layered-division-multiplexing theory and practice

Upload: aditya-joshi

Post on 01-Mar-2018

217 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    1/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    IEEE TRANSACTIONS ON BROADCASTING 1

    Layered-Division-Multiplexing: Theory and PracticeLiang Zhang, Senior Member, IEEE, Wei Li, Member, IEEE, Yiyan Wu, Fellow, IEEE,

    Xianbin Wang, Senior Member, IEEE, Sung-Ik Park, Senior Member, IEEE,

    Heung Mook Kim, Member, IEEE, Jae-Young Lee,

    Pablo Angueira, Senior Member, IEEE, and Jon Montalban

    AbstractAs the next generation digital TV (DTV) stan-dard, the ATSC 3.0 system is developed to provide significantimprovements on the spectrum efficiency, the service reliability,the system flexibility, and system forward compatibility. One ofthe top-priority requirements for the ATSC 3.0 is the capabil-ity to deliver reliable mobile TV services to a large variety ofmobile and indoor devices. Layered-division-multiplexing (LDM)is a physical-layer non-orthogonal-multiplexing technology to effi-ciently deliver multiple services with different robustness andthroughputs in one TV channel. A two-layer LDM structure isaccepted by ATSC 3.0 as a baseline physical-layer technology.This LDM system is capable of delivering robust high-definition(HD) mobile TV and ultra-HDTV services in one 6 MHz chan-nel, with a higher spectrum efficiency than the traditionaltime/frequency-division-multiplexing (T/FDM)-based DTV sys-tems. This paper presents a detailed overview on the LDMtechnology, and its application in the ATSC 3.0 systems. First,the fundamental advantages of the LDM over the traditionalTDM/FDM systems are analyzed from information theory pointof view. The performance advantages of the LDM are thenconfirmed by extensive simulations of the ATSC 3.0 system. Itis shown that, LDM can realize the potential gain offered bysuperposition coding over the TDM/FDM systems, by properlyconfiguring the transmission power, channel coding, and modu-lation, and using different multiple antenna technologies in themultiple layers. Next, the efficient implementation of LDM inthe ATSC 3.0 system is presented to show that the performance

    advantages of the LDM are obtained with small additional com-plexity. This is achieved by carefully aligning the transmissionsignal structure and the signal processing chains in the multiplelayers. Finally, we show that the LDM can be further integratedwith different multiple antenna technologies to achieve furthertransmission capacity.

    Manuscript received August 21, 2015; revised October 28, 2015 andNovember 13, 2015; accepted November 16, 2015. This work was supportedby the ICT Research and Development Program of MSIP/IITP under GrantR0101-15-294 through Development of Service and Transmission Technologyfor Convergent Realistic Broadcast.

    L. Zhang, W. Li, and Y. Wu are with the Communications Research CentreCanada, Ottawa, ON K2H 8S2, Canada (e-mail: [email protected];

    [email protected]; [email protected]).X. Wang is with the Department of Electrical and Computer

    Engineering, Western University, London, ON N6A 5B9, Canada (e-mail:[email protected] ).

    S.-I. Park, H. M. Kim, and J.-Y. Lee are with the Broadcasting SystemResearch Department, Electronics and Telecommunications Research Institute,Daejeon 305-700, Korea (e-mail: [email protected]; [email protected];

    [email protected]).P. Angueira and J. Montalban are with the Department of Communications

    Engineering, University of Basque Country, Bilbao 48013, Spain (e-mail:[email protected]; [email protected]).

    Color versions of one or more of the figures in this paper are availableonline athttp://ieeexplore.ieee.org.

    Digital Object Identifier 10.1109/TBC.2015.2505408

    Index TermsCloud transmission, layered-division-multiplexing, LDM, OFDM, UHDTV, channel estimation,successive signal cancellation, mobile service, multiple-layertransmission, MIMO, MISO, SIMO, SISO.

    I. INTRODUCTION

    THE BROADCASTING industry is facing a severe chal-

    lenge due to the reallocation of some existing terrestrial

    TV spectrum to meet the rapidly increasing demand for com-

    mercial broadband wireless services. To address this challenge,

    the Advanced Television Systems Committee (ATSC) started

    the process to develop the standard of the next generation

    digital broadcasting system, the ATSC 3.0 system, to pro-

    vide higher data capacity, more robust performance, better

    spectrum efficiency, and to enable new services [1]. One of

    the primary requirements for the next generation DTV is the

    capability of delivering high data rate multimedia services to

    mobile terminals, including mobile HDTV and non-real-time

    data streaming services. Delivering mobile services over a

    large coverage area requires the transmission be robust in both

    low Signal to Noise Ratio (SNR) and high Doppler distortion

    environments.

    To address these challenges, Layered-Division-

    Multiplexing (LDM), which was code-named Cloud

    Transmission (Cloud Txn), was proposed in [2] for next

    generation digital TV (NG-DTV) systems to achieve more

    efficient spectrum utilization and to provide more versatile

    broadcasting services [3]. Due to its significant performance

    advantages and the proposed simple implementation, a

    two-layer LDM was accepted as a Physical Layer (PHY)

    baseline technology for ATSC 3.0. Different from the

    traditional time-division-multiplexing (TDM) or frequency-

    division-multiplexing (FDM), LDM is a Non-Orthogonal

    Multiplexing (NOM) technology. In an LDM system, a

    layered transmission structure is used to simultaneously

    transmit multiple signals with different power levels androbustness for different services: mobile, multiple HDTV or

    Ultra HDTV (UHDTV). In a two-layer LDM system, the

    upper layer (UL), with higher power allocation, is used to

    deliver mobile services to indoor, portable and handheld

    receivers. The lower layer (LL) is designed to deliver high

    data rate services, such as UHDTV or multiple HDTV

    services to fixed reception terminals, where the operational

    SNR is usually high due to the large and possibly directional

    receive antennas. It has been shown in [4] that a properly

    designed LDM system can deliver in the UL a 720p or

    0018-9316 c 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.htmlfor more information.

    mailto:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]://ieeexplore.ieee.org/http://ieeexplore.ieee.org/http://ieeexplore.ieee.org/http://ieeexplore.ieee.org/http://ieeexplore.ieee.org/http://ieeexplore.ieee.org/http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://www.ieee.org/publications_standards/publications/rights/index.htmlhttp://www.ieee.org/publications_standards/publications/rights/index.htmlhttp://-/?-http://-/?-http://-/?-http://-/?-http://ieeexplore.ieee.org/mailto:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    2/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    2 IEEE TRANSACTIONS ON BROADCASTING

    1080p HDTV service to mobile/handheld receivers with

    very robust performance, and an UHDTV service in the LL

    which requires 20 to 30 dB SNR (4k or 8k HD resolution)

    for successful reception. One advantage of LDM is that it

    can co-exist with all the other emerging PHY technologies,

    such as the multiple antenna technologies, Non-Uniform

    constellations (NU-QAM), Bit-Interleaved-Coded-

    Modulation (BICM), Peak-to-Average-Power-Ratio (PAPR)

    reduction technologies, etc.

    In information theory, the concept of LDM has been des-

    cribed as a form of superposition coding [5]. However, its

    application in practical systems has been prohibited by the

    extra implementation required for successive signal cancel-

    lation. A similar concept has also be proposed as a non-

    orthogonal multiple user access technology in [6][8], where

    it is shown that the actual application of this technology is

    also limited by the implementation complexity.

    The key contribution of the proposed LDM structure for

    ATSC 3.0 is the identification of the layered system structure

    which not only achieves the significant performance improve-

    ment, but also provides an efficient implementation with lowcomplexity. Firstly, by properly configuring the transmission

    power, signal coding and modulation, and using different mul-

    tiple antenna technologies in the multiple layers, LDM can

    realize the potential capacity gain offered by superposition

    coding over the TDM/FDM systems. Secondly, by carefully

    aligning the transmission signal structure and the baseband

    signal processing in the multiple layers, highly accurate suc-

    cessive signal cancellation can be achieved with low additional

    signal processing complexity and low additional memory

    requirement. In other words, the LDM realizes the potential

    gain of the superposition coding with affordable complexity

    for consumer devices.

    This paper is organized as follows: Section II brieflydescribes the LDM system structure. In Section III, chan-

    nel capacity analysis is performed to reveal the fundamental

    advantage of the LDM systems over the TDM/FDM systems.

    SectionIV presents extensive simulation results on two-layer

    ATSC 3.0 systems to demonstrate the performance advantages

    of LDM. In Section V, the efficient implementation of LDM

    in ATSC 3.0 standard is described, including the additional

    modules required for both the transmitters and the receivers.

    In Section VI, a coverage analysis is presented to show that

    the LDM provides a different paradigm for mobile and fixed

    service coverage as compared to the traditional single-layer

    systems. SectionVII briefly discusses the application of mul-

    tiple antenna technologies in LDM systems for further capacityand robustness improvement. Section VIII gives more poten-

    tial application scenarios of LDM in ATSC 3.0 system for

    future investigation. Finally, SectionIX gives the conclusions.

