Understanding 5G Standards
SUNDEEP RANGAN, DIRECTOR, NYU WIRELESS
DEC 14, 2016
1
Outline
Motivation, vision and enabling technologies
3GPP 5G standardization process and activities
Channel models above 6 GHz
PHY & MAC design
Networking issues
Summary and outlook
2
Cellular Generations
1G 2G 3G 4G 5G
Anolog voice Digital voice Internet data Broadband data ?
1981 1991 2000 2008 2020
AMPS GSM, IS‐95 WCDMA, LTE, WiMax
CDMA2000
3
Success of 4G
Ericsson Mobility Report 2016
4
What Will 5G Do?
Many new use cases for cellular
◦ Massive connectivity
◦ AR / VR
◦ Autonomous driving
◦ …
IMT Vision – Framework and overall
objectives of the future development
of IMT for 2020 and beyond, Sept 2015
5
5G Requirements: Many Dimensions
From Roberto Padovani, “The Road to 5G”, Jack Wolf Lecture, NYU, Sept 2016.
6
Understanding Rate
Cellular user experience a distribution of rate
◦ Variability due to many factors
◦ Interference, location, blockage, …
◦ Loading, density / layout of cells, …
Various metrics for rate
◦ Peak rate
◦ Average rate
◦ Edge of cell (5%)
Rangan, New strategies for
femto‐macro cellular
interference control, 2013
7
How to Increase Rate in Cellular?
Shannon Formula
log 1
Rate per user Bandwidth per user, Spectral efficiency
antenna degrees of freedom
Current coding methods close to spectral efficiency bound
Most techniques for 5G: Increase degrees of freedom
◦ Number of cells Densification, greater bandwidth per user in each cell
◦ Bandwidth Millimeter wave
◦ Number of antennas Massive MIMO, higher spatial degrees of freedom
8
Reducing Network Latency
5G goals:
◦ 1 to 4 ms data plane latency (user already connected)
◦ 10 ms control plane (user starts in idle mode)
Total delay has multiple components
Processing Faster decoding, hardware
Queuing Simplified network,
congestion control
Rysavy Research, Transmission Higher data rate, HARQ, MAC
Mobile Broadband Explosion:
3GPP Broadband Evolution to IMT‐Advanced Bring content closer, less hops
Propagation
9
Small Cells and Densification
Cell splitting
◦ Key driver for capacity increase up to 4G
Creates heterogeneous networks
◦ Cell sizes / power
◦ Backhaul
◦ Indoor / outdoor
Considerable work in 4G / LTE‐A
◦ Intercellular interference coordination
◦ Self‐organizing networks
Practical challenges
◦ Backhaul
◦ Site acquisition
10
Massive MIMO
Massive number of antennas
◦ Typically > 100
Large number of spatial streams
◦ Tens of UEs simultaneously
◦ Spatial division multiple access
Targeted for macro base station
Emil Björnson, Radio Resource ◦ 1‐2 meter panel
Management in Massive MIMO
Communication Systems,
Lund‐NI massive MIMO prototype Linkoping Univ
128 elements, 30.72 Mbps in 20 MHz
Erik Luther, 5G Massive MIMO Testbed:
From Theory to Reality, ni.com
11
Capacity Gains via Massive MIMO
Massive MIMO trial system,
Qualcomm simulation, Woodstock, VA. PCS band
Macro site 1.7km ISD 4x12x2 elements, 52 dBm TX
46 dBm transmit power Image courtesy Blue Danube
12
Millimeter Wave
1‐10 mm wavelength = 30 to 300 GHz
Up to 100x bandwidth
Very high‐dimensional antenna arrays
This talk:
◦ MmWave = above 10 GHz
◦ (10‐30 GHz sometimes called cm‐wave)
From Khan, Pi “Millimeter Wave Mobile Broadband:
Unleashing 3‐300 GHz spectrum,” 2011
13
Massive Bandwidth with MmWave
System Duplex fc Antenna Cell throughput Cell edge rate Sim assumptions:
antenna BW (GHz) (Mbps/cell) (Mbps/user, 5%) ◦ 10 UEs per cell
DL UL DL UL ◦ Hex cell layout, ISD=200m
