Xuan
Design of Software-Defined
Down-Conversion and
Up-Conversion: An Overview
Yue Zhang 1, Li-Ke Huang 2, Carsten Maple 1, and Qing Xuan 3
(1. Department of Computer Science and Technology, University of Bedfordshire, UK;
2. Aeroflex International Ltd, Stevenage, UK;
3. Sino-European-Link Ltd, UK)
Abstract: In recent years, much attention has been paid to software-defined radio (SDR) technologies for multimode wireless systems.
SDR can be defined as a radio communication system that uses software to modulate and demodulate radio signals. This article
describes concepts, theory, and design principles for SDR down-conversion and up-conversion. Design issues in SDR
down-conversion are discussed, and two different architectures, super-heterodyne and direct-conversion, are proposed. Design
issues in SDR up-conversion are also discussed, and trade-offs in the design of filters, mixers, NCO, DAC, and signal processing are
highlighted.
S
oftware-defined radio (SDR) will front-end) is the only analog section synthesizing, digital numerical control
play a key role in future radio and includes mixer, low-noise amplifier oscillator (NCO), digital mixing, I/Q
configurations because of the (LNA), power amplifier (PA), RF demodulation, and multirate decimation
emergence of new wireless combiner, bandpass filter filtering. In digital IF processing, digital
technologies and their integration into (anti-aliasing), and antenna. The RF up-conversion (DUC) is performed in
fourth generation standards such as front-end is responsible for transmitting the reverse order of DDC. Digital
LTE-Advanced. SDR is a fundamental and receiving RF signals and systems such as FPGAs are normally
part of many radio systems, which converting RF signals to IF. Some used for IF-to-baseband conversion
include up-conversion of the discrete advanced RF front-ends allow a certain because they are able to handle tight
baseband signal into a high-resolution degree of control, for example, real-time constraints imposed by
radio signal at the transmitter and frequency tuning, through software. high-speed sampling and digital
down-conversion of the The other parts of the SDR conversion. In the back-end of the SDR
high-resolution radio signal back into architecture are digital processing architecture, the baseband process
baseband at the receiver [1]. Fig. 1 components. In the IF section, sampling mainly performs digital communications
shows an SDR receiver in and separation of IF carriers, and functions such as symbol timing
super-heterodyne architecture — a up/down frequency conversion to recovery, equalization, modulation, and
dual-stage conversion architecture in baseband is performed by digital IF channel coding. These functions are
which a radio frequency (RF) signal is processing normally implemented in usually carried out using digital signal
down-converted to intermediate field programmable gate array (FPGA). processors (DSPs) with slightly relaxed
frequency (IF) in the first stage and then Take the receiving path for example. real-time constraints.
converted from IF to baseband in the The IF signal is digitized by the SDR architecture tends to use
second stage. The hardware in the RF, analog-to-digital converter (ADC) and general-purpose digital systems.
IF, and baseband sections are then converted to baseband by digital FPGAs and DSPs in SDR allow the
controllable and reconfigurable using IF processing. Digital IF processing transmitted signal to be generated and
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P14
Design of Software-Defined Down-Conversion and Up-Conversion: An Overview Special Topic
Yue Zhang, Li-Ke Huang, Carsten Maple, and Qing Xuan
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P15
Special Topic Design of Software-Defined Down-Conversion and Up-Conversion: An Overview
Yue Zhang, Li-Ke Huang, Carsten Maple, and Qing Xuan
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P16
Design of Software-Defined Down-Conversion and Up-Conversion: An Overview Special Topic
Yue Zhang, Li-Ke Huang, Carsten Maple, and Qing Xuan
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P17
Special Topic Design of Software-Defined Down-Conversion and Up-Conversion: An Overview
Yue Zhang, Li-Ke Huang, Carsten Maple, and Qing Xuan
front-ends. Advantages and [7] B. Razavi.“Design considerations for transmitter/receiver cyclic delay diversity,”in IEEE
direct-conversion receivers,”IEEE Trans. Circ. Trans. Broadcasting, vol. 52, no. 4, Dec. 2006, pp.
disadvantages of each design have Syst. II: Analog and Digital Signal Processing, vol. 464-474.
also been analyzed. A well-designed 44, no. 6, pp. 428-435. [12] F.H. Raab, P. Asbeck, S. Cripps, P.B. Kenington, Z.
[8] R. Svitek and S. Raman,“DC offsets in direct B. Popovic, N. Pothecary, J.F. Sevic and N.O.
architecture for multiband multimode conversion receivers: Characterization and Sokal,“RF and microwave power amplifier and
receivers and transmitters should allow implications,”IEEE Microw. Mag., vol 6, no.3, transmitter technologies-Part 3,”High Frequency.
hardware resources to be optimally pp.76-86. Electronics., vol. 2, no.5, pp.34-46, Sept 2003.
[9] Fa-Long Luo (Ed.), Mobile Multimedia [13] A. Loke and F. Ali,“Direct conversion radio for
shared. The architecture should make Broadcasting Standards: Technology and Practice. digital mobile phones—Design issues, status, and
use of software-controlled components Dordrecht: The Netherlands: Springer, 2008. trends,”IEEE Trans. Microw. Theory Tech., vol. 50,
[10] Yue Zhang, J. Cosmas, K.K Loo and M. Bard, no.11, pp. 2422-2435, Nov. 2002.
and software-programmable devices. “Design and implementation of digital echo [14] G-T Gil,“Non-data-aided I/Q mismatch and DC
A direct-conversion approach was cancellation on-channel repeater in DVB-T/H offset compensation for direct-conversion
used in the transmitter and receiver networks,”in 57th Annu. IEEE Broadcast Symp., receivers,”IEEE Trans. Signal Processing, vol. 56,
Washington DC, 2007. no. 7, Jul. 2008, pp. 2662-p2668.
front-end for implementation in most [11] Yue Zhang, J Cosmas, M. Bard, and Y.H Song, [15] F. Ellinger. Radio Frequency Integrated Circuits
commercial wireless systems. “Diversity gain for DVB-H by using and Technologies. New York: Springer, 2007.
Direct-conversion is the most
appropriate architecture because it is Biographies
low cost, consumes little power, and Yue Zhang (yue_zhang@ieee.org) is a senior lecturer in the Department of Computer Science and Technology,
has a high-data-rate radio interface. University of Bedfordshire. He received his B.Eng. and M. Eng. degrees in 2001 and 2004 from Beijing University
of Post and Telecommunications. In 2008, he received his Ph.D. degree from Brunel University — where he also
However, system designers need to worked as a research engineer for the EU IST FP6 project, PLUTO. He was responsible for transmitter and receiver
pay more attention to I/Q imbalance, diversity design and measurement for DVB-T/H systems as well as RF/Digital/DSP design for on-channel
repeaters. After 2008, Dr. Zhang worked as a signal processing design engineer in Anritsu. He was responsible for
carrier leakage, and DC offset in a RF/IF, digital and DSP design for wireless communication systems.
direct-conversion receiver and
transmitter system. Li-Ke Huang (like.huang@gmail.com) is a technical expert and algorithms group leader at Aeroflex UK. He
develops test and measurement technologies for wireless system engineering and specializes in transceiver
algorithm and architecture designs for the main wireless standards. He is responsible for new product features
and new technology developments. He received his B.Sc. degree in electronic engineering from Shenzhen
University in 1998 and his Ph.D. degree in communication and signal processing from Imperial College London in
2002.
References
[1] Fa-Long Luo (Ed.), Digital Front-End in Wireless Carsten Maple (carsten.maple@beds.ac.uk) is pro-vice-chancellor of research and enterprise at the University
Communications and Broadcasting. New York: of Bedfordshire. He received his B.Sc. degree in mathematics and his Ph.D. degree in numerical analysis from the
Cambridge University Press, 2011. University of Leicester. He is a member of the IEEE, and a fellow of the FBCS and CITP. He heads the Center for
[2] V. Giannini, J. Craninckx, and A. Baschirotto, Research in Distributed Technologies (CREDIT) at the University of Bedfordshire, which has 40 staff and Ph.D.
Baseband Analog Circuits for Software Defined researchers. His research interests include information security, trust and authentication in distributed systems,
Radio. Dordrecht: The Netherlands: graph theory, and optimization techniques. He has led a number of research projects in these areas with funding
Springer-Verlag, 2008. totaling of more than-million from UK EPSRC, EU, and the Department of Trade and Industry (DTI). He has been
[3] F. Harris, Multirate Signal Processing for editor or guest editor for several international journals. He has been chairman and session chairman for a number
Communication Systems. New Jersey: Prentice- of international conferences. He has also been invited to present keynote speeches to various international
Hall, 2008. conferences. Dr. Maple has published over 70 papers internationally and has been invited to talk on security,
[4] Behzad Razavi. RF microelectronics. New Jersey: robotics, and applied computing on UK radio and television.
Prentice-Hall, 1998.
[5] A. Abidi,“The path to the software-defined radio Qing Xuan (jasx@talk2-1.com) received her Ph.D. degree in power and energy from the University of Bath in
receiver,”in IEEE J. Solid-State Circuits, vol. 42, 1995. She has more than 20 years’experience in the telecom, energy, finance, and aviation sectors. Since March
no.5, pp. 954-966, May 2007. 2009, she has co-founded two companies. From January 2005 to March 2009 she was a business strategy
[6] V.Giannini, P. Nuzzo, C.Soens, K.Vengattaramane, director for Huawei Technologies. Her responsibilities included working with the minister and regulator to develop
M.Steyaert, J.Ryckaert, M.Goffioul, B.Debaillie, J. markets and introduce investor partners to clients. She also provided supply chain consultancy services to
Van Driessche, J.Craninckx, and M.Ingels, secured-business clients, including M&A Technology Company. From 2000 to 2004, Dr. Qing Xuan was a
“A 2 MM2 0.1-to-5 GHz SDR receiver in 45 nm commercial and technical architect at Vodafone Group. From 1998 to 2000, she was as design engineer for
digital CMOS, ”in IEEE J. Solid-State Circuits, vol. Panasonic. From 1995 to 1998, she was post-doctoral researcher in the Power Group at the University of Bath.
44, no. 12, Dec. 2009, pp. 3486-3498. From 1987 to 1991, she was deputy director of the Telecom Department at the Ministry of Power and Water, China.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P18
Practical Non-Uniform Channelization for Multistandard Base Stations Special Topic
Álvaro Palomo Navarro, Rudi Villing, and Ronan J. Farrell
Practical Non-Uniform
Channelization for Multistandard
Base Stations
Álvaro Palomo Navarro, Rudi Villing, and Ronan J. Farrell
(Callan Institute, Electronic Engineering Department, National University of Ireland, Maynooth (NUIM), Ireland)
Abstract: A Multistandard software-defined radio base station must perform non-uniform channelization of multiplexed frequency
bands. Non-uniform channelization accounts for a significant portion of the digital signal processing workload in the base station
receiver and can be difficult to realize in a physical implementation. In non-uniform channelization methods based on generalized DFT
filter banks, large prototype filter orders are a significant issue for implementation. In this paper, a multistage filter design is applied to
two different non-uniform generalized DFT-based channelizers in order to reduce their filter orders. To evaluate the approach, a TETRA
and TEDS base station is used. Experimental results show that the new multistage design reduces both the number of coefficients and
operations and leads to a more feasible design and practical physical implementation.
R
adio spectrum is typically digital-to-analogue conversion is wideband uplink, which comprises all
allocated using coarse-grained performed as close as possible to the the mobile station channels, is digitally
frequency division multiplexing. antenna [2]. Most of the radio converted immediately after the RF
Different radio standards are components, now in the digital domain, front-end. Multirate digital signal
allocated independent and are implemented on a reconfigurable processing (DSP) is then used to filter
non-overlapping frequency bands that platform. Reconfigurability makes SDR and shift to baseband all the
are reserved exclusively for each particularly suitable for working with independent information channels. On
standard. This approach simplifies the multistandard systems and for the transmitter side, complementary
radio design because each standard
operates in isolation; however, available
spectrum is not used optimally because Downlink
reserved bands may be under-used MS1
x 0( n )
some of the time. A more efficient Wideband MSK 1 -1
x 1(n ) MS0
alternative is to allow multistandard Downlink
Mobile
…
Synthesis
multiplexing of the frequency band. The x K-1(n ) Standard 1 MS1
Uplink
frequency band is instead multiplexed y 0( n )
MSK 2-1
Wideband MS1
among multiple radio standards [1]. y 1(n ) Uplink
MS0
Mobile
Channels within the frequency band are
…
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P19
Special Topic Practical Non-Uniform Channelization for Multistandard Base Stations
Álvaro Palomo Navarro, Rudi Villing, and Ronan J. Farrell
processing is carried out. Extraction of techniques, parallel generalized invested heavily in TETRA V&D
the uplink channels (channelization) is a DFT-FB (GDFT-FB) and recombined networks, so it is important that any
bigger challenge than downlink GDFT-FB, by applying multistage future updates are compatible with
synthesis because more stringent filtering. We then evaluate these these legacy networks.
filtering is required to avoid adjacent modified FBs in TETRA and TEDS base For all PMR standards, including
channel interference. stations and compare their filter lengths TETRA and TEDS, the permitted
Different non-uniform channelization and computational loads with those in channel centre frequencies, fc (n ), are
techniques have been described in [4]. Finally, the effect of multistage defined by
[3]-[5]. In many cases, these design on the physical implementation
techniques are based on combining or of the channelizer is considered. fc (n ) = band _ edge + (n - 1 ) Δfc (1)
2
altering efficient uniform channelization Section 2 describes the non-uniform
methods, such as uniform complex channelization requirements of a base where n is the channel number, band
discrete Fourier transform (DFT) station, specifically, TETRA and TEDS edge is the lower-edge frequency of
modulated transmultiplexers and tree base stations and their possible the multiplexed frequency band, and Δ
quadrature mirror filter banks (TQMFB) updates. Section 3 describes the fc is the channel spacing [10]. The
[6]. In these filter bank (FB) based classic implementation of the two available spectrum is divided into
structures, a synthesis bank multiplexes non-uniform channelization methods frequency sub-bands that are equal to
the channel frequency at the transmitter analyzed in this paper. Designs for the channel spacing, and channels are
side, and an analysis bank performs for methods to be used in TETRA and centered within each sub-band.
equivalent channelization in the TEDS base stations are given. Section 4 In common hardware, the
receiver. Such a structure is symmetric; describes a new multistage design for multiplexed frequency band solution
that is, amplitude and phase distortion the two non-uniform channelization requires fixed-channel allocation.
in one half of the FB are suppressed in methods, and improvements in design Because the hardware cannot be easily
the complementary half, allowing and implementation are also described. reconfigured, different sections of the
perfect reconstruction. This approach Section 5 concludes the paper. spectrum are allocated for different
includes multicarrier techniques such channel sizes. This solution has poor
as OFDM. In contrast, the system in 2 Frequency Multiplexed spectrum efficiency because the
Fig. 1 has an asymmetric design; that channel allocation is not optimized. The
is, the uplink signal channelized by the
Spectrum for two SDR-based approaches described
analysis bank is not created by a Next-Generation PMR here allow reconfiguration so that
complementary synthesis bank. The There are two main releases of channels can be dynamically allocated,
system comprises the transmissions of TETRA: TETRA voice and data (V&D) and channels of different sizes can be
independent mobile stations that are and TEDS. The former supports 25 kHz adapted as required. This means that
assumed to use single-carrier channels that are mainly allocated in spectrum can be used more efficiently.
transceivers. The downlink is similarly the 380-400 MHz frequency band. The To test these non-uniform
asymmetrical. This structure allows the latter was approved by the ETSI in 2005 channelization structures, we created a
base station to be compatible with and supports wideband services using scenario where the standard
legacy systems. A professional mobile 50, 100, and 150 kHz channels. 380-400 MHz PMR band was shared
radio (PMR) base station using the However, the maximum throughput by TETRA 25 kHz and TEDS 25, 50, and
terrestrial trunked radio (TETRA) using TEDS is not enough for real-time 100 kHz channels. In this band, the
standard or its high-speed derivative, applications. Therefore, the addition of 5 MHz between 380-385 MHz was
TETRA enhanced data service (TEDS) a broadband communication system reserved for the uplink signal, and the
[7], is a good application of this similar to one used in commercial 4G 5 MHz between 390-395 MHz was
architecture. mobile communications is being used for the downlink signal. TEDS
In an asymmetric system, subcarriers investigated [8]. 150 kHz channels were not considered,
and aliasing overlap because of In a legacy base station, the antenna but the procedure can be extended to
down-sampling in the analysis bank. system can only cover 380-400 MHz, 150 kHz channels.
The overlap must be minimized by and adding extra antennas can be
high-order filters [4], and these high problematic because of 3 Non-Uniform
orders are a bottleneck in the design electromagnetic restrictions in different
and physical implementation of the countries [9]. As a result, a
Channelization Based on
channelizer filters. Multirate DSP multistandard multiplexed frequency GDFT-FB
techniques, such as multistage filtering, band has been proposed in which Channelization methods can be
can significantly reduce the filter orders TETRA and TEDS channels share the uniform or non-uniform depending on
for more practical implementation. In 380-400 MHz band instead of their capacity to filter channels that
this paper, we compare two occupying different bands. Prior to have equal or different bandwidths
non-uniform channelization TEDS being released, countries within the same frequency band.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P20
Practical Non-Uniform Channelization for Multistandard Base Stations Special Topic
Álvaro Palomo Navarro, Rudi Villing, and Ronan J. Farrell
Uniform channelization using where Ei are the polyphase processed through different critically
polyphase DFT-FBs has been components [6]. The rest of the decimated odd-stacked GDFT-FBs
proposed for real-world applications sub-filters, Hk (z ), expressed in (2) can operating in parallel. Each FB uniformly
that have a large number of (uniform) then be obtained from divides the frequency band according
channels. Various methods have also K-1
to specific channel spacing, and the
-(k+k 0)n 0 -ki -k 0 i -k 0k
been proposed for non-uniform Hk (z )=WK Σz -iWK WK Ei (z K·WK ). (5) FBs have overlapping frequency. In
i =0
channelization [3], [5], [11]-[12]. [15], the transmitter (synthesis) side is
However, methods based on uniform The general implementation of the described, and in [4], the receiver
DFT-DB are particularly attractive analysis GDFT-FB for even- and (analysis) side is described.
because of its low implementation odd-stacking cases is shown in Fig. 2. In the parallel GDFT-FB channelizer,
complexity. From (5), the different complex the digitized wideband signal with
Rather than using DFT-FB directly exponentials are applied to the different sample rate fS , is fed into multiple filter
[6], GDFT-FB is used. GDFT-FB gives branches in order to obtain the desired banks running in parallel. There is one
additional control over channel sub-band stacking and phase shifts. filter bank for each uniform channel
The complex exponentials, WK and
-k k
frequency allocation and phase 0
frequency plan. Narrowband channels
WK , can be directly hard-coded into
-k i
response because it has two are extracted by selecting an
0
parameters in the subfilter complex the polyphase components of the filter appropriate subset of outputs from
bank, and WK denotes the DFT
-k i
modulation [13]. In the filter bank, k each filter bank. Any permissible
algorithm. After the DFT, WK
-k nD
filters are obtained by complex is combination of channels can be
0
modulation of the lowpass prototype applied in order to present the different specified by choosing the appropriate
filter, H (z ): outputs, yk (n ), centred at DC. Finally, filter banks and channel numbers.
