Understanding 5G, A Practical Guide to Deploying and Operating 5G Networks, 5G New Radio (Part 2)

In the span of twenty-five years, the world went from a few million phones with only?voice capability at the beginning of 1990 to almost five billion phones with more than?twenty thousand exabytes of mobile data used by the end of 2015. Despite these amazing?technological feats, not everyone on Earth has a mobile phone. In 2020, we estimate that?between twenty-five and thirty percent of Earth’s citizens still have no mobile device. On?the other hand, we estimate upwards of twenty billion Machine Type Communications?(MTC) connections are active in that same year.

One key element that will enable this exponential expansion in connections and usage is?the efficient use of radio spectrum at higher and higher frequency bands.

The radio spectrum ranges from low-frequency waves at around 10 kHz up to high-frequency waves at 100 GHz. The mobile network uses predefined bands of various widths?at different frequency ranges in the radio spectrum. We can think about these frequency?bands as highways: the wider the highway, the more cars (data) go through it. However,?the higher the frequency, the more energy is needed to make it go through obstacles?(walls, glass, trees, air, etc.), which means the shorter distance it can go before dissipating.

The first generation of mobile technologies used frequencies ranging from 450 MHz to 900?Mhz. The second generation, 2G, expanded the frequency ranges used up to 1900 MHz.

With the enhancements made by the General Packet Radio Service (GPRS or 2.5G), mobile?data speeds reached upward of 114 Kbps (kilobits per second). Enhanced Data Rates for?GSM Evolution (EDGE) supported speeds of more than 200 Kbps.

Smartphones started to emerge and expand toward the end of the 2.5G era. While using?the same spectrum frequencies as its predecessor, 3G went on to efficiently use the?spectrum by splicing the data across different frequency channels, hence increasing the?total throughput per single user. In 3.5G, data rates upwards of 40 Mbps (megabits per?second) were achievable.

4G expanded the frequencies used to between 3 GHz and 4 GHz and added more?spectrum efficiency by using a technology called OFDM (Orthogonal Frequency Division?Multiplexing); speeds of 100 Mbps were attainable.

LTE Advanced (aka 4.5G) added carrier aggregation, a technology to combine two or more?separate LTE carriers into one data channel, to support speeds up to 1 Gbps (gigabits per second).

2.1 5G to Simultaneously Reduce Latency by an Order of Magnitude?and Increase Data Speeds by One to Two Orders of Magnitude over LTE

As we saw in part 1, the aspirational capabilities of 5G are ultimately driven by a desire?to create a network able to simultaneously satisfy a vast range of use cases with widely?disparate performance characteristics. The adaptable and dynamic network that results will?foster an ecosystem in which new services that were previously not possible can be created.

The resulting technological inflection may even see the emergence of entirely new industries.

The 5G system needs to satisfy various performance goals, referred to by standards?organization 3GPP as the 5G service enablers. These are the enormous data throughput?rates required by enhanced Mobile Broadband (eMBB) along with Ultra-Reliable Low?Latency Communications (URLLC), and massive Machine Type Communications (mMTC).?Satisfying each of these service enablers in isolation is challenging. Satisfying all these?together simultaneously in the same network further compounds that challenge. An?additional level of complexity is introduced by the desire to reduce the power consumption of?5G. As we shall see in this part, the new radio aspects of 5G standardized by 3GPP take?a significant step toward making these requirements a reality.

Here we examine why the performance goals being asked of 5G are so challenging, and?how the 5G NR is addressing those challenges.

?2.2 Frequency Bands for 5G

To put the challenge for 5G to deliver the eMBB service enabler into context, consider?how enormously consumers and business fixed broadband services have evolved over the?last two decades. Innovations in transmitting data at ever higher rates and over longer?distances through xDSL, cable, and fiber connections have underpinned the explosion in the streaming media industry. The penetration of fiber links into these networks and the?associated ability to transmit with less noise and signal attenuation has pushed data rates?way up and unlocked more industry verticals to the retail subscriber.

The wireless broadband challenge stems from the inherently shared nature of the medium.?In comparison to fiber, xDSL, or cable, the radio interface is an analog and shared resource?subject to harsh conditions such as interference and noise. If the eMBB service enabler is to?be delivered, the network must squeeze orders of magnitude more data through the same?radio resources. While there are numerous hurdles to this, the principal limitation is that?the spectral efficiency is reaching the Shannon limit, meaning we cannot do significantly?better just by squeezing more data through the same spectrum resources without a radical?rethink of how to utilize the resource more efficiently.

Although 5G does address this spectral efficiency limitation, it also addresses the?limitation through the allocation of a new spectrum for 5G carriers. This immediately raises the?challenge that the valuable resources below 6 GHz are in very limited supply. Older mobile?communication technologies have relied on the spectrum in this range because it has more?attractive propagation characteristics. These characteristics include an ability to achieve?non-line-of-sight propagation by combinations of reflection, scattering, or diffraction, and in some cases penetration of physical barriers. While some of this can be “re-farmed”?from older technologies to 5G, this takes the spectral efficiency only a little closer to the?Shannon limit rather than delivering the orders of magnitude improvements required by?5G. The spectrum is also still in short supply.

The solution to the capacity problem adopted by 5G is to open up the higher frequencies,?including millimeter Wave (mmWave) bands, because they have the major advantage of?the significantly more plentiful spectrum, and spectrum regulators are more willing to make?them available for use by operators of 5G services. The mmWave range also benefits from?the availability of wider contiguous blocks of spectrum able to accommodate single 5G?carriers with larger frequency bandwidth, which leads to less loss of spectral efficiency?from overheads. Although 5G supports aggregation of different carriers together for?communication to devices, larger contiguous blocks of spectrum can make for simpler?management of the spectrum and less complex devices.

The use of spectrum above traditional blocks for wireless communication and up to the?mmWave range imposes various challenges on the 5G standards. At these higher?frequencies, propagation starts to become challenging since they typically require line-of-sight. Also, transmissions in this frequency range attenuate through solid objects, meaning?that outdoor base stations cannot generally provide indoor coverage. These bands also?experience less scattering, reflection, and refraction, presenting more challenges to the 5G?network operator to plan and deliver reliable coverage.

