waveform multiplexing for new radio: numerology management...
TRANSCRIPT
-
1536-1284/18/$25.00 © 2018 IEEE IEEE Wireless Communications • October 201886
AbstrActA 5G communication system includes massive
machine-type communication, enhanced mobile broadband, and ultra-reliable low-latency commu-nication. To meet the requirements of 5G applica-tions, researchers have considered various scalable subcarrier spacings and transmission time intervals. In this article, we introduce the concept of wave-form multiplexing, which is a resource management system for numerology multiplexing that employs appropriate waveforms to various numerologies. It also possesses both a dynamic cyclic prefix and a minimum guard band, which are the key features for achieving high spectral efficiency. We verify the significant gain produced by proposed waveform multiplexing by performing extensive 3D ray-trac-ing-based system-level evaluations in realistic 3D environments. Finally, we provide guidelines on waveform multiplexing design by considering the time-domain and frequency-domain characteristics of the new radio waveform candidates.
IntroductIonIn recent years, researchers have both clarified fifth-generation (5G) communication require-ments and discussed its potential applications, including servicing the Internet of Things (IoT), gigabit wireless connectivity, and the Tactile Internet, as well as the various requirements of massive machine-type communication (mMTC), enhanced mobile broadband (eMBB), and ultra-re-liable low-latency communication (URLLC) [1]. 5G applications require different numerologies, including organizations of scalable subcarrier spacings, cyclic prefixes (CPs), and transmission time intervals (TTIs), for different service types. In trying to meet the various requirements of 5G applications, researchers should consider multiple scalable numerologies.
Multiple numerologies can be serviced either individually, under various frequency bands, or col-lectively, under a single frequency band. The use of various frequency bands to support single service types under each band has been a topic of discus-sion at 3GPP meetings. 3GPP has also discussed numerology multiplexing, which operates on a single frequency band to support the integrated devices or the central controllers of applications with different requirements.
It has not yet been determined which waveform will be used for numerology multiplexing systems
and how such systems can be designed. To estab-lish these systems, researchers must evaluate them using new radio (NR) waveform candidates. Nev-ertheless, the influence of the real propagation of the NR waveform candidates on system-level per-formance in realistic multi-cell environments based on their properties, such as delay and out-of-band emissions (OOBE), remains an open area for study.
To address these issues, we provide the fol-lowing contributions in this article: we provide insight into the numerology management of NR waveform candidates; and we present a three-di-mensional (3D) evaluation for both NR waveform candidates and numerology multiplexing systems operating in realistic environments. We first elab-orate on the resource management challenges of NR waveform candidates. Based on these challeng-es, we demonstrate the potential of a system that consists of different waveforms on various numer-ologies. We call this concept waveform multiplex-ing. Waveform multiplexing systems can operate more efficiently than both multiband systems and single-waveform-based numerology multiplexing systems. We next describe the challenges involved in evaluating waveform multiplexing in practice. We then propose a more practical 3D channel model that utilizes realistic digital maps. According to the performance trade-off of waveform designs, the performance of waveform multiplexing varies depending on the specific environment. Finally, we provide guidelines for waveform multiplexing design.
resource MAnAgeMent chAllenges for nr WAveforM cAndIdAtes
In this section, we outline an overview of NR waveform candidates and elaborate on their resource management challenges.
overvIeW of nr WAveforM cAndIdAtes for doWnlInk systeMs
In LTE systems, CP orthogonal frequency-division multiplexing (CP-OFDM) is an excellent waveform solution that not only improves the spectral effi-ciency but also mitigates inter-symbol interference (ISI). CP-OFDM, however, has some limitations that researchers are striving to overcome by cre-ating new waveforms [1]. Some of the NR wave-form candidates being considered for the design of a downlink system are discussed below.
Yeon-Geun Lim, Taehun Jung, Kwang Soon Kim, Chan-Byoung Chae, and Reinaldo A. Valenzuela
Waveform Multiplexing for New Radio: Numerology Management and 3D Evaluation
ACCEPTED FROM OPEN CALL
Yeon-Geun Lim, Taehun Jung, Kwang Soon Kim, and Chan-Byoung Chae are with Yonsei University; Reinaldo A. Valenzuela is with Nokia Bell Labs.
Digital Object Identifier:10.1109/MWC.2018.1700351
-
IEEE Wireless Communications • October 2018 87
ACCEPTED FROM OPEN CALL Orthogonal Waveform Candidates: 3GPP agreed that NR waveforms should be based on OFDM with either filtering or windowing. Both their easy implementation and back-and-for-ward compatibilities are the primary rationale for being NR waveforms. As candidates, researchers proposed filtered OFDM (f-OFDM) [2], weight-ed-overlap-and-add OFDM (WOLA-OFDM) [3], and flexibly-configured OFDM (FC-OFDM) [4]. An f-OFDM system is an integrated system that consists of multiple independent CP-OFDM sys-tems with various numerologies, and it is based on subband splitting and filtering. A properly designed baseband finite impulse response (FIR) filter can suppress the OOBE of each subband. The WOLA-OFDM system achieves low OOBE by using a time-domain windowing method. Applying soft edges to the cyclic extensions of CP-OFDM symbols provides lower complexity than a time-do-main filtering approach [3]. Researchers have also attempted to find a uniform solution for numerolo-gy multiplexing systems by investigating FC-OFDM, but these systems require offset quadrature ampli-tude modulation (QAM) operation.
Non-Orthogonal Waveform Candidates: Although OFDM-based waveform candidates will be used for 5G, some potential NR waveform candidates, including non-orthogonal waveforms, could support various applications beyond the ini-tial 5G system. Such candidates fulfill the require-ment of low OOBE and demonstrate robustness to asynchronous systems that is similar to that of OFDM-based candidates. In searching for a can-didate, researchers have studied how to enable filter-bank multicarrier (FBMC) to use QAM [5]. Its potential to be a 5G waveform, however, has been suppressed by its extremely high implementation complexity. Meanwhile, researchers proposed gen-eralized frequency-division multiplexing (GFDM), which divides time-frequency resources into blocks consisting of subsymbols and subcarriers [6]. Its OOBE can be reduced by selecting a well-localized pulse-shaping filter.
