5g technologies and advances, part ieasitc.com/slides/bs_1.pdf · introduction deployment...

IntroductionDeployment Scenarios, Numerologies, and Frame Structure

New Waveform in NRMultiple Access and Initial/Random Access

5G Technologies and Advances, Part I5G New Radio – An Overview

Borching Su 1

1Graduate Institute of Communication Engineering,National Taiwan University, Taipei, Taiwan

August 6, 2018

Graduate Institute of Comm. Engineering, National Taiwan University IEEE EASITC, August 2018 1/49



Outline

1 Introduction5G New Radio – An Overview

2 Deployment Scenarios, Numerologies, and Frame StructureDeployment ScenariosNumerologiesFrame Structure

3 New Waveform in NROFDM-Based New WaveformsDesired PropertiesVarious Techniques

4 Multiple Access and Initial/Random AccessMultiple AccessInitial/Random AccessSummary




5G New Radio – An Overview

Today’s Outline

Part I – 5G New Radio – An Overview.

Part II – A theoretical treatment in orthogonal frequency divisionmultiplexing (OFDM) systems.





Introduction

Cellular mobile networks have been deployed for several decades.

In the past, these networks were developed to optimize a particularservice primarily (e.g., voice/video streams), while other serviceswere supported additionally (e.g., Internet browsing and Internet ofThings deployment).





Introduction

Nevertheless, in the upcoming decades, manifold applications (toname a few, unmanned vehicles/ robots, intelligent transportationsystems, smart grid/buildings/cities, virtual/augmented/sensoryreality, mobile social services, and ubiquitous remote control) areurgently desired.

To empower these emerging applications with miscellaneous trafficcharacteristics, an engineering paradigm shift is needed in thedevelopment of fifth generation (5G) mobile networks.





Introduction

Instead of solely enhancing data rates to optimize transmissions of ahandful of traffic patterns, the International TelecommunicationUnion Radiocommunications Standardization Sector (ITU-R) hasannounced multifold design goals of 5G mobile networks known asInternational Mobile Telecommunications 2020 (IMT-2020), whichinclude

1 20 Gb/s peak data rate,2 100 Mb/s user experienced data rate,3 10 Mb/s/m2 area traffic capacity,4 100 devices/km2 connection density,5 1 ms latency,6 mobility up to 500 km/h,7 backward compatibility to LTE/LTE-Advanced (LTE-A), and8 forward compatibility to potential future evolution.





Introduction

To meet the above design goals, the Third Generation PartnershipProject (3GPP) started the standardization activity of 5G NewRadio (NR) in 2016, and completed the first phase (Phase I)system in 2018 (Release 15).

The second phase (Release 16) is expected to be ready in in 2020.

The following scope is considered in the 5G specificiations.1 Standalone and Non-Standalone NR Operations.2 Spectrum Below and Above 6 GHz.3 Enhanced Mobile Broadband (eMBB), Ultra-Reliable and Low

Latency Communications (URLCC) and Massive Machine-TypeCommunications (mMTC).





Standalone and Non-Standalone NR Operations

Standalone operation implies that full control plane and data planefunctions are provided in NR.

Non-standalone operation indicates that the control plane functionsof LTE1 and LTE-A2 are utilized as an anchor for NR.

1LTE: Release 8, Mar. 20092LTE-Advanced: Release 10, June 2011





Spectrum Below and Above 6 GHz

Subject to existing fixed spectrum allocation policies, it is achallenge to obtain available spectrum with a sufficiently widebandwidth from frequency range below 6 GHz.

Consequently, spectrum above 6 GHz turns out to be critical.

On the other hand, accessing the radio resources below 6 GHz is stillnecessary to fulfill diverse deployment scenarios required byoperators.





eMBB, URLCC, mMTC

In Release 15, three major use cases are emphasized:3

Enhanced Mobile Broadband (eMBB),Ultra-Reliable and Low Latency Communications (URLCC) andMassive Machine-Type Communications (mMTC):

33GPP TR 38.913: Study on Scenarios and Requirements for Next GenerationAccess Technologies. (Dec. 2015 -)Graduate Institute of Comm. Engineering, National Taiwan University IEEE EASITC, August 2018 10/49




eMBB, URLCC, mMTC

Enhanced Mobile Broadband (eMBB),

Ultra-Reliable and Low Latency Communications (URLCC) and

Massive Machine-Type Communications (mMTC):

Offering urgent data delivery with ultra low latency and massivepacket transmissions are of crucial importance for NR.