WK
-(k +k )n
-(k+k 0)n 0 (k+k 0)
0 0
is applied to the outputs for Changing channel allocation in
H k (z ) =W K H (zW K ) (2) phase shift purposes. Depending on real-time does not require redesign or
where Wk = exp(j 2π/K ), and K is the the values of the decimation factor, D, re-optimization of the filter bank
number of sub-bands of the analysis the filter bank can be critically sampled structure. Only selection of appropriate
bank. The parameter k 0 determines the (K =D) or oversampled (K = LD), where outputs needs to be adapted.
way different sub-bands are stacked L denotes the oversampling factor. The Fig. 3 shows the parallel GDFT-FB
on the spectrum, and n 0 determines the main benefit of an oversampled FB is for the 5 MHz uplink band of a
phases of the different channels. If that aliasing due to decimation is TETRA/TEDS base station. In this band,
k 0 = 0 and n 0 = 0, the sub-band significantly reduced [14]. However, an there are 200 channels for TETRA V&D
spectrum allocation is even-stacked, oversampled FB has additional and TEDS 25 kHz. For TEDS 50 kHz,
and the first sub-band is centred at computational load because it runs at a there are 100 channels, and for TEDS
DC. In this case, the GDFT-FB higher sample rate (by a factor of L). 100 kHz, there are 50 channels. DFT
structure is reduced to the classic modulation is implemented using a
DFT-FB. If k 0 = 1/2 and n 0 = 0, the 3.1 Parallel GDFT-FB power-of-2 FFT for efficiency; thus, the
spectrum allocation is odd-stacked; A parallel GDFT-FB is proposed in filter bank sub-bands cover a
that is, the sub-bands are shifted half [4], [15] as a non-uniform bandwidth larger than the frequency
the channel spacing, and no channel is channelization solution for base band containing the information
centred at DC. An odd-stacked stations. The wideband signal is channels. The excess sub-bands
approach is used in the proposed
non-uniform designs because it meets WK -k nD WK
0 -(0+k )n
0 0
DFT y 2(n )
Δf = fs /K (3) D E 2(z LWK -k K ) 0
WK -k nD WK
-(k-1+k )n
decomposed into its polyphase
0 0 0
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P21
Special Topic Practical Non-Uniform Channelization for Multistandard Base Stations
Álvaro Palomo Navarro, Rudi Villing, and Ronan J. Farrell
1 Nulls
TETRA
…
28 Critically Decimated
and Channels
TEDS 30 Sub-Band #29
…
25 kHz
29
256-Band 228
GDTF-FB 229
Nulls
…
256 -12.5 kHz 0 12.5 kHz
Non-Uniform Input
1 Nulls
…
TEDS 14
29 30 16 9 50 kHz 15
Nulls Nulls 16
…
Sub-Band #16
…
128
1 Nulls
…
TEDS 7 9
100 kHz 8 Sub-Band #19
64 Parallel GDFT-FB
channelization structure.
outside the 5 MHz bands are some constraints, the recombined output of the GDFT-FB onwards:
permanently null. GDFT-FB structure is a more flexible
R -1
For the input signal in Fig. 3, starting non-uniform channelization method. Y k,R (z ) = rΣ=0 e jφ Yk+r (z M )HM (ze-jβ ).
r r
(6)
from the lower edge
(-2.5 MHz), the first two TETRA/TEDS 3.2 Recombined GDFT-FB Therefore, channels are recombined
25 kHz channels are extracted from A non-uniform recombination FB first by interpolating each of the R
branches 29 and 30, respectively, of divides a signal into uniformly spaced sub-bands by a factor M, frequency
the first (25 kHz) FB. The next channel sub-bands and then recombines shifting by βr to the correct center
in the multiplexed spectrum, a 50 kHz certain groups of sub-bands to form frequency, and phase correcting by φr
TEDS channel, is selected as branch 16 wider bandwidths that are an integer in order to combine these shifted
of the second (50 kHz) FB. Branch 15 multiple of the uniform spacing in-phase channels. To minimize
refers to the same frequency range as (Fig. 4a). The sub-band bandwidth is amplitude distortion in the recombined
branches 29 and 30 of the 25 kHz FB. the granularity band, and chosen channels, an
Similarly, the next channel in the according to the application amplitude-complementary prototype
multiplexed spectrum, a TEDS 100 kHz requirements. filter is required [19].
channel, is selected as branch 9 of the Non-uniform, recombined FBs have Because the GDFT-FB outputs are
third (100 kHz) FB. been proposed in literature on audio already oversampled by a factor L, the
In the parallel GDFT-FB, the design and speech processing [16]. In these interpolation factor can be M =R/L.
can be critically decimated or applications, the FBs are critically Hence, the frequency and phase shifts
oversampled. The critically decimated decimated, so a perfect reconstruction are, respectively,
design has more aliasing from adjacent algorithm with parameter optimization is
bands, whereas an oversampled needed to cancel the resulting aliasing. βr = π+ π (2r +1) for r = 0, ..., R -1 (7)
R
design has a higher computational However, because of the asymmetric
load. The level of aliasing interference configuration shown in Fig. 1, such an φr = -( MN + NM ) βr for r = 0, ..., R -1 (8)
that can be tolerated depends on the algorithm is not possible. 2D 2D
radio standard specifications. Oversampling ensures that aliasing in where NM is the order of the
Parallel configuration is not the best the transition bands of the sub-band interpolation filter, and N is the order of
solution when the channel spectrum is bandpass filters is less than that in the the GDFT-FB prototype filter. To
not allocated according to (1), that is, critically sampled case. Non-uniform improve the structure in Fig. 4(b), the
when the possible channel center channelization using recombined, phase shift, φr , can occur before the
frequencies are not constrained to be a oversampled FBs has been proposed interpolation. This means the phase
multiple of the channel spacing. In this in [4], [17], [18]. Recombination is shift is carried out at a lower sample
case, additional GDFT-FBs must be carried out by the structure shown in rate than the anti-image filtering and
added to the parallel structure so that Fig. 4(b). A recombined signal, denoted frequency shift.
each center frequency can be covered Y k,R (z ), is formed by R contiguous The FB is designed to cover an uplink
by at least one GDFT-FB. Even with sub-bands allocated from the k th frequency band of 5 MHz and deliver
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P22
Practical Non-Uniform Channelization for Multistandard Base Stations Special Topic
Álvaro Palomo Navarro, Rudi Villing, and Ronan J. Farrell
GDFT-FB Recomb.
k +R GDFT-FB is best for such as the radix-2 and radix-4, are
y k +R +1 (n ) schemes with few possible preferred for practical implementation
y k +R +2 (n ) channel bandwidths and on FPGAs [21].
alignment patterns; for For the parallel GDFT-FB, the
…
yk+(R -1)(n ) e jβ n
R -1
e jφ
R -1
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P23
Special Topic Practical Non-Uniform Channelization for Multistandard Base Stations
Álvaro Palomo Navarro, Rudi Villing, and Ronan J. Farrell
specifications. This aliasing is the filter transfer function [25]. In an FIR Component filters have more relaxed
especially significant in the critically filter, these changes can lead to specifications than the original filter.
sampled case [14]. deviations in the magnitude response Therefore, the number of coefficients in
To approximate the order of the (particularly the stop-band each component filter is smaller (often
different filters, the following equation is attenuation), and this renders the filter much smaller) than the original filter.
used [24]: unsuitable for certain applications. The The design of each component filter is
problem is greater with high-order also simplified, and the total
-20log10( δpδs )-13
N≈ . (13) filters because quantization errors computational load is often less than
14.6[(ω stop -ω pass )/2π]
affect the position of filter zeros in the the original filter.
The filter order is a function of the z-domain, and the distance between Here, we show that multistage
passband ripple, δp , stop-band zeros is reduced. Therefore, the filtering can be extended to modulated
attenuation, δs , and normalized frequency response of high-order FBs and, in particular, to the GDFT-FB.
transition bandwidth (ω stop , - ω pass) filters is more sensitive to small In FB literature, the term multistage
where ω = 2πf /fs .Consequently, for the changes in the zero positions than the typically refers to a structure in which
same values of δp , δs , f pass, and f stop , the frequency response of low-order filters. multiple FB stages are cascaded to
sample frequency, fs , determines the form a complete FB [5], [27]. In the
normalized transition band and filter approach presented here, the
order. For channelizers with a large
4 Modified GDFT-FB for multistage technique is applied to the
number of channels, the sample Filter Coefficient Reduction prototype filter of only one FB.
frequency of the wideband signal is Although the parallel GDFT-FB and There are various ways in which the
much larger than the normalized recombined GDFT-FB are flexible and multistage technique could be applied
transition band, and this leads to very more efficient than other structures, to the prototype filter. The specification
high orders. high filter orders are still required for of the prototype filter, H (z ), and its
When applied to the TETRA and TETRA/TEDS. For high-order polyphase components, is relaxed, and
TEDS standards, the 25 kHz prototype channelization structures and, more a half-band filter,HB (z ), is applied to
filter for both channelizer structures has generally, in communication every sub-band output to obtain the
an order of 8085 taps. For the parallel applications, FIR filters are chosen original filtering specifications. This
GDFT-FB, the required prototype filter because of their linear phase response. design is shown in Fig. 5.
orders for the 50 kHz and 100 kHz FBs To achieve a linear phase response in In the single-stage design, the
are 3584 and 1444 taps, respectively. an FIR filter of length N +1, where N is prototype filter transition band is narrow
However, these filter orders are only the order, (N +1)/2 coefficients and centered at π/K. In the multistage
theoretical; the actual filter orders may contribute to the magnitude response of design, the prototype filter transition
be higher [4] because interchannel the filter, and the other half of the band is shifted so that it starts at π/K
interference and aliasing produced by coefficients provide the linear phase and its width is increased so that it
decimation in the analysis bank causes property [22]. As a consequence, an extends to 2π/K and includes
the filters’frequency response to FIR filter design generally requires a frequency components from adjacent
deteriorate [14]. As the number of larger number of coefficients compared channels. The half-band filter on each
channels increases, aliasing also with other designs. For applications sub-band output provides the original
increases, and additional filter where perfect linear phase response is sharp transition band and also
overdesign is required. not required, a minimum-phase FIR or eliminates the undesired frequency
Each GDFT-FB requires only one infinite impulse response (IIR) filter can components from the adjacent
prototype filter design; however, in the provide a functionally equivalent channels passed by the relaxed
case of TETRA/TEDS, the large filter magnitude response with a reduced prototype filter. This filtering process is
orders make designing the filter and number of filter coefficients. shown in Fig. 6.
implementing the corresponding When linear phase response is For a multistage GDFT-FB, the FB
channelizer structures impractical. required, multistage filtering [26] is a cannot be critically decimated, but it
Large filter orders can be expected for useful technique that can be applied to has to be oversampled by 2, that is,
other radio standards with similar FIR filter design to reduce the total K = 2D. Oversampling prevents the
specifications. number of filter coefficients. Multistage undesired adjacent-channel
This impracticality arises because the filtering is most commonly applied to frequencies from aliasing with the
efficient implementation of a filter in a interpolators and decimators that have channel of interest. Also, oversampling
reconfigurable hardware platform large sample rate conversion factors. In by 2 allows the use of half-band filters.
typically requires fixed-point a multistage filter design, an original This is attractive because half-band
representation of filter coefficients. filter is factorized into multiple filters have desirable impulse and
Coefficients must be quantized to the component filters which, when frequency response properties in
word length of the device, and the cascaded, produce the original filter Fig. 7(a). Half-band filters have a
resulting quantization error may change magnitude and phase response. symmetric impulse response where
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P24
Practical Non-Uniform Channelization for Multistandard Base Stations Special Topic
Álvaro Palomo Navarro, Rudi Villing, and Ronan J. Farrell
x (n ) y 0(n ) z 0(n )
• D E 0(z LWK -k K )
0 HB (z ) complex input for the channelizer is
now determined by
z -1 WK -k nD WK
0 -(1+k )n 0 0
y 1(n ) z 1(n ) L K
D E 1(z LWK -k K ) HB (z ) μM -GDFT = [4(N +1)+4( (log2(K )-1))
K 2
0
(N +1)
z -1 +4K+2K B ] (14)
WK -k nD WK
-(2+k )n
8
0 0 0
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P25
Special Topic Practical Non-Uniform Channelization for Multistandard Base Stations
Álvaro Palomo Navarro, Rudi Villing, and Ronan J. Farrell
▼Table 1. Filter orders calculated for TETRA/TEDS non-uniform channelizers using multistage equivalent parallel GDFT-FB in every
GDFT-FB situation. The reason for this is that
GDFT-FB Output Single-Stage Number multistage Number multistage filtering has a greater effect
Sub-Band Bandwidth (kHz) of Coefficients of Coefficients on the recombined GDFT-FB.
25 N = 8085 N = 1294, N B = 64 For the multistage channelizers, the
Parallel GDFT-FB 50 N = 3584 N = 595, N B = 58 filter orders were again obtained using
theoretical calculations (section 3.3).
100 N = 1444 N = 291, N B = 46
However, real implementations require
Recombined GDFT-FB 25 N = 8085 N = 1294, N B = 64 higher filter orders because of
GDFT-FB: generalized discrete Fourier transform filter bank undesired effects, such as aliasing, in
the FBs (Fig. 9). In Fig. 9, a channelizer
▼Table 2. Number of coefficients and computational load of the TETRA/TEDS non-uniform for eight 25 kHz channels was
multistage channelizers compared with single-stage design designed using the single-stage and
multistage multistage multistage GDFT-FBs. The theoretical
Parallel GDFT-FB Recombined GDFT-FB order for the single-stage prototype
Percentage of Coefficients Compared with Single-Stage Design 16.7% 16.0% filter was 253, whereas that for the
Percentage of Real Multiplications Compared with Single-Stage Design 70.6% 35.3%
multistage prototype filter was 42, and
the half-band filter order was 64. All
Percentage of Real Additions Compared with Single-Stage Design 92.3% 45.0%
filters were designed as optimal
GDFT-FB: generalized discrete Fourier transform filter bank equiripple filters.
Because of aliasing, the magnitude
two non-uniform channelization The test results are shown in Fig. 8. response does not exhibit 55 dB of
methods are used here. The theoretical The parallel GDFT-FB has a constant attenuation in the stopband for the
filter orders are calculated again using computational load that is independent single-stage design. For the multistage
(13). Even though the multistage design of the sub-band allocation pattern. On design, the cascade of the two filters
has two filters instead of one, it has two the other hand, the computational load produces decay in the stopband that
main advantages: its prototype filter is of the recombined GDFT-FB varies provides the desired attenuation in
designed with a much larger transition according to the number and type of most of the stopband. In the passband,
band, and the half-band filter is recombined channels. The load is at a both filters provide a ripple within the
designed for a much lower sample rate. minimum when the whole frequency 0.1 dB limits. To conform to the required
Both factors contribute to reducing the band is covered by 25 kHz channels specifications, an overdesign of the
filter orders shown in Table 1. because no recombining is required. ideal prototype filter is needed to
A reduction in the number of The load is at a maximum when the compensate for the aliasing. This
coefficients also implies a reduction in frequency band is fully occupied by usually leads to filter orders that are
the number of operations per complex TEDS 100 kHz channels; that is, larger than the theoretical orders,
input sample (Table 2). For the recombination is applied to every especially in the single-stage design.
multistage recombined GDFT-FB in GDFT-FB output sub-band. In contrast Optimization techniques are outside the
Table 2, the figures represent only the to the results obtained in [4], the scope of this paper, but
GDFT-FB part of the channelizer. The recombined GDFT-FB using multistage Kaiser-Window filters could possibly
structure used to recombine the TEDS filtering is more efficient than the be used instead of optimal equiripple
50 kHz and 100 kHz channels is the
same for both single and multistage 280
cases and is not included for that 260
reason. The saving on multistage
Real Multiplications per Input Sample
: Parallel FB
filtering computational load is greatest 240 : Recombined FB 25 kHz only
: Recombined FB 50 kHz only
in the recombined GDFT-FB. 220 : Recombined FB 50:50 split
The performance of the two : Recombined 100 kHz only
non-uniform channelizers can be 200
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P26
Practical Non-Uniform Channelization for Multistandard Base Stations Special Topic
Álvaro Palomo Navarro, Rudi Villing, and Ronan J. Farrell
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P27
Special Topic Practical Non-Uniform Channelization for Multistandard Base Stations
Álvaro Palomo Navarro, Rudi Villing, and Ronan J. Farrell
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P28
Crest Factor Reduction for OFDM Using Selective Subcarrier Degradation Special Topic
R. Neil Braithwaite
Abstract: This paper describes a crest factor reduction (CFR) method that reduces peaks in the time domain by modifying selected data
subcarriers within an OFDM signal. The data subcarriers selected for modification vary with each symbol interval and are limited to those
subcarriers whose data elements are mapped onto the outer boundary of the constellation. In the proposed method, a set of peaks are
identified within an OFDM symbol interval. Data subcarriers whose data element has a positive or negative correlation with the set peak
are selected. For a subcarrier with an outer element and a significant positive correlation, a bit error (reversal) is intentionally introduced.
This moves the data element to the opposite side of the constellation. Outer elements on negatively-correlated subcarriers are increased
in magnitude along the real or imaginary axis. Experimental results show that selecting the correct subcarriers for bit reversals and
outward enhancements reduces the peak-to-average power ratio (PAPR) of the OFDM signal to a target value and limits in-band
degradation measured by bit error rate (BER) and error vector magnitude (EVM).
Keywords: crest factor reduction; OFDM; PAPR; wireless communication system; digital transmitter
O
rthogonal frequency-division (WiMAX) 802.16 [1]. This peak power is measured by the BER and error
multiplexing (OFDM) waveforms bound allows the power amplifier (PA) vector magnitude (EVM).
can have large in the transmitter to be designed for
peak-to-average power ratios greater power-added efficiency (PAE) 1.1 OFDM Transmit Overview
(PAPR). Crest factor reduction (CFR) [2]. It may be necessary to enforce a Creation of an OFDM symbol for
reduces peaks at the expense of signal second PAPR bound if the dynamic radio frequency (RF) transmission is
quality. Usually, degradation is range of the digital-to-analog shown in Fig. 1. The symbol is part of a
distributed throughout the signal’ s converter (DAC) cannot prevent data stream that has been 1) encoded,
frequency bandwidth in the form of clipping for all possible signals. 2) modulated, 3) converted from serial
spectral leakage and in-band errors. The remainder of this introduction to parallel as a 256 sample block, 4)
However, because the OFDM signal is describes OFDM signal generation as converted to the time domain using an
created in the Fourier domain, well as the subcarrier phase alignments inverse fast Fourier transform (IFFT), 5)
degradation can be concentrated to that create peaks. Section 2 contains a extended using a cyclic prefix (CP) to
specific subcarriers (frequency bins of review of CFR methods. Section 3 256 + CP samples, and 6) converted
the FFT). introduces new CFR methods for back into a serial data stream. The data
In the proposed method, CFR is used intentionally degrading selected stream is then converted from DAC,
sparingly to bound the PAPR when subcarriers whose data elements are lowpass filtered, up-converted to RF,
transmitting at high power levels. This correlated (phase aligned) with the then amplified by a PA. The signals Y(k)
differs from the usual approach of peaks of the OFDM signal. Section 4 and y (n) are important for the crest
maximizing PAPR reduction. The bound contains experimental results that show factor method presented in this paper.
is met, and bit error rate (BER) is limited a WiMAX signal can be crest-factor These symbols correspond to the
without deviating from the OFDM reduced to a target PAPR by Fourier domain and time domain data
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P29
Special Topic Crest Factor Reduction for OFDM Using Selective Subcarrier Degradation
R. Neil Braithwaite
To Digital
QPSK, it varies between constellation
Data
Y (k ) y (n ) Transmitter elements for 16-QAM and 64-QAM.
Modulator:
Input
Encoder
1 BPSK, QPSK, 1 Serial 256
to IFFT
256 Append 256+CP Parallel
to
1 Higher magnitudes are found at the
16-QAM, or Cyclic
64-QAM Parallel Prefix Serial outer elements of the 16-QAM and
64-QAM constellations. As a result, it
can be assumed that many of these
outer constellation elements are
Digital
Transmitter DAC Lowpass PA
RF present in the data subcarriers when a
Filter Transmission
Input large peak appears in the time domain.
LO
BPSK: binary phase-shift keying IFFT: inverse fast Fourier transform QAM: quadrature amplitude modulation 2 Crest Factor Reduction
CP: cyclic prefix LO: Local oscillator QPSK: quadrature phase-shift keying Large peaks cause problems
DAC: digital-to-analog converter PA: power amplifier
because PAs become less efficient as
▲Figure 1. RF transmission of one OFDM symbol [3]. the PAPR of the RF signal increases.
Limiting the PAPR is necessary for a
blocks, respectively, of the OFDM down-link (DL) subframe. This more efficient transmitter design. This
symbol. subframe comprises a preamble, frame process is called CFR. Past CFR
The encoding includes coders and control header (FCH), and DL bursts. methods for OFDM signals include clip
interleavers to allow for error correction QPSK is used for the preamble and and filter, partial transmit sequence
at the receiver. The modulator allows BPSK for the FCH. BPSK, QPSK, (PTS), selective mapping (SLM), tone
rate changes, and the rate is selected 16-QAM, or 64-QAM (except for the reservation, and constellation
according to the received BPSK pilots) is used for the DL bursts. extension. The following is a summary
signal-to-noise ratio (SNR) and BER at The preamble and FCH are sent first, of these CFR methods and their
the receiver. The rates from lowest to and the DL bursts are sent in order of suitability for OFDM signals.
highest are binary phase-shift keying the modulation rates (lower rates are The direct CFR method involves
(BPSK), quadrature phase-shift keying sent first). clipping peaks of the time signal y (n )
(QPSK), 16-quadrature amplitude The OFDM symbol at the output is a when they exceed a specified level, L.
modulation (16-QAM), and 64-QAM. time-domain data sequence. Although The clipped signal y clip(n ) is
In the serial-to-parallel conversion, a the individual data subcarriers are
frequency domain representation is transmitted using simple modulation y (n )
L· for y (n ) >L
created, and the data elements are mappings, the magnitude in the time y clip(n ) = y (n ) (2)
assigned to different subcarriers of the domain varies significantly. This is due y (n ) otherwise.