WAveforM resource MAnAgeMent consIderAtIons In the tIMe And frequency doMAIns
As mentioned above, researchers have nominated various NR waveform candidates. Figures 1a–c illustrate that there are performance trade-offs between the time and frequency domains that are caused by various waveform design methods. Resource management for waveform numerol-ogies has become an issue due to the variance of the time and frequency properties of both the performance requirements and the channel envi-ronments for each 5G service type. Waveform multiplexing systems represent a potential solu-tion, as shown in Fig. 1d, and will be discussed later. Researchers must address this issue in either waveform multiplexing systems or their chosen alternatives by considering the followings.
Constraint from Numerology and System Designs: Without loss of generality, the numbers of subsymbols and subcarriers (i.e., the number of samples in a symbol or resource grids in the time and frequency domains) are both dependent on and limited by the given system parameters, including the subcarrier spacing, sampling rate, and bandwidth. Also, both determining and scaling
numerologies depend on users’ requirements and channel characteristics [7].
Performance Trade-Offs of Waveform Designs: A waveform’s performance trade-offs result from either its filter/window design or its CP/slot design, as shown in Figs. 1a and b. These trade-offs can be summarized as follows:• In general, for a well-chosen prototype filter/
window, a longer filter/window length in the time domain results in lower OOBE. Thus, low OOBE challenges both the time resource over-heads and the ISI, which affects the overall rate performance.
• It is common for URLLC applications that require short TTIs to support wide subcarrier numerology. Based on both the fixed number of symbols per slot and the fixed CP length (for a given maximum delay spread), the CP overhead increases as the TTI shrinks. also, the delay yielded from the tails is critical for short TTIs.Figure 1c shows that the f-OFDM exhibits lower
OOBE than the others, because f-OFDM systems typically adopt a long filter length, half the symbol duration. In contrast, WOLA-OFDM edges can be viewed as filter tails with a small rolloff factor; their length is shorter than twice the CP length. Other-wise, adopting a circular pulse-shaping filter, GFDM does not present filter tails. To enhance the OOBE performance of GFDM, an additional time-domain procedure, such as windowing and inserting guard subsymbols, is required. Meanwhile, since the tail overhead of f-OFDM is longer than the guard peri-od in time division duplex systems, it adopts a burst truncation that results in higher OOBE.
Figure 1b shows the two slot designs used to reduce the CP overhead when the mobility requirement of the URLLC users is low. In the first case, a slot is organized for the back-and-forward compatibilities using a small number (e.g., one) of OFDM-based waveform symbols, although a nar-row subcarrier numerology is used. For this case, researchers should investigate a frequency-domain slot design. The drawbacks are that the tail delay is non-negligible for URLLC users and that the tail overhead increases (especially during burst trans-missions). In the other case, the subband for the short TTI uses GFDM, and an LTE-like slot design can be applied. For example, one GFDM symbol consisting of eight subsymbols can be viewed as one slot consisting of eight OFDM symbols.
Performance-Limiting Interference of Unrelated Numerology: The OOBE represents the level of inter-band interference, which is unavoidable in numer-ology multiplexing [7]. The improvement of spectral efficiency by reducing interband interference and the guard band provides the primary motivation for using NR waveforms. Nevertheless, in practical 5G systems, interband interference has a marginal thresh-old of performance degradation [4], as shown in Fig. 1c, because some sources of performance-limiting interference exist. These are:• Under multicell scenarios, intercell interference
is one of the dominant factors in the user-ex-perienced signal-to-interference-plus-noise-ratio (SINR) because of the high network density.
• In multiple-input multiple-output (MIMO) sys-tems, inter-stream interference is a perfor-mance-limiting factor if the level of channel estimation error is non-negligible. This problem
To enhance the OOBE performance of GFDM,
an additional time-do-main procedure, such
as windowing and inserting guard subsym-bols, is required. Mean-
while, since the tail overhead of f-OFDM is
longer than the guard period in time division
duplex systems, it adopts a burst trun-cation that results in
higher OOBE.
-
IEEE Wireless Communications • October 201888
also appears in agile spectrum-sharing systems if the accuracy of the channel estimation or the spectrum sensing is low.
WAveforM MultIpleXIngWAveforM MultIpleXIng systeM concept
While 3GPP has discussed numerology multiplex-ing on a single frequency band, there remain the numerology management issues; the performance of NR waveforms are aff ected by their numerol-ogies, filter/window design, slot design, and the performance-limiting interference as mentioned above. In this article, we present a waveform mul-tiplexing technique as an efficient numerology multiplexing operating system. We define wave-form multiplexing as a subband system covering the following characteristics.1
Multiplexing of Waveforms on Subbands with Scalable Subcarrier Spacing: Figure 1d illustrates that a waveform multiplexing system can support various applications that demand other subcarri-
er spacings, TTIs, and specific waveforms. Each subband consists of diff erent waveforms with vari-ous numerologies. Additionally, the overall perfor-mance can be improved by allocating appropriate numerologies to users with various mobilities. This method exploits the fact that long symbol dura-tion is vulnerable to the Doppler eff ect (especially at a high frequency). The operator should adopt a suitable set of waveforms to enhance the spec-trum utilization per subband with consideration for resource management.
Dynamic CP: Thanks to the low OOBE of 5G waveform candidates, the proposed waveform multiplexing can optimize a multiband system by employing diff erent CP lengths. For high network density in 5G communications, researchers have considered deploying a large number of small cells. This method shortens and diversifi es the maximum delay spreads per cell, permitting the reduction of the CP length according to cell radius and the channel characteristics of each subband. Also, the users in the same cell can be supported with
FIGURE 1. Waveform resource management considerations: a) time-domain performance due to fi lter tails of WOLA-OFDM and f-OFDM; b) time-domain performance due to slot designs for the low mobility URLLC users; c) frequency-domain performance of NR waveform candidates and performance-limiting interference of unrelated numerology. This illustration shows the power spec-tral density of NR waveform candidates where nsc = 1024 and N = 2048, and example levels of threshold are shown; d) a concep-tual illustration of a waveform multiplexing system that is a potential solution for numerology management issues.