This greater supply of spectrum is coupled with other technological solutions in the 5G?standards to address the propagation problems at these bands. Received power reduces?with the square of the wavelength (multiplied by λ2/4π) owing to the smaller antenna?aperture at lower wavelengths. The higher transmission losses at these bands also need?to be addressed. Highly directive radiation of energy can overcome this problem and is?a feature of massive MIMO beamforming that will be covered later in this part. In?practice, although physical antenna dimensions can be smaller for these wavelengths,?larger antennas are required to capture sufficient energy.

The 5G standards separate the spectrum supported into two main portions. These are?consistent with the bands identified in the International Mobile Telecommunication-2020?(IMT-2020) of the International Telecommunication Union (ITU). This defines Frequency Range 1 (FR1) which occupies the range 450 MHz to 6 GHz. The frequency bands for FR1?defined by 3GPP are shown in Figure 1 below.

Figure 1. Snapshot of frequency bands defined for FR1 by 3GPP in release 15

Figure 2.?Snapshot of frequency bands defined for FR2 by 3GPP in release 15

There are also frequency bands defined as FR2, shown in Figure 2, occupying a spectrum?between 24.25 GHz and 52.6 GHz, although no bands above 40 GHz are defined in Release?15 of the standards.

This latter range is almost entirely in the mmWave range. Real-life deployments will?generally depend on where the available spectrum is in the associated jurisdiction. The?relatively high availability of mmWave spectrum together with technological solutions for?the problems of transmission at these frequencies means that the spectrum strategy of 5G?is a major cornerstone of the ability of the standards to deliver eMBB use cases.

?

2.3 5G Waveforms, Numerologies, and Frame Structure

The waveform choices for 5G were determined by balancing many factors. The resulting?system had to be capable of delivering the massive data rates required by eMBB. This?meant it had to be sufficiently compatible with massive MIMO. It also had to perform?well overall the frequency bands under consideration for 5G. Transmitter and receiver?complexity was also a consideration, even over large bandwidths.

The 5G waveform is based on OFDM. The downlink and uplink use Cyclic Prefix-Orthogonal?Frequency Division Multiplexing (CP-OFDM) while the uplink (UL) can optionally use?Discrete Fourier Transform-Spread-OFDM (DFT-S-OFDM). While CP-OFDM is inherited from LTE of which it was a cornerstone, DFT-S-OFDM is included in 5G because it improves?uplink coverage. DFT-S-OFDM benefits from low peak to average power ratio (PAPR). This?means that the RF amplifier can be simpler and require less power while avoiding the?distortion associated with a large dynamic range.

OFDM has a high PAPR. The high dynamic range restricts the efficiency that the amplifier?can achieve. It also depends on subcarriers that are stable and aligned concerning each?other in the frequency domain. Deviation from this will lead to some loss of orthogonality?between subcarriers. Despite these disadvantages, the OFDM system has many significant?advantages. Because subcarriers across a wide bandwidth can be selected dynamically for?transmissions, it has the resilience to frequency selective fading. For example, interference that?is limited to part of the carrier bandwidth will only affect part of the transmission and in?some cases can be avoided. Because the symbols in OFDM are relatively long, they are?more resilient to inter-symbol interference than other systems with shorter symbols.

2.3.1 Symbols and Modulation

In 5G NR, data is transmitted in symbols, each of which carries one or more bits of?information in a single tone. The number of bits conveyed is determined by the modulation?scheme. The quadrature phase-shift keying (QPSK) 16 quadrature amplitude modulation?(16QAM), 64QAM, and 256QAM modulation schemes are supported by CP-OFDM for downlink (DL) and UL, and the DFT-S-OFDM in the UL. These modulation schemes convey?two, four, six, and eight bits of information respectively. The inclusion of the 256QAM?modulation scheme allowing eight bits to be transmitted per symbol in the best radio?conditions supports the eMBB service enabler.

The DFT-S-OFDM UL additionally supports the π/2-binary phase-shift keying (BPSK)?modulation scheme, which conveys a single bit of information for each symbol. The benefit of?this modulation scheme is that the phase shift changes at each symbol transition, irrespective?of whether the underlying conveyed bit is the same as or different from the previous bit.?This makes it easier for the receiver to maintain phase lock and keep track of the boundaries?between symbols. This can be particularly advantageous in complex propagation environments?with delay spreading from multipath propagation. π/2-BPSK also benefits from a lower PAPR,?which leads to increased efficiency of the power amplifier, particularly for lower data rates.

The supported modulation schemes are shown in Table 1 and the constellation diagrams?are illustrated in Figure 3.

Table 1. Supported modulation schemes in 5GNR

Figure 3. Constellation diagrams for supported modulation schemes in 5G NR

2.3.2 Subcarriers

OFDM systems pack many subcarriers close together in frequency. OFDM demodulators?depend on fast Fourier transforms to separate the various sub-carriers so that they?can be demodulated. However, the different subcarriers are not orthogonal. If the lack?of orthogonality is not addressed, it will impair the ability to demodulate the different subcarriers successfully. This orthogonality is at a maximum at multiples of λscs Hz from the?subcarrier center frequency where λscs is the reciprocal of the time over which the symbol?tone is transmitted or symbol length for brevity. In the case of the 5G NR, the standard?symbol length is 66.67 μs. The reciprocal of this is 15 kHz, which is the standard subcarrier?spacing, identical to the subcarrier spacing for LTE.

The 66.67 μs symbol duration is preceded by a 4.7 μs cyclic prefix, identical to the approach?for wideband transmission used in LTE. As in LTE, the purpose of the cyclic prefix is?twofold. The main reason is to provide separation between adjacent symbols so that the?transmission can be resilient to delay spread from the multipath transmission. If delay spread?does occur, then as long as it does not exceed the length of the cyclic prefix, it will not?interfere with the next symbol. In practice, in 5G and LTE, the cyclic prefix is not devoid?of transmission. Rather, the end of the symbol about to be transmitted is replicated and?transmitted in the cyclic prefix. In practice, this means that the transmitted tone lasts a?little longer than it would otherwise. This has the beneficial side effect that if the delay?spread is minimal, then the receiver can use the transmission in the cyclic prefix to support?its effort to demodulate the symbol. This adds to the resilience of the receiver, lending?support to the requirement for ultra-reliability.

