INDUSTRY PERSPECTIVES

5G Air Interface System Design Principles

Naga Bhushan, Tingfang Ji, Ozge Koymen, John Smee, Joseph Soriaga, Sundar Subramanian, and Yongbin Wei
Qualcomm Technologies, Inc.

The fifth generation (5G) of mobile networks is on

the horizon and promises to be a new scalable frame- Outdoor Subcarrier spacing 2n scaling of
macro coverage e.g., 15 kHz subcarrier spacing
work that can efficiently multiplex envisioned and e.g., FDD 700 MHz
e.g., 1, 5, 10, and 20 MHz
future 5G services. This kind of network will not only
interconnect people, but also interconnect and control Outdoor macro and Subcarrier spacing
machines, objects, and devices. 5G should be a platform small cell e.g., 30 kHz
e.g.,TDD 3–5 GHz
for innovations that can enable new services, empower e.g., 100 MHz
new user experiences, and connect new industries. The
Indoor
design of 5G is focused not only on enhancing and low- wideband
Subcarrier spacing
e.g., 60 kHz
ering the cost per bit for mobile broadband services, e.g., unlicensed 6 GHz
e.g., 160 MHz
but also enabling mission-critical control through ultra-re-
liable, low-latency communication (URLLC) links and Subcarrier spacing, e.g., 120 kHz
mmWave
connecting the massive Internet of Things (IoT). e.g., 28 GHz
At the heart of 5G is a new, more capable, unified e.g., 400 MHz
air interface — 5G New Radio (NR). It adopts an opti-
mized orthogonal frequency-division multiplexing (OFD- FIGURE 1. Example usage models and channel bandwidths.
M)-based family of waveforms and multiple access
techniques, as well as a common, flexible framework that enables subcarrier spacing provides larger cyclic prefix, which can be used
efficient multiplexing of various services and provides the forward to support broadcast service (Fig. 2), in the same carrier. The 5G
compatibility required to future-proof 5G. This article explores these NR unified air interface will also allow carrier aggregation across
key foundational elements that are integral to the 5G NR air inter- numerologies, such as aggregating mmWave and sub-6 GHz car-
face design. 5G NR will also incorporate a plethora of advanced riers to bring more robust and higher-performance connectivity.
wireless technologies that will bring new levels of performance and
efficiency. We explore some of these advanced techniques from our AF F F
lexible ramework with orward ompatibility C
viewpoint, with a focus on the key features targeted for Third Gen- Supporting the wide range of 5G services and devices requires more
eration Partnership Project (3GPP) 5G NR Release 15 and beyond. than optimized waveforms and multiple access. 5G NR is being
designed with a flexible frame structure to efficiently multiplex diverse
D
esigning 5G NR U A I
nified ir nterface 5G services and provide forward compatibility for future ones. This
One of the foremost decisions for designing a unified 5G air equates to flexibility not only in the frequency domain, as discussed
interface is the choice of radio waveforms and multiple access earlier with scalable OFDM numerology, but also in the time domain.
techniques. Not only will they need to deliver high performance The 5G NR framework will be able to support the diverse services,
at low complexity, but they must also be capable of supporting features, and deployment scenarios envisioned for 5G. Some high-lev-