    I I . LAYERED-D IVISION-M ULTIPLEXING(LDM)

    The block diagrams of the transmitters and the receivers

    in a two-layer LDM system are plotted in Fig. 1. Different

    from the traditional TDM/FDM systems, in the LDM sys-

    tems, multiple digital broadcasting/multicasting services are

    transmitted in different signal layers within one RF channel,

    Fig. 1. Two-Layer LDM Transmitter and Receiver.

    while each layer occupies the full frequency spectrum at full

    time duration. Therefore, this structure inherently provides full

    frequency diversity and time diversity to the signals in all

    layers.In the currently proposed ATSC 3.0 system, the transmis-

    sion signals in different layers use OFDM-based physical-layer

    waveforms, and all layers share the same OFDM signal struc-

    ture. These include the OFDM FFT size, the cyclic prefix (CP)

    duration, as well as the pilot structure. While using the

    same OFDM structure is not mandatory for LDM systems

    in general, it can greatly simplify the signal detection at the

    LDM receivers and therefore allows a low-complexity and

    power-efficient receiver implementation.

    For the rest of this paper, for all layers in an LDM sys-

    tem, the signals are assumed to use an OFDM structure with

    N subchannels, among which Kactive subchannels carry sig-

    nal power and the others are left empty (null subchannels)to provide adjacent channel interference protection. Among

    the K active subchannels, M subchannels are allocated for

    data transmission and P subchannels are used to carry pilot

    symbols.

    In an LDM system, transmissions in different layers have

    different characteristics for the delivery of different services.

    In general, the UL signal is designed to deliver robust

    mobile broadcasting services to portable, handheld and indoor

    receivers. Delivering reliable mobile broadcasting services has

    proven very challenging in the past. First of all, there is lim-

    ited antenna gain and directivity for most portable receivers.

    Secondly, mobile devices can frequently move into shadowed

    or indoor areas. Both result in very low received signal power.Secondly, fast moving mobile receivers experience fast time-

    varying channels. Designing receivers for reliable detection

    in fast-fading channels has always been a challenging task,

    especially for OFDM-based systems.

    To deliver robust mobile services with decent coverage

    areas, the UL signal in an LDM system is designed to have

    higher transmission power and use very strong channel cod-

    ing and modulation, e.g., rate- 14

    Low-Density-Parity-Check

    (LDPC) code with QPSK modulation. This signal is optimized

    for robustness in harsh wireless channel environments and can

    provide good detection at very low SNR (lower than 0 dB),

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    3/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    ZHANG et al.: LAYERED-DIVISION-MULTIPLEXING: THEORY AND PRACTICE 3

    and therefore provides a large coverage area and less service

    outages that are caused by multipath distortion and shadow-

    ing. It can be further shown that, designing a robust receiver

    for the UL signal in fast fading channels is no longer a signif-

    icant challenge [9]. In ATSC 3.0 systems, new LDPC codes

    were designed to provide even better performance specifically

    under low SNR conditions [10][13].

    The LL is usually designed to deliver UHDTV or multiple

    full HDTV services, which usually require much higher SNRs

    for successful reception. These services are intended for fixed

    receivers, which have large antennas at high locations (e.g.,

    rooftop), and sometimes directional antennas that are capable

    of providing significantly higher antenna gain as compared to

    mobile devices. Therefore, the LL is usually designed to have

    lower transmission power, weaker channel coding and larger

    signal constellation (i.e., 256QAM or higher). When desired,

    more transmission layers can be added to deliver other types

    of services.

    As shown in Fig. 1.(a), for a two-layer LDM system, the

    data of each service is first processed by its own physical-

    layer signal processing modules, including channel encod-ing, interleaving etc.. The interleaved bit sequence is then

    modulated into QPSK/mQAM symbols. A block of M sym-

    bols are allocated to the data subchannels of one OFDM

    symbol, where pilot symbols are inserted into the pilot

    subchannels.

    The frequency-domain (FD) LDM signal is generated as the

    superposition of the FD signals from the two layers as,

    X(k)= XUL(k)+gXLL(k) (1)

    where XUL(k) and XLL(k) are FD symbols from the UL and

    the LL, X(k) is the combined LDM symbol, and kis the sub-

    channel index. The injection level, g

    , defines the power levelof the LL signal relative to the UL signal.

    The injection level, g, determines the power allocation

    among the two layers. Since the UL signal is designed to

    have higher power, g has a real value in [0, 1), where a g = 0

    results in a single-layer system.

    The block diagram of an LDM receiver is shown

    in Fig. 1.(b). The received LDM signal can be expressed as,

    Y(k) = XUL(k)H(k)+gXLL(k)H(k)+N(k) (2)

    where Y(k) is the received symbol in the kth subchannel, and

    N(k) contains AWGN noise and other additive interferences.

    To decode the UL signal, the lower-power LL service is

    treated as an additional interference. The impact of this inter-

    ference is controllable by using different injection levels. For

    example, a 5 dB injection level sets the LL signal 5 dB lower

    than the UL signal. The injection level of an LDM system is

    usually selected to meet the service requirements of the two

    layers.

    To decode the LL signal, the receiver first needs to cancel

    the UL signal. From (2), the decision symbols of the LL signal

    can be obtained as,

    YLL(k) = 1

    g

    Y(k) XUL(k) H(k)

    (3)

    where XUL(k) is the estimate of the UL transmission symbol

    in the kth subchannel, and H(k) is the channel estimate.

    To perform the signal cancellation in (3), the receiver

    needs to obtain the estimates of the UL transmission symbols,XUL(k). This is achieved by performing a regular UL signal

    detection including equalization, demodulation, deinterleav-

    ing, and channel decoding, generating an accurate decision bit

    sequence. The receiver then performs channel encoding, inter-

    leaving and modulation to re-construct the UL transmission

    symbols. Although this cancellation process involves complex-

    ity to perform UL channel decoding, it can provide the most

    reliable UL symbol estimates.

    The LL is typically designed to deliver high data rate ser-

    vices to fixed receivers at high SNRs, which should easily

    guarantee perfect UL signal detection, i.e., XUL(k)= XUL(k).

    Furthermore, at the high SNRs required for LL service detec-

    tion, the LDPC decoding for the UL services can usually

    achieve the perfect detection with small number of iterations,

    which results in very low additional complexity to perform

    UL signal detection [14].

    Simpler signal cancellation can be performed by mak-ing hard decisions on the UL transmission symbols without

    involving the channel decoding process. This, however, will

    inevitably introduce cross-layer interference (CLI) since the

    error performance of the UL signal without channel decoding

    will not be perfect even at very high SNRs.

    III. ACHIEVABLE C APACITIES OF THEM ULTIPLEL AYERS

    IN LDM-BASEDS YSTEMSI NFORMATION

    THEORETICAL A NALYSIS

    In this section, we conduct an information theoretical ana-

    lysis on the achievable transmission capacities of the different

    layers in an LDM system, and prove the fundamental advan-tage of the LDM system over the TDM/FDM systems when

    delivering mixed multiple services with different requirements

    on the throughput and the robustness.

    The fundamental difference of the LDM from the

    TDM/FDM is to use power allocation to the multiple layers to

    perform the service multiplexing, where all layers occupy the

    entire time-frequency resource of the channel. This concept

    is called superposition coding in the literature [5] and [15],

    where it was proved that, for two transmissions over one chan-

    nel, the best performance is achieved by using a superposition

    code operating in the full available band (or time duration),

    rather than by using separate superposition codes operating

    in disjoint smaller subbands (i.e., FDM) or sub-time-durations(i.e., TDM). This shows that LDM strictly outperforms the

    TDM/FDM in delivering multiple mixed TV services in one

    channel.

    In [16], the authors investigated the achievable rates of the

    mobile and fixed services in an LDM system with non-capacity

    achieving transmissions. It is shown that the performance

    of LDM can be worse than those of TDM/FDM systems.

    However, this only happens when the mobile service in the

    UL has an SNR threshold close to that of the fixed service

    in the LL. We would like to point out that, for practi-

    cal ATSC 3.0 system deployment, these scenarios will be

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    4/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    4 IEEE TRANSACTIONS ON BROADCASTING

    Fig. 2. Equivalent Single-Layer Signal Model of a Two-Layer LDM System.

    strictly avoided since they require more complicated signal

    cancellation techniques and result in large signal cancellation

    noise for LL signal detection.

    A. Achievable Capacities of the Multiple Layers in

    LDM-Based DTV Systems in Typical Selective

    Fading Channels

    In this section, we present the achievable transmission

    capacities in the different layers in LDM-based DTV systems.The analysis assumes the selective fading channels as more

    typical scenarios as comparing to AWGN channel.

    For a two-layer LDM system, which delivers a mobile ser-

    vice in the UL and a fixed service in the LL, the mobile

    receivers decode the UL signal treating the LL signal as addi-

    tional interference, while the fixed receivers need to first cancel

    the UL signal component from the received signal and then

    perform the LL service detection.

    When assuming perfect UL signal cancellation, the trans-

    missions in the two layers of the LDM system can be modeled

    as two parallel independent transmissions as shown in Fig. 2.

    The SNRs for the UL and LL signals are calculated as,

    UL = pUL |H|2

    pLL |H|2 +pN

    LL = pLL |H|

    2

    pN(4)

    where pUL and pLL are the transmission powers of the two

    layers, His the channel response, and pN is the noise power.