Figure 4. OFDM waveforms showing orthogonality between subcarriers with perfect waveforms.

Each subcarrier coincides with a minimum of inter-subcarrier interference.

2.3.3 Guard Bands and Spectral Efficiency

OFDM subcarriers cannot utilize the whole spectrum band. The power of the subcarriers?will spread out around the center frequency and tail off. If this is ignored, then the?power can leak out of the band in use whether licensed or unlicensed. OFDM, like many

transmission schemes, requires a guard band at the edge of a carrier where no subcarriers?are transmitted. While LTE was able to utilize ninety percent of the carrier for subcarriers,?5G does better than this, using techniques such as windowing and filtering to contain?the transmissions. The guard band is a fixed overhead per carrier. The guard band must?also be used at the edge of each carrier and eliminates otherwise useful spectrum from

being used for subcarriers. Very large carriers are introduced in 5G NR. Carriers of up to 100?MHz are supported for FR1. In FR2 carriers of up to 400 MHz are supported as well. These?carriers span large frequency ranges, meaning that the effect of the guard band overhead?is reduced, and spectral efficiency is improved.

2.3.4 Flexible Subcarrier Spacing and Numerologies

The subcarrier spacing is chosen not only to defeat the interference between subcarriers?but also to fix the problem arising from phase noise. This is caused by imperfections in?the oscillators, meaning that the modulated tone is not perfect. This also means that the?transmitted energy will be spread over an interval centered on the center frequency of the subcarrier with sidebands of phase noise interference. As the phase noise phenomenon?increases, it can spread out over the frequency domain to the extent that it interferes with?nearby subcarriers.

The problem of phase noise becomes more significant at higher frequencies and the width?of the sideband interference increases. Therefore, phase noise becomes more serious in?mmWave parts of the spectrum. If the 15 kHz channel spacing were maintained across all 5G carriers, phase noise would become significant and, coupled with the challenging?propagation at higher frequencies, would result in a serious loss of capacity in mmWave.

To address the problem of phase noise, 5G NR deviates from its predecessors and?introduces flexible subcarrier spacing. This technique allows the subcarrier spacing to be 15?kHz multiples of 2μ where μ of 0, 1, 2, 3, or 4 results in subcarrier spacings of 15, 30, 60, 120,?or 240 kHz. These are known as numerologies, with numerologies having a larger power of?two being referred to as higher numerologies.

But increasing the subcarrier spacing brings a side effect. The symbol length must be?reduced with higher numerologies owing to the need for the subcarrier spacing to be the?reciprocal of the symbol length. If this were not the case, maximum orthogonality between?subcarriers in the FFT as described above would not be maintained. The symbol length for a subcarrier spacing of 240 kHz is only 4.17 μs, which is 1/16th of the symbol length for 15 kHz subcarrier spacing. This appears to be a problem. As multipath propagation?spreads the signal, the symbols can bleed into one another. At first glance, these much?shorter symbols appear to be far more vulnerable to inter-symbol interference. In practice,?the length of the symbol that can be used is dependent on the frequency band. As the frequency increases, the shorter symbols can be used. But this coincides with the?frequencies that are less subject to inter-symbol interference. Delay spread becomes less?significant at higher frequencies as propagation is predominantly line-of-sight and so the?very shortest symbols can be used for very high frequencies.

The use of higher numerologies also has a secondary advantage. The mmWave carriers?can span hundreds of MHz. This is demanding for the transmitter and receiver if the?normal 15 kHz subcarrier spacing is maintained as larger Inverse Fast Fourier Transforms?(IFFTs) must be used, leading to more complex devices. Spacing the subcarriers further?apart and shortening the symbol length, therefore, results in resilience to phase noise?while maintaining resilience to inter-symbol interference and the orthogonality between?subcarriers in the receiver. It also demands less complexity in the transmitter and receiver?for the widest carriers. Wider subcarrier spacing and the associated shorter symbols have implications for low latency communications. As we shall see, communication is broken up?into self-contained units called slots, normally consisting of fourteen symbols. These slots?become shorter in time as the numerology increases and the subcarrier spacing becomes?larger, meaning that there are more frequent opportunities for scheduling data.

Some complexity arises with carrier bandwidth parts (BWP). These will be introduced fully?later. Different BWPs can have different numerologies with varying subcarrier spacings and?symbol lengths. If two BWPs adjacent in the frequency domain has different subcarrier?spacing, the orthogonality between subcarriers in the receiver will be broken. This requires?the insertion of a guard band between the bandwidth parts, which impairs the efficient use of the spectrum. An alternative is filtering in the receivers which supports BWP to maintain?better spectral efficiency.

The 5G system must be able to support a wide range of environments including those?with complex propagation and specifically those with high degrees of delay spread. As the?complexity of the propagation grows, so must the length of the cyclic prefix to ensure that?the receiver can reliably demodulate the symbol. Rather than set a cyclic prefix length long?enough for every conceivable propagation environment that the system must support, 5G?NR has the concept of the extended cyclic prefix for complex environments supporting the?normal cyclic prefix for regular environments.

2.3.5 Frame Structure and Slots

The physical radio resources can be thought of as a set of subcarriers in the frequency?domain and a set of opportunities for modulated symbols in the time domain. A single?modulated symbol on one subcarrier is called a resource element. These resource elements?are grouped into logical structures that can be used for transmission and?reception. As in LTE, there are ten subframes per frame and each subframe is broken down?into a variable number of slots such that the number of slots per subframe is dependent?on the numerology, as discussed above. Various numerology options are illustrated in?Figure 5 for the lower numerologies and Figure 6 for the higher numerologies.

In the frequency domain, the resource elements are arranged in groups of twelve?subcarriers that are called resource blocks. This is illustrated in Figure 7. Standard slots?have fourteen symbols per slot, but mini-slots can comprise seven, four, or two symbols.?These mini-slots are designed for low latency applications and allow transmissions to be?rapidly acknowledged. Traffic pre-emption means that they can be inserted when required?for low latency applications.