(and multiplexing) the initial and future 5G use cases efficiently. el features of this new flexible framework are described in this section.
Based on extensive studies [1], we believe that the OFDM fam-
ily is the right choice for 5G enhanced mobile broadband (eMBB) S T T I
calable ransmission ime nterval
and beyond. OFDM allows for enhancements such as waveform In the time domain, 5G NR will enable scalability to latencies
windowing/filtering, which can effectively minimize in-band and out- much lower than what is possible in today’s LTE networks. Today,
of-band spurious emissions — critical for 5G service multiplexing. The LTE supports a fixed transmission time interval (TTI) of 1 ms. In
OFDM family can also efficiently coexist with other waveforms and the LTE evolution path, there is an ongoing Work Item on latency
multiple access schemes in the same framework, such as supporting reduction in 3GPP. Although the technical details are still under
asynchronous, grant-free transmissions (e.g., RSMA1) for connecting discussion, the design target for hybrid automatic repeat request
IoT devices and enabling mission-critical control communications. (HARQ) retransmission time is likely greater than 8 shortened
TTIs (sTTIs) for 1.14 ms with 143 ms sTTI. To better support ser-
S calableOFDM N S S
umerology with caling of ubcarrier pacing S vice with short latency requirement, the 5G NR flexible frame
Today, LTE supports carrier bandwidths up to 20 MHz with structure is being designed with TTI that will scale up and down,
mostly a fixed OFDM numerology — 15 kHz spacing between depending on specific service requirements. This flexibility allows
OFDM tones or subcarriers.2 5G NR, on the other hand, will the air interface to optimize for lower latencies using shorter TTI
introduce scalable OFDM numerology to support diverse spec- (e.g., hundreds of microseconds) or trading off for higher spectral
trum bands/types and deployment models. For example, 5G efficiency for delay-tolerant use cases with a longer TTI. Scalable
NR must be able to operate in millimeter-wave (mmWave) frame structure is also designed to cover all the frequency bands,
bands that have extremely wide channel widths (e.g., hundreds from sub-6GHz all the way to mmWave bands.
of megahertz). It is critical that the OFDM subcarrier spacing is In addition to a scalable TTI, 5G NR will also support ser-
able to scale with the channel width, so the fast Fourier trans- vice-aware TTI multiplexing on the same frequency, which allows
form (FFT) size scales such that processing complexity does not transmissions with different TTIs to start on integer symbol bound-
increase exponentially for wider bandwidths. aries instead of a subframe boundary (i.e., 1 ms). For instance,
In addition to supporting different channel widths for different a high-quality of service (QoS) mobile broadband service may
deployment types, as seen in Fig. 1, 5G NR is also being designed 1 Resource-spread multiple access.
to accommodate different/scalable numerologies, allowing ser-
vices that use different bandwidths to efficiently multiplex in the 2 Some exceptions, such as narrowband IoT (NB-IoT) as defined in 3GPP Release
same network. For example, using scalable numerologies, smaller 13, can support transmissions with 3.75 kHz subcarrier spacing.