1 eMBB supports high capacity and high mobility (up to 500 km/h)radio access (with 4 ms user plane latency).

2 URLCC provides urgent and reliable data exchange (with 0.5 ms userplane latency).

3 NR also supports infrequent, massive, and small packet transmissionsfor mMTC (with 10 s latency).





5G New Radio Topics

To integrate the above features, agile radio resource management isessential to achieve optimized network performance.

The following things should be developed first.1 deployment scenarios,2 numerologies,3 frame structure,4 new waveform,5 multiple access,6 initial/random access, and7 enhanced carrier aggregation (CA)

In this section4, we provide insightful overview knowledge to thestate-of-the-art standardization of NR.

The performance in terms of random access (RA) latency ofenhanced CA is also demonstrated, as a performance benchmark tofacilitate future engineering practice.

4Much of the content is excerpted from our recent paper, Lien et al., “5G New Radio: Waveform, Frame Structure, Multiple Access,

and Initial Access,” IEEE Comm. Magazine, vol. 55, no. 6, June 2017.




Deployment ScenariosNumerologiesFrame Structure

Deployment Scenarios (1/5)

For backward compatibility with LTE/LTE-A, the architecture of NRis required to closely interwork with LTE/LTE-A. For thisrequirement, cells of LTE/LTE-A and NR can have differentcoverage or the same coverage, and the following deploymentscenarios are feasible.

1 LTE/LTE-A eNB is a master node:2 NR gNB is a master node3 eLTE eNB5 is a master node

5An eLTE eNB is evolved eNodeB that can support connectivity to EPC as well asnext-generation core network(NG-CN)






1 LTE/LTE-A eNB is a master node: An LTE/ LTE-A eNB offers ananchor carrier (in both control and user planes), while an NR gNBoffers a booster carrier. Data flow aggregates across an eNB and agNB via the evolved packet core (EPC).






2 NR gNB is a master node: A standalone NR gNB offers wirelessservices (in both control and user planes) via the next generationcore. A collocated enhanced LTE (eLTE) eNB is able to additionallyprovide booster carriers for dual connections (Fig 1c).






2 eLTE eNB is a master node: A standalone eLTE eNB offers wirelessservices (in both control and user planes) via the next generationcore, or a collocated NR gNB is able to provide booster carriers.






Inter-Radio Access Technology (RAT) Handover betweenLTE/LTE-A/eLTE eNB and NR gNB: An LTE/ LTE-A eNBconnects to the EPC, and an NR gNB connects to the nextgeneration core to support handover between eNB and gNB.

An eLTE eNB can also connect to the next generation core, andhandover between eNB and gNB can be fully managed through thenext generation core.





Numerologies of NR (1/4)

The above scenarios reveal a heterogeneous deployment of NR withdifferent coverage.

Further considering user equipment (UE) mobility up to 500 km/h,multiple cyclic prefix (CP) lengths should be adopted in NR.

In practice, the carrier frequency and subcarrier bandwidth may alsoaffect the adopted CP length.

Therefore, there can be multiple combinations of physicaltransmission parameters in NR, such as

1 subcarrier spacings,2 orthogonal frequency-division multiplexing (OFDM) symbol

durations,3 CP lengths, and so on.

These physical transmission parameters are collectively referred to asnumerologies in NR.






In NR, transmitters and receivers may enjoy a wider bandwidth athigh frequency bands.

In this case, the subcarrier spacing can be extended (larger than 15kHz as adopted by LTE/LTE-A, and potentially up to 960 kHz).

In addition, high carrier frequencies are also vulnerable to theDoppler effect, and a large subcarrier spacing may facilitateinter-carrier interference (ICI) mitigation.






On the other hand, NR should also support a small subcarrierspacing, such as 3.75 kHz as supported by narrowband Internet ofThings (NB-IoT), to enjoy better power efficiency at low frequencybands.

Consequently, subcarrier spacings in NR are scalable as a subset orsuperset of 15 kHz.

Feasible subcarrier spacings can be 15 kHz × 2m, where m can be apositive/ negative integer or zero.

For each subcarrier spacing value, multiple CP lengths can beinserted to adapt to different levels of inter-symbol interference (ISI)at different carrier frequencies and mobility.