mmW 1 GHz 28 4x4 UE 1514 1468 28.5 19.9 ◦ LTE estimates for 36.814
TDD 8x8 eNB
73 8x8 UE 1435 1465 24.8 19.8
Further gains with spatial
8x8 eNB mux, subband scheduling
and wider bandwidths
Current 20+20 2.5 (2x2 DL, 53.8 47.2 1.80 1.94
LTE MHz 2x4 UL)
FDD
Akdeniz, Mustafa Riza, et al. "Millimeter wave
channel modeling and cellular capacity
evaluation." IEEE JSAC, 2014 ~ 25x gain ~ 10x gain
14
NFV and SDN
Network function virtualization
Software defined networking
Reconfigurable resources
Move content closer to edge
◦ Reduce latency
◦ Reduce backhaul
Distributed mobility
Multiplexing of resources
From 3GPP 32.842
15
Outline
Motivation, vision and enabling technologies
3GPP 5G standardization process and activities
Channel models above 6 GHz
PHY & MAC design
Networking issues
Summary and outlook
16
5G “New Radio”
Single unified framework for diverse applications
Not backward compatible with LTE
Phase 1 (Rel 15, 2018)
◦ Non‐standalone
◦ Focus below 40 GHz
Phase 2 (Rel 16, 2019)
◦ Standalone
◦ Include above 40 GHz
From Qualcomm blog, Acceleration of the 5G NR global
standard gains industry momentum, Sept 2016
17
Timelines
Very aggressive schedule
Three key studies
◦ Requirements
◦ Architecture
◦ RAN
SA1‐RAN interaction
From Giovanni Romano, TIM, 3GPP progress on “5G”, 2016
18
Selected Use Cases in 3GPP 38.913
Use case Carrier (GHz) Bandwidth (GHz) Layout
<6 >6 <6 >6
Indoor hotspot 4 30, 70 200M 1 ISD 20m, 20 UEs per TRP
Dense urban 4 30 200M 1 ISD 200m, micro+macro, 10‐20 UEs per TRP
Rural 2, 4 20,200M ISD 1732, 5000m, mobility
Urban macro 4 30 200M 1G ISD 500m, Focus on ubiquitous coverage
Extreme rural <3 40M 100km cell range, up to 160km/h
Massive connection <3 TBD 1732, 500m, Connection density TBD
Highway < 6 TBD Inter‐RSU 100m, macro 500m
Urban grid for < 6 TBD RSU at each intersection, Macro 500m
connected car
19
Example: Urban Connected Car
Macro only or Macro + road‐side unit (RSU)
Currently focused below 6 GHz
Attributes Values or assumptions
Carrier Frequency Macro only: Below 6 GHz (around 6 GHz)
NOTE1 Macro + RSUs NOTE2:
1) For BS to RSU: Below 6 GHz (around 6 GHz) NOTE3
2) RSU to vehicles or among vehicles: below 6 GHz
Aggregated system [TBD] MHz (DL+UL)
bandwidth NOTE4
Layout Option 1: Macro only
Option 2: Macro + RSUs NOTE2
ISD Macro cell: ISD = 500m
Inter‐RSU distance = [100m] NOTE5
BS antenna elements Tx: Up to [32 Tx]
Rx: Up to [32 Rx]
UE antenna elements RSU Tx: Up to [32 Tx]
RSU Rx: Up to [32 Rx]
Vehicle Tx: Up to [8 Tx]
Vehicle Rx: Up to [8 Rx]
User distribution and 100% in vehicles
UE speed Average inter‐vehicle distance (between two vehicles’ center) in the same lane is [1sec * average
vehicle speed] (average speed: [100‐300km/h])
Traffic model [50 messages] per 1 second with absolute average speed of [100‐250 km/h] (relative speed: 200 –
500km/h)
20
Outline
Motivation, vision and enabling technologies
3GPP 5G standardization process and activities
Channel models above 6 GHz
PHY & MAC design
Networking issues
Summary and outlook
21
Initial NYU MmWave Measurements
Millimeter wave: It can work!
◦ First measurements in urban canyon environment
◦ Distances up to 200m
◦ Propagation via reflections
Sufficient for cellular system at current density
◦ Measurements made urban macro‐cell type deployment
◦ Rooftops 2‐5 stories to street‐level
Rappaport, Theodore S., et al. "Millimeter wave mobile communications for 5G
cellular: It will work!." IEEE access 1 (2013): 335‐349.