6 IEEE Wireless Communications • October 2017

 INDUSTRY PERSPECTIVES

choose to utilize a 500 ms TTI instead y

og
of a standard, LTE-compatible 1 ms erol Blank resources Integrated framework Mission-critical transmissions
TTI/subframe, while another laten- num
ble
cy-sensitive service further shortens its Sca
la
TTI to ~140 ms with scaled numerol- D2D
ogy. Instead of requiring the second,
latency-sensitive transmission to wait Scalable TTI MBB Multicast
until the beginning of the next sub- DL DL UL UL UL
frame (i.e., 500 ms later), it can begin
Scalable transmission time Self-contained integrated Dynamic uplink/downlink
as soon as the previous transmission is interval (TTI) subframe
completed on the symbol boundary,
thereby eliminating a waiting period. FIGURE 2. Scalable framework that can efficiently multiplex envisioned and future 5G services.

Self-Contained Integrated Subframe Adaptive

The self-contained integrated subframe is another key enabler Unlicensed
UL/DL
spectrum
for lower latency, forward compatibility, and many new 5G
NR features. The lower latency is achieved by having the data
transmission and its acknowledgment all contained in the same Add’l Ctrl Data ACK

period
Guard
headers (Tx) (Tx) (Rx)
subframe. Figure 3 shows an example of a time-division duplex
(TDD) downlink-centric subframe, where data transmission is
from the network to the device, and the acknowledgment is D2D, mesh
and relay Massive
sent by the device back to the network in the same subframe. MIMO
With the 5G NR self-contained integrated subframe, each TTI is
now a modular transaction (e.g., DL grant  DL data  guard FIGURE 3. Self-contained integrated subframe design (e.g., TDD
period  UL ACK) that gets completed within that time period. downlink).
The modular aspect of the self-contained integrated subframe
design also allows for different types of subframes to be multiplexed Massive MIMO and mmWave for Extreme Capacity
for new services that are introduced in the future. This, along with
the ability for the 5G NR framework to support blank subframes
and Uniform User Experience
and blank frequency resources, enables a forward-compatible 5G One key innovation area in wireless communications is advanced
NR design for easily adding future features/services to be deployed antenna technologies. By using more antennas intelligently, one
in the same frequency in synchronous and asynchronous manners. can improve both network capacity and coverage. That is, more
The subframe can also contain additional precursor headers spatial data streams can significantly increase spectral efficiency
that can be used to provide additional information for the trans- (e.g., with MU-MIMO3), allowing more bits to be transmitted per
mission. For example, operating in unlicensed or shared spectrum Hertz, and smart beamforming can extend the reach of base sta-
typically requires the support of Listen-Before-Talk (LBT) to ensure tions by focusing RF energy in specific directions. In LTE today, net-
fair sharing across different users. These LBT headers can include works are evolving from 2  2 to 4  4 MIMO, with even more
measurement gaps used by downlink devices to assess channel antennas in sight. However, there is an intrinsic limitation on how
availability for data transmission as well as precursor transmissions many antennas one can realistically fit onto a device, especially at
to signal channel occupancy. And when used in device-to-device low frequencies where the antennas are large due to long wave-
communications, these headers can indicate the link direction and length. One way to further increase capacity without adding more
provide scheduling information for the directly connected devices. device antennas is to have more antennas at the base stations.
The self-contained subframe also plays an integral role in enabling By continuing to evolve FD-MIMO,4 which was first intro-
advanced 5G NR antenna techniques, such as massive multiple-in- duced in LTE Advanced Pro, 5G NR will support massive
put multiple-output (MIMO). On the downlink, in order for a trans- MIMO [2, 3], which can utilize an even larger number of anten-
mitting cell to more efficiently direct RF energy to a device, it needs na elements, supporting up to 256 as currently defined. The 2D
to continuously evaluate link quality and make necessary beam- antenna arrays are capable of 3D beamforming in both the azi-
forming adjustments. The feedback mechanism is provided by the muth and elevation planes, and test results have shown signifi-
transmission of common uplink burst, which carries uplink control cant gains in both capacity and cell edge user throughput — key
information (e.g., the ACK) and the uplink sounding reference sig- to bringing a more uniform mobile broadband user experience.
nals. More accurate and timely knowledge of downlink channel can Massive MIMO is also a key enabler for opening up the
be obtained at the base stations thanks to channel reciprocity, which high-frequency bands in the sub-6 GHz spectrum. With intel-
enables the use of uplink sounding reference signal for downlink ligent beamforming and beam-tracking, it is possible to reuse
channel estimate in a TDD system. Note that in addition to the existing cell sites (e.g., at 2 GHz) and transmit power for new
uplink control information and sounding signals, common uplink macrocell networks that operate at higher frequencies (e.g.,
burst can also be used to carry time critical data (e.g., TCP ACK). at 4 GHz). Simulation results demonstrated more than 4
An uplink-centric subframe has a similar structure, with sim- improvement in average capacity going from a 2  4 MIMO
ilar downlink and uplink control bursts at the beginning and to a 24  4 massive MIMO setup, and almost the same level of
the end of the subframe, but with an uplink data burst that fol- throughput enhancements for users at the cell edge.
lows a guard period. This design reduces downlink and uplink 5G NR will not only enable the use of higher frequencies in
control channel interferences by requiring all control bursts to the 3 to 6 GHz band for macro/small cell deployments, but will
be transmitted in the same direction across neighboring cells, also open up new mmWave opportunities for mobile broadband.
thereby allowing more robust link direction switching. The The abundant spectrum available at these high frequencies is
dynamic configuration of downlink- or uplink-centric subframes
3 Multi-user MIMO.
increases overall network efficiency and capacity by allowing
faster switching based on network traffic conditions. 4 Full-dimensional MIMO.