    Following (4), the achievable capacities for the UL and LL

    signals are calculated as,

    CUL = E

    log2

    1+

    PUL |H|2

    PLL |H|2 +Pn

    CLL = E

    log2

    1+PLL |H|

    2

    Pn

    (5)

    For the LDM system delivering mobile services in the UL

    and fixed services in the LL, the UL and LL capacities are

    the channel capacities available for mobile and fixed services,

    i.e., Cm = CUL and Cf = CLL. For this scenario, (5) shows

    that, the achievable capacities of the mobile and fixed services

    in an LDM system are nonlinearly dependent on the power

    allocation of the two services.

    In existing DTV systems, multiplexing of fixed and mobile

    services in one RF channel is implemented by TDM in the

    DVB-T2/NGH systems [17], [18], and by FDM in the ISDB-T

    systems. For these systems, the channel capacity available for

    one service is directly proportional to its time or frequency

    allocation. For a TDM system to deliver a mobile service and

    a fixed service, the capacity allocated for the two services are

    calculated as,

    Cm = CTm

    Tt

    Cf = CTf

    Tt(6)

    where Cm and Cf are the available capacities for the mobile

    and fixed services, Tm and Tf are the times allocated for the

    mobile and fixed services, and Ttis the total time duration.

    For selective fading channels, the total channel capacity, C,

    is calculated as [19]

    C= E

    log2

    1+

    Ps |H|2

    Pn

    (7)

    where Ps and Pn are the signal and noise powers.

    It is easy to understand that(7) is directly applicable to FDMsystems, where the corresponding time-durations (Tm,Tf,Tt)

    are replaced by the spectrum occupancies (Fm, Ff, Ft).

    Eq. (6) shows that the achievable capacities of the mobile

    and fixed services in TDM/FDM systems are linearly depen-

    dent on the power allocation of the two services.

    Lets assume a DTV system simultaneously delivering a

    mobile service and a fixed service in one RF channel. The

    mobile service is configured to have robust performance with

    a low SNR threshold of 0 dB, while the fixed service is con-

    figured to have a high SNR threshold of 20 dB, which is a

    reasonable value for existing fixed receivers usually equipped

    with roof-top antennas. In Fig. 3, we plotted the achievable

    capacities for the mobile and fixed services for both LDMand TDM/FDM systems based on (5) a n d (6). The total

    transmission power is constant and shared between the two

    services. Fig. 3 shows the achievable transmission rates of

    the mobile and fixed services with different power allocations.

    The left-most point of each curve represents a single-service

    transmission where all the power is allocated for the mobile

    service, while the right-most point for the single-service sce-

    nario where only the fixed service is delivered with all the

    power. A similar curve for AWGN channel was shown in [16].

    Due to the difference in the SNR thresholds of the two

    services, Fig. 3 show an uneven tradeoff between the mobile

    and fixed service transmission rates, i.e., the fixed service can

    obtain higher additional capacity with the same amount ofpower increase. With all the transmission power, the system

    can provide a single mobile service of 0.86 b/s/Hz, or a single

    fixed service of 5.9 b/s/Hz.

    When comparing the two curves for LDM and TDM/FDM,

    it is clearly shown that the LDM always provides better trans-

    mission capacities than the TDM/FDM. The advantage of

    LDM is further illustrated by the two scenarios in Fig. 3,

    corresponding to system requirements for mobile service

    throughputs of 0.5 b/s/Hz and 0.7 b/s/Hz. For the mobile

    capacity of 0.5 b/s/Hz, using LDM provides an additional

    capacity of 2.0 b/s/Hz for the fixed service over the TDM/FDM

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    5/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    ZHANG et al.: LAYERED-DIVISION-MULTIPLEXING: THEORY AND PRACTICE 5

    Fig. 3. Achievable Mobile and Fixed Service Capacities for LDM VersusTDM/FDM.

    Fig. 4. Achievable Service Capacities with Different Fixed Service SNR

    Thresholds.

    systems; while for the higher 0.7 b/s/Hz mobile capacity, this

    advantage becomes more than 2.5 b/s/Hz.

    In Fig. 4, we compare the achievable data rates for the

    mobile and fixed services of the LDM and TDM/FDM sys-

    tems for a constant mobile service SNR threshold of 0 dB, and

    different fixed service SNR thresholds from 0 dB to 30 dB.

    It is observed that LDM offers better performance than

    TDM/FDM in all scenarios. The advantage of LDM over

    TDM/FDM is quite small when the fixed service has sim-

    ilar SNR threshold as the mobile service. The LDM curve

    almost overlaps the TDM/FDM curve for the fixed serviceSNR threshold of 0 dB; while for the fixed service SNR thresh-

    old of 30 dB, significant capacity advantage is observed. The

    higher the SNR threshold of the fixed service, the larger the

    advantage of the LDM systems. For example, for a mobile

    throughput of 0.5 b/s/Hz, LDM provides additional capacity

    of {0.04, 0.6, 2.0, 3.8} b/s/Hz for fixed service delivery, for

    the fixed service SNR thresholds of {0, 10, 20, 30} dB.

    One likely deployment scenario for the ATSC 3.0 system is

    to first deliver a fixed DTV service which has similar or bet-

    ter quality as comparing to existing DTV services delivered

    by the ATSC 1.0 system. The rest of the system capacity is

    Fig. 5. Mobile Service Capacity Advantage of LDM Over TDM/FDM withDesired Fixed Service Throughputs.

    then allocated to deliver one or multiple mobile DTV services.

    For this scenario, the advantage of LDM over TDM/FDM sys-

    tem is measured by comparing the achievable mobile servicedata rates for specific fixed service data rates. In Fig. 5, the

    extra mobile service capacities offered by the LDM system

    over the TDM/FDM systems are plotted with three desirable

    fixed service data rates: {2.0, 3.0, 4.0} b/s/Hz, correspond-

    ing to about {10, 15, 20} Mbps fixed service throughputs. An

    SNR threshold of 15 dB is assumed for the fixed services.

    The x-axis is the mobile service SNR threshold, and the y-

    axis is the additional mobile service capacity offered by the

    LDM system over the TDM/FDM systems, i.e., Mobile =

    Cm(LDM)Cm(TDM/FDM) b/s/Hz.

    It is shown in Fig. 5 that, for all fixed service throughputs,

    using LDM always provides higher mobile service through-

    put than that for TDM/FDM. The advantage of LDM is thehighest for the mobile service SNR thresholds from 4 dB to

    6 dB. For moderate fixed service data rates (i.e., 2-3 b/s/Hz),

    LDM provides over 50% mobile service throughput increase

    when comparing to TDM/FDM. The advantage of LDM over

    TDM/FDM becomes smaller when the fixed service data rate

    is closer to the limit, 4.3 b/s/Hz, which is achieved when all

    the power is allocated to the fixed service.

    For scenarios where providing mobile services is of higher

    priority, the system is deployed to first guarantee a desirable

    mobile service throughput and allocate the rest of the capacity

    to fixed services. The performance advantage of the LDM sys-

    tem is then measured by the additional fixed service capacity

    offered by LDM over the TDM/FDM systems. In Fig. 6, theadditional fixed service capacities for the LDM system over

    the TDM/FDM systems are plotted for four desirable mobile

    service data rates: {0.2, 0.4, 0.6, 0.8} b/s/Hz, corresponding

    to about {1.0, 2.0, 3.0, 4.0} Mbps mobile services. For each

    curve, the x-axis is the fixed service SNR threshold, and the y-

    axis is the additional fixed data rate of LDM over TDM/FDM

    systems, i.e., Fix= Cf(LDM)Cf(TDM/FDM) b/s/Hz.

    The first observation is that, for all mobile data rates, the

    advantage of LDM becomes more significant for fixed ser-

    vices with higher SNR thresholds. Secondly, the advantage of

    LDM is small for low mobile service data rate, and it becomes

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    6/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    6 IEEE TRANSACTIONS ON BROADCASTING

    Fig. 6. Fixed Service Capacity Advantage of LDM over TDM/FDM withDesired Mobile Service Throughputs.

    larger for higher mobile service data rates. When the desired

    mobile service data rate approaches the limit (0.9 b/s/Hz as

    shown in Fig. 4), the advantage of LDM becomes smaller

    again. This is consistent with the achievable data rates of LDMand TDM/FDM systems shown in Fig.4.

    B. Channel Capacity Distribution for LDM With

    Specific Power Allocation

    In this section, we take a closer look at the channel capacity

    distributions between the two layers of an LDM system when

    specific power allocation. Considering the general analysis in

    previous section, a specific power allocation corresponds to

    one specific point in the achievable mobile/fixed capacity pairs

    in Fig. 3.

    The power allocation of a two-layer LDM system is con-

    trolled by the injection level. For practical LDM systems, suchas the ATSC 3.0 system, it is important to select a proper

    inject level. First of all, the injection level needs to be prop-

    erly configured so that the UL can provide a desired mobile

    service throughput with strong robustness in the challenging

    mobile channels faced by the large variety of mobile receivers.

    Secondly, the LL injection level needs to be low enough to

    guarantee that a sufficiently low cross-layer-interference (CLI)

    can be achieved with low complexity from signal cancellation.

    In the current proposals, proper injection levels are defined

    from 4 dB to 10 dB, based on the extensive simulation

    results.

    In this section, we show the capabilities of the LDM sys-

    tem to deliver mixed mobile and fixed service with the typicalinjection level of5 dB, and compare them to the TDM/FDM

    systems. These results will directly confirm the simulation per-

    formances and the comparisons to the TDM/FDM counterpart

    systems as shown in [3].