22
Key Challenges for mmWave
Directionality http://www.miwaves.eu/
◦ High isotropic path loss
◦ Compensated by directional beams
◦ Impacts all aspects of cellular design
Blockage
◦ MmWave signals blocked by many common materials
◦ Brick > 80 dB, human body > 25 dB
◦ Leads to highly intermittent channels
23
3GPP 38.900: Initial Channel Model Spec
Massive industry effort at wide range of bands, scenarios
Focus on four key scenarios:
◦ Urban macro
◦ Urban micro
◦ Indoor Hotspot (open and mixed office)
◦ Rural macro (up to 7 GHz supported)
Wide range of bands
Some use cases may need further study
◦ Vehicular (including below 6 GHz)
https://www.siradel.com/portfolio‐item/hetnet‐
◦ Massive connection, … deployment‐strategy/
See discussion in Ericsson, Telstra, Vodafone, CMCC,
5G channel modeling way forward,RP161179, June 2016
24
Path Loss and Fading
Path loss, propagation, I‐O penetration
Antenna models
Extends 3GPP spatial cluster channel model
◦ Captures spatial characteristics of the channel
◦ Essential for high‐dimensional arrays
U Mi pathloss models
Sun et al, Propagation Path Loss Models for
5G Urban Microand Macro‐Cellular
Scenarios, IEEE VTC 2016
From TR 25.996
25
LOS and Outage
Three state link models:
◦ LOS, NLOS and outage
Captures loss of signals from blockage
3GPP has detailed LOS models
◦ Various scenarios
◦ Includes spatial consistency
Outstanding issues:
◦ Correlations in multiple cells
◦ Required for macro‐diversity
Akdeniz, Mustafa Riza, et al. "Millimeter wave channel modeling
and cellular capacity evaluation." IEEE journal on selected areas
in communications32.6 (2014): 1164‐1179.
26
Blockage and Channel Dynamics
MmWave signals blocked by many materials
◦ Body, hand, cars, …
Key cause of intermittency
Several new studies to understand time scales
Integrated to 3GPP 38.900
◦ Analytic models (e.g. knife edge diffraction)
◦ Simplified models with mobility
41 42 43 44 5 m
45 46 47 48 49 G. R. MacCartney, Jr., S. Deng, S. Sun, and T. S. Rappaport, “73 GHz Millimeter‐
Wave Human Blockage and Dynamic Measurements,” IEEE 84th Vehicular
0.5 m 1.0 m 1.5 m 2.0 m 2.5 m 3.0 m 3.5 m 4.0 m 4.5 m Technology Conference Fall (VTC 2016‐Fall), Sept. 2016.
TX RX
27
Outline
Motivation, vision and enabling technologies
3GPP 5G standardization process and activities
Channel models above 6 GHz
PHY & MAC design
Networking issues
Summary and outlook
28
OFDM Waveform Options
Many flavors of OFDM considered:
◦ CP‐OFDM
◦ WOLA (windowed overlap and add)
◦ UFMC (universal filter multi‐carrier)
◦ GFDM
◦ …
Key issues
◦ CP overhead flexibility
◦ Out of band / adjacent carrier
◦ PAPR
◦ Multiplexing flexibility
◦ Equalization complexity
Qualcomm, R1‐162199, “Waveform candidates”, Apr 2016
29
DFT‐Spread OFDM
DFT followed by IDFT
Effectively signals in time‐domain
Reduce PAPR
◦ For QPSK modulation
◦ Important in mmWave
Low PA efficiency
But, reduced multiplexing flexibility
Equalize in frequency‐domain
Used in LTE uplink
◦ control and data channels
Image from “Single‐Carrier FDMA”,
https://en.wikipedia.org/wiki/Single‐carrier_FDMA
30
Frame Structure
Flexible frame structure
◦ Scalable TTI
◦ Scalable subcarrier spacing
(15 kHz x 2^n)
◦ Common framework
WAN, D2D, Multicast, …
◦ Puncturing for short control
NYU studies
◦ Can achieve <1 ms airlink latency
◦ Significant benefit control signaling
1. Qualcomm, NYU Talk, Sept 2016
2. Ford et al, Achieving Ultra‐Low Latency in 5G Millimeter Wave
Cellular Networks, 2016
3. Dutta et al, MAC layer frame design for millimeter wave cellular
system, 2016
31
Directionality and MIMO Architectures
Directional transmissions essential for mmWave
Need high‐dimensional arrays