OFDM signal. The IFFT transforms the to the IFFT operation that forms each
data from the frequency domain to the time-sample from a sum of 200 The excess peak waveform
time domain. The cyclic prefix is a copy random-phase variables (56 of the 256 y peaks(n ) becomes
of the tail of the time domain block and subcarriers are null subcarriers). Phase
is appended to the beginning of the alignment of subcarriers in the y peaks(n ) = y (n ) - y clip(n ). (3)
time domain block. The cyclic prefix frequency domain results in large Clipping moves the constellation
protects against intersymbol peaks in the time domain. elements from their assigned positions.
interference (ISI) caused by multipath Peak-forming phase alignment in the The difference between the actual and
RF propagation. frequency domain varies according to assigned positions in the IQ space is
Within an OFDM symbol, the the position of the peak after the IFFT called the constellation error or EVM.
subcarriers may be data subcarriers, within the time block. A peak at time The allowable relative constellation
pilot subcarriers, or null subcarriers. No t peak within the interval t = [0, 255] is error for WiMAX, averaged over
data elements are mapped onto the null maximized by the following subcarrier subcarriers, frames, and packets,
subcarriers, which include the outer phases: depends on the rate modulation where
guard-band frequencies and the DC the most difficult specification is -31.0
θ align(t_peak)(k ) = -k·Δω·t peak +θ (t peak) (1)
subcarrier. The pilots are dB for the 3/4 rate 64-QAM. Clipping
2π
BPSK-modulated and assigned to where Δω = (N = 256), k is the tends to distribute excess peak energy
specific subcarriers. The remaining
N
over the 256 subcarriers, including the
subcarriers are used for data subcarrier frequency index (DC = 0), null and pilot subcarriers. As well as
transmission, which may be modulated and θ (t peak) is the phase of the complex EVM limits on the data subcarriers,
using BPSK, QPSK, 16-QAM, or time sample at t peak. there are also limits on spurious
64-QAM. The magnitude of the subcarriers emissions and adjacent channel
The OFDM-transmitted signal is a also has an effect on peaking. Although leakage ratio (ACLR). These limits are
sequence of symbols sent as a the magnitude is constant for BPSK and specifically for the null subcarriers
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P30
Crest Factor Reduction for OFDM Using Selective Subcarrier Degradation Special Topic
R. Neil Braithwaite
within the guard-band frequencies. that points on the outer boundary of the is used so that opposing points on the
The energy from the clipped peaks constellation have no neighbors in the constellation, dIQ and -dIQ, represent the
must be constrained to reside primarily outward-expanding direction. As a same data at the receiver. The
on the in-band subcarriers. result, EVM need not be limited in the advantage of constellation extension is
Filtering attenuates the excess peak outward direction because it does not that π radian phase shifts can be
energy present in the null subcarriers cause bit errors at the receiver. introduced onto some of the subcarriers
(except for the DC subcarrier). That is, Outward expansion is also used in the with the goal of disrupting
proposed CFR and is discussed in peak-forming phase alignments. The
Δy (n ) = Στ y peaks(n -τ)·h (τ) (4)
section 3. downside of this method is that one bit
where h (τ ) is a filter kernel. The In a different class of CFR methods, is lost in the constellation mapping,
crest-factor-reduced signal being CFR disrupts the subcarrier phase which reduces throughput for QPSK,
transmitted, denoted y CFR(n ), becomes alignments that create peaks. These 16-QAM, and 64-QAM to 1/2, 3/4, and
methods include partial transmit 5/6 of the original value, respectively.
y CFR(n ) = y (n ) - α·Δy (n ) (5) sequence (PTS) [9]-[11] and selective Also, the Gray code mapping specified
where α is a scaling term used to mapping (SLM) [12]. The subcarriers in the standard must be abandoned.
control the EVM introduced by the CFR are multiplied by a set of different The proposed CFR method similarly
[4], [5]. In some clip and filter phase-shift vectors that produce a set introduces large phase shifts onto
implementations [6], the excess peak of potential time sequences. The time selected subcarriers; however, the
waveform y peaks(n ) is replaced by a sequence with the lowest PAPR is phase shifts are applied to the standard
sequence of delta functions. Each delta transmitted. Information about the Gray code mapping instead of using
function is located at a peak and phase-shift vector used must be sent the redundant constellation mapping
assigned a magnitude and phase that to the receiver to allow demodulation. already described.
match the excess value of the peak. This is considered a disadvantage of
This sequence is filtered using (4). the method. A PTS method specifically
Using α in (5) for controlling EVM of for LTE and not requiring the 3 Proposed CFR Method
OFDM signals is not effective because phase-shift vector to be sent to the The proposed CFR method does not
the EVM limit varies according to the receiver is described in [13]. Neither alter the WiMAX (or other OFDM)
function of the subcarrier and the rate PTS nor SLM is used in the proposed standard nor does it require additional
modulation of the data (BPSK, QPSK, CFR method. information to be sent to the receiver.
16-QAM, or 64-QAM). It is beneficial A third CFR method involves The accuracy of the null and pilot
for the receiver to have accurate pilot reserving some of the subcarriers as subcarriers is preserved by restricting
information, which means the peak-reducers; that is, the reserved constellation errors to selected data
CFR-induced EVM should be zero for subcarriers do not carry any data. This subcarriers. CFR modifications are
pilot subcarriers. Multiaccess OFDM, method is called tone reservation [14], applied only to subcarriers whose data
such as OFDMA and LTE, may have [15]. Once a peak is detected in the elements are mapped onto the outer
different rate modulations on different time domain, the magnitudes and positions of the constellation. For BPSK
subcarrier blocks. Because the phases of the reserved subcarriers are and QPSK, the outer elements include
allowable EVM is larger for BPSK and selected to reduce the peak. the entire constellation. For 16-QAM
QPSK than for 16-QAM and 64-QAM, Unfortunately, data throughput is and 64-QAM, there are 12 (of 16) and
some researchers [7] convert the reduced because fewer data 30 (of 64) outer elements, respectively.
excess peak signal back into the subcarriers are available. It is possible The outer elements have special
frequency domain and apply the to use tone reservation as a form of clip properties in terms of constellation
filtering and rate-dependent EVM and filter with EVM control. The EVM is errors and BER. These properties are
control by weighting the subcarriers. set to zero for all data subcarriers that exploited in the proposed CFR method.
The result is then converted back into are not reserved for peak reduction. In a 16-QAM constellation, the IQ
the time domain to produce the desired The scale factor, α , may have to be mapping and decision boundaries
Δy (n ) and y CFR(n ). increased above unity for such an used by the receiver define the
The EVM control for a data subcarrier implementation in order to compensate relationship between constellation
depends on the modulation type. CFR for the sparseness of the reserved errors and BER. The 16-QAM mapping
attempts to introduce EVM on data subcarriers in the frequency domain. is a Gray code (Fig. 2). Also shown in
subcarriers to reduce peaks without Reserved tones are not used in the Fig. 2 are the decision boundaries of
increasing the BER at the receiver. proposed CFR method. the receiver. It is of interest to determine
Thus, QPSK allows more EVM than A fourth CFR method involves altering a) the largest constellation error that
64-QAM because the distance the constellation mapping so that can be tolerated without causing a bit
between neighboring points in the elements are not unique. This is error, and b) the largest constellation
constellation is greater. However, some referred to as constellation extension error caused by a single bit reversal.
researchers [8] have exploited the fact [16]. A redundant mapping of input bits The number of bit reversals between
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P31
Special Topic Crest Factor Reduction for OFDM Using Selective Subcarrier Degradation
R. Neil Braithwaite
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P32
Crest Factor Reduction for OFDM Using Selective Subcarrier Degradation Special Topic
R. Neil Braithwaite
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P33
Special Topic Crest Factor Reduction for OFDM Using Selective Subcarrier Degradation
R. Neil Braithwaite
4 Results CFR: crest factor reduction CSC: correlated subcarrier components IFFT: inverse fast Fourier transform
The CFR approach in Fig. 4 is
applied to a WiMAX DL subframe ▲Figure 6. Serial implementation of the correlated subcarrier component search, suitable for
comprising two QSPK symbols for the CFR using bit reversals only [3].
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P34
Crest Factor Reduction for OFDM Using Selective Subcarrier Degradation Special Topic
R. Neil Braithwaite
-0.5 1730460 bits sent). The EVM, including communications systems,”US patent 7,095,798,
Aug. 22, 2006.
-1.0 the bit reversals, is 0.0725, which is [6] G. A. Awater, R. D. de Wild, A. Hendrik,
-1.5
high for 64-QAM. The limit for 3/4 rate “Transmission system and method employing peak
cancellation to reduce the peak-to-average power
-2.0
64-QAM is 0.0282. However, when ratio,”US patent 6,175,551, Jan. 16, 2001.
Original using the easier measure that excludes [7] M. J. Hunton,“OFDM communications system
-2.5
β CCDF the contribution from bit reversals, EVM employing crest factor reduction with ISI control,”
-3.0 US patent application 20070058743, March 15,
is 0.0119. This EVM level is considered 2007.
-3.5
acceptable. Most of the EVM for this [8] B. S. Krongold and D. L. Jones,“PAR reduction in
-4.0 CFR OFDM via active constellation extension,”IEEE
CCDF second measure is due to the outward Trans. Broadcasting, vol. 49, no. 3, pp. 258-268,
-4.5 expansion of the constellation, which Sept. 2003.
-5.0 does not increase BER at the receiver. [9] S. H. Han and J. H. Lee,“PAPR reduction of OFDM
2 3 4 5 6 7 8 9 10 11 signals using a reduced complexity PTS
PAPR (dB) This approach does not involve trying technique,”IEEE Signal Proc. Lett., vol. 10, no. 11,
CCDF: complementary cumulative distribution function to achieve the lowest CFR. Instead, a pp. 887-890, Nov. 2004.
[10] L. J. Cimini, Jr. and N. R. Sollenberger,“OFDM
CFR: crest factor reduction bounded CFR is created to ease the communication system and method having a
PAPR: peak-to-average power ratio design of the PA and digital circuitry reduced peak-to-average power ratio,”US
and to generate a low BER. With this patent 6928084, Aug. 9, 2005.
[11] A. Alavi and C. Tellambura,“PAPR reduction of
▲Figure 7. CCDF for the original and crest type of CFR, 90% of the CFR symbols OFDM signals using partial transmit sequence: an
factor reduced WiMAX time waveforms. are transmitted without modification. optimal approach using sphere decoding,”IEEE
The vertical axis represents the 10β Commun. Lett., vol. 9, no. 11, pp. 982-984, Nov.
probability that the signal power |y (n )|2 For such pass-through cases, only the 2005.
exceeds the level Po = PAPR*Pave, where PAPR measurement is required, and no [12] C.-L. Wang and Y. Ouyang,“Low-complexity
selected mapping schemes for peak-to-average
Pave is the average signal power. additional IFFTs are computed. power ratio reduction in OFDM systems,”IEEE
Trans. Signal Processing, vol. 53, no. 12, pp.
after filtering. In practice, however, the 4652-4660, Dec. 2004.
[13] R. N. Braithwaite,“Crest factor reduction for
PAPR is defined using the digital time 5 Conclusion down-link LTE by transmitting phase shifted
domain signal y (n ) and is based on the A CFR method that is suitable for resource blocks without side information,”
European Wireless Conf., EuWIT 2009, Rome, Italy,
signal power statistics rather than the WiMAX and other OFDM signals has pp. 13-16.
absolute peak. The practical peak is the been presented. Degradation [14] S. Hosokawa, S. Ohno, K. D. Teo, and T.
level Po , for which the signal power associated with CFR is restricted to Hinamoto,“Pilot tone design for peak-to-average
power ratio reduction in OFDM,”IEEE Int. Sym.
y (n ) 2 has a 10β probability of selected data subcarriers whose data Circ. Syst. (ISCAS 2005), Kobe, Japan, pp.
exceeding this peak. The elements are mapped onto the outer 6014-6017.
[15] L. Guan and A. Zhu,“Gaussian pulse-based
complementary cumulative distribution boundary of the IQ constellation. two-threshold parallel scaling tone reservation for
function (CCDF) of y (n ) 2 , in which β is Subcarrier components correlated to PAPR reduction of OFDM signals,’Int. J. Digital
plotted as a function of Po, is a useful the phase profiles of the primary and Multimedia Broadcasting, vol. 2011, Article ID
470310, 9 pages, 2011. doi:10.1155/2011/470310.
description of the signal. In this paper, secondary peaks are identified for [16] Y. J. Kou, W.-S. Lu, and A. Antoniou,
two probability thresholds are selected modification by outward expansions “Peak-to-average power ratio reduction
algorithms for OFDM systems via constellation
for defining the signal peak: β= -4 and bit reversals. The extension,”IEEE Int. Sym. Circ. Syst. (ISCAS
and -5. peak-to-average power for the WiMAX 2005), Kobe, Japan, pp. 2615-2618.
The CCDFs of the original and signal is reduced to a target level, and [17] R. N. Braithwaite,“General principles and design
overview of digital predistortion,”chapter in Digital
crest-factor-reduced OFDM time this allows the PA in the transmitter to Processing for Front End in Wireless
sequences are shown in Fig. 7. Using be designed for high PAE. CFR is Communication and Broadcasting, F. Luo (Ed.),
Cambridge Univ. Press, 2011.
the 10-4 probability threshold for the achieved with an acceptable amount of
CFR OFDM time sequence, PAPR is in-band BER and EVM degradation.
8.42 dB, which is a reduction of 1.2 dB
from the 9.62 dB PAPR of the original
signal. Using the 10-5 probability
References Biography
threshold, PAPR of the CFR and original [1] IEEE 802.16-2009, IEEE Standard for Local and
OFDM time sequences are 8.56 dB and Metropolitan Area Networks Part 16: Air Interface for R. Neil Braithwaite (nbraithwaite@pwav.com)
Broadband Wireless Access Systems, received his B.Sc. degree in electrical engineering
10.54 dB, respectively. That is, CFR from the University of Calgary in 1985. He received
P802.16Rev2/D9, Jan. 2009.
reduces PAPR by 2 dB when the 10-5 [2] S. C. Cripps, RF Power Amplifiers for Wireless his M.Sc. and Ph.D. degrees from the University of
British Columbia in 1989 and 1992. From 1992 to
threshold is used. The Communications, Norwood, MA: Artech House,
1999. 1995, he conducted postdoctoral research at the
crest-factor-reduced waveform meets [3] R. N. Braithwaite,“Crest factor reduction system University of California, Riverside. From 1985 to
the target PAPR of 8.5 dB for the 10-4 and method for OFDM transmission systems using 1987 and 1995 to 2002, he worked for Computing
selective sub-carrier degradation,”US patent Devices Company (Canada), Nortel (Canada), and
probability threshold, which is used application 20070140367, June 21, 2007. Agilent Laboratories (USA). Since 2002, he has
more commonly in PA design than the [4] M. J. Hunton,“System and method for peak power been working for Powerwave Technologies (USA).
He is the author of several papers and patents, as
10-5 threshold. reduction in spread spectrum communications
well as a recent book chapter on digital
systems,”US patent 6449362, Sept. 10, 2002.
The BER introduced by the CFR is [5] M. J. Hunton,“System and method for post filtering predistortion in RF power amplifiers [17].
0.00019 (327 bit reversals from peak power reduction in multi-carrier
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P35
Special Topic An Antenna Diversity Scheme for Digital Front-End with OFDM Technology
Fa-Long Luo, Ward Williams, and Bruce Gladstone
Abstract: In this paper, we propose a new antenna diversity scheme for OFDM-based wireless communication and digital broadcasting
applications. Compared with existing schemes, such as post-fast Fourier transform (FFT), pre-FFT, and polyphase-based filter-bank,
the proposed scheme performs optimally and has very low computational complexity. It offers a better compromise between
performance, power consumption, and complexity in real-time implementation of the receivers of broadband communication and digital
broadcasting systems.
I
n modern wireless broadband such as one antenna’ s output and then give 5-10% gain in bit-error reduction
communication and digital forwarded for further FFT processing,
broadcasting, orthogonal which is required in any OFDM-based x 1(n )
w 1(n )
Z (1)
frequency-division multiplexing receiver. These weights are designed
Z (2)
(OFDM)-based modulation schemes to meet an optimization criterion, and x 2( n )
w 2( n ) +
y (n )
FFT
are usually used. However, an OFDM the most common criterion is the Z (N )
system has very poor reception when maximum ratio combination (MRC).
…
there is noise, interference, or moving With MRC, the optimized weight vector, xM (n )
wM (n )
objects. To solve this problem, many W (n ) = [W1(n ), W2(n ), ..., WM (n )], can
FFT: fast Fourier transform
technologies have been proposed and be obtained by
used in real applications [1], [2]. W opt (n ) = E [X (n )×Y (n )] (1) ▲Figure 1. Pre-FFT scheme [1].
Among these technologies,
beamforming-based diversity
technology is the most promising. It Filter
Analysis
uses multiple antennas at the receiver x 1(n ) Bank 1
side and spatial filtering to optimize Z (1)
reception in noisy and mobile Filter y (n ) Z (2)
Analysis Synthesis FFT
environments. Figs. 1, 2, and 3 show x 2( n ) Bank 2
Z (N )
three representative solutions: pre-fast
Fourier transform (pre-FFT), polyphase
filter-bank, and post-FFT [1], [4], [5],
…
[7], [8].
In Fig. 1, X1(n ), X2(n ), ... , XM (n ) are
the received signals in one of M Figure 2.▶ Analysis
Filter
xM (n ) Bank L
antennas, and W1(n ), W2(n ), ... , WM (n ) Polyphase bank FFT: fast Fourier transform
are the weights applied to these scheme.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P36
An Antenna Diversity Scheme for Digital Front-End with OFDM Technology Special Topic
Fa-Long Luo, Ward Williams, and Bruce Gladstone
Multiplications with Weights for scheme that provides a Now we determine the N +M weights
Different Frequency Bins better compromise in proposed scheme.
between performance, WE, X h, and X t denote the weight
FFT complexity, and power vector of the eigenfiltering, the
x 1(n )
+
Z (1) consumption is highly sequence of the head-guided interval,
desired. and the tail of the symbols,
respectively. If the norm is a constant
(unity, for instance), the weight vector is
FFT +
Z (2) 2 The Proposed obtained in such a way that the error is
x 2( n )
Scheme minimized. That is,
Here, we propose a
new scheme in which minWE E‖W HX h - W HX t‖2
…
Z (N )
+ the number of variables St.‖W H W‖2 = 1. (3)
computed is reduced According to beamforming matrix
xM (n )
FFT from N ×M (in post-FFT) theory [3], the solution to the
to N+M. The scheme optimization problem is the eigenvector
FFT: fast Fourier transform shown in Fig. 4 performs (minor component) corresponding to
just as well as the the smallest eigenvalue of the
▲Figure 3. Post-FFT scheme [1]. post-FFT scheme. correlation matrix R, which is
M antenna signals are weighted by
and is not complex in implementation. the corresponding weight as the E [(Xh -Xt ) (Xh -Xt )H ]. (4)
However, this processing is done only processing made in the pre-FFT This is also the reason that the
in the time domain, and the same algorithm in Fig. 1. These M weighted filtering in Fig. 4 is called eigenfiltering.
weights are used over the entire signals are added together to pass In practical implementation, the minor
frequency band (all-carriers). Because through an N-tap finite impulse component in the following adaptive
of the shortcomings of simple pre-FFT, response (FIR) eigenfilter, after which algorithm can be updated:
a polyphase filter bank scheme is they are input for FFT processing. The
proposed (Fig. 2). number of unknown variables in the W (t +1) =W (t ) + γs (t ) (X h -Xt -s (t )W (t ))
In a polyphase scheme, each proposed scheme is N +M, which is s (t ) = W H(t ) (X h -X t ) (5)
antenna signal is divided into a number significantly less than the N ×M needed where γ is a constant. This algorithm
of frequency band signals, and different in post-FFT. Furthermore, only one FFT is as complex as LMS algorithm, and
weights apply to different bands. This operation is needed in the proposed the required multiplications are only on
scheme can further improve scheme instead of M FFT operations in the order of N. An algorithm, such as
performance at the cost of increased the post-FFT scheme. Here we will MRC used in pre-FFT, can be used to
computational complexity that arises as prove that the post-FFT scheme in Fig. determine the weights applied to each
a result of multichannel antenna signal 3 and the proposed scheme in Fig. 4 antenna before eigenfiltering.
decomposition (analysis) and give the same frequency-domain
multiplication of weights in each outputs, Z (1), Z (2),...,Z (N ). Using the
channel. Different weights in different optimization criterion in post-FFT, the
3 Comparison and
frequency bins and bands can be used variable matrix, WN ×M, in Fig. 3 can be Discussion
in the frequency domain, and this is rewritten as Compared with existing solutions, the
called post-FFT (Fig. 3).