    1) Channel Capacity Distribution in AWGN Channels:

    The channel capacity (b/s/Hz) of an AWGN channel is

    calculated as,

    C= log2

    1+

    Ps

    Pn

    (8)

    where Ps and Pn are the signal and noise powers.

    Fig. 7. Mobile Channel Capacities, AWGN Channel.

    The channel capacities of the two layers are calculated as,

    CUL =log2

    1+

    pUL

    pLL+ pn

    CLL =log2

    1+ pLLpn

    (9)

    Taking (8) and (9), it is easily verified that,

    CUL+ CLL = C (10)

    Eq. (10) shows that, in an LDM system, the distribution

    of the capacity in the multiple layers is solely controlled by

    the LL injection level. Furthermore, in the LDM system, the

    transmission capacity of each layer is a non-linear function

    of its power allocation, which occurs inside the logarithm

    calculation in(9).

    To demonstrate the impact of the non-linear capacity dis-

    tribution in the different layers, the channel capacities ofthe UL and LL are calculated for a two-layer LDM system

    and compared to the TDM/FDM-based systems which carry

    approximately the same mobile and fixed service throughputs.

    A two-layer ATSC 3.0 system is assumed with an LL injec-

    tion level of5 dB, i.e., the LL has a power level 5 dB lower

    than the UL, which means that 76% of the transmission power

    is allocated to the UL mobile service and 24% to the LL fixed

    service.

    For fair comparison, the TDM/FDM-based systems are con-

    figured to carry one mobile service and one fixed service in a

    single RF channel. Three TDM/FDM systems are considered,

    with time/frequency allocations of {25%, 33%, 50%} for the

    mobile service.The TDM/FDM scenario with 75% time/frequency allocated

    to the mobile service has the same power allocation as the

    LDM system with5 dB injection. However, this TDM/FDM

    system is extremely inefficient in delivering mobile services

    with reasonable throughput. It is therefore ignored in our

    comparisons.

    The channel capacities allocated to the mobile and fixed ser-

    vices in the two-layer LDM and the single-layer TDM/FDM

    systems are calculated and plotted in Fig.7. It is shown that,

    at low SNRs from 10 to 10 dB, the LDM system allo-

    cates higher channel capacity to mobile service than the three

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    7/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    ZHANG et al.: LAYERED-DIVISION-MULTIPLEXING: THEORY AND PRACTICE 7

    Fig. 8. Fixed Channel Capacities, AWGN Channel.

    TDM/FDM systems. For a bandwidth efficiency of 0.4 b/s/Hz,

    the SNR threshold of the mobile service in the LDM system

    is {1.8, 4.0, 6.0} dB better than those of the FDM/TDM sys-

    tems with {25%, 33%, 50%} mobile service allocation. For

    0.8 b/s/Hz, the SNR advantages of the LDM-based systembecome {1.8, 4.5, 7.5} dB.

    The channel capacities allocated for the fixed services in

    the LDM system and in the TDM/FDM systems are plotted

    in Fig. 8. It is shown that, the LDM system provides higher

    fixed service capacity than the TDM/FDM systems at SNRs

    higher than {10, 17, 25} dB for the {50%, 67%, 75%} fixed

    service allocations.

    In Fig.7,it is shown that the TDM/FDM system with 50%

    mobile service allocation suffers only 0.2 b/s/Hz lower mobile

    service capacity than the LDM system. However, Fig. 8 shows

    that the same TDM/FDM systems suffer fixed service capac-

    ities losses of {0.5, 1.2, 2.0} b/s/Hz at SNRs of {15, 20,

    25} dB, respectively.2) Capacity Distribution in Selective Fading Channels:

    For wireless fading channels, when the channel selectivity

    is significant in time-domain, frequency-domain or time-

    frequency-domains, one can average over many independent

    channel fades by performing channel decoding over a large

    number of coherent time or frequency intervals. Therefore, it

    is always possible to achieve reliable communication at the

    average channel capacity over independently fading channel

    intervals. Detailed analysis on the channel capacity of fast

    fading channels can be found in [19].

    This scenario is equivalent to an OFDM transmission in an

    ideal frequency-selective channel where each subchannel expe-

    riences an independent fading process. The channel capacity

    of the kth subchannel is calculated as,

    C(k)= log2(1+k)

    = log2

    1+

    ps

    pn|H(k)|2

    (11)

    where H(k) is the channel response.

    The average capacity is calculated as the mean of all

    subchannels,

    C= E

    log2

    1+

    ps

    pn|H(k)|2

    (12)

    Fig. 9. Mobile Channel Capacities, Selective Fading Channel.

    Fig. 10. Fixed Channel Capacities, Selective Fading Channel.

    Assuming a Wide Sense Stationary (WSS) Rayleigh fading

    channel, H(k) can be approximated as a complex Gaussianrandom variable with unit power, i.e.,

    H(k) = Hr(k)+jHi(k) (13)

    where Hr/Hi N(0, 1/2).

    This analysis assumes independent fading in different

    time-frequency units, which is not the case of most practi-

    cal scenarios, where the multipath fading channels usually

    have both non-zero coherent bandwidth and coherent time.

    Therefore, there is usually strong correlation between two

    channel responses closely located in either time or frequency

    domain.

    On the other hand, most of the wireless communications

    systems are designed with an interleaver and forward errorcorrection (FEC) coding. An interleaver with sufficient inter-

    leaving depth makes the received symbols appear to have

    passed a channel with random fading in both time and fre-

    quency domains before they are passed to the FEC decoder.

    With a powerful FEC, such as the LDPC code, the chan-

    nel capacities shown in (11) and (12) and using (13) can be

    approached.

    The capacity allocations to the mobile and fixed services

    in selective fading channels are plotted in Fig. 9and Fig. 10,

    for the same two-layer LDM and the three TDM/FDM-base

    systems as described in last section. Similar to the observation

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    8/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    8 IEEE TRANSACTIONS ON BROADCASTING

    in AWGN channels, the LDM-based system allocates higher

    channel capacity to the mobile services at low SNRs and to

    the fixed services at high SNRs.

    For a mobile service with a bandwidth efficiency of

    0.4 b/s/Hz, i.e., 2.2 Mbps in a 6 MHz channel, the LDM

    system provides {2.0, 5.0, 7.0} dB better performance over

    the TDM/FDM systems allocating {25%, 33%, 50%} power

    to the mobile service. For a mobile capacity of 0.8 b/s/Hz,

    using LDM brings improvements of {1.8, 5.3, 8.3} dB over

    the three TDM/FDM systems.

    For fixed services, the LDM-based system provides higher

    capacity than the TDM/FDM systems with {50%, 67%, 75%}

    fixed service allocation at SNRs higher than {10, 19, 27} dB.

    3) Capacity Distribution in Slow Fading Channels: In

    slow-fading channels, the instantaneous channel capacity is

    calculated as,

    Cinst= log2

    1+|H|2

    (14)

    where H is the instantaneous channel response and = pspn

    is

    the SNR.

    Since the channel response, H, has non-zero probabilityof being close to zero, there is no definite capacity as the

    maximum rate of reliable communications supported by the

    channel.

    When the transmitter does not know the channel state

    information (CSI), which is the typical scenario in digital

    broadcasting systems, the transmitter encodes data at a rate

    of R (b/s/Hz). There is a non-zero probability for the chan-

    nel gain, H, to have a value that is too small to make the

    detection error probability arbitrarily small. When the instan-

    taneous channel capacity is lower than the encoded data rate R,

    the system is called to be in outage.

    For slow-fading channels, the outage probability is calcu-

    lated as,

    Pout(R)= P

    log2

    1+ |H|2

    (15)

    The transmission capacity of the slow-fading channels is

    measured by channel outage capacity, which is defined as the

    transmission rate, RC, at which the outage probability, Pout, is

    equal to a pre-defined threshold, , i.e.,

    Pout(RC)= (16)

    Taking (14) and (15), we have,

    Pout(RC)= P|H|2 = 2RC 1

    = P

    H2r +H2i =

    2RC 1

    = P

    rH=

    2RC 1

    = (17)

    where for Rayleigh fading channels, Hr and Hi are Gaussian

    random variables N

    0, 12

    , and rH is a Rayleigh random

    variable.

    Fig. 11. Mobile Channel Capacities, Slow-Fading Channels.

    Fig. 12. Fixed Channel Capacities, Slow-Fading Channels.

    The -Outage capacity can be calculated as,

    RC() = log2

    1+

    1()

    2 (18)

    where (x)= P {rHx}.

    For a two-layer LDM-based system, the instantaneous chan-

    nel capacities for the UL and LL are calculated as,

    CULinst =log2

    1+

    |H|2 pUL

    |H|2 pLL+ pn

    CLLinst =log2

    1+

    |H|2 pLL

    pn

    (19)

    The -Outage capacity for the UL is now calculated as,

    RULC () = log2

    1+

    pUL1()

    2

    pn+ pLL1()

    2

    (20)

    and for LL,

    RLLC () = log2

    1+

    1()

    2pLL

    pn

    (21)

    For an outage probability of = 0.1, the -Outage capaci-

    ties allocated for the mobile and fixed services in slow fading

    channels are plotted in Fig. 11 and Fig. 12, for the same

    LDM and TDM/FDM systems. Similar to AWGN and selec-

    tive fading channels, the LDM system provides higher capacity

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    9/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    ZHANG et al.: LAYERED-DIVISION-MULTIPLEXING: THEORY AND PRACTICE 9

    for mobile services at low SNRs and for fixed services at

    high SNRs.