Multiple OFDM numerologies are supported as given by Table 4.2-1where µ and the cyclic prefix for a bandwidth part are given by thehigher-layer parameters DL-BWP-mu and DL-BWP-cp for thedownlink and UL-BWP-mu and ULBWP- cp for the uplink.

Figure 1: Supported transmission numerologies – From TS 38.211, Rel-15





Frame Structure of NR (1/7)

In the time domain, the subframe length of NR is 1 ms, which iscomposed of 14 OFDM symbols using 15 kHz subcarrier spacing andnormal CP.A subframe is composed of an integer number of slots, and each slotconsists of 14 OFDM symbols.Each slot can carry control signals/channels at the beginning and/orending OFDM symbol(s).This design enables a gNB to immediately allocate resources forURLLC when urgent data arrives.






OFDM symbols in a slot are able to be all downlink, all uplink, or atleast one downlink part and at least one uplink part.

Therefore, the time-division multiplexing (TDM) scheme in NR ismore flexible than that in LTE.

To further support small size packet transmissions, mini-slots areadditionally adopted in NR, where each mini-slot is composed ofz < y OFDM symbols.






Each mini-slot is also able to carry control signals/channels at thebeginning and/or ending OFDM symbol(s).

A mini-slot is the minimum unit for resource allocation/scheduling.






In NR, different subcarrier spacings with the same CP overhead canbe multiplexed within a subframe.

To maintain 1 ms subframe length, there should be symbolboundary alignment within a subframe.






For subcarrier spacing(s) larger than 15 kHz, the sum of theseOFDM symbol durations (including CP length) should equal onesymbol duration of 15 kHz subcarrier.

On the other hand, the sum of OFDM symbol durations of 15 kHzsubcarriers should equal one symbol duration of subcarrier spacingsmaller than 15 kHz.






In the frequency domain, the basic scheduling unit in NR is aphysical resource block (PRB), which is composed of 12 subcarriers.

All subcarriers within a PRB are of the same spacing and CPoverhead.

Since NR should support multiple subcarrier spacings, NR supportsPRBs of different bandwidth ranges.

When PRBs of different bandwidth ranges are multiplexed in thetime domain, boundaries of PRBs should be aligned.






For this purpose, multiple PRBs of the same bandwidth should forma PRB grid.

A PRB grid formed by subcarriers with spacing 15kHz × 2m, wherem is a positive (resp. negative) integer, should be a superset (resp.subset) of PRB grids formed by subcarriers with spacing 15 kHz.




OFDM-Based New WaveformsDesired PropertiesVarious Techniques

New Waveform in NR

There have been considerable discussions on whether a new type oftransmission waveforms, on top of the incumbent CP aided OFDM(CP-OFDM), shall be used in NR.

Schemes alternative to conventional OFDM, including

filterbank multicarrier (FBMC),generalized frequency- division multiplexing (GFDM), and so on,

have been studied for years.

Many of them called for advantages in terms of1 increase of bandwidth efficiency,2 relaxed synchronization requirements,3 reduced inter-user interference, etc.

But there are also challenges in increased transceiver complexity,difficulties in multiple-input multiple-output (MIMO) integration,and specification impacts.





OFDM-Based New Waveforms

OFDM is a mature technology broadly adopted in manifold productsdue to its several merits such as low complexity, easy integrationwith MIMO, plain channel estimation, and so on.

It thus strongly motivates 5G NR still choosing OFDM as the basisof new waveform design.

Distinct from OFDM, new waveforms usually possess additionalfunctionalities to deal with two challenging but crucial issues.

1 Spectral Containment.2 Peak-to-Average Power Ratio (PAPR).





Spectral Containment

One of the major desired properties is to offer enhanced spectralcontainment, that is, lowered out-of-band emission (OOBE).

A waveform with low OOBE may provide the following virtues.

First, as NR will support different numerologies, the interferenceincurred due to orthogonality loss might be severe, which, however,could be mitigated.Second, it is now possible to relax stringent synchronizationrequirements.This merit may facilitate grant-free asynchronous transmissions.In addition, bandwidth utilization might be much more efficient thanthat of LTE, since the amount of guard band would be greatlydecreased.





Peak-to-Average Power Ratio (PAPR)

Another major desired property is low PAPR, more specifically, thetransmit signal quality under the consideration of power amplifier(PA) nonlinearity.

OFDM modulation has been known to possess rather high PAPR,and demands large power backoff to maintain the operation in thePA linear region.