IEEE Wireless Communications • October 2017 7

INDUSTRY PERSPECTIVES

capable of delivering extreme data speeds and capacity that code (TBCC). In fact, the encoding structure of these codes can
will reshape the mobile experience. However, mobilizing the be thought of as generalizations of Reed Muller codes, which
mmWave comes with its own set of challenges. Transmissions are also employed at the smaller blocklengths.
in these higher bands suffer from significantly higher path loss as
well as susceptibility to blockage, while meeting the power and Carrier Aggregation across Spectrum Bands and Types
form-factor requirements of mobile devices has also proven to be 5G will build on the solid LTE carrier aggregation (CA) founda-
challenging. Further, antennas for a given aperture get progres- tion to support aggregation across frequency bands (e.g., low
sively more directive with frequency. Thus, traditional mmWave band FDD, mid-band TDD, and mmWave TDD) and types (e.g.,
implementations have been limited to mostly stationary applica- licensed, shared, and unlicensed).
tions such as shorter-range wireless docking, enabled by technol- CA delivers many benefits; for example, adding a supple-
ogies like 802.11ad that operates in the 60 GHz band. mental low-band FDD uplink to mid-band TDD can improve
With the recent advancements in signal processing, mmWave system performance for both NSA and SA5 deployments. For
components, and antenna technologies, the idea of mobilizing NSA, the low-band LTE or supplemental NR uplink can enhance
the mmWave is no longer out of reach. By utilizing a large num- the data rate and range of the deployment. For SA, using low-
ber of antenna elements in both the base station and the device, band NR to carry uplink control and data for cell edge users
along with intelligent beamforming and beam-tracking algorithms can improve robustness of the mid-band system.
[4], 5G mmWave can provide increased coverage, reduced inter-
ference, and a continuous connectivity experience even for non- Device-Centric Mobility to Improve Energy
line-of-sight (NLOS) communications and device mobility. and Overhead Efficiency
In order to provide continuous connectivity with high data rate In addition to downlink-centric mobility, 5G NR will consider uplink-
in mmWave bands, it is imperative to have mechanisms that can or device-centric mobility, which allows the device to send out peri-
provide the mmWave base station and user equipment (UE) with odic reference signals for the access network to monitor; with this
the right beamforming patterns at any given time. The beams cho- information, the network can trigger cell reselection or handover
sen should provide near-optimal link gain as well as be robust to based on the uplink signal strength measurement. This offloads the
sudden changes or mobility at the UE or in the channel. However, device from monitoring and processing reference signals from all
traditional techniques of learning the full channel matrix (NT  NR) nearby cells, thereby decreasing both signaling and processing over-
and computing the right Eigenmode beamformers for the channel head. This is in contrast to LTE, where device mobility is driven by
do not work well for mmWave channels as: a more overhead-heavy process that involves the device measuring
• Learning the full channel matrix is time- and resource-intensive. the signal strength of a downlink reference signal sent by the access
• The Eigenmode-based beamformers are very sensitive to network, which requires overhead processing on the device. To
channel changes. optimize even further, the 5G network will also reduce the amount
• The channel is likely to be primarily composed of a few spec- of broadcasts. Instead of sending system information regardless
ular/strong directional components [8]. of device presence, the 5G network will only send out minimum
Initial directional search based on periodic directional sweeps system information periodically, and on-demand system information
and further refinement and on-demand tracking around primary upon requests from devices. The reduction in signaling and process-