WN ×M = [WN,1, WN,2, ..., WN,M]
A polyphase scheme is a special w (n ) 1
== WC [C 1, C 2, ..., C M] (2)
case of post-FFT. Post-FFT performs x (n ) 1
the best but has the greatest where WN,i is the i th column of the Z (1)
computational complexity. Pre-FFT has N ×M matrix and denotes the w (n ) 2
FIR Z (2)
the least computational complexity but weights after FFT processing of the x (n ) 2
Eigenfiltering
FFT
Z (N )
improves performance the least. i th antenna in Fig. 3. WC is an
Another approach is to divide M N-dimensional vector, and Ci is a
antennas into L groups. Each group scalar (corresponding to each
…
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P37
Special Topic An Antenna Diversity Scheme for Digital Front-End with OFDM Technology
Fa-Long Luo, Ward Williams, and Bruce Gladstone
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P38
A Histogram-Based Static-Error Correction Technique for Flash ADCs Special Topic
Armin Jalili, J Jacob Wikner, Sayed Masoud Sayedi, and Rasoul Dehghani
A Histogram-Based Static-Error
Correction Technique for Flash ADCs
Armin Jalili 1, J Jacob Wikner 2, Sayed Masoud Sayedi 1, and Rasoul Dehghani 1
(1. Department of Electrical and Computer Engineering, Isfahan University of Technology, Isfahan 84156-83111, Iran;
2. Department of Electrical Engineering, Linköping University, SE-581 83 Linköping, Sweden)
Abstract: High-speed, high-accuracy data converters are attractive for use in most RF applications. Such converters allow direct
conversion to occur between the digital baseband and the antenna. However, high speed and high accuracy make the analog
components in a converter more complex, and this complexity causes more power to be dissipated than if a traditional approach were
taken. A static calibration technique for flash analog-to-digital converters (ADCs) is discussed in this paper. The calibration is based on
histogram test methods, and equivalent errors in the flash ADC comparators are estimated in the digital domain without any significant
changes being made to the ADC comparators. In the trimming process, reference voltages are adjusted to compensate for static errors.
Behavioral-level simulations of a moderate-resolution 8-bit flash ADC show that, for typical errors, ADC performance is considerably
improved by the proposed technique. As a result of calibration, the differential nonlinearities (DNLs) are reduced on average from 4 LSB
to 0.5 LSB, and the integral nonlinearities (INLs) are reduced on average from 4.2 LSB to 0.35 LSB. Implementation issues for this
proposed technique are discussed in our subsequent paper,“A Histogram-Based Static-Error Correction Technique for Flash ADCs:
Implementation Aspects.”
Keywords: Calibration; flash ADC; offset; trimming; uniform distribution
A
n analog-to-digital converter discrete elements, such as high-Q technologies. However, this comes at
(ADC) is an essential part of an filters, and lowering the number of the cost of increased digital complexity
RF receiver, but in many analog cascaded stages in because a larger part of the ADC is
systems, an ADC limits down-conversion. Reduction in analog digital rather than analog.
performance. complexity allows the system to be In this paper, we focus on flash ADCs
To relax the dynamic range more integrated and makes it more with low to moderate resolutions, that is,
requirements of an ADC, a voltage gain compatible with a system-on-chip 3 to 8 bits. These resolutions are used
amplifier (VGA) and filters (for (SOC) design. extensively as the sub-ADC part of
adjacent-channel blocking) are used An important factor in ADC design is sigma-delta ADCs, in pipelined ADCs,
before the ADC. Typically, linear analog the power-accuracy trade-off. This is or in applications ultrawideband where
filters with sharp transition bands and more critical in cellular applications, low resolution and very high-speed
good VGA designs are used. Such where portability significantly affects converters are needed. In a standard
filters are difficult to implement, power consumption, especially that of DCS-1800 for a low-IF receiver, a
especially in an integrated system and the converter. A highly dynamic range high-resolution, low-bandwidth
in newer CMOS technologies. The in the converter is sometimes converter is required. Such a converter
design of these blocks should be achievable at the cost of high power can be realized using a sigma-delta
simplified, and more should be consumption. Calibration optimizes the ADC.
required of the ADCs. In today’ s power-accuracy trade-off, and the The accuracy of a flash ADC is
cellular applications, the main desired accuracy can be obtained with mainly affected by the offsets of its
challenge is to reduce analog reduced power consumption. By comparators and references. To
complexity as much as possible and reducing analog complexity and improve accuracy, components that are
shift it to the digital part of the system increasing digital complexity, well-designed and comparatively large
that relies on digital signal processing calibration makes the converter more in terms of area and power are
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P39
Special Topic A Histogram-Based Static-Error Correction Technique for Flash ADCs
Armin Jalili, J Jacob Wikner, Sayed Masoud Sayedi, and Rasoul Dehghani
commonly used. This results in higher voltages of the MOS transistor. given. Section 7 concludes the paper.
cost and, perhaps more importantly,
longer design time and limited 1.2 Algorithm Example
process-scaling if the ADC is A histogram-based calibration 2 Examples of Static Errors
embedded in a system on chip (SOC). technique for flash ADCs is an example In an n-bit flash ADC, 2n -1 reference
In addition to the material presented of an error-correction technique for levels are commonly generated by a
in this paper, we give a more detailed front-end ADCs. The theory of resistor ladder. For each reference
analysis of the histogram-based histogram test methods is well-known level, a comparator is used. The
static-error correction technique in [1], [7]-[13], and in this paper, a histogram comparators compare the input signal
taking implementation issues into method is used to extract the equivalent with the reference levels, and
account. We study the input-referred offset of each combined, the comparators generate a
transistor-implementation of a 5-bit, comparator in the flash ADC. This is an thermometer code at their outputs. A
1 GHz ADC in a 1.2 V, 65 nm CMOS estimation process. Trimming is then digital decoder that filters out sparkle
process, and we evaluate the performed by adjusting the reference errors is usually used. Static errors
improvement in performance brought voltage levels in order to compensate affect the accuracy of the decision
about by the calibration technique. We for the static error sources. The levels in the converter. Errors in the
also suggest further improvements to estimation process is fully digital, and reference levels are mainly caused by
the analog design so that analog no changes are made to the resistor mismatch and comparator
complexity is further reduced in favor of comparator structure. Trimming is offsets. These two errors can be
digital complexity. performed by adding arrays of analog modeled (and combined) as a voltage
switches that are connected to the source, Vos , at the input of each
1.1 Calibration reference taps of the comparators so comparator, as shown in Fig. 1(a). For a
Calibration techniques for flash ADCs that the high-speed signal path of the low-resolution ADC, the matching of
have been reported in [2]-[6]. In most converter is not affected. the resistors is usually much better than
of these techniques [2]-[4], [6], the To illustrate the calibration technique, the nominal resolution of the flash
analog part — especially the behavioral-level simulations can be converter, so the reference voltages
high-speed analog signal path — of performed. Individual error sources can from the resistor ladder can be treated
the flash converter is affected by the then be isolated, and their effects can as ideal values [6]. However, this is not
calibration circuitry. In some of these be analyzed separately. In [1], a the case for higher-resolution flash
techniques [3], [4], offset compensation transistor-level ADC that uses the ADCs.
(trimming) is performed inside the calibration technique is described, and In Fig. 1(a), the comparator is
comparator structures. In [3], trimming practical implementation issues are assumed to be ideal, and the error
is performed by adjusting the internal discussed. voltage, Vos , in practice would be
resistive loads of the comparator In section 2, main sources of static described by statistical variation over
preamplifiers. However, these loads are error are modeled. In section 3, the many individuals. This is shown in Fig. 1
often difficult to trim accurately because calibration technique is described, and (b), where Vr,i is the ideal reference
of their high gain and nonlinearity, and in section 4, the estimation process is level of the ith comparator, and Vos,i is
trimming might be inefficient. In [2] and described in mathematical detail. In the corresponding offset voltage source
[6], linear trimming is performed section 5, the trimming process and of the ideal reference. The ith interval
efficiently by adjusting the voltages related practical issues are discussed. between two adjacent reference levels
generated by the reference ladder. In In section 6, simulation results are is denoted Ii. In an ideal converter, the
[2], an accurate digital-to-analog
converter (DAC) generates a calibration
signal, but the design of such a DAC Vr,i Vr,i
still has a high degree of analog
complexity. In [6], each comparator is Vr,i - VOS,i
VOS,i
chosen sequentially and calibrated. - + Vr,i -1 - VOS, i -1
+ Ii
This affects the main ADC structure; Vr,i - VOS,i
specifically, extra calibration circuitry
•
Vr,i - Vr,i -1 - VOS, i -1
affects the high-speed input signal
Vr,i -1 Vr,i -1
path. In [5], trimming is performed by
adjusting the bulk voltages of the
comparator input transistors to keep the
high-speed signal path intact. (a) (b) (c)
However, the trimming is nonlinear
because of the nonlinear relationship ▲Figure 1. (a) Static errors with voltage sources, Vos,i, and their effects for (b) small
between the bulk and threshold and (c) large offsets.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P40
A Histogram-Based Static-Error Correction Technique for Flash ADCs Special Topic
Armin Jalili, J Jacob Wikner, Sayed Masoud Sayedi, and Rasoul Dehghani
Digital
…
V in
T/H
input signal characteristics, as is calibration. The outputs PDF
commonly done in histogram-based of the array are Generator Trimming Estimation
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P41
Special Topic A Histogram-Based Static-Error Correction Technique for Flash ADCs
Armin Jalili, J Jacob Wikner, Sayed Masoud Sayedi, and Rasoul Dehghani
estimated using the initial condition (5), time during the calibration, 16
Vr, 1 Vr, 1
and offset errors related to comparators comparators are chosen and their
Vr, 1 - VOS, 1 8 to 14 are estimated using initial offsets are estimated. The results are
I1 condition (6). stored in a RAM. Then, another set of
For large offset values, as shown in 16 comparators is selected for the
V bot =-VR
V bot =-VR Fig. 1(c), Ii may disappear, creating a estimation. This procedure continues
Vr, 1 - VOS, 1
negative interval. In the case of an until all offsets of all comparators have
(a) (b) overlap, (4) must be modified so that been estimated and stored in the RAM.
the estimation can handle errors given Only one state, corresponding to a
▲Figure 4. The effect of static error values on by standard 4-bit flash ADC, contains
interval Ii for (a) small and (b) large errors. ~ i - 1 comparators. This is the state
Ni + Ni when the first input of each multiplexer
Vos,i = Vos,i -1 +2VR· . (7)
(which grows exponentially with ADC N is chosen.
resolution) so that the critical area does This modified version of (4) is used
not increase too much. instead of (4). The case of small offset
values (Vos,i) in comparator 1 is shown in 5 ADC Trimming Process
4.1 Estimation for a Low-Resolution Fig. 4(a), where (5) is valid. For large We take the trimming approach
ADC error values, the reference level may be suggested by Chen et al. in [6].
If the PDF circuit generates an analog pushed below the bottom voltage -VR, Besides having single LSB steps for the
signal with a uniform distribution as in Fig. 4(b). In this case, Vos,1 is ideal reference voltages, the resistor
(constant PDF) in the range [-VR, VR ], decreased during trimming until the ladder also generates smaller steps,
then for an ideal 4-bit converter, and situation shown in Fig. 4(a) is reached. that is, fractions of an LSB to allow the
for every sufficiently large batch of N Again, (5) is valid and can be used in reference voltages to be more finely
input samples, on average the estimation process. The same adjusted. This implies that trimming is
strategy is used for the upper-end done without any changes being made
Ni = N/16 (2)
using a modified version of (6). to the comparator.
samples in each interval i = 1, 2,..., 16. The approach is shown in Fig. 6. An
In a non-ideal 4-bit converter with 4.2 Extending to Higher appropriate voltage tap is selected by
interval Ii, as in Fig. 1(b), the expected Resolutions the trimming circuit in order to
number of samples is To extend the low-resolution compensate for the offset value. In the
technique to a general n -bit converter, resistor ladder, the coarse taps
2VR
~ -Vos,i + Vos,i -1 a segmented structure can be used generate the ideal reference voltages,
Ni = 16 ×N. (3) (Fig. 5). This structure
2VR
saves hardware for +
- l
By rearranging (3), a recursive higher-resolution ADCs MUX
…
N
…
exponentially. These +
- l
Equation (4) can now be used to benefits are described in MUX
…
#i
Vos,0 = Vos,16 = 0. According to (4), In the segmented +
-
1 RAM
comparators 1 and 15 give two initial approach, m = 2n - 1 •
…
…
MUX
(6) last group, which contains
+
-
N 16
i - 1 comparators. Sixteen +
-
2 #16
'1' 1
Either one of (5) or (6) is a sufficient multiplexers are used to
•
starting condition for the calibration choose the output of one
Select times
algorithm. However, to minimize comparator from each
MUX: multiplexers
computational error accumulation group and connect these to
during error estimation, the offset errors the 16 -decision-level ▲Figure 5. Extending 4-bit ADC calibration to an n -bit
related to comparator 2 to 7 are estimation block. At any converter with l = 2n - 4 comparators in each group.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P42
A Histogram-Based Static-Error Correction Technique for Flash ADCs Special Topic
Armin Jalili, J Jacob Wikner, Sayed Masoud Sayedi, and Rasoul Dehghani
…
…
to the sensitive values of the parasitic capacitors, Cgs
…
Taps
for an ordinary decoupling capacitance is, by
V r, i + 2δV r •
differential-pair implication, added to different taps in
comparator input the resistor ladder. Adding trimming
V r, i +δV r •
stage. circuitry does affect kickback noise but
Coarse Tap V r, i • … Trimming To simulate the in a positive way.
effects of kickback An increased time constant is an
…
…
…
0.25 V. The number of sub-LSB steps time, T conf, which is the time the system
Vi+1
is M = 4, see (8). The amplitude takes to settle after the trimming
distribution for simulated kickback in command has been applied (Fig. 9).
Vi • V in
different comparators is shown in Fig. 8. The perturbation is the effect of the
Vi+1 •
The effect of the kickback noise kickback noise. The amplitude of the
depends on the reference level and the kickback noise is Vn, and the residual
…
that generate M - 1 fine taps. Each fine (9) state is fixed. Depending on the
2
tap generates a voltage of Vr,i +δVr , resistance of the resistor ladder and the
where where i = 1, 2,...,2n - 1 is the tap size of the switches, the configuration
number, and R is the unit resistance of time differs from a few clock cycles to
Vr,i +1-Vr,i
δVr = (8) the reference ladder. According to (9), several. The design of the circuits must
M
a symmetrical shape can be expected. include a time constant that is
is the fine-level step — a fraction of the However, the amount of kickback noise negligible compared with the overall
original coarse LSB step. also depends on the input signal level calibration time. A more detailed picture
Our aim is to create a calibration as well as the reference level created of the effect of kickback noise on the
process that does not adversely affect by the voltage and capacitance calibration technique is given by the
the performance of the ADC operating differences between the gate and drain transistor-level simulation in [1]. The
in normal mode. However, in practice,
: Ideal Curve
this is not possible. Any calibration 1.0
Normalized Kickback Noise Amplitude
: Actual Curve
changes or influences the ADC 0.8
structure, and this can affect the overall Figure 8. ▶
performance. Kickback noise for 0.6
Because more switches are added to different tap numbers.
The solid line represents 0.4
the ADC’s design (Fig. 6), kickback
the actual curve and the
noise, a well-known problem, emerges. dotted line is the
0.2
In a clocked comparator, the large predicted curve when 0
variations on the regeneration nodes in impedance variations 0 10 20 30 40 50 60
the latch circuit are fed back through a are taken into account. Tap Number
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P43
Special Topic A Histogram-Based Static-Error Correction Technique for Flash ADCs
Armin Jalili, J Jacob Wikner, Sayed Masoud Sayedi, and Rasoul Dehghani
532 slice flip flops, and 1280 four-input and trimming is performed by adjusting
lookup tables were also used. the comparator reference voltages
Vr, j One hundred Monte Carlo analyses through arrays of analog switches.
…
ΔVn were run, and the performance of the A behavioral-level 8-bit flash ADC
Vn
ADC before and after calibration was was designed, and typical errors were
measured. The static performance in added to the comparators, resistor
~
~ terms of DNL and INL is shown in Figs. ladder, and PDF generator. The results
10(a) and (b), respectively. For about of 100 Monte Carlo analysis show that
Vr, j T conf 98% of the Monte Carlo runs, the converter performance was improved
converter’ s performance
▲Figure 9. Settling behavior of the resistor considerably improved. After ▼Table 1. Design parameters
ladder in response to changing from
calibration, the DNL was reduced
reference level V r,i to V r, j (as a result of Parameter Value
inserting a trimming command) also to on average from 4 to 0.5 LSB, and
kickback noise. the INL was reduced on average ADC Resolution 8
from 4.2 to 0.35 LSB. The failing Offset Distribution (σ ) os 25 mV
simulation uses a 65 nm CMOS with 2% were cases where the offset Offset Distribution of TSG Comparators (σ ) 5 mVos,tsg
4
sampling frequency of the ADC was a
modest 400 MHz. 3
The system design parameters and 2
applied-error sources are shown in
1
Table 1. Each calibration cycle contains
the estimation of equivalent offset 0
0 10 20 30 40 50 60 70 80 90 100
voltages for a set of 16 comparators Simulation Index
chosen by the multiplexer array. In the (a)
case of interval overlap, one more 7 : Before Calibration
calibration cycle is required. Further on, : After Calibration
two calibration cycles are required to 6
4
converter is being studied, there are 16
sets of comparators to trim, and 64 3
cycles are needed to fully estimate and 2
trim all the error sources in the ADC
structure. 1
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P44
A Histogram-Based Static-Error Correction Technique for Flash ADCs Special Topic
Armin Jalili, J Jacob Wikner, Sayed Masoud Sayedi, and Rasoul Dehghani
considerably in 98% of the runs. Instrum. Meas., vol. 47, no. 6, pp. 1448-1456, 1998. [11] Degang Chen, Xin Dai and R. Geiger,“A
[8] J. Doernberg, H.-S. Lee, and D.A. Hodges, cost-effective histogram test-based algorithm for
This work can be used as a “Full-speed testing of A/D converters,”IEEE J. digital calibration of high precision pipelined
grounding for the study and Solid-State Circ., vol. 19, no. 6, pp. 820- 827, Dec. ADCs,”in Proc. IEEE Int. Symp.Circ.Syst., Kobe,
1984. Japan, May 2005, pp. 4831-4834.
implementation of ADC calibration [9] Y. Betrand, F. Azais, S. Bernard, and M. Renovell, [12] Degang Chen Le Jin and R. Geiger,“A digital self
techniques. Prospects and limitations of “Towards an ADC BIST scheme using the calibration algorithm for ADCs based on histogram
the technique are also highlighted in histogram test technique,”in IEEE Proc. European test using low-linearity input signals,”in IEEE Int.
Test Workshop, Cascais, Portugal, May 2000, pp. Symp. Circ. Syst., Kobe, Japan, May 2005, pp.
this work. An IC implementation of an 53-58. 1378-1381.
ADC using the proposed calibration [10] U. Eduri and F. Maloberti,“Online calibration of a [13] J. Elbornsson and J.-E. Eklund,“Histogram based
Nyquist-rate analog-to-digital converter using correction of matching errors in subranged ADC,”
technique is detailed in [1]. output code-density histograms,”IEEE Trans. in Proc. 27th European Solid-State Circ. Conf.,
Circ. Syst. I, vol. 51, no. 1, pp. 15-24, Jan. 2004. Villach, Austria, 2001, pp. 555-558.
References
[1] J. Wikner, A. Jalili, S. M. Sayedi, and R. Dehghani,
“A histogram-based static error correction Biographies
technique for flash ADCs: Implementation aspects,”
accepted for publication in ZTE Communications,
Armin Jalili (arminj@ec.iut.ac.ir) received his B.Sc. and M.Sc. degrees in electrical engineering from Isfahan
Mar. 2012.
University of Technology (IUT) in 2004 and 2006. He is currently working towards his Ph.D. degree in electrical
[2] H. Yu and M.-C. F. Chang,“A 1 V 1.25 GS/s 8-bit
engineering at IUT. His research interests include ADC design.
self-calibrated flash ADC in 90 nm digital CMOS,”
IEEE Trans. Circ.. Syst.. II, vol. 55, no. 7, pp.
J. Jacob Wikner (jacob.wikner@liu.se) received his Ph.D. from the Department of Electrical Engineering,
668-672, Jul. 2008.
Linköping University, Sweden, in 2001. He has worked as a research engineer at Ericsson Microelectronics, senior
[3] C.-W. Lin Y.-Z. Lin and S.-J. Chang,“A 5-bit 3.2
analog design engineer at Infineon Technologies, and senior design engineer and chip architect at Sicon
GS/s flash ADC with a digital offset calibration
Semiconductor. Dr. Wikner has been an associate professor at Linköping University since 2009. His research
scheme,”IEEE Trans. Very Large Scale Integration
interests include biologically inspired architectures, high-speed ADC and DAC, and general analog and
(VLSI) Syst., vol. 18, no.3, pp. 509-513, Mar. 2010.
mixed-signal design. He holds six patents, has published 40 scientific papers, and has co-authored“CMOS Data
[4] V. Srinivas, S. Pavan, A. Lachhwani, and N.
Converters for Telecommunications.”He is the co-founder of CogniCatus and AnaCatum Design.