There are various engineering compromises in the choice of frame structure and how the?resources are managed. The resources must be utilized efficiently. The management must?be flexible and accommodate different mixes of DL and UL data demands. Ultra-Reliable?Low Latency must be supported with data being scheduled when required without delay?and acknowledged immediately.

Figure 5. 5G NR lower numerologies

Figure 6. 5G NR higher numerologies

Figure 7. Resource grid, resource blocks, and resource elements within frames for numerology 1

For example, network slicing is a key capability for the 5G network. This capability allows?different classes of subscribers with wildly different quality of service requirements to be?satisfied simultaneously with the same network. The frame must support network slicing?and other mechanisms for separating users with different requirements such that the?radio bandwidth can be controlled by different logical network functions simultaneously,?potentially with different functional splits in operation.

Some 5G use cases require dynamic alternating demands on the downlink and the uplink.?For example, there may be a predominance of eMBB during working hours and with massive?Machine Type Communications (mMTC) devices synchronizing during the night. The eMBB?usage would tend to have significantly more DL than UL. The mMTC devices would have?the converse requirement. 5G allows for symbols in each subframe to be allocated flexibly?to the DL or the UL depending on the need. This applies to the time division duplex?(TDD) scheme, where the same carrier is shared for DL and UL at different times. Flexible?allocation is performed using the Slot Format Indication (SFI). This can be statically allocated?or even performed dynamically to respond immediately to changing relative needs?of network slices. The flexible slot formats are shown in Table 2 and Table 3.

Table 2: Flexible slot formats showing which slots are for DL, UL, or flexible (part 1)

Table 3: Flexible slot formats showing which slots are for DL, UL, or flexible (part 2)

2.3.6 Bandwidth Parts

Carrier bandwidth parts (BWP) mentioned above allow for different parts of the?bandwidth of a carrier to have different numerologies. This allows for different use cases?to be supported simultaneously. Network slicing is also facilitated by this arrangement. It?also permits high-complexity devices to enjoy the elevated throughputs of low subcarrier?spacing while low-complexity devices can achieve a service with high subcarrier spacing. Single user equipment (UE) can have up to four bandwidth parts allocated to it on DL and?UL with one active at any given moment. This allows for multiple use cases to be achieved?by a single device, or for the device to be on multiple network slices.

2.3.7 Blank Slots

Not every slot needs to be schedulable. With a view to forwarding compatibility, blank slots?can be reserved. This can allow a mix of 3GPP releases to operate in the same carrier.

The new architectural options and network slicing means that this is entirely possible as?different virtualized network functions could be creating the transmissions for different?parts of the carrier for different network slices. The reservation of blank slots removes?the requirement for all to be on the same release, allowing the operators to invest early?with the knowledge that early investment will not become quickly obsolete by lack of?forwarding compatibility.

2.4 Channel Coding

Channel coding must be able to deal with the massive data rates of 5G in such a way that?it is efficient. In other words, it must be able to achieve good throughput for a given?coding rate (level of protection). It must also be able to do this with a sufficiently low?complexity encoder/decoder and do this with low latency. The Multi-Edge Low-Density?Parity-Check Code (ME-LDPC) coding scheme for eMBB data can achieve these goals and is?capable of throughput rates that are significantly better than the turbo codes used in LTE, ?especially at higher coding rates. ME-LDPC also benefits from parallel decoding in hardware,?which results in higher throughput.

Many control channels use the CRC-Aided Polar (CA-Polar) channel coding. These benefit?from having low algorithmic complexity. They also have no error floor, which is a?phenomenon in other channel coders where there is a plateau in the improvement in error?rate as the signal-to-noise ratio improves.

?

2.5 NR Medium Access Control (MAC) layer and HARQ for flexibly?balancing latency, reliability, and energy efficiency requirements

The MAC layer in a wireless communication system provides data transfer and radio?resource allocation services to the upper layers and provides data transfer, signaling of?Hybrid Automatic Repeat Request (HARQ) feedback, signaling of scheduling request and?measurements services, for example, channel quality indication (CQI) to the physical layer.

In 5G NR the different requirements of the use case scenarios eMMB, URLLC, and mMTC?call for a flexible and configurable approach to the MAC. In the sections below we?introduce the HARQ function which may be considered the heart of the MAC. We then?analyze the latency contributions of the different signaling and processing functions within?the MAC layer and describe how these may be configured flexibly to meet?diverse use case scenarios.

It should be noted that, as MAC is a software and signaling control layer, these enhancements may be ported back to LTE, albeit with gains limited by the constraints of the air?interface. For example, short Transmission Time Interval (TTI) has been introduced to LTE.

5G seeks to make operation with an FDD and a TDD type frame as similar as possible,?consequently, both air interface frame structures can be configured to support the range?of use cases. However, FDD is more able to support a mix of different configurations?simultaneously as low latency operation in a TDD frame requires frequent UL/DL switch?points which is not compatible with the use of different switch point configurations in the?same cell or different cells.

2.5.1 An Introduction to HARQ: Reliably sending a signal through a noisy channel

Sending signals over the mobile radio propagation channel is messy and subject to failure.?The transmitted signal is scattered and obstructed or shadowed by all manner of objects?including buildings, street furniture, and people, which results in a received signal that?consists of multiple attenuated and reflected copies of that transmitted. Other mobiles in?the same cell or different cells, or even attached to different systems, transmit at the same?time causing interference. Not to forget interference and intermodulation from other base?stations in the same channel and base stations in adjacent channels. Moreover, background?noise arises from thermal fluctuations in the circuitry of the receiver, impulse noise from?car ignitions, cosmic background radiation, and many other sources including, in at least one?case, electrical noise from a malfunctioning beer fridge. Consequently, the received signal is?composed of multiple attenuated and phase-shifted copies of the transmitted signal which?add-in and out of phase as the mobile moves, or as scatterers in the radio environment?move. This results in fast or Rayleigh fading, with the imposition of slower shadow fading?caused by obstructions and all underlaid with background interference and noise.