directions identified in periodic sweeps have been observed to be ing overhead will enable longer battery life on the devices as well
more robust and close to optimality. In [4], a few simulation results as make the network more energy-efficient; in addition, the new
and analytical results comparing different approaches to beam device-centric mobility design also contributes to supporting seam-
search and tracking are provided. Moreover, 5G NR will also lever- less handovers at higher speeds (e.g., up to 500 km/h).
age multi-connectivity with 5G sub-6 GHz and/or gigabit LTE to
improve overall link robustness and to help achieve faster system Conclusion
acquisition. Additional design considerations have also been incor- In summary, 5G is a unifying connectivity fabric that will expand the
porated to enable easy deployment of dense mmWave small cells, value of mobile networks to take on a much larger role than previ-
such as with the support for integrated access and backhaul. ous generations, empowering many new connected services across
an array of world-changing use cases. At the heart of 5G is the new
Advanced Channel Coding 5G New Radio unified air interface that is being designed to meet
Along with the scalable numerology and flexible framework for the expanding connectivity needs in the next decade and beyond.
5G NR services, the physical layer design should include an effi- References
cient channel coding scheme that can provide robust perfor- [1] “5G Waveform & Multiple Access Techniques Whitepaper,” Qualcomm Tech-
mance and flexibility. Although turbo codes were the basis for nologies, Inc.; https://www.qualcomm.com/documents/5g-research-wave-
3G and 4G technologies, low-density parity check (LDPC) codes form-and-multiple-access-techniques
[2] T. L. Marzetta, “Noncooperative Cellular Wireless with Unlimited Numbers of Base
[5] have demonstrated advantages from both the complexity and Station Antennas.” IEEE Trans. Wireless Commun., Sept. 11,, 2010, pp. 3590–600.
implementation standpoints when scaling to very high through- [3] C. Shepard et al., “Argos: Practical Many-Antenna Base Stations,” Proc. 18th
puts and larger blocklengths. In particular, an advanced LDPC ACM Annual Int’l. Conf. Mobile Computing and Net., 2012.
channel coding scheme with a flexible parity check matrix struc- [4] V. Raghavan et al. “Beamforming Tradeoffs for Initial UE Discovery in Millimeter-Wave
MIMO Systems,” IEEE J. Selected Topics Signal Proc., Oct. 3, 2016, pp. 543–59.
ture can provide 5G NR with full rate compatibility including [5] T. Richardson and U. Rüdiger, “Multi-Edge Type LDPC Codes,” workshop honor-
incremental redundancy (IR) HARQ, and flexibility for blocklength ing Prof. Bob McEliece on his 60th Birthday, Caltech, Pasadena, California, 2002.
scaling and fine rate granularity, all while allowing for an easily [6] E. Arıkan, “Channel Polarization: A Method for Constructing Capacity-Achiev-
parallelizable decoder design that can be scaled to achieve high- ing Codes for Symmetric Binary-Input Memoryless Channels,” IEEE Trans. Info.
Theory, vol. 55, no. 7, July 2009, pp. 3051–73.
er throughputs at low complexity. Such codes can also provide [7] I. Tal and A. Vardy, “List Decoding of Polar Codes,” IEEE Trans. Info. Theory, vol.
an effective solution for mission-critical and massive IoT traffic, 61, no. 5, May 2015, pp. 2213–26.
efficiently reusing the hardware from mobile broadband services. [8] 3GPP Tech. Spec. Group Radio Access Network, “Channel Model for Frequen-
For very small blocklengths such as 80 bits or less, which cy Spectrum above 6 GHz, TR 38.900, June 2016.
are often used in physical control channels, polar codes [6, 7] 5 Non-standalone(NSA) 5G NR utilizes LTE as an anchor (e.g., to support mobility
provide an encoding structure that can allow 5G NR to achieve management and other signaling). Standalone (SA) 5G NR operates as an inde-
performance gains over the existing LTE tail biting convolutional pendent system without an LTE anchor.

8 IEEE Wireless Communications • October 2017