Waveform Resource Management Considerations
Numerology manager
Subband
〮〮〮 〮〮〮 〮〮〮 〮〮〮 〮〮〮 〮〮〮
Scalable subcarrier spacing sets
Minimum guard bande.g., zero
Dynamic CP sets CP
Different waveforms e.g., GFDM e.g., f-OFDM e.g., WOLA
CP CP
URLLC
Tactile Internet, Smart Grid
eMBB
Smartphones, 3D Video Streaming
mMTC
Wearables, Automated Vehicles
e.g. 67.5, 16.875, 135 kHz
Different scaling factor
OFDM-based Slot in LTE
1CP/slot
8CP/slot
Slot Length (Short TTI)
OFDM-basedOne Symbol Slot
GFDM-basedOne Symbol Slot
Slot Length (Short TTI)
Slot Length (Short TTI)
: Control Channel : Data
Slot Length (Short TTI)with Filter/Window
Additional Transmission Delay0 8.64 25.92 34.56 43.217.28
Bandwidth (MHz)
-250
-200
-150
-100
-50
0
50
Pow
er (d
B)
GFDM f-OFDM WOLA-OFDM
Example level of threshold forinterference-limited environments
Example level of threshold fornoise-limited environments
(c)
Tail Overhead
…
CP
…
Edge Extension & Weighting Overlap & Add
WOLA-OFDM
f-OFDM
(a)
(b)
Tail Overhead & Delay
(d)
Time-Domain Property Frequency-Domain Property
1 The proposed concepts can cover several special cases for numerology multiplexing as single waveform multiplexing. The signals of the f-OFDM sys-tem in [2] and the FC-OFDM system in [4] with different subcarrier spacings on multi-ple subbands are, respectively, mixed into the whole of each system. These are viewed as single waveform multiplexing.
-
IEEE Wireless Communications • October 2018 89
different CP lengths, which are shorter than the cell-specifi c CP length, depending on the user-spe-cific delay spread by splitting the subband of the cell into many subbands.
Minimum Guard Band between Subbands: NR waveforms promise higher spectrum usage than OFDM, as the guard band between the subbands can be minimized because of the low OOBE. At the 3GPP meeting, zero guard band has been considered for numerology multiplexing.
WAveforM MultIpleXIng gAInThe numerology managements and characteris-tics of waveform multiplexing for improving the spectral efficiency of multiband systems can be summarized as follows:• A different waveform on each subband can
optimize the time and frequency characteristics according to the user’s requirements and its SINR on the subbands.
• CP overhead can be reduced by adopting dynamic CP and slot designs.
• A minimum guard band enhances spectrum usage.We define the gains from the above spectral
enhancements as waveform multiplexing gain. Later we verify the signifi cant waveform multiplexing gain compared with the conventional system.
unIfIed WAveforM MultIpleXIng trAnsceIver desIgnIn this section, we propose how to design a generalized and unifi ed version of the CP-based and QAM-based waveform multiplexing systems. For simplicity, we employ the two representa-tive waveform candidates that respectively have strong points in the time and frequency domains, but employing other waveform candidates is straightforward. One is f-OFDM, which provides much lower OOBE performance, so it can signifi -cantly reduce interband interference in the fre-quency domain. The other is GFDM, which has no fi lter/window tail in a slot, so it improves spec-tral effi ciency in the time domain.2
Figure 2 illustrates a block diagram of the pro-posed waveform multiplexing transmitter and its receivers.3 In the ith subband transmitter, nsc,i × Mi m-QAM symbols are mapped to the Mi desired time grids and the nsc,i desired frequency grids where m is the modulation order. Each subband uses an appropriate subcarrier filter of Ni size. If f-OFDM is used on the ith subband, the subcarrier fi lter is a rectangular pulse easily implemented by an inverse fast Fourier transform (FFT) fi lter where Mi = 1 and Ni becomes FFT size. If GFDM is used on the ith subband, each sample passes through a corresponding subcarrier filter. These are time-
and frequency-shifted versions of a prototype fi lter. A CP with ncp,i length is added to the waveform symbol of the ith subband. Next, it passes through a subband filter, which is a bandpass FIR filter in the f-OFDM case or a coeffi cient 1 in the OFDM and GFDM cases with a time-domain convolu-tion. Finally, the baseband waveform signals are arranged according to the center frequencies of their subbands. The procedures at the receiver are processed inversely. The GFDM receiver uses a post-processing block such as zero-forcing to mitigate inter-carrier interference with subcarri-er filtering. The number of subbands is given by nsubband = k=1nSss nScp,k, where nSss and nScp,k are the numbers of subcarrier spacing sets and CP sets (including TTI designs) for the kth subcarrier spac-ing set, respectively. The specific parameters of the subbands are determined under the system constraints in the previous section. Moreover, the main concern with designing a GFDM’s numerol-ogy is that Mi has to be an odd number so that its transmit fi lter matrix can be invertible.
3d perforMAnce evAluAtIonkey chAllenges for
nr WAveforMs’ perforMAnce evAluAtIonAn important issue is devising a performance evaluation method for NR waveforms. In ear-lier work, researchers have concentrated on performance evaluation based on link-level sim-ulations, such as the performance of the bit error rate, OOBE, and overheads due to the filter or CP. What remains an open problem, though, is the system-level simulation in practical envi-ronments. Such simulations are performed by efficiently reflecting link-level simulation results considering the numerology management issues. Practical path loss, moreover, does not mono-tonically increase as the frequency of the subcar-rier increases, while free space path loss does, because each subcarrier goes through reflec-tion, attenuation, and diff raction independently. Thus, to obtain a precise evaluation, one should consider the channel of each waveform’s sub-carrier, including the interference level per sub-carrier from the OOBE of another subband’s waveforms.