An analogy is trying to communicate over the hubbub of a school dining hall; conversing?over any distance in such a place relies on repetition. That is, one party indicates to the?other a negative acknowledgment (NACK) when they don’t get the message and the other party tries again until they receive a positive acknowledgment (ACK). This strategy,?Automatic Repeat Request (ARQ) is employed in radio systems.

However, it is possible to do better than simple repetition using a strategy called Hybrid?ARQ (HARQ). In this case, the listening party stores a representation of the received signal?they failed to decode when indicating a NACK. The counterparty sends a new version of?the message and the recipient combines this with one or more stored versions creating a?composite version more resilient to channel impairments and the process is repeated until?successful ACK, or the maximum number of attempts is reached and a higher layer ARQ?starts the process again or declares failure.

Different versions of HARQ are defined in the literature. Type I HARQ always sends the?same sequence of data on each re-transmission; erroneous data may be discarded, or the?soft decision data may be stored and combined with the subsequent transmission. This?exploits time diversity between the subsequent transmissions. Type II HARQ sends a different?subset of the same coded bits on each re-transmission with the effect, after combining, of?increasing the effective coding rate (or redundancy) with each subsequent transmission.

For example, if the 1st transmission has a coding rate of 1, the 1st and 2nd transmission together?will have a coding rate of 1/2, and combining all three will give coding rate 1/3. Type III HARQ is?like type II, but each transmission is individually decodable, so decoding is possible even if?one of the transmissions is lost. LTE and 5G NR use type II HARQ.

HARQ is very beneficial for the efficient sending of digital data over the mobile radio channel?as rather than having to ensure an acceptable frame error rate, say 1 in 1000 (1x103), with a single transmission the system can tolerate an error rate of 10% or more knowing that?subsequent re-transmissions will ensure a low frame error rate for those that didn’t make it?the first time. For example, 3 repetitions, if required, will take the error rate down to 1x10-4.?In this way, the system can strike a balance between the coding rate and frame error rate to?optimally use the available downlink power (albeit with limited granularity).

There is a trade-off between power efficiency and latency. Latency is lowest if a message?is delivered in a single attempt, but this necessitates the use of very robust modulation?and/or higher power which compromises power efficiency and the overall capacity of?the system. Additionally, when success is not assured with a single transmission, one or?more retransmissions will be required. The speed at which transmission, feedback, and subsequent retransmission can be accomplished is determined by the round-trip time (RTT)?between the UE and the gNB. RTT is affected by processing time in the UE and the eNB?and by the configuration and time duration of the messages used to send the data and?exchange feedback.

At the end of this section, we illustrate how HARQ has evolved from LTE to NR. However,?we first examine the other aspects of UE and gNB processing, MAC signaling, and frame?structure that have been made more configurable to facilitate this trade-off between?latency, reliability, and power efficiency.

2.5.2 Creating a configurable 5G NR MAC layer to balance reliability, latency, and?power efficiency

The cornerstone of 5G NR, as set generally throughout this tutorial, is flexibility and?configurability. Not surprisingly, 5G NR has taken many steps to make the air interface?suitably configurable to allow a dynamic trade-off between reliability, latency, and?power efficiency.

Figure 8. Components of DL latency

Figure 8, inspired by “New services & applications with 5G Ultra-Reliable Low Latency?Communications, 5G Americas, November 2018”, illustrates the key factors that determine?air interface latency. The Figure assumes semi-persistent scheduling (SPS) is used so a?single grant may be used for both data transmissions in the Figure; it is up to the MAC?as to whether new data or retransmitted data is sent. Many of the issues addressed are?common for the UL and DL, but UL-specific issues are dealt with below.

?

Some of the configuration options that 5G NR introduces to minimize latency are?deleterious to other aspects of performance such as reliability and power efficiency.?Therefore, the approach taken in the following sections is to present, concerning the?highlighted callouts in the Figure, how choices may be made to enhance or optimally?configure the air interface to trade-off latency, reliability, and energy efficiency to best suit?the use-case family at hand.

?

2.5.3 Exploiting gNB and UE processing time capability

LTE HARQ process imposed a hard-wired delay of 3ms to allow for processing and signaling?opportunities in the eNB and UE which created a minimum RTT of 8ms. Albeit, backporting of the short TTI feature to LTE allows RTT to be reduced to 2ms. 5G NR removes this hard-wired restriction allowing device capability and channel configuration to be exploited to reduce latency.

The processing delay in the gNB is not defined in the Standard and may be as short as?implementation allows. The value is accounted for in the HARQ scheduling processes in?both UL and DL.

As is clear from Figure 8, UE and gNB processing times are elements in MAC latency.?Thus, the gNB must know what the UE processing capability is to make?appropriate scheduling decisions. Consequently, the UE capability is standardized. Two?levels of UE capability are specified in Rel-15. The minimum required processing delay is?dependent on the configuration of the channels that the UE receives and transmits on as?well as the reported capability; it is calculated using look-up tables referenced from the?parameters described in Table 4.

Table 4. parameters reflecting UE capability to respond to a grant or produce a HARQ-ACK

2.5.4 Configuring PDSCH to trade-off latency, reliability, and power efficiency

5G NR provides the capability to configure the data channel to minimize processing delay,?affecting the latencies in Figure 8, where the gNB takes account of the UE minimum?processing delay through parameters K0 and K1. Alternative configurations are also?supported where performance in reliability or power efficiency is of greater importance?than latency.

A demodulation reference signal (DMRS): LTE had both cell reference signals and UE-specific?DMRS. However, to support power efficiency cell reference signals are removed in 5G NR?which makes the air interface less chatty and more “lean.”

5G NR supports front-loading of the DMRS so they arrive at the start of the transmission?which allows the receiver to begin channel estimation immediately on reception which?supports low latency.

However, this configuration compromises performance when the channel is changing?rapidly, for example with high-velocity UE, and reduces time diversity in the channel?estimate which may compromise reliability. Therefore, additional DMRS may be configured?across the channel.

Removal of time-domain interleaving: In LTE data bits are interleaved in both time and?frequency domains. Time-domain interleaving provides time diversity which improves?resilience against time-varying fading and interference. In 5G NR interleaving is only?done across the frequency domain which allows the receiver to perform de-interleaving?on a symbol-by-symbol basis without having to wait for all the interleaved symbols to?be received.