Sasidhar,“A distortion compensating flash analog
to-digital conversion technique,”IEEE J. Solid
State Circ., vol. 41, no. 9, pp. 1959-1969, Sep. Sayed Masoud Sayedi (m_sayedi@cc.iut.ac.ir) received his B.Sc. and M.Sc. degrees in electrical engineering
2006. from Isfahan University of Technology (IUT), Iran, in 1986 and 1988. He received his Ph.D. degree in electronics
[5] Jin Liu Junjie Yao and Hoi Lee,“Bulk voltage from Concordia University in 1996. From 1988 to 1992, and since 1997, he has worked at IUT, where he is currently
trimming offset calibration for high-speed flash an associate professor in the Department of Electrical and Computer Engineering. His research interests include
ADCs,”IEEE Trans. Circ. Syst. II, vol. 57, no. 2, pp. VLSI fabrication processes, low power VLSI circuits, vision sensors, and data converters.
110-114, Feb. 2010.
[6] Chun-Ying Chen; M. Q. Le and Kwang Young Kim, Rasoul Dehghani (dehghani@cc.iut.ac.ir ) received his B.S.E.E., M.Sc., and PhD degrees in electrical
“A low power 6-bit flash ADC with reference engineering from Sharif University of Technology (SUT), Iran, in 1988, 1991 and 2004. From 1987 to 1991, he
voltage and common-mode calibration,”IEEE J. worked on design and implementation of different electronic circuits and systems at SUT. From 1991 to 1998, he
Solid-State Circ., vol. 44, no. 4, pp. 1041-1046, was involved in implementing various electronic circuits focused on industrial applications. From 1998 to 2004, he
Apr. 2009. worked with Emad Co. in Tehran and Jaalaa Company in Kuala Lumpur as a senior design engineer. Since 2006,
[7] J. Larrabee, F.H. Irons and D.M. Hummels,“Using he has been an assistant professor at IUT. His research interests include RF IC design for wireless communication,
sine wave histograms to estimate analog-to digital frequency synthesis, and low-voltage low-power circuits.
converter dynamic error functions,”IEEE Trans.
⬅ From P.34 N. Kikuma, S. Hori, T. Wada.“Experimental study of Trans. Broadcasting, vol. 53, no.1, pp. 161-170,
adaptive array antenna system for ISDB-T 2007.
elemental computing array (ECA), high-speed mobile reception,”Proc. IEEE Int. [8] M. Chari, F. Ling, A. Mantravadi, R. Krishnamoorthi,
which is designed to achieve a better Symp. Antenna and Propag., pp.1697-1700, R. Vijayan, G. Walker, R. Chandhok,“FLO physical
Hawaii, 2007. layer: An overview,”IEEE Trans. Broadcasting, vol.
compromise between performance, [6] Fa-Long Luo and R. Unbehauen,“A minor 53, no. 1 pp. 145-160, Mar. 2007.
flexibility, cost, and power efficiency. subspace analysis algorithm,”in IEEE Trans. Neural [9] Fa-Long Luo, Digital Front-End in Wireless
With ECA, the FGPA scheme can have Netw., vol. 8, no.5, 1997, pp1149-1155. Communications and Broadcasting, Circuits and
[7] M. Kornfeld, G. May,“DVB-H and IP Signal Processing. Cambridge: Cambridge
performance, power, and cost similar to datacast-broadcast to handheld devices,”IEEE University Press, 2011.
ASICs. ECA is far more flexible than a Biographies
DSP with accelerators. It is also much
Fa-Long Luo (falong.luo@elementcxi.com) is chief scientist at Element CXI. He has been the editor-in-chief of
cheaper and consumes far less power the International Journal of Digital Multimedia Broadcasting since 2007. Fa-Long Luo is now the chairman of the
than FPGA devices. IEEE Industry DSP Standing Committee and technical board member of the IEEE Signal Processing Society. He
has 28 years of research and industry experience in multimedia, communications, and broadcasting with
real-time applications and standards. Fa-Log Luo has received worldwide recognition. He has authored and
References edited 4 books, more than 100 technical papers and 18 patents.
[1] D. H. Pham, J. Gao, T. Tabata, H. Asato, S. Hori, T.
Wada,“Implementation of joint pre-FFT adaptive Ward Williams (ward.williams@elementcxi.com) is vice president of marketing for Element CXI. Ward has broad
array antenna and post-FFT space diversity technology marketing and business development experience in senior management positions. He has worked for
combining for mobile ISDB-T receiver,”IEICE companies such as Xilinx, Chips and Systems, S3/SonciBLUE, and Hewlett-Packard. He frequently speaks at
Trans., vol. E91-B, no.1, pp.127-137, 2008. international technology shows or forums such as IEEE ICASSP and Global Mobile Congress. His research
[2] Fa-Long Luo, Multimedia Mobile Broadcasting interests include digital front-end, software-defined radio, single-chip solutions, SoC-based hardware, and
Standards: Technology and Practice Springer, 2008. software IPs.
[3] Fa-Long Luo, R. Unbehauen, Applied Neural
Networks for Signal Processing. Cambridge: Bruce Gladstone (bruce.gladstone@elementcxi.com) has worked in the semiconductor and EDA industries for
Cambridge University Press, 1999. over 20 years. After graduating from the University of Illinois, he went to Silicon Valley to work in IC design,
[4] I. Nobuo, T. Kenichi,“HDTV Mobile Reception in applications, and technical marketing positions. Bruce has worked for companies such as AMD, Western Digital,
Automobiles,”Proc. IEEE, vol.94, no.1, Synopsys, and other companies in California and Japan. As a senior director at Element CXI, he is focused on
pp.274-280, 2006. putting new-generation programmable semiconductors into the next-generation communications products.
[5] T. Tabata, H. Asato, Dang Hai Pham, M. Fujimoto,
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P45
Guest Editorial
Sean Cai Li Mo
T
his special issue of ZTE Communications focuses on information in order to improve overall network performance.
recent advances in mobile data communications for the In SCML, the combined burst rate of a UE can be significantly
ICT and telecommunications industries. The improved between neighboring cells.
ever-increasing amount of mobile data traffic has been In“Uplink Power Control for MIMO-OFDMA Cellular
the subject of many studies. This research area is widely Systems,”by Rongzhen Yang, uplink power control
applicable to contemporary technology and network algorithms, including FPC used in 3GPP LTE and
optimization techniques. LTE-Advanced, are compared. A novel power control
The priority of most operators is to further expand their algorithm is proposed for MIMO-OFDMA systems. A
existing 3G networks and focus on new 3G technologies simplified maximum sector throughput (SMST) algorithm is
being developed to meet ever-increasing data traffic and also proposed to maximize the sector throughput by adjusting
data speed. Optimization techniques of 3G and 4G networks uplink transmit power.
are being explored in order to improve network performance. In“Mobile Backhaul Solutions,”by Li Mo et al., mobile
Today, operators are concerned about adopting the right backhaul requirements are reviewed, and backhaul solutions
backhaul technologies, leveraging existing Wi-Fi networks to are proposed with timing synchronization, multiprotocol label
offload data traffic, building and evolving the core network, switching (MPLS); operation administration, and maintenance
and introducing mobile cloud computing in order to support (OAM), and protection taken into consideration.
expanded business in the future. We thank all the authors for their excellent papers and the
In this special issue of ZTE Communications, we called for reviewers that had helped us with their careful and detailed
papers that provide greater insight into improved mobile data critical reviews. We are very grateful to the editorial board for
networks. Inter cell interference coordination (ICIC), their helpful suggestions in improving the quality of the papers
coordinated multipoint transmission (CoMP), multiple input in this issue.
multiple output (MIMO), adaptive beamforming, and network
load balancing (NLB) are introduced. These techniques may
Biographies
require operators to review the way they engineer and deploy
their networks. Techniques to further improve network Sean Cai (SeanCai@zte.com.cn) is the vice president of wireless at ZTE.
performance are discussed in this issue. High-level backhaul He is responsible for developing 4G technologies as well as strategic
technologies and core network evolution paths are also marketing and technical support for business development. He leads a
technical team in USA, working closely with leading operators. He is
discussed in this issue.
actively involved in the working group activities of various standard
In“Enhanced Cell Edge Performance with Transmit Power bodies. He has more than 20 years’professional experience in
Shaping and Multipoint/Multiflow Techniques,”by Philip telecommunications, and he specializes in advanced wireless products
Pietraski et al., a new concept called Fuzzy Cells is introduced research and development. Prior to joining ZTE, he held various senior
to improve cell coverage. Fuzzy Cells enable user equipment positions in infrastructure ASIC architecture design, CDMA, and optical
(UE) to access the full system bandwidth in a sector by using development. He has been awarded multiple international patents in
a component carrier (CC) transmitted at higher power than the telecommunications. He was recognized as one of the“Mobile Movers &
Shakers”in 2008 by RCR Wireless News for his leadership in the
other CC frequencies. If at least three CC frequencies are
development of new wireless technology.
used, three sets of cell boundary locations can be created so
that no UE is simultaneously in the cell-edge region for every Li Mo (mo.li@zte.com.cn) received his B.Eng. degree from the
CC. Cell coverage can be improved using the techniques Department of Electrical Engineering, Queen’ s University, in 1989. After
proposed in the article. graduation, he worked at IBM, Nortel, and Fujitsu before joining ZTE in
In“Spatial Load Balancing in Wide-Area Wireless 2001. Currently, Dr. Mo is the chief architect for ZTE USA and works as a
marketing specialist for ZTE headquarters in Shenzhen. Dr. Mo has
Networks,”by Kambiz Azarian et al., NLB and single-carrier
worked in the networking industry for more than 20 years, and has
multilink (SCML) are compared. These techniques are used to numerous publications and U.S. patents. His research interests include
increase network capacity and improve user experience. In fixed and mobile core networks, session control, QoS, and routing. Dr. Mo
NLB, each sector-carrier periodically provides its loading is also an active member of IEEE, ITU, ETSI, and IETF.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P46
Enhanced Cell-Edge Performance with Transmit Power-Shaping and Multipoint, Multiflow Techniques Special Topic
Philip Pietraski, Gregg Charlton, Rui Yang and Carl Wang
Abstract: In this paper, we present a technique called“fuzzy cells ”that builds on the multicarrier features of Long Term
Evolution-Advanced (LTE-A) and high-speed packet access (HSPA). Multiple carriers are aggregated to create a larger system
bandwidth, and these carriers are transmitted at different powers by each sector antenna. This creates a set of cell-edge locations that
differ from one frequency to the next. System-level simulations are performed to estimate individual user and average throughput for a
hexagonal deployment of 3-sector base stations. For moderately high loads, a fuzzy cell deployment can improve tenth percentile
(cell-edge) user throughput by 100% and can improve average throughput by about 30% compared with a reuse 1 scheme. Fuzzy cells
reduce inter-cell interference in the same way as higher-order reuse schemes and allow users to access the full system bandwidth.
Keywords: LTE; fractional frequency reuse; multicarrier; flow aggregation; system simulation
hexagonal deployment, this cell-edge More importantly, some UEs will not
1 Introduction condition affects much of the have access to the full system
I
n current and evolving cellular deployment area. In an OFDM system, bandwidth. In the FFR case shown in
systems (LTE is the example in this such as LTE Release 8, the cell-edge Fig.2, the spectrum is split into two
paper) [1], it is generally very problem can be mitigated to some parts: reuse 3 and reuse 1. Some of the
difficult to create a uniform user extent by organizing UEs into groups UEs are classified as cell edge, and
experience because throughput at the according to different parts of the others are classified as cell center. The
cell edge is limited by interference from spectrum and using different transmit classification determines which parts of
other cells. In standard frequency powers for these groups. the spectrum are available to the UEs.
reuse 1, the cell-edge downlink (DL) Cell-edge UEs are grouped The details depend on the particular
signal to interference plus noise ratio logically. They are assigned a specific FFR scheme, but the cell-edge UEs
(SINR) can be less than 0 dB, which part of the spectrum and are allocated are expected to have improved SINR in
limits user throughput even if there is a a potentially higher power, as in soft the red, blue, or green frequency areas.
large number of resource blocks (RBs) frequency reuse (SFR), or higher Improved SINR comes at the price of
for UEs. The cell-edge problem is frequency reuse transmission, as in reduced accessible bandwidth for UEs.
shown in Fig.1. fractional frequency reuse (FFR).
For UEs close to an E-UTRAN Although the data signal received by
Node B (eNB), the DL SINR is quite cell-edge UEs may be improved, this
high. However, as the UE moves away approach has limitations. Take, for SINR
from the eNB, the DL SINR decreases example, the cell coverage and control
until the UE is midway between the channel. The control channel spans the
eNBs (at the cell edge). A handover system bandwidth and is shared
User Location
(HO) then occurs; the UE passes to the among UEs. Thus, the same extended
SINR: signal to interference plus noise ratio
next eNB, and SINR starts to increase range that a frequency selective,
again. SINR is lowest near the cell UE-specific data channel might attain ▲Figure 1. Cell edge in a one-dimensional
edge. In a typical two-dimensional is not possible for UEs at the cell edge. deployment.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P47
Special Topic Enhanced Cell-Edge Performance with Transmit Power-Shaping and Multipoint, Multiflow Techniques
Philip Pietraski, Gregg Charlton, Rui Yang and Carl Wang
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P48
Enhanced Cell-Edge Performance with Transmit Power-Shaping and Multipoint, Multiflow Techniques Special Topic
Philip Pietraski, Gregg Charlton, Rui Yang and Carl Wang
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P49
Special Topic Enhanced Cell-Edge Performance with Transmit Power-Shaping and Multipoint, Multiflow Techniques
Philip Pietraski, Gregg Charlton, Rui Yang and Carl Wang
However, for cells that are realistically eNB1 finishes transmitting part 1 before Core Network
loaded, it is possible to improve both eNB finishes transmitting part 2, then at (MME/S-GW)
cell-edge and average throughput. that point, only eNB 2 can transmit to 2
X2
The CCs providing the largest-biased the eUE. 1 2
SINRs in each frequency are assigned The packet size for the traffic model eNB1 eNB2
to each UE. The bias applied to the in [2] is 2 MB. For a 2 MB packet and an
SINR involves optimizing one of the assumed X2 delay of 10-20 ms, the
system parameters so that the tenth delay is negligible and not modeled in MME: mobility management entity
S-GW: serving gateway
percentile user throughput is the simulations.
maximized for each average ▲Figure 9. Data- flow splitting at eNB.
throughput value. 2.7 Estimating Performance
Because of the complexity of 100
2.6 Impact of Backhaul Data Flow cellular systems, it is difficult to
Split on Performance predict gains without extensive 50
Fig. 9 shows the splitting of the simulation; however, some 30
downlink data flow in a fuzzy cell estimation is possible. For lightly 20
deployment. The split occurs in the loaded cells, UEs do not contend
CDF (%)
eNB. The application data packet is much for resources in any given 10
downloaded from the core network to transmission time interval (TTI).In
5
the serving eNB (eNB1) via S1 this case, the throughput : Baseline
: Fuzzy, 10 dB
interface. The eNB1 decides how to distribution can be largely 3
: Fuzzy, 100 dB
split received data for each UE, and the predicted by the eSINR 2
data payload is split in two. Part 1 is distribution, and the eSINR can
1
sent to the UE directly from the eNB1, be mapped to an achievable 0 10 20 30 40 50 60
Throughput (Mbit/s)
and part 2 is forwarded to the throughput. If we consider only
CDF: cumulative distribution function
cooperating eNB (eNB2) via X2 the long-term SINR in each CC
interface before being sent to the UE by frequency in the system, we can ▲Figure 10. eSINR based achievable throughput CDF
the eNB2. Status signals are frequently compute the achievable for a 10 dB power difference.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P50
Enhanced Cell-Edge Performance with Transmit Power-Shaping and Multipoint, Multiflow Techniques Special Topic
Philip Pietraski, Gregg Charlton, Rui Yang and Carl Wang
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P51
Special Topic Enhanced Cell-Edge Performance with Transmit Power-Shaping and Multipoint, Multiflow Techniques
Philip Pietraski, Gregg Charlton, Rui Yang and Carl Wang
▲Figure 12. Comparison of algorithmic schemes for low traffic with shadowing. Inter-Site Distance (ISD) 500 m
Total Number of Sites 19
as CoMP, investment needs to be made macro cell no line-of- sight. UEs are Number of UEs per Site 30 (10 per Sector Antenna)
in low-latency, high-capacity dropped randomly with uniform UE Distribution Uniform Distribution
Random Drop
backhaul, for example, fiber optics. distribution. The simulation parameters Channel Model -
However, fuzzy cells can be realized are summarized in Table 1. Two site PL = 161.04 - 7.1 log10 (W ) +
with existing backhaul. Fuzzy cells can rings are populated with users, and 7.5 log10 (h ) - [24.37- 3.7(h/hBS )2]
log10 (hBS ) + [43.42 -
be realized before other advanced statistics are collected only from UEs Path Loss 3.1 log10 (hBS )] [log10 (d )-3] +
20 log10(fc ) -
techniques and without investment in near the center site. {3.2 [(log10 (11.75 hUT ) ] 2 - 4.97}
new low-latency backhaul. Street Width (W) 20 m
References
Appendix [1] Further advancements for E-UTRA physical layer Avg. Building Height (h) 20
aspects,3GPP TR 36.814, 2010.
A system-level simulator using [2] R. Chang, Z. Tao, J. Zhang, C-C Jay Kuo,“A graph
hBS 25
hexagonal cellular layout is used to approach to dynamic fractional frequency reuse hUT 1.5 m
(FFR) in multi-cell OFDMA networks,”Proc. IEEE
validate our concept. The simulation International Conf. Commun., Dresden, Jun. 2009,
Shadow Fading -
Gregg Charlton obtained his BSEE from Carnegie-Mellon University in 1982 and his MSEE in EE systems from Front/Back Ratio (Am) 20 dB
the University of Michigan, Ann Arbor in 1986. While at Michigan, he received a research fellowship and worked as Proportional Fair. Common Pool
a research assistant in the EECS department. Mr. Charlton has been with InterDigital since 2000 and is currently a Scheduler Buffer. Perfect CSI Information.
member of technical staff in the advanced air interfaces organization. Prior to working at InterDigital, he worked for 1 RB Granularity.
Lockheed Martin Global Telecommunications, General Electric Aerospace, AT&T Bell Laboratories, and TRW. His Beta Optimized in Simulation
research interests include analysis of mobile network deployments, wireless communication theory, and satellite
communications. Link Performance Model EESM
AWGN with Power Based on
Rui Yang received his M.S. and Ph.D. degrees in electrical engineering from the University of Maryland in 1987 Interference Model PL and per CC Tx Power
and 1992. He joined InterDigital Communications in 2000 and has 11 years experience researching and Setting
developing wireless communication systems. His interests include digital signal processing and physical layer X2 Latency 0 ms
design for wireless devices. He has more than 10 patents has publiushed several papers. He is currently an
Traffic Model 3GPP FTP Model 1
engineering manager at InterDigital Communications.
Packet Size (S) 1 MB
Carl Wang obtained his BSEE from Polytechnic University, Brooklyn, NY, in 1994. He received his MSE in 0.00063 (high load) or
computer engineering from the Carnegie-Mellon University in 2006. Mr. Wang has been with InterDigital since 0.00013 (low load) Packets/
1996 and is currently a member of technical staff in the advanced air interfaces organization. Prior to working at User Arrival Rate (λ) TTI/UE Corresponds to
InterDigital, he worked for Raytheon Company in digital signal data processing. His research interests include Approximately 50% and 10%
analysis of mobile network deployments, wireless communication theory, and communication protocol and Resource Utilization
architecture. Simulation Duration 50 s/drop
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P52
Spatial Load Balancing in Wide-Area Wireless Networks Special Topic
Kambiz Azarian, Ravindra Patwardhan, Chris Lott, Donna Ghosh, Radhika Gowaikar, and Rashid Attar
Abstract: Load balancing is typically used in the frequency domain of cellular wireless networks to balance paging, access, and traffic
load across the available bandwidth. In this paper, we extend load balancing into the spatial domain, and we develop two
approaches—network load balancing and single-carrier multilink—for spatial load balancing. Although these techniques are mostly
applied to cellular wireless networks and Wi-Fi networks, we show how they can be applied to EV-DO, a 3G cellular data network. When a
device has more than one candidate server, these techniques can be used to determine the quality of the channel between a server and
the device and to determine the load on each server. The proposed techniques leverage the advantages of existing EV-DO network
architecture and are fully backward compatible. Network operators can substantially increase network capacity and improve user
experience by using these techniques. Combining load balancing in the frequency and spatial domains improves connectivity within a
network and allows resources to be optimally allocated according to the p-fair criterion. Combined load balancing further improves
performance.