This issue is especially important to uplink transmissions at highcarrier frequencies, since the corresponding impacts on battery lifeand coverage of user equipment (UE) are quite noticeable.

Handling high PAPR also results in spectral regrowth thatdeteriorates the expected spectral containment property.





Various Techniques in New Waveforms

Most waveforms studied by 3GPP for NR can be described as aspecial case of the following figure.

Based on the inverse fast Fourier transform (IFFT) foundation,additional filtering, windowing, or precoding, are considered toachieve the desired enhancements.





Filtering

Filtering: Filtering is a straightforward way to suppress OOBE byapplying a digital filter with pre-specified frequency response.

Candidate waveforms like filtered OFDM (f-OFDM) and universalfiltered OFDM (UF-OFDM) belong to this category.

However, the delay spread of the equivalent composite channel mayeat up CP budget and guard period (GP) in time-division duplexing(TDD) mode, which leads to ISI and imposes burdens ondownlink-to-uplink switch, respectively.

Furthermore, the promised OOBE performance may degradesignificantly when PA nonlinearity exists.

At the cost of increased PAPR, filtering techniques are generallyknown to be unfriendly to communication at high carrier frequencies.





Windowing

Windowing is to prevent steep changes between two OFDM symbolsso as to confine OOBE.Multiplying the time domain samples residing in the extendedsymbol edges by raised-cosine coefficients is a widely usedactualization as chosen by windowed OFDM (W-OFDM) andweighted overlap-andadd (WOLA) OFDM waveforms.This technique generally has little or no PAPR overhead and alsolower complexity compared to that of filtering techniques.Nevertheless, the detection performance might be degraded becauseof ISI caused by symbol extension.





Precoding (1/2)

A linear processing of input data before IFFT is usually known asprecoding, and may be helpful to improve OOBE and PAPR.

One representative example is discrete Fourier transform spreadOFDM (DFT-S-OFDM) waveform that has been adopted in LTEuplink transmissions because of its low PAPR.

Numerous variants of DFT-S-OFDM have been proposed for NR.

Zero-tail (ZT) DFT-S-OFDM aims at omitting CP by letting the tailsamples approximate to zero.Guard interval (GI) DFT-S-OFDM superposes a Zadoff-Chusequence to the tail samples for synchronization purposes.Unique word (UW) DFT-S-OFDM replaces zeros in front of the DFTby certain fixed values to adaptively control waveform properties.





Precoding (2/2)

On the other hand, single carrier circularly pulse shaped (SC-CPS)and generalized precoded OFDMA (GPO) waveforms usepre-specified frequency domain shaping after the DFT for furtherPAPR reduction at the cost of excess bandwidth.

CPS-OFDM can be regarded as a generalized framework thatflexibly supports multiple shaped subcarriers in a subband.

DFT-S-OFDM-based waveforms, in contrast to filter- basedwaveforms, usually make it much easier to maintain PA linearoperation with less deterioration from lowering OOBE.

Moreover, an appropriate modification of modulation schemes, suchas p/2 binary phase shift keyint (BPSK), can greatly assist suchwaveforms in achieving an extremely low PAPR.

Note that in the absence of redundant intervals, ISI still occurs.

From DFT-based precoding techniques, other types of precodingmatrices often have undesirable complexity and compatibility issues.




Multiple AccessInitial/Random AccessSummary

Multiple Access in NR

Previous generations of communication standards rely on orthogonalmultiple access (OMA).

Each time/frequency resource block is exclusively assigned to one ofthe users to ensure no inter-user interference.

Toward NR, synchronous/scheduling- based OMA continues to playan important role for both DL and UL transmissions.

Non-orthogonal multiple access (NOMA) transmission, which allowsmultiple users to share the same time/frequency resource, wasrecently proposed to enhance the system capacity and accommodatemassive connectivity.

Unlike OMA, multiple NOMA users’ signals are multiplexed by usingdifferent power allocation coefficients or different signatures such ascodebook/codeword, sequence, interleaver, and preamble.






The fundamental theory of NOMA has been intensively studied innetwork information theory for decades.

Theoretically, uplink and downlink NOMA can be modeled as amultiple access channel (MAC) and a broadcast channel (BC),respectively, with the capacity region shown in the following figure.






The capacity region of the Gaussian BC can be achieved by powerdomain superstition coding with a successive interferencecancellation (SIC) receiver.