Frequency-first mapping: Additionally, 5G NR data bits are mapped to resource elements?following a frequency-first mapping across the active Bandwidth Part (BWP) that further?supports the ability to perform symbol-by-symbol processing at the receiver.

Polar channel coding: As discussed elsewhere in this part, 5G NR replaces the Turbo?coding employed by LTE with LDPC coding which is more amenable to parallel processing?implementations that facilitate lower latency processing in both gNB and UE.

?

2.5.5 Configuring PDCCH monitoring to trade-off latency, reliability, power efficiency

The downlink control channel PDCCH sends the downlink control information (DCI) to?schedule UL and DL transmission. As illustrated in Figure 8, an element of latency is the?time that the eNB must wait for an opportunity to send the DCI to the UE. 5G NR uses?several strategies to optimize this wait-time.

Defining frequent PDCCH opportunities that the mobile must monitor minimizes wait?time. However, mobile battery life is one of the most precious resources in a mobile?communication system as perhaps the most compelling selling proposition for a mobile?device is being able to be in contact and contactable at any time. However, for the longest?time, mobiles are in stand-by mode monitoring to see if there is any data for it to receive?or to send. Thus, power consumption in standby mode is critical. Consequently, there is a?direct trade-off between the frequency of the opportunities that the mobile is required to?monitor and battery consumption.

5G NR can be configured to create PDCCH monitoring opportunities every few symbols, for?example with mini-slot structures. Such a configuration clearly supports the low latency

needed by some kinds of URLLC services. On the other hand, PDCCH opportunities may be?configured less frequently and also monitored sporadically using a discontinuous receive?(DRX) process that minimizes UE battery consumption at the expense of latency.

For the lowest latency use cases, the duration of the PDCCH (As illustrated by C in Figure?8) is important and it may be sent in a control-resource set (CORESET) spanning just

a single symbol period. However, other use cases require a highly reliable PDCCH, and?care should always be taken that the control channels are at least as reliable as the data?channels that they seek to control. Correspondingly, in 5G NR, as with LTE, the resources?used to send PDCCH may be aggregated. For example, rather than using a single control

channel element (CCE) at aggregation level 1 to send the PDCCH, the message may be sent?at aggregation level 2 using 2 CCEs and so on. In 5G NR aggregation levels up to 16 CCEs?are supported.

2.5.6 Configuring PDSCH duration to trade-off latency, reliability, power efficiency

Figure 8 F, PDSCH may be flexibly configured with mini-slot configurations allowing?durations as low as two symbols in the DL to meet tight latency constraints where?transmission duration is a significant issue. Relatively high bandwidth may still be achieved?by allocating resources across the available bandwidth. However, as discussed above, such?an approach achieves good latency performance at the expense of robustness.

For less delay-sensitive use cases, PDSCH may be allocated across the remainder of the?14-symbol slots that NR supports. In addition, the PDSCH may be repeated, adding time diversity to the process. The number of repetitions is set by the PUSCH-Aggregation Factor?and is monitored by a single HARQ process. This is touched on again at the end of this?Section where several examples of NR HARQ are presented.

Although the PUSCH is not illustrated in Figure 8, it is sufficient here to note that the?flexibility of the mapping and the use of repetition is defined by a PUSCH-Aggregation?Factor concerning latency/reliability/power efficiency apply similarly for PUSCH.

?

2.5.7 Configuring PUCCH duration to trade-off latency, reliability, power efficiency

The Physical Uplink Control Channel (PUCCH) (I in Figure 8) is used to carry Uplink Control Information (UCI) including HARQ feedback and Channel State Information (CSI) as well Scheduling Request (SR).

In LTE the location, duration, and timing are fixed. However, in 5G NR PUCCH may be flexibly configured in time, frequency and duration. PUCCH may be mapped to between 1 and 14 symbols and may use time or frequency multiplexing. The configuration options allow a trade-off between latency and robustness to enable the supported use case.

?

2.5.8 Configuring PUSCH duration to trade-off latency, reliability, power efficiency

Figure 9 shows the components of latency for example UL data flow to illustrate how the elements of the NR MAC layer can be configured to manage latency concerning reliability and power efficiency for the UL.

Figure 9. Components of UL latency

There is a degree of similarity between the components of latency in DL and UL with similar configuration optimizations that may be made to trade-off latency, reliability, and power efficiency. Consequently, similar types of optimization may be made for PUCCH and PDCCH frequency of opportunities/monitoring occasions and duration and PUSCH duration. For example, sending messages in a single symbol or spanning multiple symbol periods. And, as noted above, PUSCH also supports aggregation/repetition to further increase resilience.

?

2.5.8.1 Grant-free operation for UL transmission

A significant difference between DL and UL operation, which can be seen comparing Figure 8 and Figure 9, is the necessity to send a scheduling request (SR) to the gNB to request UL resources. And, following the SR, there is a wait to receive a PDCCH with a DCI indicating a grant of UL resources. The most extreme strategy for reduction of latency associated with uplink data transmission is to use grant-free operation which eliminates the need for SR and waiting to receive a grant, which eliminates steps A through F in Figure 9. The UE will still have to process the data on arrival and wait for the pre-configured grant opportunity. A pre-configured grant is defined, for example, by using broadcast system information.

?2.5.8.2 Pre-emption of UL and DL resources

The resource reservation associated with grant-free operation reduces efficiency as resources are earmarked for use even if they are not used. 5G NR ameliorates this effect by allowing the use of the ear-marked resources while reserving the right to pre-empt them. The gNB may indicate that ongoing UL resources will be pre-empted allowing UEs to suspend transmission, reducing power consumption and interference.

Pre-emption is also possible in the downlink. UEs whose resources were pre-empted are informed in the following TTI allowing them to discard, rather than decode and store for use in incremental redundancy, any data apparently received during the pre-emption period. The gNB may then send the portion of the data that was not sent due to pre-emption. This strategy is suitable for the most latency-sensitive use cases. Additionally, it is beneficial for UE power consumption as it eliminates at least one set of messages that the mobile would otherwise need to monitor frequently.