Keywords: CDMA 2000; EV-DO; DO-RevC; high rate packet data (HRPD); network load balancing (NLB); single-carrier multilink (SCML)
loading (per-sector forward link (FL) are lightly loaded. Operators add
1 Introduction slot use) during peak hour in a hardware by splitting cells near the
D
emand for data in wireless metropolitan commercial EV-DO chokepoint sectors or increase
networks is spatially network where sectors with low slot use bandwidth by adding carriers to the
non-uniform, which leads to
chokepoint sectors, and it is also 0.40 1
time-varying, which means that
chokepoint sector locations change. 0.9
0.35
Chokepoint sectors operate at or close
to the maximum load; therefore, they 0.8
0.30
determine network performance, user
0.7
perception of network performance
0.25
and, ultimately, user experience. In a 0.6
Probability < X
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P53
Special Topic Spatial Load Balancing in Wide-Area Wireless Networks
Kambiz Azarian, Ravindra Patwardhan, Chris Lott, Donna Ghosh, Radhika Gowaikar, and Rashid Attar
chokepoint sectors. Such actions lead server with the best channel quality. queues at the sector carrier. It is
to networks that are designed for (Here, hysteresis based on channel calculated for each slot (which takes
worst-case scenarios and are therefore quality and time is not discussed for the 1.66 ms) and is fed to a filter with a
underused most of the time. sake of simplicity.) In EV-DO, the AT time-constant of a few seconds to
The spatial load-balancing selects the best forward layer (FL) generate the Neff metric.
techniques discussed in this paper are server from the candidate servers that Networks are at times
network load balancing (NLB) and have a corresponding reverse link (RL) backhaul-limited, and this leads to the
single-carrier multilink (SCML). of acceptable quality. If the DRCLock air link at the BTS being underused. A
In NLB, a device is opportunistically bit on, the RL quality is acceptable; if backhaul-limited sector-carrier may
reassigned from a heavily loaded the DRCLock bit is off, the RL quality is not appear to be loaded because of
sector to a lightly loaded sector, even if unacceptable. The DRCLock bit itself is queue under-run. Therefore, backhaul
the channel quality of the heavily set according to the quality of the RL limit must be accounted for when
loaded sector is better. The end result overhead channels received at the calculating loading level at each
is increased capacity for the chokepoint base station. sector-carrier. A metric for backhaul
sector from offloading the most limit is the number of unfilled flow
expensive users (poor channel quality 2.2 Multicarrier Operation control requests for a sector-carrier
means that the network has to allocate When an AT in EV-DO is assigned that has non-empty AT queues at the
more resources for a given amount of multiple carriers, those carriers may BSC.
traffic) to the neighbor sector. originate from the same sector or
With SCML, any device that can different sectors (one forward-link 2.4 Operator Deployments
simultaneously process two or more serving sector for each carrier) [1], [2]. Operators deploy wireless data
independent data streams can achieve Commercial EV-DO uses independent networks according to data demand.
the benefits of a multicarrier network, schedulers on each sector carrier, and Hence, operator deployments are not
even though the device may be in the multicarrier EV-DO uses multilink radio uniform and have spatial variations in
hand-off region of a single-carrier link protocol (RLP) on each sector the number of carriers deployed.
deployment. carrier. Multilink RLP enables the Typically, a large area in an operator
With multicarrier EV-DO, the EV-DO network to split a data stream at the network has three or less deployed
carriers that serve a mobile device use transmitter across multiple independent carriers. In regions of the network with
different carrier frequencies and may schedulers and to combine the data just one carrier, multicarrier EV-DO
originate from the same cell or different stream at the receiver while ensuring cannot improve performance of ATs
cells. With SCML, the same carrier early detection of packet loss and with multicarrier capability.
frequency serves the mobile device in-order delivery of RLP packets [3].
with independent data streams and When multiple links (or carriers) are 2.5 Smart Carrier Management
from different sectors. NLB improves assigned to an AT, the network EV-DO access networks (ANs)
performance when the chokepoint distributes forward-link data across assign the best-available carrier to an
sector is loaded, and SCML improves available links. Data throughput on AT by transporting the load levels
performance at all load levels. In this each link can be different, and the across carriers (on the forward and
paper, we describe how these throughput varies over time. Enhanced reverse links) and the channel quality
techniques augment load-balancing in flow control (EFC) ensures FL data is from the ATs to each of the candidate
the frequency domain. distributed across links according to carriers.
In Section 2, we provide necessary link throughput and that the data
background information. In sections 3 distribution adapts to changes in link
and 4, we discuss the concepts, throughputs. 3 Network Load Balancing
algorithms, and implementations of Each forward link in EV-DO requires NLB is shown in Fig. 2. PN (b) is the
NLB and SCML. In section 5, we reverse-link feedback to indicate FL best FL sector for the user; however,
perform simulations to determine the channel quality and acknowledgements PN (b) is heavily loaded and the
performance of NLB and SCML. In (ACK/NACK) for physical-layer HARQ. neighboring sector, PN (a), is lightly
section 6, we make concluding remarks EV-DO supports multiple feedback loaded. If PN (b) is selected as the
and discuss future work. multiplexing modes. RL overhead of serving sector, it has better
multiple FL carriers can be transmitted signal-to-interference plus noise ratio
over a single RL carrier to minimize the (SINR) than PN (a) but is scheduled
2 Background RL pilot overhead. much less time than in PN (a) because
of heavy loading. Thus, even at the cost
2.1 Server Selection 2.3 FL Load Metrics of FL SINR degradation, the AT’ s
When the access terminal (AT) in a We introduce a new metric for the throughput can increase when the FL
wireless network has a choice of loading at each sector. This metric is loading differential is sufficient to offset
serving sectors, it typically selects the the average number of non-empty the SINR degradation.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P54
Spatial Load Balancing in Wide-Area Wireless Networks Special Topic
Kambiz Azarian, Ravindra Patwardhan, Chris Lott, Donna Ghosh, Radhika Gowaikar, and Rashid Attar
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P55
Special Topic Spatial Load Balancing in Wide-Area Wireless Networks
Kambiz Azarian, Ravindra Patwardhan, Chris Lott, Donna Ghosh, Radhika Gowaikar, and Rashid Attar
50 channel quality and FL load. versus rate curve for the scenario is
● : 1 Rx ● This is similar to the NLB then formed from all such rate-delay
● : 2 Rx algorithm for new ATs. The pairs. Both single and dual-antenna
40 AN has multiple paths that it devices are simulated. Single and
can use to serve FL data, multicarrier devices are also simulated.
●
and DNS allows network
NLB Gain (%)
30
● throughput to be optimized. 5.1 Load-Balancing Gain for
Single-Carrier Devices
20 ●
Fig. 4 shows the tail page delay
● 5 Performance versus rate curves for single-carrier
●
10 Simulations reveal the devices with one or two antennas. Load
performance gained using balancing provides significant gain only
0 NLB and SCML. The when the load level is sufficiently large,
1 2 3 4 5 6 7 8 performance results are valid that is, when the solid red and blue
Demand Ratio
for a seven-cell layout with a curves (which represent the baseline
NLB: network load balancing
center-cell demand model; and new NLB cases for two-antenna
▲Figure 3. NLB gain as a function of demand ratio. that is, the center cell has devices, respectively) fall on top of
multiple times the number of each other at low rates. Because of the
SCML AT receives interference from users of neighboring cells. The ratio of low load differential at low load levels,
another link(s). To achieve gains from the number of users in the center cell to no user is offloaded to another sector
multiple links, the AT must overcome users in the neighboring cell is the and there is no load balancing gain.
the interference. Some EV-DO demand ratio. Except for Fig. 3, which Also in Fig. 4, the load-balancing gains
receiving diversity implementations can shows load-balancing gain as a for single-antenna devices (new or
spatially null a strong interferer, and this function of demand ratio, all other legacy) are not as large as those for
is the key to SCML. The network should figures are valid for a demand ratio of dual-antenna devices. Once a user is
limit SCML assignment to receiving three. Users have an http traffic model; offloaded from a heavily loaded sector
diversity ATs with spatial nulling that is, they repeat downloading a to a lightly loaded one, the heavily
capabilities. ATs with more Rx antennas sample http webpage according to a loaded sector appears as a strong
are typically capable of superior spatial random arrival process with a fixed interferer. This occurs because the user
nulling of strong interferers, and rate. To determine performance in a receives a stronger signal from the
therefore, SCML performance improves particular scenario, for example, heavily loaded sector than from the
with an increase in number of AT Rx dual-antenna or dual-carrier devices lightly loaded one. Dual-antenna
antennas. with load balancing enabled, devices can spatially null some of the
simulations are performed for a range interference. Single-antenna devices
4.4 Smart Carrier Management of download rates. Each simulation is lack such capability and thus achieve
The smart carrier management then characterized in terms of lower load balancing gains. Multiple
algorithm can be modified to take the download rate (the average number of receiving antennas and special null of a
SCML capability of a device into webpages downloaded
account. The network allocates and per minute per Tail of Page Delay
removes SCML based on signal high-demand ●
●
gain. If the AT can support more links downloaded http page is ●
10 1
than assigned carriers, AN uses SCML the time between when the ● ● : 2 R ,Baseline x
: 2 R ,Legacy NLB
according to available FL control queued in the BSC to the ●
●
● x
● : 1 R ,Baseline
resources and RL load on that carrier.
x
●
time the last byte is ●
● ● : 1 R ,New NLB x
The additional links provide trunking downloaded by the user. ● : 1 R ,Legacy NLB x
gain in the spatial dimension. Once the Then, the page delay CDF
AT server selection algorithm has been (across all users and all
5 10 15 20 25
Web-Page (minute/sector-carrier)
assigned multiple links by the webpages) is formed, and NLB: network load balancing
smart-carrier management algorithm, the ninety percentile delay
the AT server-selection algorithm is chosen as the tail page ▲Figure 4. NLB gain for single-carrier devices with one and
chooses the best link(s) by using the FL delay. The tail page delay two antennas (600 kB per web page).
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P56
Spatial Load Balancing in Wide-Area Wireless Networks Special Topic
Kambiz Azarian, Ravindra Patwardhan, Chris Lott, Donna Ghosh, Radhika Gowaikar, and Rashid Attar
●
Seconds
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P57
Special Topic Spatial Load Balancing in Wide-Area Wireless Networks
Kambiz Azarian, Ravindra Patwardhan, Chris Lott, Donna Ghosh, Radhika Gowaikar, and Rashid Attar
●
● ● ● Azarian and R. Attar,“Distributed Scheduling for
● ● Wireless Networks,”in Proc. IEEE ISIT, 2011.
● ● ●
1.5
● ●
Biographies
●
1.0 Kambiz Azarian (kambiza@qualcomm.com)
● : Baseline received his Ph.D. in electrical engineering from
● : SCML Ohio State University. He joined Qualcomm in 2007
and has since been working on 3G and 4G cellular
systems.
0.5 ◀Figure 8.
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Burst rate for a Ravindra Patwardhan joined Qualcomm in 1998.
Distance from Center Cell (km) single-carrier user moving He has worked on ISDN, Globalstar, Telematics,
1xRTT, 1xEV-DO, and UMB.
SCML: single-carrier multilink between two neighboring
cells. Christopher Lott received his Ph.D. in control
theory from the University of Michigan. He has
show that load balancing in worked on 1xEV-DO systems design for more than
10 years and is currently the DO systems lead at
Tail of Page Delay the spatial dimension and Qualcomm. He has designed aspects of
frequency dimension can communications systems, including LTE, GPS, and
● : Baseline Inmarsat.
● : NLB ●
be combined to amplify
●
● : SCML gains in networks where Donna Ghosh received her M.S. and Ph.D.
10 1
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P58
Uplink Power Control for MIMO-OFDMA Cellular Systems Special Topic
Rongzhen Yang and Hujun Yin
Abstract: In this paper, we propose a novel uplink power control algorithm, SMST, for multiple-input multiple-output orthogonal
frequency-division multiple access (MIMO-OFDMA).We perform an extensive system-level simulation to compare different uplink
power control algorithms, including the FPC adopted in 3GPP LTE and LTE-Advanced. Simulations show that SMST adopted in IEEE
802.16m outperforms other algorithms in terms of spectral efficiency, cell-edge performance, interference control, and trade-off control
between sector-accumulated throughput and cell-edge user throughput. The SMST performance gain over FPC can be more than 40%.
U
plink power control is a critical user experience (UE), even in control (OLPC), or CLPC-OLPC
feature of CDMA cellular broadband wireless systems based on combined power control. In CLPC,
systems. It is used to alleviate OFDMA[4]-[7]. As opposed to 3G, power control is centralized at the base
the near-far problem caused by where closed-loop power control is station (BS). The mobile station (MS)
intracell interference. The new typically used to control both intracell provides feedback on link quality, and
generation of broadband wireless and intercell interference, the main the BS calculates the uplink transmit
technologies, including WiMAX and objective of uplink power control in power level for the MS and instructs the
LTE, are designed to provide higher OFDMA-based systems is to control MS to transmit at that level. In OLPC,
bandwidth, higher peak throughput, intercell interference [8],[9]. the MS measures link quality and
and higher spectral efficiency [1],[2]. In this paper, we investigate control calculates the uplink transmit power
These wireless technologies are based of uplink transmit power in order to level (based on predetermined
on orthogonal frequency-division maximize sector and cell-edge equations) in a distributed manner. The
multiple access (OFDMA), in which all spectral efficiency. These are essential BS may influence the MS by adjusting
the uplink transmissions are orthogonal parameters for next-generation certain parameters in the equation, but
within one cell [3]. Therefore, intracell broadband wireless systems [10]-[13]. the BS does not directly control the MS
interference is minimal compared with A simplified maximum-sector transmit power.
technologies based on 3G CDMA. throughput (SMST) algorithm is
Fourth-generation WiMAX and LTE proposed that maximizes sector 2.1.1 CLPC
wireless standards are designed to throughput by adjusting uplink transmit In a CDMA system, CLPC allows
support up to 20 MHz channel power. commands to be sent from the BS so
bandwidth, high-order 4 × 2 and 4 × 4 that power can be quickly adjusted. In
multiple-input multiple-output (MIMO), an OFDMA system, intracell
and aggressive frequency reuse
2 Uplink Power Control interference is not significant, so
(reuse 1) to improve spectral efficiency Mechanisms adaptive modulation and coding (AMC)
and user throughput. Especially for and hybrid automatic repeat request
cell-edge users, intercell interference 2.1 Background (HARQ) are used to provide fast link
is a significant problem caused by Power control has been studied adaptation for the data channel. CLPC
aggressive frequency reuse. Uplink extensively since the introduction of is mainly used for fixed-rate control
power control is an important cellular systems. A power-control channel when channel fading exceeds
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P59
Special Topic Uplink Power Control for MIMO-OFDMA Cellular Systems
Rongzhen Yang and Hujun Yin
the power margin. In CLPC in the BS, control (FPC) in LTE is an SNR-based If we assume that all BSs occupy the
the received signal quality is monitored algorithm. same bandwidth, (1) can be modified:
and power-adjustment commands are
sent. 2.2.2 SINR-Based Algorithm R Aggregated =∑BW ×SEi =BW ×∑SEi
i i
→
The goal of an SINR-based algorithm SE Aggregated =∑SEi (2)
2.1.2 OLPC is to maintain the desired received i
OLPC is mainly implemented in the SINR level in the BS while taking into For uplink power control of each MS,
MS. The MS measures downlink signal account the uplink interference power aggregated spectrum efficiency SE is
status, compensates the uplink path level. SMST used in IEEE 802.16m is an prioritized as the target of the algorithm,
loss, and controls interference to SINR-based algorithm that defines the which can be expressed as
neighboring BSs. With OLPC, the MS desired received SINR goal for each
needs some static/semi-static MS and takes into account the Ptx =argmax∑SEi (3)
Ptx i
configuration parameters to be measured downlink
signaled by the BS and does not signal-to-interference ratio (SIR), To solve (3) for each MS, the
require short-term inputs. This saves uplink MIMO mode, and BS antenna following process is used:
overall signaling overhead. configuration. 1) PS is the transmission power of one
subcarrier that is initialized to 0
2.1.3 CLPC-OLPC Combined Power 2.2.3 IoT-Based algorithm 2) A power increase, ΔPS , is assumed
Control The goal of an IoT-based algorithm in order to calculate
In current 4G wireless standards, is to maintain the desired uplink i) SE gain : as the power increases, this
both CLPC and OLPC are combined to interference level in the BS. In general, is the spectrum efficiency gain that the
balance flexibility with signaling an IoT-based algorithm can stabilize MS can achieve in the home BS on the
overhead. OLPC is mostly used to save the interference level in the uplink to subcarrier.
signaling overhead. help control interference and ii) SE loss : as the power increases, this
modulation-coding-scheme (MCS) is the spectrum efficiency loss the MS
2.2 Uplink Power Control Algorithms level estimation. inflicts on the neighboring BSs on the
Depending on the goal of the power subcarrier.
control mechanism, the uplink 3) SE gain and SE loss are compared
power-control algorithm can be based
3 Maximizing Throughput i) If SE gain > SE loss, PS = PS + ΔPS , go
on signal-to-noise ratio (SNR), with Power Control back to 2
signal-to-interference plus noise ratio ii) If SE gain ≤ SE loss, then the PS is
(SINR), or Internet of things (IoT). 3.1 System Modeling optimum on the subcarrier for the MS.
Fig. 1 shows an uplink interference This process is used to evaluate the
2.2.1 SNR-Based Algorithm model for one MS in a typical SE gain and SE loss for each power
The goal of an SNR-based algorithm MIMO-OFDMA cellular system. increase of ΔPS .
is to maintain the desired received At any time, the MS uses the uplink to At any time, when a power increase
signal strength in the BS. An transmit to its home BS and causes of ΔPS is assumed, the SE gain of the MS
SNR-based algorithm does not take interference to neighboring BSs. To on the subcarrier can be obtained by
into consideration the uplink-received simplify the analysis, simple input
interference power level. This type of simple output (SISO) is assumed, and
SE gain =log(1+SINR (Home)New )
- log(1+ SINR (Home)Original )) (4)
algorithm is simple to implement and the parameters for the model are
always convergent. Fractional power •Ps : The MS uplink transmission where
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P60
Uplink Power Control for MIMO-OFDMA Cellular Systems Special Topic
Rongzhen Yang and Hujun Yin
PS
CL 0
SINR (Home)Original = (5) Home Channel Virtual Cell Channel
NI 0 Home Cell Loss: CLH Loss: CLi
(PS +ΔPS )
SINR (Home)New = CL 0 (6)
NI 0
Combining (4), (5), and (6), SEgain can Virtual Neighbor Cell
be expressed as
ΔPS
SE gain = log 1+( CL 0
NI 0 + PS
). (7) ▲Figure 2. Uplink interference from one MS.
CL 0 theory and is impossible to be BS (an MRC receiver is assumed).
The SE loss on the subcarrier in one implemented in practice. The algorithm It is further assumed that each
neighbor BSi (i = 1-N ) can be requires the home BS or MS to receiving antenna on the BS suffers
expressed as accurately know the received signal similar noise and interference level
power and interference power on all NIH,Ant , and the path loss from
SE loss(i ) = log(1+ SINR (i ) Original )
neighbor BSs on each subcarrier at the transmission antenna to each receiving
- log(1+ SINR (i ) New ) . (8)
time of uplink transmission. Making antenna is similar to CLH , then (14) can
Si
where SINR (i ) Original = , some simple assumptions, we be re-written as
NIi
developed a practical SMST algorithm.
Si Nr ×(PSD Data+ΔPSD Data)
SINR (i )New = , First, we modeled one virtual neighbor 1+
NIi +ΔIi CLH ×NI H,Ant
BS (or sector) that accounts for all SE gain = log
Si is the received signal power on the interference impact on SEloss (Fig. 2). Nr ×PSD Data
1+
ΔPS We assume that all BSs have the CLH ×NI H,Ant
subcarrier of BSi , and ΔI i = is
CLi same downlink transmission power ΔPSD Data
the increased inference power caused level. The total downlink reference CLH
by ΔPS. interference power from the virtual BS = log 1+ . (15)
Nr ×PSD Data
From (8), the total SEloss on the to the MS is NI H,Ant +
CLH
subcarrier in all neighbor BSs can be
PDL_Preamble N
PDL_Preamble
expressed as PDL_I = =∑ (11) Similarly, the new SE loss on the virtual
CLI i=1 CLi
N neighbor BS can be expressed as
SE loss = ∑SE loss(i ) . (9) where PDL_Preamble is the downlink
(
SE loss = log 1+SINR IOriginal,MRC )
I
i=1
preamble power used as the reference 1+SINR New,MRC
Therefore, the optimum power on the signal to measure downlink path loss,
Nr ×SNR I,Ant ×PNoice,Ant
subcarrier can be calculated by and CL I is the virtual downlink path 1+
NII,Ant
increasing P S of the subcarrier by ΔPS loss. The downlink reference signal = log (16)
steps from 0 until the SE gain and SE loss on power of the MS is Nr ×SNR I,Ant ×PNoice,Ant
1+
this subcarrier no longer satisfies ΔPSDData
PDL_Preamble NI I,Ant +
PDL_s = . (12) CLI
SE gain ≥ SE loss . (10) CLH
where NI I,Ant , is the noise plus
If the MS is assigned resources for Downlink SIR can be measured as interference for each antenna on the
uplink transmission by the BS, optimal P virtual BS, PNoice,Ant , is the white noise for
SIRDL = DL_ S , so that
transmission power for all assigned PDL_ I each antenna on the virtual BS, SNR I,Ant
uplink subcarriers can be calculated in is the average SNR level for each
CLI
order to achieve the overall optimum SIRDL= . (13) antenna on the virtual BS, and
CLH
cell throughput. This algorithm is called SNR I,Ant × PNoise,Ant , is used to estimate
the maximum sector throughput (MST) The SE gain can be expressed as received signal PSD for each antenna
algorithm. on the virtual BS.