Three dominant architectures
◦ Analog BF: Low power, but “look” in one direction
◦ Digital BF: High power, but most flexible
◦ Hybrid: Combination of both
Significant impact in PHY and MAC
◦ Channel tracking,
◦ Cell search
◦ Control signals
Sun, Shu, et al. "MIMO for millimeter‐wave wireless communications:
beamforming, spatial multiplexing, IEEE ComMag, 2014
32
Directional Initial Access
Initial access:
UE
BS cell
◦ Establish connection from idle mode
Challenges for mmWave:
Sync signal Detects BS ◦ Must find directions of communications
Learns
Random ◦ More widely‐used
direction
access Radio link failure, handover, idle mode to save power
Detects UE
Learns
direction UL grant
Also central problem in massive connection / IoT
Scheduled Latency Airlink RTT Current Target for 5G
transmission measurement LTE
Data plane UE in connected mode 22 ms 1 to 4 ms
Control UE begins in idle mode 80 ms 10 ms
33
Fast Search with Fully Digital
Sync Delay
Low resolution fully digital
◦ One ADC per element
◦ Compensate power via low
resolution (2‐3 bits)
Dramatically better performance
◦ Cell search
◦ Control signal multiplexing
◦ Channel tracking
MIMO Sync delay RA delay Infinite
SNR w/ resolution
Analog BF only 32 ms 128 ms
quantization
Low power digital 4 ms 2 ms
Finite
C. N. Barati et al., "Initial Access in Millimeter Wave Cellular Systems," IEEE Transactions on resolution
Wireless Communications, Dec. 2016. SNR
34
Outline
Motivation, vision and enabling technologies
3GPP 5G standardization process and activities
Channel models above 6 GHz
PHY & MAC design
Networking issues
Summary and outlook
35
5G Network Architecture
Harmonization across multiple RANs
◦ 4G, 5G,…
◦ WiFi
Heterogeneous services
◦ Cellular, IoT, …
Flexible architecture
◦ Network virtualization
◦ Flexible deployment of services
◦ Caching, edge services
Ford et al, Achieving Ultra‐Low Latency in 5G Millimeter Wave Cellular
Networks, 2016
36
MmWave ns3 Module
First, open‐source mmWave module
End‐to‐end
◦ Detailed channel models (statistical, ray tracing, …)
◦ Customizable MAC including adaptive HARQ, …
◦ RLC, PDCP, realistic RRC
◦ Configurable core networking
◦ Handover (inter‐5G and 4G/5G)
https://github.com/mmezzavilla/ns3‐mmwave
Mezzavilla, Marco, et al. "5G mmwave module for the ns‐3 network
simulator." Proceedings of the 18th ACM International Conference on Modeling,
Analysis and Simulation of Wireless and Mobile Systems. ACM, 2015.
37
Insights from Simulations
Statistical Models Ray Tracing models Issues for 5G
◦ Can TCP adapt in mmWave?
Courtesy Andy Nix, U Bristol
◦ Architectures
◦ Traffic patterns in new applications
Simulations reveal several issues
◦ Buffer bloat
◦ TCP start lag
◦ Control / ACK overhead
◦ CN delay
M. Zhang et al., "Transport layer performance in 5G mmWave
cellular," INFOCOM WKSHPS, April 2016
Outline
Motivation and vision
3GPP 5G standardization process and activities
Channel models
PHY & MAC design
Networking
Summary and Perspectives
39
Summary
5G will enable large numbers of use cases:
◦ Massive mobile broadband, vehicular, AR/VR
◦ But, we still don’t know what will be the killer app
Builds on the massive success of earlier systems
Many new technologies
◦ Millimeter wave, Massive MIMO, core network evolution, densification
Significant research but no forseeable show stoppers
40
People
Faculty:
◦ Ted Rappaport, Elza Erkip, Shiv Panwar, Pei Liu
◦ Michele Zorzi (U Padova)
Postdocs: Marco Mezzavilla, Aditya Dhananjay
Students:
◦ Sourjya Dutta, Parisa Amir Eliasi, Russell Ford, George McCartney, Oner Orhan, Shu Sun, Menglei Zhang
U Bristol ray tracing:
◦ Evangelos Mellios, Di Kong, Andrew Nix
41
NYU WIRELESS Industrial Affiliates
42