?2.5.8.3 Code Block Group-based re-transmission

In LTE transmission and re-transmission is on a Transport Block (TB) basis. However, in cases where a TB is quite long, and errors occur only in a small portion, this leads to inefficient operation. 5G NR introduces the idea of Code Block Group (CBG) based re-transmission which, at the expense of a little extra signaling, allows large TB to be sent, but only the corrupted CBG re-transmitted.

CBG based re-transmission is particularly suited to the operation of eMBB, which demands large TBs, when combined with URLLC service employing pre-emption as only the CBG affected by the pre-emption need be sent.

Figure 10. LTE FDD HARQ operation DL example

Figure 10 illustrates the operation of HARQ using the example of LTE FDD downlink data?transmission. The eNB sends a data block B1 to the UE which arrives propagation delay?(PD) later at the UE. The UE has a fixed timing budget of 3ms (3 timeslots) less timing?advance (TA) to process the block and determine if it was correctly received. Figure?B1 is not correctly decoded on the first transmission. The result of the decoding event is?stored and a HARQ NACK is sent to the eNB at a time 3ms – TA later. The eNB has a time?budget of 3ms (3-time slots) to process the UE response and, in the case of HARQ NACK, to?generate a second version of the data block B1’ with a different puncturing to support?incremental redundancy. The second transmission is possible in the 8th timeslot after the?initial transmission. Consequently, the HARQ round trip time (RTT) is 8 timeslots which are?equal to 8ms for the LTE frame structure (without short TTI). In the Figure, the UE can combine the initial decoding attempt of B1 with that of B1’ and correctly decode the?data. Consequently, a HARQ ACK is sent to the eNB and in response, the eNB sends the next?data B2 block after the appropriate processing time. In LTE, there is a fixed processing/?waiting time built into the HARQ process for both UL and DL. In addition, the UL uses an asynchronous HARQ process that enforces a fixed timing arrangement between a downlink?transmission and its corresponding HARQ-ACK. This reduces signaling at the expense of losing some scheduling flexibility. As described above, LTE HARQ has an RTT of 8-time slots.

Consequently, to avoid gaps in transmission while waiting for a data block to be ACKed LTE?supports 8-interlaced HARQ processes.

5G NR allows greater flexibility and may be configured where transmission, acknowledgment, and retransmission can be accomplished within a TTI. However, this is not always the preferred action, particularly if there is an additional latency that the air interface must?deal with, such as in the case of non-terrestrial access that is being standardized in Rel-16.?Consequently, LTE Rel-15 allows up to 16 simultaneous HARQ processes.

Figure 11 provides 3 exemplary configurations. A is a configuration suitable for low latency use cases such as URLLC with transmissions confined to durations of a few symbols.

Reliability in this configuration relies on the diversity provided by a wide bandwidth in the frequency domain. The configuration may use a pre-configured grant to eliminate the SR/grant delay. The resources reserved for the pre-configured grant may be shared with other applications pre-empting when necessary. For UL transmission, the gNB may indicate to mobiles whose resources will be pre-empted allowing them to suspend transmission, and in the DL, the gNB may indicate in the following TTI that part of the earlier transmission has been pre-empted and should be discarded. B is a configuration suited to high reliability or low signal level configurations that use an aggregation level of 4 for the PDSCH at the expense of lower efficiency as HARQ feedback is only provided after all re-transmissions are received, while C is a configuration for similar requirements with aggregation level of 3 for the PUSCH.

Figure 11. Examples of 5G NR FDD HARQ configuration: A self-contained slot for URLLC, B PDSCH with aggregation level 4, C PUSCH with aggregation level 3.

The Figure highlights that a wide range of RTT values may be achieved through flexible configurations and indicates that each configuration represents a trade-off between latency, reliability, and power efficiency. An additional element of this trade-off that is not dealt with here but is dealt with elsewhere in the part is the effect of variable numerology that allows the reduction of TTI.

Figure 12 is a qualitative illustration of the trade-off achievable with the 5G NR configuration parameters. It is provided for the case of the URLLC use case for the dimensions of capacity, latency, and bandwidth. The Figure shows that URLLC capacity?may be increased by reducing the latency target or by increasing the available bandwidth,?whereas capacity is considerably reduced by making the reliability requirement more?stringent based (Qualcomm reference https://www.qualcomm.com/media/documents/files/ ?expanding-the-5g-nr-ecosystem-and-roadmap-in-3gpp-rel-16-beyond.pdf)

Figure 12. The trade-off of latency, capacity, and bandwidth

Dual connectivity will allow NR carriers to be used simultaneously with LTE carriers.?These support the eMBB use case by providing potentially massive bandwidths. Carrier?aggregation is managed in terms of component carriers (CCs). Multiple CCs can be aggregated together where each CC can span up to 400 MHz and comprise a maximum of?3,300 active subcarriers. The actually allowed combinations of CCs are specified by 3GPP. In?some cases, the duplex mode can vary between the LTE and NR carrier components, with?LTE on FDD and NR on TDD, for example.

?

2.6 MIMO, Massive MIMO, and Beamforming

MIMO is a well-known cornerstone of 5G. It has wide implications, particularly for?achieving eMBB. Here we set out to explain MIMO in terms of the underlying principles,?the evolution through different flavors of multiple antenna technology, its similarities and?differences from beamforming, and how it will play its part in delivering the promise of 5G.

?

2.6.1 History of Spectrum Reuse in Wireless Access

In GSM, neighboring cells using the same frequency would have resulted in substantial?interference and dramatic loss of spectral efficiency. Therefore, the available spectrum?had to be divided into chunks, and nearby cells allocated different frequencies. This reuse?was wasteful as the entire available bandwidth could not be used in multi-cell systems.

In WCDMA, neighboring cells could use the same frequency bandwidths, but scrambling?codes were used to separate the spread spectrum signals from different nearby cells. In this?case, the users at the edge of the cell would sometimes suffer poor signal-to-noise ratios?and thus have to use more robust modulation and coding schemes which in turn lowered?the spectral efficiency.