SE gain = log (1+1+SINR
SINR New,MRC
Original, MRC
). (14) The optimum overall SE can be
obtained when the following condition is
4 Simplified Maximum
One virtual neighbor model is used, met:
Sector Throughput and one MS keeps the same
max (SEH +SEI ) ⇒ ΔSE
Algorithm transmission power spectral density ΔPSDData →0
(PSD) for all data tones (PSD Data). Only = SE gain-SE loss = 0 (ΔPSDData →0 ) . (17)
4.1 MST Algorithm Simplified Form one transmission stream is used, and
The MST algorithm is only useful in there are Nr receiving antennas at the From (13), (15), (16) and (17), we can
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P61
Special Topic Uplink Power Control for MIMO-OFDMA Cellular Systems
Rongzhen Yang and Hujun Yin
derive and is converted into a linear value to Equation (23) is the final derivation of
align with the linear result of (19). SMST adopted into IEEE 802.16m [3]
NI I,Ant 1
SINR H,Ant = ×(1+ )× Equation (21) is a core part of the IEEE for data-channel power control. 16 m
NI H,Ant Nr ×SINR I,Ant
802.16m [3] uplink power-control control-channel power-control design
SIR DL- 1 (18) algorithm. was discussed in [14].
Nr
Equation (21) is the SMST algorithm
where, SINR H,Ant is the received average combined with minimum target SINR
SINR for each antenna on the home BS, threshold. It expresses the suitable
5 Algorithm Evaluation and
and uplink target SINR in each receiving Comparison
antenna on a home BS. If OLPC is
SIN I,Ant×P Noise,Ant
SINR I,Ant = . applied, the PSD of each data 5.1 Evaluation Considerations
NII,Ant
subcarrier used by the MS is Before evaluating uplink power
Equation (18) can be further control algorithms, the following general
simplified: PSD Tx =L+SINRTarget (dB)+NL+Offset (22) points for system-level simulation need
where L is the average downlink path to be considered:
SINR H,Ant =γ ×SIR DL- 1 (19) loss measured by MS base on BS •sector-cumulated throughput and
Nr
transmission power level and received cell-edge throughput. These are the
where γ is a derived parameter to signal power level, NI is the average metrics of overall performance
control the interference to other cells: noise and interference level for each determined by uplink power-control
subcarrier at the BS antenna, the algorithms.
γ=
NII,Ant
NIH,Ant (
× 1+
1
Nr×SINR I,Ant
. ) (20) information is broadcasted from BS to
MS, and Offset is the MS-specific
•fairness control curve. In general,
the uplink power control algorithm
The parameter,γ , is linearly related power offset decided by the BS. provides a trade-off between
to the ratio of virtual-cell average NI If (21) is used for conventional CLPC sector-cumulated throughput and
level to home-cell average NI level. design, MS needs to report the cell-edge throughput. One new-form
High γ results in high interference over measured downlink SIR value SIRDL to curve is used to show the sector SE and
thermal (IoT), and low γ results in low BS periodically, and BS then uses (21) cell-edge SE (Figs. 3, 4, 8 and 9).
IoT. to calculate the target SINR. Compared •IoT control. IoT is the key metric of
with the measured SINR, the difference uplink interference in a system
4.2 Limitation of Minimum Transmission is compensated by CLPC commands. evaluation. Effective IoT control needs
Rate Equation (21) is the solution for uplink to be proven for an uplink power control
Equation (19) provides the optimal single stream. In IEEE 802.16m [11], algorithm.
solution for the SMST algorithm. The two options for uplink multistream As (22) and (23) show, there are
resulting target SINR on each antenna power control were discussed. The some key parameters that need to be
is expressed as a linear value. From power level of each stream in uplink evaluated and discussed for real
(19), some MSs could have negative multistreams can be kept the same as implementation:
target SINR because of the very low the single stream or the power level of •γ plays a key role in IoT and
measured downlink SIR. For these MSs, each stream in uplink multistreams can fairness. In the evaluation, the results of
the results of (19) indicate that any be reduced to keep the sum of different γ values are discussed
power assigned to these MSs reduces power/interference similar to the single •SINR MIN (dB) is the minimum
overall throughput. Because all active stream. Both options improve threshold for cell-edge user SINR.
MSs must support a minimum performance in different cell sizes and •L is the average downlink
transmission rate to keep them online, in different scenarios, so there is one path-loss measured by the MS. This is
the minimum target SINR (to support additional control parameter added into key to deciding the transmission power
the minimum transmission rate) should (18) to support both options: level. For different product
be set as the threshold. Then, (19) is implementations, the long-term
SINR MIN
modified as follows: average or short-term average have
SINR Target =10log10(max(10 10 ,
cause different effects.
SINR MIN 1
SINRTarget =10log10(max(10 10 , γ ×SIRDL- )) •The open-loop power adjustment
Nr
is decided by the MS. The rate of power
-β ×10log10(TNS ) (23)
γ ×SIRDL- 1 )) . (21) control may be used in product
Nr where β is the newly added parameter implementation. Fast (per-frame) or
In (21), the resulting target SINR is and can be set by the BS as 0 slow (per 50 frames) control rates are
converted to decibels so that the (disabled) or 1 (enabled) for evaluated.
transmission power can be environment performance tuning. TNS
conveniently calculated. The minimum is the total number of uplink 5.2 Results and Discussion
SINR threshold is expressed in decibels multistreams. The evaluation scenario in [15]
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P62
Uplink Power Control for MIMO-OFDMA Cellular Systems Special Topic
Rongzhen Yang and Hujun Yin
▼Table 1. Key parameters for IEEE 802.16 m ▼Table 2. Evaluation results of different gamma values
ULPC evaluation scenario
Gamma SINR MIN (dB) Sector Throughput(Mbit/s) Cell-Edge Throughput(kbit/s) Sector SE Cell-Edge SE
Parameter Value 0 0 2.7083 256.5689 0.7222 0.0684
Hexagonal Grid,19 3-Sector
Site Layout 0.2 0 3.2387 262.6901 0.8636 0.0701
Sites, Wrap-Around
Carrier Frequency(GHz) 2.5 0.4 0 3.7738 252.3136 1.0064 0.0673
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P63
Special Topic Uplink Power Control for MIMO-OFDMA Cellular Systems
Rongzhen Yang and Hujun Yin
0.5
The average path loss shows minor loss :γ=0.6
in a cell-edge SE because :γ=0.8
0.4
instantaneous path-loss estimation :γ=1.0
allows the MS to perform fast-link 0.3 :γ=1.2
adaption by OLPC power change. :γ=1.4
Similarly, Fig. 4 shows that a slow 0.2
power-updating rate of 50 has some
gain in maximum sector SE and some 0.1
loss in cell-edge SE.
0
Other aspects of uplink power 0 500 1000 1500
control, such as user throughput User Throughput (kbit/s)
distribution, IoT control, and power CDF: cumulative distribution function
distribution, are shown in one of the
following four cases, because it is ▲Figure 5. User throughput CDF by control factor γ.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P64
Uplink Power Control for MIMO-OFDMA Cellular Systems Special Topic
Rongzhen Yang and Hujun Yin
0.5
:γ=0.6 dBm with fixed α = 0.8, the sector SE is
0.4 :γ=0.8 0.9375, and cell edge SE is 0.0447.
:γ=1.0 Compared with the SMST algorithm
0.3 :γ=1.2
results without optimum searching for
:γ=1.4
SINR MIN ( the value is just set as 0 dB),
0.2 there is still big performance gap. If
SMST maintains the same sector SE as
0.1 FPC optimum point of 0.9375, the SMST
can provide cell-edge SE of 0.0687,
0.0
0 5 10 15 20 which is a 53.69% gain over 0.0447 of
IoT (dB) FPC optimum cell-edge SE. If SMST
CDF: cumulative distribution function maintains the same cell-edge SE as
FPC optimum point of 0.0447, the SMST
▲Figure 6. IoT CDF by control factor γ. can provide sector SE of 1.2404, which
is a 32.31% gain over the 0.9375 of FPC
incorporated into 3GPP E-UTRA after large range. optimum-sector SE. The average
Release8. For the scenario defined in Table 1, performance gap between SMST and
During the development of 3GPP LTE the results of SMST and FPC are shown FPC is 43%. In the second method of P0
specifications, FPC was proposed on Fig. 8. selection, P0 is transformed fromα value
[18]-[20] and used [17] as an OLPC The performance trade-off of FPC is and cell-edge target SNR, as
method for uplink data channel physical located in the range of P0 larger suggested in [23] and given by
uplink shared channel (PUSCH). than -79 dBm. If the P0 is not well
The key to FPC is to compensate the selected, for example, when P0 < -79 P0=α ×(SNR Target +PSDn)+(1-α )×PSD Max (26)
fraction of the path loss by control factor
α, which can be expressed as PSD of γ=0.2
each subcarrier, the same as in (22): 1.0
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P65
Special Topic Uplink Power Control for MIMO-OFDMA Cellular Systems
Rongzhen Yang and Hujun Yin
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P66
Mobile Backhaul Solutions Special Topic
Li Mo, Fei Yuan, and Jian Yang
Abstract: In this paper, we give an overview of mobile backhaul solutions and propose an MPLS-centered solution that takes into
account timing synchronization, OAM, and protection. We also propose an evolved protection bandwidth allocation mechanism that
makes the transport network as efficient as possible.
Keywords: mobile backhaul; multiprotocol label switching (MPLS); optical networking; protection
R
ecent advances in cellular •timing synchronization protection path for an MPLS-based
technology have necessitated •frequency and phase network needs to be articulated.
significant improvement in the synchronization for optimal handover •OAM techniques of data-centered
mobile backhaul network. With •coordinated multipoint networks need to be incorporated into
increased use of smartphones and transmission (CoMP) and interference traditional transport networks.
laptops, mobile communication has cancellation in LTE-Advanced •Traffic aggregation needs to be
been moving from voice-centered •enhanced OAM performed closer to the base stations.
infrastructure to data-centered •multiple transport modes that In this paper, a new trend in mobile
infrastructure. support data traffic and TDM traffic for radio access network architecture is
New mobile technologies, such as legacy base stations briefly discussed. In this architecture,
high-speed packet access (HSPA+) •symmetric bandwidth requirements called cloud radio access network
and LTE, have significantly increased for the up-stream and down-stream (C-RAN), baseband processing of the
downlink and uplink speeds over the between the central office and base radio signal is centralized. Although
radio link. Improvements over the radio stations. there are many advantages to a
access link demand similar In traditional wired broadband centralized approach in terms of power
improvement in bandwidth on the access, bandwidth is usually consumption and project engineering,
backhaul network. asymmetrical. As the use of cloud among other things, C-RAN also
However, current backhaul networks services increases, symmetrical requires a very different transport
are based on legacy technologies that bandwidth in both directions becomes network between the radio amplifiers
are optimized for voice transport. To more important when designing the and antenna, and the central location
benefit from advances in radio access network. where baseband signals are processed.
link technologies, the backhaul network In this paper, an
needs to be transformed from a slow, Ethernet/MPLS-based solution is
TDM-centered network into a proposed to satisfy mobile backhaul 2 Basic Requirements
high-speed, data-centered network. requirements. The solution addresses During mobile communication,
Driven by bandwidth demand of the the following issues: connectivity from the base station to the
Internet, engineers have been building •Depending on the situation, the central office occurs over the mobile
high-speed, data-centered networks migration path from the backhaul backhaul network.
for many years. However, building a network (based on legacy technology) In this backhaul network, traffic for
mobile backhaul network is very to the new infrastructure (capable of different generations of the radio
different from building a commercial supporting the new cellular access network is transported. Different
data network. A mobile backhaul technologies such as LTE) may be generations of cellular technologies
network requires different. have different architectural names to
•QoS-based traffic with strict •High availability necessitates indicate cell-site equipment and the
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P67
Special Topic Mobile Backhaul Solutions
Li Mo, Fei Yuan, and Jian Yang
an MPLS-based network.
Gateway Another contender for mobile
CO:central office
MSC-S: message switch center-server CO Site ◀Figure 1. backhaul could be optical transport
Mobile backhaul network. network (OTN) with ODU-0 to transport
gigabit Ethernet signals. But an OTN
corresponding equipment in the central carrier-sections, is around network is TDM in nature and cannot be
office (CO). In this paper, the cell-site 300-400 Mbit/s. In LTE, the physical as flexibly deployed as an MPLS
equipment is called base station unless layer can be gigabit Ethernet, 10G network.
otherwise noted (e.g. in LTE, the base Ethernet (in sharing mode), GPON, or
station is eNodeB). The central office 10G EPON. 3.1 QoS and CoS
equipment is called message switch Normal EPON is not included in the To make the network scalable so that
center-server (MSC-S) for voice, and physical layer because the sharing it can provide many different types of
gateway for data (e.g. SAE-GW for capability of the link is a maximum of services, CoS is usually used with
EPC, SGSN for UTRAN). The backhaul 1G. 10G GPON is also omitted because queuing and discard techniques. A
network is shown in Fig. 1. current 10G GPON has asymmetric mobile backhaul network needs to
Connectivity between the base bandwidth. queue and discard priorities from the
station and the CO site is determined Fig. 2 shows the basic requirements radio access network.
by the technology used in the RAN, the for a mobile backhaul for advanced In almost all radio access
location of the cell site, the bandwidth mobile technologies such as LTE and technologies, the relative queuing and/
requirements, and local regulations. LTE-Advanced. or discard priorities are marked. After
In a GSM (voice and limited data baseband processing of the radio
traffic) and UMTS/CDMA 2000 signals at the entrance of the mobile
(voice-centered with increasing data 3 MPLS-Based Solutions backhaul network, the markings on the
traffic) mobile network, many cell sites In this section, the specifics of using radio signals need to be remapped
use microwave as backhaul multiprotocol label switching (MPLS) onto the technologies used in the
technology. However, in the LTE era of technology for mobile backhaul backhaul network.
wide spectrum and high data rate, the application are articulated. The core technology used in the
microwave-based backhaul network Apart from the final access link mobile backhaul network is MPLS over
does not have adequate bandwidth. In (which can be xPON-based), inside Ethernet. For layer-2 (Ethernet), the
this paper, we discuss a fiber-based the mobile backhaul network, the 802.3p bits can indicate the priority of
mobile backhaul network. layer-2 technology is Ethernet. With the packet. At the starting point of the
The current practice for building Ethernet, technologies such as MPLS label switched path (LSP), the
mobile backhaul networks is a flat Layer-2/Layer-3 virtual private network EXP bits can also indicate the priority of
IP-centered architecture. To support (VPN), pseudo-wire for circuit the packet.
legacy mobile technologies, one of the emulation in supporting TDM traffic,
key requirements of a mobile backhaul and MPLS can be easily supported. 3.2 Timing Synchronization
network is support for a native TDM For mobile backhaul applications, a In a mobile backhaul network with
interface. Support for such an interface pure IP-based network with no MPLS Ethernet and MPLS, timing needs to be
usually comes in the form of circuit can also provide adequate service with synchronized for better handoff
emulation, for example, PWE3 increases in delay and jitter. These support, better voice/video quality,
emulation, in the flat IP network.
Different generations of mobile
MSC-S
network also have different delay and
jitter requirements. For LTE, Next Base Station
Generation Mobile Network Alliance
(NGMN) specifies a maximum delay of 万
万 万万 万 万万
万
TDM and IP Backhaul Network
Access Technologies
specific jitter requirements. Gigabit Ethernet
The common requirement for LTE Figure 2. ▶ 10 G Ethernet Gateway
bandwidth, which also depends on Mobile backhaul network GPON, 10 G EOPN CO Site
spectrum width and the number of access technologies. CO:central office MSC-S: message switch center-server
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P68
Mobile Backhaul Solutions Special Topic
Li Mo, Fei Yuan, and Jian Yang
Protection path with Bandwidth Reservation ◀Figure 3. assurance. When high availability is
Gateway Protection path and required, bandwidth also needs to be
CO:central office
CO Site
bandwidth reservation in assured when the network experiences
MSC-S: message switch center-server the MPLS network. a limited fault, such as a single fiber cut
or a single MPLS transport nodal failure.
lower interference, and better also minimizes service interruptions The current method of supporting
bandwidth (e.g. CoMP). and operational repair time. high-availability is to provide a working
There are a number of ways of •enhancing network availability by path and a protection path. The
supporting timing synchronization, ensuring that defects, such as those required bandwidth is allocated on both
each of which has different advantages resulting in misdirected customer the working path and protection path. If
and disadvantages: traffic, are detected, diagnosed, and the working path experiences a failure,
•GPS based timing. This is accurate dealt with before a customer reports the the protection path is used. Because
for frequency and can also be used for problem. the bandwidth is also allocated on the
phase synchronization. The drawback •meeting service and performance protection path, bandwidth is assured
is that GPS is often unavailable objectives. OAM allows service-level for the backhaul network. This
because of a poor environment for agreements (SLAs) to be verified in a arrangement is shown in Fig. 3.
acquiring GPS signals or because the multimaintenance environment and The protection bandwidth reservation
U.S. government has encrypted the allows service degradation caused by scheme assures bandwidth when there
signals. packet delay or packet loss to be is limited network failure, but shared
•packet-based timing determined. protection between many different cell
synchronization (IEEE1588v2). This A backhaul solution should support sites is not taken into account. This
synchronization mechanism has been end-to-end OAM in a multivendor critical flaw is shown by the simple ring
widely evaluated over the past a few environment and simplify network topology in Fig. 4.
years. It is adequate for synchronization operations with OAM tools. In the arrangement in Fig. 4, the
in universal mobile telecommunications Currently, the telecommunication protection bandwidth reserved on the
system (UMTS) networks. A CDMA standardization sector (ITU-T) SG15 link between nodes 5 and 6 (same as
network uses GPS timing. The and Internet engineering task force other links) is the sum of the bandwidth
performance of this mechanism in (IETF) are cooperating on MPLS for the working paths A and B. If we
synchronizing timing for more transport profile (MPLS-TP). Compared assume a single failure on any node
time-stringent requirements, such as with MPLS, OAM is a very important and take into consideration the ring
CoMP, still requires further study. characteristic of MPLS-TP. Fault structure, the maximum bandwidth
•synchronous Ethernet. This management (FM) OAM functions, such required for protection is the maximum
scheme only works if Ethernet is as continuity check (CC), continuity bandwidth for working paths A and B.
provided end-to-end. In this case, the verification (CV), and
last access link also needs to be remote defect indication
Ethernet, and an xPON-based solution (RDI), can automatically
is excluded. Synchronous Ethernet can
be deployed if the carrier is not using
detect and localize
defects that occur in
1
xPON-based technologies for the network. Performance Protection Path B
access link between the backhaul
network and the base station.
management (PM) OAM
functions, such as packet
6 Working Path A
2
loss measurement (LM), Working
3.3 OAM for MPLS-Based Network packet delay Path B
OAM is an important and measurement (DM), and
fundamental in transport networks[1]. It
contributes to
throughput measurement,
can diagnose service
5 Protection Path A 3
•reducing operational complexity
and costs because it allows for efficient
degradation. OAM
functionality is also the
4
and automatic detection, localization, key for network
handling, and diagnosis of defects. It survivability and ▲Figure 4. Current working and protection-reservation method.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P69
Special Topic Mobile Backhaul Solutions
Li Mo, Fei Yuan, and Jian Yang
BN (Y (k)) is the sum of the equivalent Ba (Y (k),J (i )), and BP (Y (k),J (i )), where By using this method for reserving
bandwidths of all the working paths 0≤ i ≤KJ for any 1≤k≤K y , are protection bandwidth, the mobile
over Y (k). available once the working path and the backhaul network is efficient and
2) Bandwidth may be absent on Y (k) protection scheme is determined. competitive. A tighter upper bound
because of the failure of link i of node J With this information, we are ready to based on topology and traffic flow is
(i.e. J (i ) failure). For any working path define a quantity also possible for multiple failures, and
over Y (k) and J (i ), when there is J (i ) this will be left for future discussion.
failure, the working-path traffic is P (Y (k),J (i )),Y≠J (1)
absent if the working path goes over where P (Y (k),J (i )) = BP (Y (k),J (i ))-
J (i ) before Y (k). There is a difference Ba (Y (k),J (i )),and 0 ≤ i ≤ KJ. 5 New RAN Architecture:
between 1+1 protection [4] and MPLS If there are N nodes in the network, C-RAN
fast reroute protection [5]. The working for all values of J, where 1≤J≤N,J ≠Y, To this point, there has been no
path traffic is absent for 1+1 protection and for all values of i, where 0≤ i ≤KJ , change to the traditional mobile
and may be absent for MPLS fast a list in descending order can be network architecture, where base
reroute, depending on topology and the constructed based on P(Y(k),*) for a stations consist of the radio unit (RRU)
detour path. Depending on the particular Y (k), 1≤k≤K y. This list is and baseband processing unit (BBU).
protection scheme, the working-path given by The RRU and BBU are connected
traffic may be absent if the working with fiber optics. The RRU is located
P (Y (k),J 1(i 1)),P (Y (k),J 2(i 2)),
path goes over before Y (k). In this with the antenna on top of the cell
…P (Y (k),JM (iM )) (2)
paper, this bandwidth is Ba (Y (k)),J (i )), tower, and the BBU sits on the ground.