So far, researchers have compared techniques based on system-level evaluation in a two-dimen-sional (2D) environment following a hexagonal layout with a wrap-around configuration [8]. As cell sizes have been shrinking, it could be crucial to measure the signal propagation at the elevation angles and in a realistic network layout when evalu-ating a 5G system [9].
FIGURE 2. A block diagram of the unifi ed version of a waveform multiplexing transmitter and its receivers at downlink.
Waveform Multiplexing Signal
( sc ,1 1)
+CP( cp ,1)
SubbandFilter( tx ,1)
( sc ,2 2)
+CP( cp ,2)
SubbandFilter( tx ,2)
( sc , )
+CP( cp ,i)
SubbandFilter( tx ,i)
m-QAM Symbols
Freq.Time
Resource Mapping
SubcarrierFilter
Freq.
Freq.
Freq.
Time
Time
TimeFreq.
Center Freq.
Other waveforms
Frequency Arrangement
&RF
( sc ,1 1)
-CP( cp ,1)
SubbandFilter( tx ,1)
( sc ,2 2)
-CP( cp ,2)
SubbandFilter( tx ,2)
( sc , )
-CP( cp ,i)
SubbandFilter( tx ,i)
m-QAM Symbols
RF Subcarrier Filter& Equalization
Subcarrier Filter& Equalization
Subcarrier Filter& Equalization
WirelessChannel
RF
RF
A user on the th subband
Freq.Time
Resource Mapping
Time
Time
Time
Other waveforms
2 We choose f-OFDM instead of WOLA-OFDM since it has the lower OOBE to off er insight into resource management. Besides, GFDM with windowing and guard subsymbol are not con-sidered for the best resource utilization in the time domain.
3 The proposed waveform multiplexing is one of the waveform multiplexing systems, that is, any design satisfying the requirements explained previously could be referred to waveform multiplexing.
-
IEEE Wireless Communications • October 201890
evAluAtIon envIronMentsIn this section, for the more practical system-level evaluation, we develop a 3D channel model using realistic digital maps based on a 3D ray-tracing tool and present two environments. The proposed 3D channel model can refl ect the key challenges for NR waveforms’ performance evaluation.
3D Channel Model: We implemented a digital map of the buildings surrounding GangNam Sta-tion, Seoul, South Korea, as shown in Figs. 3a–c. The coverages of each cell vary due to the assort-ed shapes and heights of the buildings. Besides, the buildings have detailed interiors consisting of glass (white), metal (brown), concrete (yellow), sheet-rock, and wood. Such materials possess different attenuation factors.
We consider two outdoor scenarios: one is urban micro, and the other is heterogeneous net-works (HetNets). The urban micro depicts a sev-
en-cell hexagonal layout deploying four micro base stations (BSs) with a three-sector beam antenna, as illustrated in Fig. 3b. In the HetNets scenario, seven small BSs, not sectorized, are appropriately located between the micro BSs at the middle of the build-ing to enhance the SINR of the cell-boundary users and the sum rate per area, as illustrated in Fig. 3c.
We measure the received powers, OOBE propa-gations, angle parameters, and power delay profi les (PDPs) from the BSs to the user per subcarrier using a 3D ray-tracing tool, Wireless System Engineering (WiSE) developed by Bell Laboratories [10]. The 3D channels of each subcarrier of each user are gen-erated by utilizing these measurements applying the clustered channel model equation (e.g., the equation in [11]), which ensures the proposed 3D channel model is realistic.
Simulation Procedure: We present the wave-form multiplexing of three subbands. In considering
FIGURE 3. A procedure of 3D system-level simulation for waveform multiplexing: a) a real environment, GangNam Station, Seoul, South Korea the picture of GangNam Station is taken from the website (http://map.naver.com); b) the conventional 2D outdoor scenario (seven-cell hexagonal layout); c) the digital maps of GangNam Station of the urban micro (left) and HetNets (right) sce-narios. Various applications are serviced under these scenarios; d) a block diagram of the 3D system-level simulator.
: Micro BS
: Sector Beam Direction
Interior View
Smart Phone
Drone Communications
Traffic Control
E-health
(c)
(a) (b)
: Small BS
3D R
ealis
tic
Mod
elin
g
2D Regular Modeling
Real
Env
iron
men
t
Generate 3D Channel
3D Ray-tracing Simulation
Delay Parameters
ReceivedPower
Calculate Spectral
Efficiency
MobilitySubband
WaveformMapping
(d)
User Droppingand
Scheduling
DetermineCP Length
Angle Parameters
Link Level Simulator
Bit Error Probability OOBE PerformanceCP/Slot Design
3D DigitalMap
3D System-Level Simulation Platform
-
IEEE Wireless Communications • October 2018 91
a downlink system, we assume perfect channel estimation, frequency localization, and time syn-chronization to evaluate the best performance.
We assume that the BSs and the users are connected with the point-to-point communication under the full buffer traffic scenario. The low mobil-ity users are uniformly distributed on the ground and every floor of the building. The users who require long symbol duration (e.g., LTE applica-tions) are supported on the second subband, while the users requiring short symbol duration (e.g., URLLC users) and the in-between applications are serviced on the third and the first subbands, respec-tively. We also assume one CP set for all subcarrier spacing sets (i.e., nSss = 3, nScp,k = 1 (∀ k)) that is determined from the maximum delay spreads on the corresponding subbands.4 The system over-head from the frame structure is based on the LTE.
We organize a total of five sets, made up of three single waveform multiplexing sets and two waveform multiplexing sets. For the best time-do-main efficiency, one slot of GFDM for the first, the second, and the third subbands consists of 9, 8, and 7 subsymbols (3, 8, and 1 symbols) similar to the LTE, respectively. Detailed system parameters are given in Table 1.