    For a mobile service of 0.4 b/s/Hz (about 2.2 Mbps in a

    6 MHz channel), the LDM system provides {2.0, 5.5, 7.5} dB

    better performance over the TDM/FDM systems with {25%,

    33%, 50%} power allocations for the mobile service. For a

    mobile service of 0.8 b/s/Hz (about 4.4 Mbps), the LDM

    brings improvements of {2.0, 5.3, 8.3} dB over the three

    TDM/FDM systems.

    In slow-fading channels, the LDM system provides higher

    fixed service capacity than the TDM/FDM systems with

    {50%, 67%, 75%} fixed service allocation at SNRs higher

    than {19, 27.5, 35} dB.

    4) Summary on Capacity Distributions of LDM vs

    TDM/FDM Systems: The most important observation here is

    that the LDM system simultaneously provides higher trans-

    mission capacity for both mobile service at low SNR and

    fixed service at high SNR. Therefore, the LDM system pro-

    vides more efficient use of the spectrum in the scenario of

    simultaneously delivering a robust mobile HDTV service and

    a high-data-rate fixed UHDTV service (or multiple HDTVservices) in one RF channel.

    This advantage of the LDM system is mainly because of

    the nonlinear distribution of the channel capacity with respect

    to the power distribution. As shown in (5), the capacity of

    different layers is calculated with signal power distribution

    inside the logarithm operation; while for TDM/FDM systems,

    the capacities of different services are calculated in (6) with

    the signal power distribution outside the logarithm operation.

    An intuitive explanation is given as following. For

    TDM/FDM systems, the SNR for the mobile receivers is

    designed for the edge of the coverage area. All mobile

    receivers inside the coverage area have higher SNR, where

    the additional power is wasted, resulting in the waste of chan-nel capacity. On the other hand, in the LDM system, this

    wasted capacity is re-allocated to the fixed service by con-

    trolling the SNR for the mobile service at a constant value

    (i.e., the injection level) inside the coverage area.

    C. Cross-Layer Interference (CLI)

    Thus far, perfect UL signal cancellation is assumed in all the

    channel capacity analysis. This assumption has been shown to

    be directly applicable to the typical scenarios of the two-layer

    ATSC 3.0 system with LDM, where the UL delivers a mobile

    service with strong channel coding and modulation scheme

    and the LL delivers fixed services to receivers with rooftop

    antennas. When decoding the LL services, the signal SNR

    is usually quite high, where the UL signal has even higher

    SNR due to the stronger transmission power, which is usu-

    ally sufficient to guarantees perfect UL signal detection and

    re-construction.

    In this case, the decision symbols for the LL signal becomes,

    YLL(k) = XLL(k)H(k)

    + 1

    g

    XUL(k)H(k)XUL(k)H(k)

    +1

    gN(k)

    =XLL(k)H(k)+IUL(k)+1

    gN(k) (22)

    where IUL(k) is the CLI from the UL into the LL decision

    symbols, which is calculated as,

    IUL(k) = 1

    gXUL(k)

    H(k) H(k)

    =

    1

    gXUL(k)H(k) (23)

    In (23), the XUL is a random process containing random

    modulation symbols. The H is the channel estimation error,

    which can be well approximated by a Gaussian-distributed

    random process with zero mean and a variance depending

    on the channel estimation algorithm. Since the XUL and the

    Hare independent, the CLI can be modeled as a Gaussian-

    distributed random process, with zero mean and a variance

    that is calculated as,

    2I = 2H

    g2 (24)

    assuming 2X=1. The 2

    His the mean square error (MSE) of

    the channel estimates.

    Eq. (24) shows that the CLI power is directly proportionalto the MSE of the channel estimates. In [20], it was shown that

    a properly designed channel estimation modules can provide a

    an estimate MSE lower than 30 dB. Assuming a LL injection

    level of 5 dB, the CLI is 25 dB lower than the LL signal,

    which is much lower than the noise thresholds of most typical

    LL services, and can be safely ignored. This 30 dB cancella-

    tion performance is obtained in the Typical Urban (TU) [21]

    channels, which are much severer than the typical channels

    encountered by most fixed receivers with rooftop antennas.

    In most practical cases with Line-Of-Sight (LOS) or strong

    Rician channels, the CLI power should be much lower.

    In summary, for typical ATSC 3.0 scenarios, the CLI can

    be safely ignored in performance analysis. This justifies thatthe above capacity analysis is directly applicable for ATSC

    3.0 system performance estimation.

    IV. PERFORMANCE A DVANTAGES OF

    LDM-BASEDATSC 3.0 SYSTEMS

    In this section, we present the performance advantage of

    LDM for delivering mixed mobile and fixed services in one TV

    channel by extensive computer simulations of the ATSC 3.0

    system. Specifically, the simulation results show that LDM is

    especially effective in providing high-data-rate robust mobile

    services in severe mobile fading channels, including both the

    slow-fading for pedestrian-speed receivers and the very fastfading channels for high vehicle-speed receivers.

    A. Simulation Performances of LDM and TDM/FDM for

    Mixed Services Delivery

    In this section, we show the performance of LDM-based

    ATSC 3.0 system to deliver mixed services with different sys-

    tem configurations. The simulated ATSC 3.0 system has two

    layers and delivers one robust mobile service and one high

    data-rate fixed service. The injection level is 5 dB. Different

    channel coding and modulation schemes were considered

    in the simulations to achieve different service throughput

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    10/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    10 IEEE TRANSACTIONS ON BROADCASTING

    TABLE ISNR REQUIREMENTS FOR G OO DP ERFORMANCE, LDM V S TDM/FDM

    and robustness. The simulations were performed with 16k-FFT

    mode, a CP ratio of 1/16, and an in-band pilot pattern

    following the PP2 defined for the DVB-T2 system [17].

    The LDM performance is compared to those of TDM sys-

    tems, which use 32k mode for fixed service and 8k mode for

    mobile service. Three TDM systems are considered, which

    allocate {55%, 40%, 30%} of time for mobile service delivery.

    Table I summarizes the simulation performances of LDM

    and TDM systems for different mobile and fixed service

    throughputs. The performances are characterized by the

    required SNRs for error-free detection performance.

    These simulation results confirm that the LDM system

    indeed provides better performance for both fixed and mobile

    services. It is further shown that, while LDM provides signif-

    icant performance improvement relative to the TDM systems

    for mobile service delivery, its advantage for fixed service

    delivery is less obvious. A TDM system can be configured to

    provide similar mobile service quality as comparing to LDM.

    This, however, results in significant loss in its performance for

    fixed service delivery.

    B. Advantage of LDM for Robust Mobile Services

    The ATSC 3.0 system uses OFDM to more efficiently over-

    come the multipath fading and to provide easy deployment of

    single-frequency-network (SFN). One well known vulnerabil-

    ity of OFDM systems is their sensitivity to fast time-varying

    channels, which are usually encountered by fast moving

    receivers. The channel time-variation causes Doppler spread

    in the received signal, resulting in a power leakage among

    the signals carried in adjacent subchannels. This power leak-

    age becomes an additional interference for the signal detection

    and is called inter-carrier-interference (ICI). The impact of the

    Fig. 13. Impact of Doppler for Mobile Receivers.

    ICI is a bit error rate (BER) floor at the receiver which can-

    not be reduced simply by increasing the transmission power.

    This puts a limitation on the receiver mobility and results

    in a much smaller mobile service coverage for fast moving

    receivers [22], [23].The power of the ICI is directly proportional to the normal-

    ized Doppler spread, fd/fU, where fd is the Doppler spread

    and fUis the subchannel bandwidth. The Doppler spread can

    be calculated as,

    fd= v

    cfc (25)

    where v is the receiver speed, c = 3 108 m/s is the light

    speed, and fc is the carrier frequency.

    A tight upper-bound of the ICI power is derived in [24] as,

    2

    ICI

    1

    122 fd

    fU (26)

    where 1 = 12

    for the Classical Jakes Doppler spectrum.

    Eq. (26) tells us that the larger the subchannel bandwidth,

    the lower the ICI power. Therefore, to avoid severe ICI, the

    existing DVB-T/T2 systems deliver the mobile services using

    OFDM transmission modes with large subchannel bandwidth.

    In[25], we have shown that the impact of ICI can be demon-

    strated as a deterioration in the SNR of the received signal,

    which can be translated into an increase in the SNR thresh-

    old for successful signal detection. In Fig. 13, the increases

    of the SNR threshold due to the ICI are plotted for three

    OFDM transmission modes, {8k, 16k, 32k}, for which the

    subcarrier spacings are {837, 419, 209} Hz, respectively.A 6 MHz TV channel at 600 MHz is assumed. In addition,

    two mobile services are considered with significantly different

    SNR thresholds, 0 dB and 10 dB.