N
where 1≤ k ≤K y and 1≤ i ≤KJ. where M =∑l =1, l≠Y Kl +N-1. The notation The BBUs for many different cell sites
3) Bandwidth may be absent on J x is defined as J x∈(1, ...,N ), can be co-located to provide
Y (k) due to failure of J. For a complete J x≠Y and 1≤ x ≤M. centralized baseband processing. The
failure of J, the bandwidth absent on In the ordered list, we have resulting architecture is called cloud
link Y (k) is Ba (Y (k)),J (0)). In this case, radio access network (C-RAN) and is
P (Y (k),J 1(i 1))≥P (Y (k),J 2(i 2))≥…≥ shown in Fig. 5.
the notation is the same as that for
…P (Y (k),JM (iM )) (3)
bandwidth absent because of link Because the modulated radio signals
failure, and J (0) denotes the nodal For a single link or node failure in the are transported between RRU and
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P70
Mobile Backhaul Solutions Special Topic
Li Mo, Fei Yuan, and Jian Yang
BBU, the requirements in terms of current and future mobile access Biographies
bandwidth, delay, and jitter are much technologies. An MPLS-based mobile
Li Mo (mo.li@zte.com.cn)received his B.Eng.
more stringent than those in mobile backhaul solution is proposed, and degree from the Department of Electrical
backhaul networks. The preferred timing synchronization, MPLS OAM, Engineering, Queen’ s University, in 1989. After
graduation, he worked at IBM, Nortel, and Fujitsu
method of transport between the BBU and protection have been discussed. A before joining ZTE in 2001. Currently, Dr. Mo is the
and RRU are direct fiber connection or new mechanism for reserving chief architect for ZTE USA and works as marketing
wavelength division multiplexing (WDM) protection bandwidth is described. This specialist for ZTE headquarters in Shenzhen. Dr. Mo
has worked in the networking industry for more than
if conservation of fiber links is required. mechanism ensures that protection 20 years, and has numerous publications and U.S.
Because of fiber connectivity in bandwidth is efficiently reserved for a patents. His research interests include fixed and
mobile core networks, session control, QoS, and
advanced mobile technology, single fault and that an upper-bound routing. Dr. Mo is also an active member of IEEE,
variations, such as C-RAN, on protection bandwidth estimation ITU, ETSI, and IETF.
traditional RAN architecture become mechanism is provided for multiple
Fei Yuan (yuan.fei@zte.com.cn) is the acting vice
feasible. faults. general manager of the Department of Standard
C-RAN architecture is still in its Development and Industry Relations at ZTE
Corporation. He has been working in
infancy. Further division of the workload References telecommunications for eighteen years and has a
between BBU and RRU according to [1] Requirements for Operations, Administration, and strong technical background in a range of fields. Fei
Maintenance (OAM) in MPLS Transport Networks, Yuan guides the standardization and advanced
transport requirements warrants further RFC5860, 2010. research teams with innovative strategies and ideas.
study. [2]ChengShang Chang, Performance guarantees in
communication networks, Springer, 2000. Jian Yang (yang.jian90@zte.com.cn) joined ZTE
[3] Jean-Yves Le Boudec and Patrick Thiran, Network after receiving her bachelor's degree from XiDian
Calculus, Springer, 2001. University. She is now the vice general standard
6 Conclusion [4] Linear Protection Switching for Transport MPLS engineer of bearer networks and has made many
Networks, ITU-T G.8131, 2006. contibutions to MPLS-TP standard development,
In this paper, we have reviewed [5] Fast Reroute Extensions to RSVP-TE for LSP both in the ITU-T and IETF.
mobile backhaul requirement for Tunnels, RFC 4090, 2005.
⬅ From P.62 Kwak,“Uplink Power Control Rule for Uplink (E-UTRA); Radio Resource Control (RRC);
Multi-Streams Transmission.”[Online]. Available: Protocol specification, 3GPP TS 36.331 v9.2.0.
[Online]. Available: http://www.ieee802.org/16/tgm/contrib/ [23]“Performance of Uplink Factional Power Control in
http://www.ieee802.org/16/tgm/contrib/ C80216m-09_1596r1.ppt UTRAN LTE” , VTC Spring 2008, IEEE.
C80216m-09_0634.doc [12] Rongzhen Yang, Hujun Yin, Yang-seok Choi and [24] Anil M.Rao,“Reverse Link Power Control for
[5] Rongzhen Yang, Ali T. Koc, Papathanassiou Apostolos Papathanassiou,“Discussion of CLPC Managing Inter-cell Interference in Orthogonal
Apostolos, W. G, Hujun Yin, Nageen Himayat, by PC-A-MAP IE for TDD and FDD.”[Online]. Multiple Access System” , VTC-2007 Fall, IEEE.
Yang-seok Choi and Shilpa Talwar,“Additional Available: http://www.ieee802.org/16/tgm/contrib/
material related to the Recommended AWD Text C80216m-09_2661.ppt
Proposal section 11.1 in C802.16m-09/0546 (Power [13] Rongzhen Yang, Hujun Yin, Yang-seok Choi and
Control).”[Online]. Available: http://www.ieee802. Apostolos Papathanassiou,“Uplink Power Control Biographies
org/16/tgm/contrib/C80216m-09_0703.ppt Offset Design Discussion.”[Online]. Available:
[6] Jeongho Park, Jaehee Cho, Heewon Kang, Hokyu http://www.ieee802.org/16/tgm/contrib/ Rongzhen Yang (rongzhen.yang@intel.com)
Choi, Jihyung Kim, Wooram Shin and Dong Seung C80216m-09_2660.ppt received his B.S. and M.S.degrees in electrical
Kwon,“Proposed Text of Power Control Section for [14] Rongzhen Yang, Xinrong Wang, Hongmei Sun, Y. engineering from Shanghai Jiaotong University in
the IEEE 802.16m Amendment Working Document Zhu, Yi Hsuan, Alexei Davydov, Hujun Yin, 1997 and 2000. He was a software engineer at
(revision1).”[Online]. Available: Yang-seok Choi, Jeongho Park, Wookbong Lee Intel's China Software Lab from 2000 to 2002, a
http://www.ieee802.org/16/tgm/contrib/ and Wilson Tim,“Discussion of Control Channel researcher at Intel Research from 2003 to 2004, and
C80216m-09_0612r1.pdf Power Control Design.”[Online]. Available: a wireless security architect at Intel APAC Wireless
[7] P. Wang, A. Boariu, Joon Chun, Xiaoyi Wang, http://www.ieee802.org/16/tgm/contrib/ Security from 2005 to 2006. He is currently a system
Zexian Li,“Proposed Text on Power Control Section S80216m-09_2846.ppt engineer of advanced wireless technology at Intel
in Amendment Text.”[Online]. Available: http://www. [15] Wookbong Lee, jeongho. park, Yang Rongzhen, Corporation. Rangzhen Yang’s research interests
ieee802.org/16/tgm/contrib/C80216m-09_0545.doc Koc, Ali T; Zeira Eldad, W. Fan, robson domingos, include technology development for advanced
[8] Rongzhen Yang, Apostolos Papathanassiou, W. minoh, HanGyu Cho; Dong-cheol Kim; wireless standards in personal, local, and wide area
Guan, Ali T. Koc, Hujun Yin and Yang-seok Choi, JongYoung Han; Fred Vook, Mark Cudak, Andre wireless networks. He has written more than 10
“Supporting material for Uplink OLPC Proposal barreto, and Xiaoyi.Wang,“Evaluation journal and conference articles and has more than
C80216m-09/0844.”[Online]. Available: http://www. Methodology for Uplink Power Control Algorithm.” 40 patents.
ieee802.org/16/tgm/contrib/C80216m-09_0845.ppt [Online]. Available:
[9] Jeongho Park, Suryong Jeong, Jaehee Cho, http://www.ieee802.org/16/tgm/contrib/ Hujun Yin (hujin.yin@intel.com) received his B.S.
Heewon Kang and Hokyu Choi,“Performance C80216m-09_1167.doc and an M.S. degrees from Shanghai Jiaotong
comparison between various Open Loop Power [16]“IEEE 802.16m Evaluation Methodology University in 1995 and 1998. He received his Ph.D.
Control schemes.”[Online]. Available: http://www. Document.”[Online].Available: http://wirelessman. in electrical engineering from the University of
ieee802.org/16/tgm/contrib/C80216m-09_1042.pdf org/tgm/core.html#08_004 Washington in 2001. From 2001 to 2002, he was a
[10] Rongzhen Yang, Apostolos Papathanassiou, W. [17] Evolved Universal Terrestrial Radio Access senior researcher at AT&T Research, where he
Guan, Ali T. Koc, Hujun Yin, Yang-seok Choi, (E-UTRA); Physical layer procedures, 3GPP TS focused on wireless network management and
Jeongho Park, Suryong Jeong, Jaehee Cho, 36.213 v9.1.0. optimization research. From 2002 to 2004, he was a
Heewon Kang, Hokyu Choi, Dong-cheol Kim, [18] Evaluation of Slow Power Control Techniques on senior wireless architect at ViVATO, where he was
Wookbong Lee, HanGyu Cho, Jihyung Kim, Dong the System Performance of Uplink SC-FDMA, responsible for smart antenna enhanced 802.11
Seung Kwon, Xiaoyi Wang and Zexian Li,“IEEE 3GPP R1-061754. WLAN network solutions. Dr. Yin is currently a
802.16m Amendment Text Proposal for Uplink [19] Uplink Power Control for E-UTRA, 3GPP principal engineer and director of advanced
Open-Loop Power Control.”[Online]. Available: R1-062612. wireless technology at Intel Corporation. He leads
http://www.ieee802.org/16/tgm/contrib/ [20] Uplink Power Control for E-UTRA, 3GPP technology development for advanced wireless
C80216m-09_0844r4.doc R1-062861. standards in personal, local, and wide area wireless
[11] Rongzhen Yang, Apostolos Papathanassiou, W. [21] Uplink Power Control for E-UTRA - Range and networks. Dr.Yin has more than 20 journal and
Guan, Hujun Yin, Yang-seok Choi, Dong-cheol Representation of P0, 3GPP R1-074850. conference articles and has over 50 patents.
Kim, Wookbong Lee, HanGuy Cho and Jin Sam [22] Evolved Universal Terrestrial Radio Access
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P71
Lecture Series The Internet of Things and Ubiquitous Intelligence (4)
Dongliang Xie and Yu Wang
A
distributed CPU is the core term“cloud”vividly describes the model must be very simple, and
component in cloud computing characteristics of cloud computing: complicated parallel execution and task
and plays a pivotal role in the virtualization, transparency, scalability, scheduling must be hidden in the
development of the Internet of and elasticity. It is an apt term for background. The programming model
things. Combining cloud computing describing the abstraction of underlying must be transparent, enabling users to
and the IoT can boost the entire infrastructure. develop programs for specific
industry and value chain if capability Cloud computing integrates purposes.
and resource sharing, quick service distributed computing, parallel Most cloud computing systems adopt
deployment, expansion of new computing and grid computing. At its Google’ s MapReduce [2]. This is a
human-thing interactive services, and core, it virtualizes computing, resources distributed parallel programming model
deep data mining are optimized. When of large data centers and provides and also an efficient task-scheduling
the Internet of things reaches a certain these to users as a service. model.
level, its dependence on cloud
computing will be greater. 8.2 Key Technologies 8.3 Combining with the Internet of Things
Cloud computing is a form of Cloud computing and IoT are
8.1 Concept and Current Status data-intensive super computing. It has complementary. Development of IoT
With the development of electronics, unique data storage and management requires the powerful processing and
communications, computers, and technologies and a unique storage capabilities of cloud
network technologies, cloud computing programming model. computing. If cloud computing
is an inevitable stage in the evolution of infrastructure is used to mine, process,
turing computing to network computing. 8.2.1 Virtualization and analyze mass data collected on the
Based on the Internet, network Virtualization is at the core of cloud ubiquitous sensing layer, IoT can
computing provides applications with computing, and is the basis for all cloud quickly, accurately, and intelligently
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P72
The Internet of Things and Ubiquitous Intelligence (4) Lecture Series
Dongliang Xie and Yu Wang
manage and control the physical world process-relevant and time interconnected device layers, monitor
and provide technical support for space-focused), enabling the discrete errors in user applications and physical
ubiquitous services. The IoT will also be computing process to interact and components, and quickly recover CPS
the largest user of cloud computing tightly couple with the continuous from a system failure. Electricity CPS is
services, which will, in turn, lay the physical process. a critical infrastructure, and
foundation for more successful cloud •It is adaptive to dynamic changes guaranteeing its security is an important
computing. in the physical world; it has powerful issue.
reorganization and reconstruction
capabilities; it can esaily; and it is easy
9 Cyber-Physical System to connect with other CPS subsystems. 10 Summary
Cyber-physical system (CPS) was •Because it has multidimensional The IoT, cloud computing, and CPS
first proposed by the US National time-space complexity, CPS should be interact and couple with each other. IoT
Science Foundation (NSF) in 2006. CPS an open, trustable, and predictable is the intuitive application of CPS, and
is expected to become the third wave of embedded system. The embedded cloud computing provides technical
information technology following the computing system of CPS support for IoT information service. The
computer and the Internet. CPS interconnects and interoperates with demands of economy and society on
ubiquitous sensing, control, and other cyber systems via the Internet. the Internet of things and CPS go far
computing and the close integration of Furthermore, CPS has stepped into beyond existing applications.
human, machines, and things are the fields that are closely related to national Therefore, from the perspective of
ultimate goals of IoT. infrastructure and people’ s daily lives. economic development and technical
It is sensitive to security, and its innovation, IoT and CPS are a trend in
9.1 CPS Overview technologies and products must be the development of global information
CPS is complicated a precise and highly reliable. technologies and industry. They will
multidimensional, embedded system have a profound effect on existing
that is integrated with computing, 9.3 Challenges industrial structure, become the core
network and physical environments. To realize the goals of CPS, there are competence in the new industrial
With the integration of 3C (computation, some technical challenges to be structure, and cultivate new economic
communication, and control) addressed. A good architecture is growth.
technologies, CPS delivers real-time critical in maintaining scalability,
sensing, dynamic control, and durability, diversity and continuous
information services for large technical innovation in CPS [3]. It is also References
[1] Q.Chen and Qianni Deng,“Cloud computing and
engineering systems. Through the key to customer investment. its key technologies,”in Journal of Computer
interactive, cyclic feedback in To ensure sustainable development Applications, vol.29,no.9,pp.2562-2567,2009.
[2] J.Dean and S. Ghemawat,“MapReduce:Simplied
computing and physical processes, of CPS, the architecture should be Data Processing on Large Clusters,” in Proc. 6th
CPS closely integrates these processes stable, a clear roadmap should be USENIX Symp. on Operation Syst. Design and
and adds or expands new functions in devised that can lead to a targeted Implementation (OSDI'04), New York,2004,
pp.137150.
real time. In this way, a physical entity study based on core CPS architecture, [3] T.Aldridge, G.Allee and A.Gorius,“Syst.
can be monitored and controlled in a and a development schedule should be Architecture and Industry Structure as Interrelated
Foundations for Progress in Cyber-physical Energy
secure, reliable, efficient, and real-time in place so that technical achievements Syst.,”in Proc.National Workshop on Research
manner [5]. are planned and predicted. The natural Directions for Future Cyber-physical Energy Syst.,
development cycle should be balanced. Baltimore,2009.
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P73
ZTE Communications
Table of Contents Volume 9, Numbers 1-4, 2011
Volume-Number-Page
GUEST PAPER
Chinese Tri-Network Convergence: Characteristics and Challenges………………………………………………Hequan Wu 9-1-01
SPECIAL TOPICS
Mobile Cloud Computing and Applications
Guest Editorial ……………………………………………………………………………………………………… Chengzhong Xu 9-1-03
A Survey of Mobile Cloud Computing ………………………………………… Xiaopeng Fan, Jiannong Cao, and Haixia Mao 9-1-04
Mirroring Smartphones for Good: A Feasibility Study ……… Bo Zhao, Zhi Xu, Caixia Chi, Sencun Zhu, and Guohong Cao 9-1-09
A Cloud-Based Virtualized Execution Environment for Mobile Applications…………………………………Shih-Hao Hung,
Tei-Wei Kuo, Chi-Sheng Shih, Jeng-Peng Shieh, Chen-Pang Lee, Che-Wei Chang, and Jie-Wen Wei 9-1-15
Building a Platform to Bridge Low End Mobile Phones
and Cloud Computing Services……………………………………………Fung Po Tso, Lin Cui, Lizhuo Zhang, and Weijia Jia 9-1-22
WiFace: A Secure Geosocial Networking System
Using Wi-Fi Based Multihop MANET …………………… Lan Zhang, Xuan Ding, Zhiguo Wan, Ming Gu, and Xiangyang Li 9-1-27
A Case for Cloud-Based Mobile Search …………………… Yan Gao, Li Fu, Zhenwei Zhang, Shengmei Luo, and Ping Lu 9-1-33
An On-Demand Security Mechanism for Cloud-Based
Telecommunications Services ………………………… Zhaoji Lin, Ping Lu, Shengmei Luo, Feng Gao, and Jianyong Chen 9-1-37
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P74
ZTE Communications
Table of Contents Volume 9, Numbers 1-4, 2011
Volume-Number-Page
Design of Software-Defined Down-Conversion and Up-Conversion:
An Overview………………………………………………………… Yue Zhang, Li-Ke Huang, Carsten Maple, and Qing Xuan 9-4-10
Practical Non-Uniform Channelization for Multi-Standard
Base Stations…………………………………………………………Álvaro Palomo Navarro, Rudi Villing, and Ronan J. Farrell 9-4-15
Crest Factor Reduction for OFDM Using Selective Subcarrier Degradation ……………………………… R. Neil Braithwaite 9-4-25
An Antenna Diversity Scheme for Digital Front-End
with OFDM Technology…………………………………………………… Fa-Long Luo, Ward Williams, and Bruce Gladstone 9-4-32
A Histogram-Based Static-Error Correction Technique
for Flash ADCs ………………………………… Armin Jalili, J Jacob Wikner, Sayed Masoud Sayedi, and Rasoul Dehghani 9-4-35
RESEARCH PAPERS
Multifrequency Networking Solution for TD-SCDMA ……………………………… Min Jin, Wenbo Wang, and Mugen Peng 9-1-41
ZP-CI/OFDM: A Power Efficient Wireless Transmission Technology …………………… Pei Gao, Xiaohu Chen, Jun Wang 9-1-45
Research on the Next Generation Naming System …………………………………………………Fuhong Lin, Changjia Chen 9-1-49
Privacy-Preserving Protocol for Data Stored in the Cloud …………………………… Hongyi Su, Geng Yang, and Dawei Li 9-2-36
A Mobility Management Solution Based on ID/Locator Separation ………… Yuhong Li, Yunjing Hou, and Shiduan Cheng 9-2-39
Self-Adaptive QoS Control in Cognitive Networks That Is Based
on Service Awareness………………………………………………………………Chengjie Gu, Shunyi Zhang, and Yanfei Sun 9-2-44
Security Service Technology for Mobile Networks ………………………………………… Aiqun Hu, Tao Li, and Mingfu Xue 9-3-49
DEVELOPMENT FIELDS
Adaptive Multiantenna Technology …………………………………………………Huahua Xiao, Dengkui Zhu, and Liujun Hu 9-1-54
A P2PSIP System with Intelligent Routing Function
on the Media Plane ……………………………………………… Yongsheng Hu, Zhenwu Hao, Jun Wang, and Naibao Zhou 9-2-49
Research on LTE Network Coverage Planning ………………………………………………………… Jun Gu and Ren Sheng 9-3-55
OPERATIONAL APPLICATIONS
Green Base Station Solutions and Technology ………………………………………………… Zhiping Chen and Licun Wang 9-1-58
Architecture and Key Technology of Distributed Intelligent
Open Systems ………………………………………………………………… Xiaoyu Tong, Yunyong Zhang, and Bingyi Fang 9-2-53
Cloud Computing in Mobile Communication Networks……………………………………………………………Xinzhi Ouyang 9-3-59
LECTURE SERIES
The Internet of Things and Ubiquitous Intelligence (1) ……………………………………………… Dongliang Xie, Yu Wang 9-1-62
The Internet of Things and Ubiquitous Intelligence (2) ……………………………………………… Dongliang Xie, Yu Wang 9-2-58
The Internet of Things and Ubiquitous Intelligence (3) ……………………………………………… Dongliang Xie, Yu Wang 9-3-63
The Internet of Things and Ubiquitous Intelligence (4) ……………………………………………… Dongliang Xie, Yu Wang 9-4-68
D:\EMAG\2010-05-26/VOL8\COVER.VFT——5PPS/P75