Figure 3d illustrates a block diagram of the 3D system-level simulator. Channel parameters are measured from the proposed 3D channel model. The long and short CPs are determined from the maximum delay spreads of the associated BSs in the urban micro and the HetNets, respectively. The probability of bit error of the waveforms is mea-sured under the generated 3D channels. Finally, we calculate spectral efficiency using the Shannon formula with backoff calibration, the SINR terms of which are calculated by combining such mea-surements as received power, the probability of bit error, OOBE, and so on.
3d systeM-level evAluAtIon resultsToy Example of Waveform Multiplexing Gain:
We verify the gains that can be obtained by the proposed waveform multiplexing compared with the conventional system. We assume that the upper bound of waveform multiplexing consists of idle waveforms with no tails and zero OOBE. The one-symbol slot designs are chosen for minimum CP overhead. Figure 4a illustrates the potential waveform multiplexing gain in the urban micro. The conventional multiband system can achieve high-er spectral efficiency with various types of numer-ology management. Waveform multiplexing can achieve the highest performance with full utiliza-tion of its characteristics. From the result, we con-firm that the proposed waveform multiplexing has 1.67 times more gain than the conventional system at the median cumulative density function (CDF). Consider a practical full-duplex system that has 1.9 times the gain [12], so the waveform multiplexing system can be an efficient resource utilization tech-nique for NR.
Comparison with Single Waveform Multi-plexing Sets: We show three evaluation scenarios according to the level of intercell interference to investigate how the time and frequency character-istics of NR waveforms affect the spectral efficiency from the following resource management perspec-tives. We also show the advantages of waveform multiplexing with different waveforms. A summary of
these evaluation results and conclusions along with comments from below can be found in Fig. 4h.
Urban Micro without Intercell Interference Management: In an interference-limited environ-ment, the time-domain performance significantly affects the spectral efficiency, while the influence of the interband interference is less than that from the intercell. Thus, the performances of SWM-GFDM, WM-1, and WM-2 are close to the upper bound of the proposed waveform multiplexing, as shown in Fig. 4b.
In URLLC numerology (the third subband), the slot design with multiple symbols has the largest CP overhead as well as a long filter tail will impact ISI
4 It is hard to solve the optimi-zation problem for determining the number of subbands of a system because it is dependent on many parameters such as the sets of delay spreads, mobilities and so on. We leave this problem for future work.
TABLE 1. Simulation parameters.
The number of subbands (nsubband) 3
Sampling rate per subband (MHz) 34.56
System bandwidth (MHz) 51.84
Guard band (Hz) 0
Center frequency (GHz) 5
Transmit power of the micro/small BS (dBm) 49/35 [11]
Antenna pattern of the micro BS Katherine antenna
Antenna pattern of the small BS Isotropic antenna (0 dBi)
Height of the micro BS 1.5 m above the roof of the building
Height of the small BS 12.5 m above the ground
Inter site distances of the micro BSs Approximately 200 m
Maximum height of the users (m) 32.5 (on the 9th floor) [11]
Noise floor (dBm/Hz) –174
Backoff calibration (dB) 5
Channel model Proposed 3D channel
Channel estimation Idle
f-OFDM filter Raised-cosine window with 2.5 guard tones [2]
GFDM filterRoot-raised-cosine filter with
0.1 of rolloff/zero-forcing
Subband index 1st 2nd 3rd
Subcarrier spacing (kHz) 67.5 16.875 135
Number of subcarriers (Ki) 512 2048 256
Number of allocated subcarriers (nSC,i) 256 1024 128
Subband bandwidth (MHz) (Di) 17.28 17.28 17.28
Center frequency of subband (GHz) (Fi) 4.98272 5.00000 5.01728
Long/short CP length (nCP,i) 71/22 76/15 73/49
Waveform multiplexing set {1st subband, 2nd subband, 3rd subband}
Subband filter order/number of subsymbols (Li/Mi)
SWM-OFDM {OFDM, OFDM, OFDM} 0/1 0/1 0/1
SWM-f-OFDM {f-OFDM, f-OFDM, f-OFDM} 256/1 1024/1 128/1
SWM-GFDM {GFDM, GFDM, GFDM} 0/3 0/1 0/7
WM-1 {GFDM, f-OFDM, GFDM} 0/3 1024/1 0/7
WM-2 {f-OFDM, f-OFDM, GFDM} 256/1 1024/1 0/7
SWM: single waveform multiplexing set WM: waveform multiplexing set
-
IEEE Wireless Communications • October 201892
within short symbol duration. Thus, SWM-f-OFDM with the LTE-based slot design has significant per-formance degradation in an interference-limited environment for URLLC users, as shown in Fig. 4c. In contrast, SWM-f-OFDM with a one-symbol slot design has reasonable performance but has the lon-gest delay for one slot transmission due to the lon-gest fi lter tail, which will be undesirable to the URLLC users. From the best time-domain performance and the lowest OOBE from the second subband, WM-2 exhibits both the best spectral effi ciency and latency performance in URLLC scenarios.
Urban Micro with Perfect Intercell Interference Management: In a noise-limited environment, Fig. 4d illustrates that SWM-f-OFDM exhibits the best performance. This result implies that in the noise-lim-ited environment, it may be more effective to choose the waveform candidate having the lowest OOBE than the one having the better time-domain
performance. WM-1 and WM-2 could, nonetheless, still serve as good options, considering how imprac-tical too high-order QAM techniques are.
Figure 4e depicts the CDF of spectral effi ciency when interference from other sources (e.g., inter-stream interference, interference from the agile spectrum-sharing technologies) exists. The level of these interferences is dependent on the accuracy of channel estimation or spectrum sensing. As the level of other interference increases, the trend of the CDF becomes similar to the CDF in the inter-ference-limited environment. In particular, when the level of other interference is higher than –15 dB, SWM-f-OFDM underperforms versus the oth-ers, while WM-2 outperforms the others.