    The first observation in Fig. 13 is that, wider subcarrier

    spacing brings more robustness performance to fast moving

    receivers, i.e., 8k mode provides the best performance and

    32k mode brings the worst. Lets define the 3-dB speed limit

    of the receiver as the speed that causes a 3 dB increase in the

    SNR threshold. The speed limits for {8k, 16k, 32k} modes

    are {320, 160, 80} km/h for the mobile service with an SNR

    threshold of 10 dB.

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    11/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    ZHANG et al.: LAYERED-DIVISION-MULTIPLEXING: THEORY AND PRACTICE 11

    The second observation is that the impact of ICI is much

    less for mobile services designed with lower SNR thresholds.

    In Fig.13, the impact of ICI to the mobile service with an SNR

    threshold of 0 dB is much less than that for the service with

    10 dB threshold. The 3-dB speed limits for this mobile service

    are {> 400, 330, 165} km/h for the {8k, 16k, 32k} modes.

    Therefore, this mobile service is significantly more robust for

    high speed receivers.

    Fig. 13 gives a comparison between the mobile

    services in LDM and traditional TDM systems

    (DVB-T2/NGH) [26], [27]. In DVB-T2/NGH systems,

    multiple mobile services are transmitted, and each is allocated

    with a small percentage of time. This makes it necessary to

    use weaker channel codes and higher modulation to deliver

    a decent service data rate. However, this results in higher

    SNR thresholds, such as the 10 dB curve in Fig.13. Because

    the higher SNR threshold makes the mobile service more

    sensitive to ICI, the most robust 8k mode has always been

    the choice for DVB-T2/NGH for mobile service delivery in

    the past.

    On the other hand, for the ATSC 3.0 system with LDM, themobile services are delivered in the UL, with high transmission

    power, strong channel codes and robust QPSK modulation,

    which results in very low SNR thresholds. The typical value

    is the 0 dB shown in Fig. 13.Therefore, LDM provides much

    more robust performance and larger coverage for fast moving

    receivers. To provide good mobility while keeping high spec-

    trum efficiency (small GI percentage), 16k mode is currently

    recommended to delivery both mobile and fixed services in

    the two layers.

    In addition, Fig.13shows that, using LDM makes it possi-

    ble to use the most vulnerable 32k mode to deliver mobile and

    fixed services in one TV channel, since the 3-dB speed limit

    is 165 km/h, which is more than sufficient for most vehicularreceivers. Detailed analysis on the performance impact of FFT

    size can be found in [28].

    In[9], the capability of the LDM system to deliver mobile

    services were investigated in detail. In Fig.14,the performance

    of an LDM-UL mobile service is plotted, with QPSK and

    rate- 415

    LDPC code, which provides a throughput of 2.7 Mbps.

    The LL injection level is 5 dB. The performance is com-

    pared to those of TDM systems with different LDPC code

    rates and different modulation schemes to provide similar

    throughput.

    It is clearly shown that, due to the strong coding and

    modulation, the LDM UL provides robust performances for

    receivers at all speeds up to 300 km/h. For the pedestrianspeed at around 3 km/h, the required SNR is 3.5 dB for good

    service detection, while it only requires an SNR of 5 dB for a

    very fast receiver moving at 300 km/h. For the medium-speed

    receivers moving at 20 to 150 km/h, the required SNRs are

    much lower, close to 2.0 dB.

    For a TDM system which allocates 50% of the time to

    deliver a mobile service of 2.7 Mbps (QPSK and rate- 815

    LDPC

    code), the required SNR is 6.0 dB for the receivers moving at

    pedestrian speed, 5.0 dB for the medium-speed receivers, and

    19.0 dB for the high-speed receivers at 300 km/h. For a TDM

    system allocating 40% time for a mobile service with the same

    Fig. 14. Mobile Performances of LDM vs TDM Systems, 16k.

    throughput, the required SNRs for the pedestrian speeds and

    the medium speeds are [8.5, 6.8] dB. The receivers moving

    at 250 km/h or faster can no longer obtain zero-free detec-

    tion performance. In addition, it requires an SNR of 13 dB or

    higher for receivers moving at 225 km/h. The performance

    becomes worse for the TDM systems allocating less time

    percentage for the mobile services.

    Interested readers are referred to [9] and [29] for detailed

    and comprehensive results on the mobile and indoor perfor-

    mances of LDM-based ATSC 3.0 systems.

    V. EFFICIENT I MPLEMENTATION OF

    LDM-BASEDATSC 3.0 SYSTEMS

    Although the concept of LDM has existed as superposition

    coding for a long time, its application in practical systemshas been prohibited by the extra complexity required for suc-

    cessive signal cancellation. The key of the proposed LDM

    structure for ATSC 3.0 lies in the simple implementation,

    which makes the benefit of superposition coding feasible for

    consumer-level devices. In this section, we describe the imple-

    mentation modules specifically for LDM in the ATSC 3.0

    system and show that there is relatively small additional

    complexity required to achieve the significant performance

    improvement.

    The following sections will briefly present the critical imple-

    mentation challenges and their low complexity realizations.

    A thorough and comprehensive description and analysis on

    the low-complexity implementation of LDM in the ATSC 3.0

    system is presented in [14].

    A. Additional Signal Processing Complexity

    in the LDM Transmitters

    In the currently proposed ATSC 3.0 system, the two layers

    of the LDM system share the same OFDM signal struc-

    ture, which makes it possible for the two layers to share

    the same signal processing modules in both the transmitters

    and receivers. This greatly simplifies the implementation and

    results in low-complexity design.

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    12/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    12 IEEE TRANSACTIONS ON BROADCASTING

    The basic structures of the transmitter and receiver are

    shown in Fig. 1. At the transmitter, the additional signal

    processing is only the combining process of the two lay-

    ers, which is implemented by the simple multiplication and

    addition operations. Therefore, there is little extra complex-

    ity on the transmitter side for the LDM as comparing to the

    single-layer systems.

    B. Additional Signal Processing Complexity

    in the LDM Receivers

    As shown in Fig.1,the implementation of the LDM mobile

    receivers is the same as the conventional single-layer wire-

    less receivers, consisting of the standard signal processing

    modules, e.g., synchronization, OFDM demodulation, chan-

    nel estimation and equalization, deinterleaving, and channel

    decoding.

    On the other hand, to decode the fixed services in the LL, the

    receiver needs to decode the UL signal, cancel the UL signal,

    and then perform the conventional detection of the LL signal.

    Therefore, the implementation of the LL receiver needs to

    incorporate the additional complexity for UL signal detection,

    re-construction, and cancellation.

    1) UL Signal Detection and Cancellation: In order to

    decode the LL signal, the receiver needs to decode the UL

    signal, re-construct it and cancel it from the received multi-

    layer signals. Since LL is configured to deliver fixed services

    with high throughput, the SNR for LL signal detection is usu-

    ally quite high. If the SNR threshold for the LL service is

    15 dB, with an LL injection level of 5 dB, the UL signal

    would have an SNR of 20 dB. Since the UL signal is con-

    figured to operate at very low SNR threshold (i.e., 0 dB), the

    detection at this high SNR should be perfect.

    The major additional complexity in decoding the UL signalis the LDPC decoding. The investigations in [14]showed that

    the complexity of the LDPC decoding process decreases sig-

    nificantly with the higher SNR of the received signal. While

    it requires 50 decoding iterations to obtain good performance

    at the low SNR threshold of the UL signal, it requires no

    more than 5 iterations for perfect decoding at the higher SNRs

    above 7 dB. The additional LDPC decoding complexity is less

    than 10%.

    The reconstruction of the UL signal is performed by re-

    encoding, bit-interleaving, and bit-to-symbol re-mapping. The

    re-encoding can be accomplished in the LDPC decoding

    process without added complexity. The bit-interleaving and

    bit-to-symbol re-mapping requires little complexity.To obtain the decision symbols for the LL signal, the UL

    signal cancellation is performed as (22). This only requires

    one complex multiplication and one subtraction for each

    received frequency-domain symbol. The additional complexity

    is minimal.

    2) Channel Estimation for UL Signal Cancellation:For LL

    signal detection, the decision symbols are obtained using (22).

    Clearly, the UL signal cancellation needs to be very accurate

    in order to provide the high SNR required for LL signal detec-

    tion. It is proven in Section III-C that the power of the CLI

    is directly proportional to the channel estimation accuracy.

    Fig. 15. LDM Receiver Block Diagram with Cross-Layer-Interleaver.

    Therefore, a good channel estimator is necessary for the LL

    receiver to obtain good performance.

    The channel estimation methods to achieve good UL signal

    cancellation were investigated in[20]. It was shown that a two-dimensional (2D) channel estimator with frequency-domain

    DFT-based interpolation and time-domain Wiener filtering can

    provides excellent cancellation performance over 25 dB, i.e.,

    the CLI level is 25 dB lower than the LL signal level. This CLI

    can usually be ignored when comparing to the noise threshold

    of the LL signal.

    It was further shown in[20] that the 2D channel estimator

    has relative low complexity and is rather simple to implement

    for consumer devices, especially for LL receivers which are

    not power limited.

    C. Reducing Memory Using Cross-Layer InterleaverMemory is an important consideration for DTV receiver

    manufactures, since it can significantly increase the cost of

    the implementation. For LDM LL signal detection, additional

    memory is required to store the received signal for UL signal

    cancellation.