In general, a weak user (i.e., a user with poor channel condition)tends to allocate more transmission power, so a weak user decodesits own messages by treating the co-scheduled user’s signal as noise.

On the other hand, a strong user (i.e., a user with better channelcondition) applies the SIC strategy by first decoding the informationof the weak user and then decoding its own, removing the otherusers’ information.






The possibility and feasibility of practicing the promised gains ofdownlink NOMA have drawn huge attention.

In order to improve multi-user system capacity, 3GPP approved astudy on downlink multiuser superposition transmission (MUST) forLTE in December 2014 to study NOMA and other schemes based onsuperposition coding.

Through extensive discussion in 2015 and 2016, the MUST schemesand corresponding LTE enhancements are identified through anassessment of feasibility and system-level performance evaluations,and are included in Release 14 standard.






Moving toward 5G, in addition to the orthogonal approach, NRtargets supporting UL non-orthogonal transmission to provide themassive connectivity that is desperately required for applications inmMTC as well as other scenarios.

During NR study, at least 15 companies (see TR 38.802) haveevaluated grant-free UL multiple access schemes targeting at leastmMTC.

Due to no need for a dynamic and explicit scheduling grant fromeNB, latency reduction and control signaling minimization could beexpected.






For uplink NOMA, fundamental network information theory suggeststhat CDMA with a SIC receiver provides a capacity achievingscheme.

However, securing uplink NOMA gain requires further system designenhancement.

As the number of co-scheduled users becomes large, so does thedecoding complexity of the SIC receiver.

The message passing algorithm (MPA), a more complexity-feasibledecoding algorithm, as well as other low-complexity receiver designshave recently drawn attention.






Along this research line, several code-spreading-based techniques,including

1 sparse code multiple access (SCMA),2 multi-user shard access (MUSA), and3 pattern division multiple access (PDMA),

have recently been proposed.

It has been shown that one can potentially achieve higher spectralefficiency, larger connectivity and better user fairness with NOMA.





Initial/Random Access

When a UE powers on, it needs to search for a suitable cell tolaunch initial access and the RA procedure.

In LTE/LTE-A, both non-contention- based and contention-basedRA procedures are supported.

For non-contention-based RA, the network semi-persistentlyallocates radio resources to a UE to deliver resource requests.

For contention-based RA, LTE/LTE-A adopt a four-messageexchange procedure shown in the next page





Contention-based RA in LTE/LTE-A

A UE randomly selects a preamble (known asmessage 1) and delivers the preamble to aneNB at a physical random access channel(PRACH). If multiple UEs select the samepreamble collision occurs.

Upon receiving message 1, an eNB replieswith message 2, carrying information on radioresources for UE to deliver the uplinktransmission request.

Upon receiving message 2, a UE sends theuplink transmission request (known asmessage 3) at the allocated radio resources.

At this moment, an eNB is able to identifypreamble collision. Then the eNB may replywith message 4 to grant/reject the resourcerequest.






Although NR may enjoy wider bandwidth onfrequency bands above 6 GHz,communications may suffer from severe pathloss.

Beamforming is thus an inevitable technologyin NR in both the user and control planes,which can be performed at the transmitterside (known as TX beam) or receiver side(known as RX beam).

Due to mobility, the locations of a transmitterand a receiver may change over time, andthus geographic space should be quantizedinto a number of directions.

Both a transmitter and a receiver shouldsweep TX/RX beams over all directions tocapture each other’s location direction, whichis known as beam steering.






For NR, PRACH beam direction is well known to a UE.

A UE thus only needs to transmit message 1 toward the beamdirection of a PRACH, but a gNB has to sweep an RX beam toreceive message 1.

A similar operation is also adopted when a gNB replies with message2 to a UE.

Then messages 3 and 4 can be exchanged via available directionsderived from messages 1 and 2.





Summary

In this part, we studied foundations of radio access includingdeployment scenarios sustaining LTE/NR interworking, framestructure multiplexing multiple numerologies, DFT-S-OFDM- andCP-OFDM- based new waveforms, NOM-based multiple access, RAwith beam steering, and enhanced CA for RA latency improvementare revealed.

The insights provided thus boost knowledge not only for engineeringpractice but also for further technological designs.

Nevertheless, NR is still in the process of development, and a numberof issues and optimizations still remain open for further study.


5g technologies and advances, part ieasitc.com/slides/bs_1.pdf · introduction deployment...

Documents