Figure 4f shows the CDF of spectral efficiency for URLLC users. For this case, since the frequen-cy-domain performance mainly impacts the spectral effi ciency, SWM-f-OFDM has the better spectral effi -
FIGURE 4. 3D system-level simulation results: a) CDF of the spectral effi ciency for the upper bound of the waveform multiplexing system and the LTE-based multiband OFDM in the urban micro scenario (interference-limited environment); b) CDF of the spectral effi ciency in the urban micro (interference-limited environment); c) CDF of the spectral effi ciency for URLLC users in the urban micro (interference-limited environment); d) CDF of the spectral effi ciency in the urban micro (noise-limited environment); e) CDF of the spectral effi ciency when interference from other sources exists (noise-limited environment); f) CDF of the spectral effi ciency for URLLC users in the urban micro (noise-limited environment); g) CDF of the spectral effi ciency with the diff erent CP length in the HetNets. The legends L and S represent long CP and short CP. The spectral effi ciency is normalized by the coverage area of the urban micro; h) a summary of comparisons of the waveform multiplexing sets.
Spectral Efficiency (bps/Hz/area)0 1 2 3 4 5 6
CD
F
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SWM-f-OFDMSWM-f-OFDM (1 Symbol Slot)SWM-GFDMWM-1WM-2Upper Bound
Spectral Efficiency (bps/Hz/area)0 1 2 3 4 5 6
CD
F
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SWM-OFDMSWM-f-OFDMSWM-GFDMWM-1WM-2Upper Bound
Waveform Multiplexing
WaveformMultiplexing Set
Rate Performance (bps/Hz/area) InterbandInterference(Frequnecy
Domain)
Agile Spectrum-Sharing among Subband
Urban Micro (b)(Interference Limited)
Urban Micro (d)(Noise Limited)
HetNets (g)(Short CP)
10th%ile
50th%ile
90th%ile
10th%ile
50th%ile
90th%ile
10th%ile
50th%ile
90th%ile
SWM-OFDM 0.707 1.755 2.767 1.156 6.784 6.843 0.018 4.046 5.782 High Almost impossible
SWM-f-OFDM 0.647 1.747 2.766 1.079 9.64 14.55 0.015 3.803 5.638 Low Possible but hardImplementation*
SWM-GFDM** 0.847 2.066 3.271 1.356 8.264 8.342 0.2 4.417 6.430 Moderate Possible bud lowperformance
WM-1 0.870 2.123 3.338 1.393 9.375 9.624 0.019 4.289 6.334 Moderate Possible
WM-2 0.804 2.001 3.158 1.304 10.00 12.20 0.018 4.170 6.216 Low Possible
(h)
* due to bandpass filter design** subsymbol-wise slot design and no filtering/windowing
Spectral Efficiency (bps/Hz/area)0 1 2 3 4 5 6 7 8 9
CD
F
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SWM-OFDM (L)SWM-f-OFDM (L) SWM-GFDM (L)WM-1 (L)WM-2 (L)SWM-f-OFDM (S) SWM-GFDM (S)WM-1 (S)WM-2 (S)Upper Bound
3.4 3.6 3.8 4 4.2 4.4 4.6
0.49
0.495
0.5
0.505
0.51
Due to short CP
URLLC Numerology (3rd Subband)
2 2.05 2.1
0.49
0.495
0.5
0.505
0.51
54.0 sec 68.5 sec 68.8 sec
1 Slot Transmission Delay
Spectral Efficiency (bps/Hz/area)0 1 2 3 4 5 6
CD
F
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Conventional LTE-based Multiband OFDMProposed Waveform Multiplexing (Upper Bound)
Gains from Idle Waveforms on each subband
Potential Total 1.67x Waveform Multiplexing Gain
Gains from CP/Slot Designs on each subband
Multiband System with Optimal CP overhead
Multiband System with Zero Guard Band/OOBE
Spectral Efficiency (bps/Hz/area)0 2 4 6 8 10 12 14 16 18 20
CD
F
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SWM-OFDMSWM-f-OFDMSWM-GFDMWM-1WM-2Upper Bound
Waveform Multiplexing
(a) (b) (c)
(d) (e) (f)
(g)
Spectral Efficiency (bps/Hz/area/stream)0 2 4 6 8 10 12 14 16 18 20
CD
F
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SWM-f-OFDM (-30 dB)SWM-f-OFDM (-15 dB) SWM-GFDM (-30 dB)SWM-GFDM (-15 dB) WM-2 (-30 dB)WM-2 (-15 dB)Upper Bound
-15 dB Other Interference Level
-30 dB Other Interference Level
Spectral Efficiency (bps/Hz/area)0 2 4 6 8 10 12 14 16 18 20
CD
F
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SWM-f-OFDM
SWM-f-OFDM(1 Symbol Slot)SWM-GFDMWM-1WM-2Upper Bound
URLLC Numerology (3rd Subband)
54.0 sec
68.5 sec
68.8 sec
1 Slot Transmission Delay
11.7 11.8 11.9 12
0.49
0.495
0.5
0.505
0.51
Urb
an M
icro
(Inte
rfer
ence
Lim
ited)
Urb
an M
icro
(Noi
se L
imite
d)H
etN
ets
(Inte
rfer
ence
Lim
ited)
-
IEEE Wireless Communications • October 2018 93
ciency among the single waveform multiplexing sets. Its long tail, however, blocks achieving the upper bound of the spectral effi ciency and the best latency performance. In contrast, the spectral effi ciency of WM-2 is close to the upper bound as well, while the transmission delay of that one is the shortest.
HetNets with Short or Long CP Length: Figure 4g shows that the HetNets with a short CP length have the advantage of spectral efficiency because of the shorter delay spread.
The more the cell size shrinks, the more the intercell interference increases in the inter-ference-limited environment. Thus, the effort to reduce the overhead in the time domain is more favorable than mitigating OOBE.
guIdelInes on WAveforM MultIpleXIng desIgnWAveforM selectIon
From the insights in the previous sections, we give some guidelines on how to choose an appro-priate waveform for a particular subband of the waveform multiplexing system.