    In theory, the UL and LL signal processing chains in an

    LDM system could be independent, which provides large

    flexibility on the type of services carried in the two layers.

    However, when the UL and LL use independent interleaving

    and de-interleaving processes, the LL signal detection needs

    to wait for the deinterleaving process of the UL signal. The

    receiver needs to store all the received symbols before each

    UL signal deinterleaving process is finished and then performthe signal cancellation. This could require a large amount of

    additional memory.

    To reduce the memory requirement, the cross-layer inter-

    leaver is introduced in [14]. With the cross-layer interleaver,

    the block diagram of an LDM receiver is shown in Fig. 15.

    It is shown that since the UL and LL are sharing the same

    time and cell interleavers, the de-interleaving is performed

    prior to the signal cancellation. This significantly reduce the

    size of buffers needed for signal cancellation. It has been

    shown in[14] that the additional memory requirement is less

    than 20%.

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    13/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    ZHANG et al.: LAYERED-DIVISION-MULTIPLEXING: THEORY AND PRACTICE 13

    Fig. 16. Coverage Area of Mobile and Fixed Services in LDM Systems.

    VI . COVERAGE A NALYSIS OF LDM-BASED N EX T

    GENERATIONATSC 3.0 DTV SYSTEM

    Due to the capability offered by the LDM to deliververy robust mobile services, the ATSC 3.0 system

    could have a quite different mobile/fixed service coverage

    paradigm when comparing to the traditional TDM/FDM-based

    systems.

    To deliver mobile services with traditional TDM/FDM-

    based systems, it has been shown in the past attempts that

    the major challenge is to achieve a reasonable service cov-

    erage [30]. Different from fixed receivers (e.g., roof-top

    receivers), the mobile receivers usually have low antenna

    installation height, small antenna size, and little antenna

    directivity. This results in significantly lower received signal

    power at the mobile receivers as compared to fixed receivers,

    and therefore a much smaller mobile service coverage. Itbecomes necessary to install many additional transmitters to

    obtain a reasonable mobile service coverage in a regular

    size city.

    For a good understanding of the coverage areas for mobile

    and fixed services of the ATSC 3.0 systems, the following

    assumptions are made:

    A mobile receiver with 1.5m antenna height and a fixed

    receiver with 10m antenna height, resulting in 12 dB

    difference in field strength.

    The fixed-receiver has a directional gain of 8 dB.

    With these assumptions, the received signal strength for the

    mobile receiver is about 20 dB (100 times) lower than the

    fixed receiver.The two-layer LDM system is assumed to deliver a mobile

    service in UL with rate- 14

    LDPC code and QPSK modulation,

    which provides an SNR threshold of about 1.0 dB for fixed

    receivers and 2.0 dB for high-speed mobile receivers. For the

    fixed services in the LL, the required SNR varies depend-

    ing on the required service data rate. The required SNR is

    14.4 dB for a throughput of 15 Mbps service and 22.3 dB

    for 28 Mbps. Therefore, the UL mobile receivers requires

    12.4 dB lower SNR (or signal strength) than the LL fixed ser-

    vice of 15 Mbps, and 20.3 dB lower SNR than the LL service

    of 28 Mbps.

    Fig. 17. Coverage Estimation for Toronto Area.

    Taking into account the 20 dB higher antenna gain for fixed

    receivers and the 5 dB LL injection level, it is calculatedthat [30]:

    The coverage of the UL service for mobile receivers is

    slightly smaller than the coverage area of a 15 Mbps

    LL service for fixed receivers. The difference is only a

    propagation distance of 2.6 dB.

    The coverage of the UL service for mobile receivers is

    larger than the coverage of a 28 Mbps LL service for

    fixed receivers, by a distance for propagation loss of

    5.3 dB.

    The coverage of the UL service for fixed receivers is

    significantly larger than the coverage of 15 Mbps and

    28 Mbps LL services for fixed receivers, by a distance

    for propagation loss of 20.4 dB and 28.3 dB.This shows that the LDM system can provide a mobile

    service coverage very close to that of the fixed-services. In

    addition, the UL mobile services have significantly larger

    coverage areas for fixed receivers.

    The mobile and fixed service coverage areas of the two-layer

    ATSC 3.0 system are illustrated in Fig. 16, which shows that

    the LDM system can greatly reduce the mobile/fixed service

    coverage mismatch. In addition, it delivers the services in UL

    to fixed receivers in a significantly larger area.

    A coverage estimation of the LDM-based ATSC 3.0 system

    is obtained using the CRC-COVLAB coverage prediction soft-

    ware for the Toronto area, where the transmitter is installed

    on the CN Tower. The coverage predictions for the differentATSC 3.0 services are shown in Fig. 17. It can be seen that

    the coverage prediction shown in Fig. 17 agree very well with

    those analytical results shown in Fig. 16.

    It is shown that, the LDM-based ATSC 3.0 system can

    provide a handheld HDTV service and a 4k-UHDTV service

    with the same coverage as the legacy fixed HDTV services

    which are delivered by the ATSC 1.0 system. The new mobile

    HDTV services, shown as the yellow curve, have smaller cov-

    erage. Finally, the HDTV service delivered in the UL has an

    extremely large coverage for fixed receivers, which is clearly

    identified by the blue curve (New fixed HDTV in the figure).

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    14/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    14 IEEE TRANSACTIONS ON BROADCASTING

    VII. ACHIEVING F URTHERC APACITYI NCREASE W IT H

    MULTIPLEA NTENNAT ECHNOLOGIES INLDM SYSTEMS

    Multiple antenna technologies have proven very effective

    in providing higher transmission rate, and/or higher service

    reliability [31], [32]. Depending on the number of antennas at

    the transmitter and the receiver, multiple antenna technologies

    can be categorized as

    Single-Input-Multiple-Output (SIMO): one transmitterantenna and multiple receiver antennas. SIMO provides

    more reliable detection by using receiver antenna diver-

    sity combining (e.g., maximum-ratio-combining (MRC),

    equal-gain-combining (EGC), selective combining, etc..).

    Multiple-Input-Single-Output (MISO): multiple transmit-

    ter antennas and one receiver antenna. MISO pro-

    vides more reliable signal detection by using transmitter

    antenna diversity combining techniques, including space-

    time/frequency-coding (ST/FC), etc..

    Multiple-Input-Multiple-Output (MIMO): MIMO refers

    to a system with multiple antennas at both the transmit-

    ters and the receivers. MIMO can be configured in many

    different forms to provide higher throughput or higherSNR, or both. In this document, MIMO is used specifi-

    cally for the spatial multiplexing (SM) mode to provide

    the highest throughput.

    Performance of LDM with multiple antenna technologies

    was analyzed in [16] in terms of the capacity improvement

    with pre-defined percentage of radio resources allocated to

    one layer.

    In this section, the capacities of the layers in an two-

    layer LDM system configured with different multiple-antenna

    technologies in different layers are presented based on

    information-theoretical analysis. More importantly, the funda-

    mental trade-off between the capacities in the two layers are

    demonstrated.Due to the lack of return link, the most typical multi-

    ple antenna configurations for DTV systems involves two

    antennas at the transmitters. We further limit the number

    of receive antennas to two (2) for practical implementa-

    tion. For broadcasting systems, the 2x2 dual-polarized spatial

    multiplexing (SM) was proposed in [33] which has been

    shown to provide good throughput increase when applied

    to DVB-T systems. In [18], an improved 2x2 polarized

    spatial multiplexing, enhanced Spatial Multiplexing with

    Phase Hopping (eSM-PH), was proposed for the DVB-NGH

    system, which is shown to provide better throughput

    increase.

    For LDM systems, there is the flexibility to apply dif-

    ferent multiple antenna technologies in different layers to

    achieve either higher throughput or more robustness. An

    obvious option is to deploy spatial multiplexing in the LL

    to increase the capacity for the fixed services, which usu-

    ally require high SNRs and therefore can obtain good SM

    gains; and to transmit the UL signal in both polariza-

    tions (with or without STC/SFC) to achieve the highest

    robustness.

    Because different multiple antenna technologies are used in

    the UL and LL, the UL signal cancellation for LDM receiver

    needs to be designed accordingly.

    Fig. 18. Achievable Mobile and Fixed Service Capacities of a Two-LayerLDM System, SNRm = 0 dB, SNRf =20 dB.

    When the throughput of the mobile layer becomes more

    important, a second option is to deploy the same SM

    pre-coding in both layers in the LDM system. Although it isclaimed that in general using SM does not provide significant

    multiplexing gain at low SNRs, it has been shown in [34] that a

    2x2 SM with QPSK modulation can achieve an uncoded BER

    of 0.05 at an SNR of 2 dB, which is more than sufficient for

    error free decoding performance with low-rate LDPC codes.

    Therefore, there should be certain multiplexing gain even for

    the mobile services with low SNRs.

    Following the information theoretical analysis shown in

    Section III-A, the achievable capacities of the LDM-based

    DTV systems with different multiple antenna configurations

    are obtained and plotted in Fig. 18. It is assumed that the

    SNR threshold is 0 dB for the mobile services and 20 dB

    for the fixed service. With the 2x2 MIMO, the LL capac-ity is almost doubled because of the high SNR threshold. On

    the other hand, for UL mobile service, using 1x2 SIMO with

    MRC, the capacity is increased by 70%. With ideal decod-

    ing of the 2x2 MIMO, the capacity is also almost doubled.