For eMBB Users: Intercell interference manage-ment will be prudently performed to meet the high data rate requirement. For users with intercell inter-ference management (including coordinated beam-forming and interference alignment techniques), a waveform with a lower OOBE could be a good candi-date. Meanwhile, it may be effi cient for users in severe interference environments (e.g., a dense cell with one reuse-factor, a simple application with imperfect MIMO or agile spectrum-sharing techniques) to adopt a waveform with lower time-domain overhead.
For mMTC Users: Mesh sensor networks may result in an interference-limited environment as well as low complexity processing, which is desir-able for the battery life. A waveform with modest OOBE performance, short fi lter/window tail, and a low computational complexity can be a good candidate on the subband. For example, WOLA-OFDM would be better than f-OFDM.
For URLLC Users: A waveform with a short TTI design and a short fi lter/window tail can be a good candidate. Note that fi ltering at the receiver yields additional delay.
AgIle spectruM-shArIng And MIMoTo further enhance spectral effi ciency, researchers have studied agile spectrum-sharing and MIMO technologies. As potential technologies of waveform multiplexing, these remain research challenges.
In Agile Spectrum-Sharing Systems: There are two categories, a cognitive radio with spectrum sens-ing and a non-orthogonal multiple access (NOMA) [13]. In the proposed waveform multiplexing system, the subbands overlap, so agile spectrum-sharing techniques are available. A more effi cient algorithm could be implemented due to the knowledge of allocated subcarriers, as illustrated in Fig. 5.
In MIMO Systems: Spatial multiplexing tech-niques targeting a high rate will be used for eMBB, while those increasing the number of simultane-ous serviced-users will be adopted for mMTC. In addition, diversity gain schemes enhancing reli-ability will be desirable for URLLC. Researchers should investigate both combining these differ-ent MIMO schemes into a waveform multiplexing system and enabling MIMO transmission of NR waveform candidates.
chAnnel codIng for 5g requIreMentsThe system-level simulation algorithm in [8] mea-sures the block rate by adapting the modulation and coding scheme from the block error rate and SINR, the performance of which is related to a channel coding algorithm. For NR, there are three channel coding candidates, turbo codes, low-density parity-check (LDPC) codes, and polar codes. We briefly introduce here the strengths and weaknesses of the channel coding schemes. While turbo codes have backward compatibility and fl exibility, they have a higher complexity and error fl oor. The advantages of LDPC codes include having the highest throughput and the best area efficiency; however, their disadvantages include an inferior performance at the short-block length. Polar codes perform well at the short-block length, but their higher complexity results in a long code. It has become more influential in determining appropriate sets of the waveforms and channel codings according to various requirements of 5G applications and the pros and cons of the channel codings [14]. For the data and control channel of eMBB users in 5G, for example, LDPC and polar codes, respectively, will be used.
conclusIonThis article has presented the potential of a sys-tem that consists of diff erent waveforms on vari-ous numerologies, which is a waveform subband system operated on a single frequency band. We have called this concept waveform multiplexing. Waveform multiplexing possesses three character-istics: the multiplexing of waveforms on subbands with scalable subcarrier spacing, dynamic CP, and a minimum guard band. We have also presented key challenges for performance evaluations for NR waveforms. Based on the proposed 3D chan-nel in realistic environments (GangNam Station), we have evaluated the system-level performances of the proposed waveform multiplexing. The per-formances vary in specifi c environments accord-ing to the numerology management. From the results, we have confirmed that waveform mul-tiplexing can be an efficient resource utilization technique in numerology multiplexing systems. Our future work will consist of evaluating wave-form multiplexing systems with more downlink and uplink NR waveform candidates (in asynchro-nous scenarios) and massive MIMO schemes in mMTC scenarios [15].
FIGURE 5. A example of agile spectrum-sharing (cognitive radio and NOMA).
Subband 3
Subband 2
Subband 1
Freq.
Freq.
Freq.
Cognitive Radio with Spectrum Sensing
NOMA
Bandwidth of Subband1
Bandwidth of Subband2
Bandwidth of Subband3
: Waveform Filter Size (e.g., FFT Size)
-
IEEE Wireless Communications • October 201894
AcknoWledgMentThis research was supported by the Ministry of Science and ICT (MSIT), Korea, under the “ICT Consilience Creative Program” (IITP-2018-2017-0-01015) supervised by the Institute for Information & Communications Technology Promotion (IITP) and the ICT R&D program of MSIT/IITP (2015-0-00300, Multiple Access Technique with Ultra-Low Latency and High Efficiency for Tactile Internet Services in IoT Environments). An earlier version of this article was presented, in part, at the Euro-pean Conference on Networks and Communica-tions (EuCNC) in 2017.
references[1] G. Wunde et al., “5GNOW: Non-Orthogonal, Asynchronous
Waveforms for Future Mobile Applications,” IEEE Commun. Mag., vol. 52, no. 2, Feb. 2014, pp. 97–105.
[2] 3GPP R1-165425, “f-OFDM scheme and filter design,” Hua-wei, HiSilicon, May 2016.
[3] 3GPP R1-162199, “Waveform candidates,” Qualcomm Incorporated, Apr. 2016.
[4] H. Lin, “Flexible Configured OFDM for 5G Air Interface,” IEEE Access, vol. 3, 2015, pp. 1861–70.
[5] B. Farhang-Boroujeny, “OFDM versus Filter Bank Multicar-rier,” IEEE Sig. Proc. Mag., vol. 28, no. 3, May 2011, pp. 92–112.
[6] N. Michailow et al., “Generalized Frequency Division Mul-tiplexing for 5th Generation Cellular Networks,” IEEE Trans. Commun., vol. 62, no. 9, Sept. 2014, pp. 3045–61.
[7] A. A. Zaidi et al., “Waveform and Numerology to Support 5G Services and Requirements,” IEEE Commun. Mag., vol. 54, no. 11, Nov. 2016, pp. 90–98.
[8] ITU-R Rep. M.2135-1, Guidelines for evaluation of radio interface technologies for IMT-Advanced, Dec. 2009.