    However, performing MRC is simple and easy to implement,

    while obtaining ideal MIMO decoding is relatively hard to

    achieve.

    The derivation of these curves and more investigation results

    on the capability of LDM with MIMO will be reported in a

    paper to be submitted shortly after.

    A superposition MIMO coding for layered sources is intro-

    duced in [35], which is essentially a special case of LDM with

    different multiple antenna technologies in the multiple layers.

    The analysis and simulation results confirmed the capacity

    analysis in Fig. 18.

    The current ATSC 3.0 standard does not allow the combined

    use of LDM with MIMO. However, the above investigations

    shows that the LDM-based DTV systems can be flexibly

    combined with different multiple antenna technologies in the

    different layers to achieve significant performance improve-

    ment, in terms of both robustness and transmission capacity.

    Therefore, the flexible combination of LDM with MIMO can

    be incorporated into the future releases of the ATSC systems

    for further system performance enhancement.

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 7/26/2019 Layered-Division-Multiplexing Theory and Practice

    15/17

    This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

    ZHANG et al.: LAYERED-DIVISION-MULTIPLEXING: THEORY AND PRACTICE 15

    VIII. MORE A PPLICATIONSS CENARIOS OF

    LDM I NATSC 3.0 SYSTEMS

    In this paper, the advantages of LDM technology in ATSC

    3.0 system is mainly demonstrated in terms of delivering

    mobile TV services and UHDTV fixed services simultaneously

    in one single TV channel.

    There are many other scenarios where LDM could find its

    place to offer better performance or more diversified servicesin the ATSC 3.0 systems, including

    transmitting video streams with scalable video coding to

    deliver the same service to different devices;

    adding local commercials, news and other services with a

    additional layers (i.e., LDM with more than two layers);

    performing non-orthogonal multiple access in broadband

    wireless communications systems.

    These are all the potential areas which are currently under

    investigation on the LDM technology.

    Another research topic under investigation is local con-

    tent/service insertion or local targeted advertisement. With

    LDM, one option is using upper layer to transmit mobile

    HD and 4K-UHDTV, while using lower layer to deliver localcontent or targeted advertisement. In ATSC 3.0, the strong

    modulation/coding schemes are capable of providing system

    SNR threshold of lower than 0dB (a negative value). In

    this case, the LDM lower layer can provide seamless cov-

    erage/service for each SFN transmitter, without the annoying

    coverage gaps between SFN transmitter service areas. Since

    the advertisement time is typically less that 20% of the pro-

    gram time, Non-Real Time (NRT) could be used to play-back

    the local content at 5 times the transmission bit rate for

    high-definition (audio/video) service quality.

    I X. CONCLUSION

    In this article, we provide a comprehensive review on the

    LDM technology which is accepted as one of the baseline PHY

    technologies for the next generation ATSC 3.0 DTV system.

    First, it is shown by theoretical analysis that the LDM pro-

    vides fundamental performance advantage over the traditional

    TDM/FDM systems when delivering multiple services over the

    same time-frequency resource. This advantage becomes sig-

    nificant when the multiple services have significantly different

    SNR thresholds. This corresponds to the important scenario

    to simultaneously deliver UHDTV fixed service (or multiple

    HDTV fixed services) and extremely robust mobile HDTV ser-

    vices in one TV channel. The performance advantage of LDM

    for ATSC 3.0 system is further demonstrated by the extensivesimulation results.

    With the careful design recommended for the ATSC 3.0

    standard, we further show that the performance improvement

    brought by LDM is achieved with very limited additional com-

    plexity for both transmitters and receivers. The mobile receiver

    for an LDM system is the same as those for traditional single-

    layer systems; while for fixed receivers capable of decoding

    both layers, the added complexity is less than 20% in terms

    of both signal processing and memory requirement.

    In addition, different multiple antenna technologies can be

    incorporated into the different layers in an LDM systems to

    achieve either throughput increase or higher robustness, which

    could result in more variety of coverage balancing among the

    different layers.

    REFERENCES

    [1] L. Fay, L. Michael, D. Gmez-Barquero, N. Ammar, andM. W. Caldwell, An overview of the ATSC 3.0 physical layerspecification, IEEE Trans. Broadcast., vol. 62, no. 1, Mar. 2016.

    [2] Y. Wu, B. Rong, K. Salehian, and G. Gagnon, Cloud transmission:A new spectrum-reuse friendly digital terrestrial broadcasting transmis-sion system, IEEE Trans. Broadcast., vol. 58, no. 3, pp. 329337,Sep. 2012.

    [3] J. Montalbanet al., Cloud transmission: System performance and appli-cation scenarios, IEEE Trans. Broadcast., vol. 60, no. 2, pp. 170184,Jun. 2014.

    [4] S. I. Park, H. M. Kim, and P. Angueira, Hardware implementation andcomplexity analysis of layered division multiplexing (LDM) system forATSC 3.0, in Proc. NAB, Las Vegas, NV, USA, Apr. 2015, pp. 7781.

    [5] P. P. Bergmans and T. M. Cover, Cooperative broadcasting, IEEETrans. Inf. Theory, vol. 20, no. 3, pp. 317324, May 1974.

    [6] Y. Saito et al., Non-orthogonal multiple access (NOMA) for cellu-lar future radio access, in Proc. VTC, Dresden, Germany, Jun. 2013,pp. 15.

    [7] Y. Saito, A. Benjebbour, Y. Kishiyama, and T. Nakamura, System-levelperformance evaluation of downlink non-orthogonal multiple access

    (NOMA), in Proc. PIMRC, London, U.K., Sep. 2013, pp. 611615.[8] A. Benjebbour et al., Concept and practical considerations of non-

    orthogonal multiple access (NOMA) for future radio access, in Proc.ISPACS, Naha, Japan, Nov. 2013, pp. 770774.

    [9] L. Zhang et al., Performance characterization and optimization ofmobile service delivery in LDM-based next generation DTV systems,

    IEEE Trans. Broadcast., vol. 61, no. 4, pp. 557570, Dec. 2015.[10] K.-J. Kimet al., Low-density parity-check codes for ATSC 3.0, IEEE

    Trans. Broadcast., vol. 62, no. 1, Mar. 2016.[11] S. I. Park, H. M. Kim, Y. Wu, and J. Kim, A newly designed quarter-

    rate QC-LDPC code for the cloud transmission system, IEEE Trans.Broadcast., vol. 59, no. 1, pp. 155159, Mar. 2013.

    [12] S. I. Park, Y. Wu, H. M. Kim, N. Hur, and J. Kim, Raptor-likerate compatible LDPC codes and their puncturing performance for thecloud transmission system, IEEE Trans. Broadcast., vol. 60, no. 2,pp. 239245, Jun. 2014.

    [13] S. I. Park et al., Two dimensional code based on low density parity

    check and ReedSolomon codes for the terrestrial cloud transmissionsystem, IEEE Trans. Broadcast., vol. 61, no. 1, pp. 7583, Mar. 2015.

    [14] S. Park et al., Low complexity layered division multiplexing systemfor ATSC 3.0, IEEE Trans. Broadcast., vol. 62, no. 1, Mar. 2016.

    [15] S. Gadkari and K. Rose, Time-division versus superposition codedmodulation schemes for unequal error protection, IEEE Trans.Commun., vol. 47, no. 3, pp. 370379, Mar. 1999.

    [16] D. Gmez-Barquero and O. Simeone, LDM vs. FDM/TDM for unequalerror protection in terrestrial broadcasting systems: An information-theoretic view, IEEE Trans. Broadcast., vol. 61, no. 4, pp. 571579,Dec. 2015.

    [17] Digital Video Broadcasting (DVB); Implementation Guidelines for aSecond Generation Digital Terrestrial Television Broadcasting System(DVB-T2), ETSI TS Standard 102 831 v1.1.1 (2010-10), 2010.

    [18] D. Gmez-Barquero, Next Generation Mobile Broadcasting.Boca Raton, FL, USA: CRC Press, 2013.

    [19] D. Tse and P. Viswanath, Fundamentals of Wireless Communication.

    Cambridge, U.K.: Cambridge Univ. Press, 2005.[20] L. Zhang et al., Application improved DFT-based channel estimation

    for accurate signal cancellation in cloud-Txn multi-layer broadcastingsystem, in Proc. BMSB, Beijing, China, Jun. 2014, pp. 16.

    [21] Digital land mobile radio communications - COST 207, Commiss.Eur. Communities, Final Report, Office Official Publ. Eur. Commun.,Luxembourg City, Luxembourg, 1989.

    [22] L. Zhang, Z. Hong, and L. Thibault, Robust DVB-T/H receiver infast fading channels, in Proc. VTC, Ottawa, ON, Canada, Sep. 2010,pp. 15.

    [23] L. Zhang, Z. Hong, L. Thibault, R. Boudreau, and Y. Wu, A low-complexity robust OFDM receiver for fast fading channels,IEEE Trans.

    Broadcast., vol. 60, no. 2, pp. 347357, Jun. 2014.[24] Y. Li and L. J. Cimini, Jr., Bounds on the interchannel interference of

    OFDM in time-varying impairments, IEEE Trans. Commun., vol. 49,no. 3, pp. 401404, Mar. 2001.

  • 7/26/201