[9] J. Jang et al., “Smart Small Cell with Hybrid Beamforming for 5G: Theoretical Feasibility and Prototype Results,” IEEE Wireless Commun., vol. 23, no. 6, Dec. 2016, pp. 124–31.
[10] R. Valenzuela, D. Chizhik, and J. Ling, “Measured and Pre-dicted Correlation Between Local Average Power and Small Scale Fading in Indoor Wireless Communication Channels,” Proc. IEEE VTC Spring, vol. 3, May 1988, pp. 2104–08.
[11] 3GPP TR 38.873 V12.2.0, Study on 3D channel model for LTE, June 2015.
[12] M. Chung et al., “Prototyping Real-Time Full Duplex Radi-os,” IEEE Commun. Mag., vol. 53, no. 9, Sept. 2015, pp. 56–64.
[13] H. Kim et al., “Multiple Access for 5G New Radio: Cat-egorization, Evaluation, and Challenges,” submitted to IEEE Commun. Mag., also available at arXiv preprint, arXiv: 1703.09042, Mar. 2017.
[14] 3GPP R1-162896, “Channel coding requirements for next generation radio access technologies,” Nokia, Alcatel-Lu-cent Shanghai Bell, Apr. 2016.
[15] Y.-G. Lim, C.-B. Chae, and G. Caire, “Performance Analysis of Massive MIMO for Cell-Boundary Users,” IEEE Trans. Wireless Commun., vol. 14, no. 12, Dec. 2015, pp. 6827–42.
bIogrAphIesYeon-Geun Lim [S’12] received his B.S. degree in information and communications engineering from Sungkyunkwan Univer-sity, Korea in 2012. He is now with the School of Integrated Technology, Yonsei University, and is working toward the Ph.D. degree. He was a co-recipient of the Samsung Humantech Paper Award (2018). He has been involved in several indus-trial and national projects sponsored by Samsung Electronics,
IITP, and KCA. His research interests include massive MIMO, next-generation waveforms, full-duplex, mm-Wave technologies, and system level simulation for 5G networks.
Taehun JunG [S’15] received his B.S. degree from the School of Integrated Technology, Yonsei University, Korea, in 2015, where he is currently pursuing the Ph.D. degree at the School of Integrated Technology. His research interests include drone communication for 5G, next-generation waveforms, and real-time prototyping.
KwanG Soon Kim [S’95, M’99, SM’04] received the B.S. (summa cum laude), M.S.E., and Ph.D. degrees in electrical engi-neering from KAIST, Korea. From 1999 to 2000 he was with the Department of Electrical and Computer Engineering, University of California at San Diego as a postdoctoral researcher. From 2000 to 2004 he was with the Electronics and Telecommunica-tion Research Institute, Daejeon, Korea as a senior member of research staff. Since 2004 he has been with the Department of Electrical and Electronic Engineering, Yonsei University, Korea, and is currently a professor. He served as an editor of the Jour-nal of the Korean Institute of Communications and Information Sciences (KICS) from 2006 to 2012, as the editor-in-chief of the Journal of KICS since 2013, as an editor of the Journal of Com-munications and Networks (JCN) since 2008, and as an editor of IEEE Transactions on Wireless Communications from 2009 to 2014. He was a recipient of the Jack Neubauer Memorial Award (best system paper award, IEEE Transactions on Vehicular Technology) in 2008, and the LG R&D Award: Industry-Academ-ic Cooperation Prize, LG Electronics, 2013
Chan-BYounG Chae [S’06, M’09, SM’12] is an Underwood Distinguished Professor in the School of Integrated Technology, Yonsei University, Korea. Before joining Yonsei University, he was with Bell Labs, Alcatel-Lucent, Murray Hill, New Jersey, as a member of technical staff, and Harvard University, Cambridge, Massachusetts, as a postdoctoral research fellow. He received his Ph.D. degree in electrical and computer engineering from the University of Texas at Austin in 2008. He was the recipient/co-recipient of the IEEE INFOCOM Best Demo Award (2015), the IEIE/IEEE Joint Award for Young IT Engineer of the Year (2014), the KICS Haedong Young Scholar Award (2013), the IEEE Signal Processing Magazine Best Paper Award (2013), the IEEE ComSoc AP Outstanding Young Researcher Award (2012), the IEEE Dan. E. Noble Fellowship Award (2008), and two Gold Prizes (1st) in the 14th/19th Humantech Paper Contest. He currently serves as an editor for IEEE Transactions on Wireless Communications, IEEE Communications Magazine, IEEE Wireless Communications Letters, and IEEE Transactions on Molecular, Biological, and MultiScale Communications.
ReinaLdo a. VaLenzueLa [M’85, SM’89, F’99] obtained his Bach-elor of Science from the University of Chile and his Ph.D. from the Imperial College of Science and Technology of the Univer-sity of London, England. At Bell Laboratories, he studied indoor microwave propagation and modeling, packet reservation mul-tiple access for wireless systems, and optical WDM networks. He became Manager of the Voice Research Department at Motorola Codex, involved in the implementation of integrat-ed voice and data packet systems. On returning to Bell Labo-ratories, he led a multi-disciplinary team to create a software tool for wireless system engineering (WiSE), now in widespread use by Lucent Technologies. He received the Distinguished Member of Technical Staff award and is Director of the Wire-less Communications Research Department. He is interested in microwave propagation measurements and models, intelligent antennas, third generation wireless systems, and space time systems achieving high capacities using transmit and receive antenna arrays. He has published over 80 papers and has 12 patents. He is a Fellow of the IEEE. He is an editor for IEEE Trans-actions on Communications and IEEE Transactions on Wireless Communications.
We have confirmed that waveform mul-tiplexing can be an efficient resource utilization technique in numerology multi-plexing systems. Our future work will consist of evaluating waveform multiplexing systems with more downlink and uplink NR wave-form candidates (in asynchronous scenari-os) and massive MIMO schemes in mMTC scenarios.