In LTE, the use of closely packed subcarriers allowed those parts of the full-spectrum?bandwidth that were most suited for each user, in terms of the interference experienced,?to be used for that user’s transmissions. However, in cases where there are many users?per cell, the theoretical maximum data rate must be shared among the users, limiting the?application bandwidths that can be achieved.

The holy grail of the communication system is that every user enjoys the full benefit of?the bandwidth. 5G is taking a big step towards this vision, overcoming the limitations of?previous generations. Massive MIMO and beamforming are major cornerstones of this by their ability to deliver multiplexing of multiple communication streams with fine?spatial granularity. By breaking the coverage down into narrower beams, the distances?over which the full spectral resources can be reused are shortened. While it may not be able?to deliver a beam per user, it continues the trend to the ever-smaller granularity of frequency?reuse and is a major leap towards this ideal.

?

2.6.2 Introduction to MIMO

MIMO stands for Multiple Input, Multiple Output. What does this mean? To what is the?input going and from what is the output coming? The answer to both questions is the?same: namely, the radio interface “channel” or medium between the transmitter and?receiver. This additionally includes the RF domain components in the physical equipment at?the transmitter and the receiver (e.g., cabling, RF amplifiers, and antennas).

It is common to think of RF communication taking place between a single transmit antenna and a single receive antenna. Data to be transmitted in the DL or UL is scheduled into?resource elements, encoded and modulated, then undergoes digital-to-analog conversion,?is amplified, carried to the antenna, and finally is transmitted. The receiver antenna?captures the radiated energy, then demodulates and decodes the received signal and looks?for the parts of the spectral resources (in frequency and time) that correspond to that?user. This is Single Input, Single Output (SISO) transmission since there is only one input to the channel and only one output from the channel, as illustrated in Figure 13. Here the?amount of data that can be conveyed to or from a given user is limited by the number of?spectral resources that can be devoted to that user. It is also subject to the harsh RF medium.?Transmission is complex, particularly in lower frequencies. The success of the communication?will depend on the combined effect of phenomena such as non-line-of-sight propagation,?reflection, refraction, scattering, multipath propagation, and constructive interference.

In cases where there is no line of sight between the transmitter and receiver, these?phenomena can be very helpful, depending on the carrier frequency, because reflection,?scattering, and refraction allow the signal to propagate via an indirect path.

Figure 13. Single input single-output (SISO) transmission

Transmitted signals can also be subject to less helpful effects such as shadowing and destructive interference or fading that conspire to defeat the transmission. The combination of these helpful and damaging phenomena together makes for a complex channel between the transmit antenna and the receive antenna as illustrated in Figure 14.

Figure 14. Single input single-output (SISO) transmission over a channel with non-line of sight propagation

2.6.3 SIMO and Receive Diversity

Whether or not the received signal is useful will depend in part on the location of the antenna. It may be receiving a good strong direct signal, a reflected signal, a scattered signal, or in the presence of combinations of these, maybe benefitting from constructive interference. The combined impact of these effects can vary significantly over short distances of even half a wavelength. Thus, if a second receive antenna is employed, and deployed at least half a wavelength from the first, then the chance of receiving a good signal increases. The second antenna may additionally or instead be cross-polarized, where it is rotated so it is offset ninety degrees from the first. With the two antennas, the communication attempt will then only be defeated if the signals received at both antennas are seriously affected by some combination of destructive interference. Thus, this approach increases the likelihood that communication can be maintained.

Figure 15. Single input multiple outputs (SIMO) transmission

This method of transmission is referred to as receive diversity. It is an example of Single Input, Multiple Output (SIMO) in its widest definition as there is one input to the channel at the transmitter and two (or more) outputs from the channel at the receiver. This is illustrated in Figure 15. This will increase the reliability of the communication link. The receiver can make use of the multiple signals in various ways. For example, maximal ratio combining will use the useful signal from all receive antennas and add them together for a stronger received signal. Another is switched diversity, which chooses the strongest received signal from any of the antennas.

?

2.6.4 MISO

An alternative to SIMO is to transmit the same signal from two transmit antennas to?the same user using the same spectral resources. If done naively, the two signals might?interfere with each other, the receiver would be unable to make sense of either, and nothing can be decoded. But this ignores the complex propagation environment discussed?previously. As the distance between the transmit antennas increases, the channels will start?to decorrelate. In other words, they will be subject to different combinations of non-line-?of sight, multi-path propagation, among others. This provides more of a chance to get the?signal successfully from the transmitter to the receiver.

Figure 16. Multiple input single-output (MISO) transmission

There are more chances for shadowing to be overcome if there are more transmissions. This?method of transmission is called space-time transmit diversity (STTD): sending the same?signal to the same user over the same spectral resources at the same time. It is an example?of Multiple Input Single Output (MISO) since there are multiple inputs to the channel but a?single output as illustrated in Figure 16. This can increase the reliability of the channel by?reducing the time that it is subject to the destructive characteristics of the channel. It can?also affect the rate of data transfer that can be achieved. Sometimes there will be more?destructive interference and the data rate will be reduced; at other times, there will be less?destructive interference and the data rate will be increased.

The STTD flavor of MISO can also utilize more than two transmit antennas, although the?benefit will diminish as the number of antennas is increased. STTD is not always appropriate.

Its success increases as the antenna separation increase up to a point; ideally, they need?to be at least a wavelength apart. In higher FR2 bandwidths, the channels are less complex?and dominated by line-of-sight propagation. While adding a second transmitter might just?mean that the receiver is no longer in the shadow of an obstacle by the modest?separation, this is unlikely and the benefit from using STTD will diminish.

2.6.5 MIMO

As described above, STTD transmits the same signal to the same user over multiple?antennas. Now consider the implications of transmitting a different signal (also known?as layer or stream) from each antenna to the same user over the same spectral resources?at the same time while introducing several receive antennas corresponding to (or?exceeding) the number of unique transmissions. If the nature of the propagation, or?channel state, between each of the transmit antennas and each of the receive antennas,?is sufficiently diverse, then the signals received at each receive antenna will be to some degree decorrelated. Although each of the transmitted signals will tend to interfere with?each other at reception, if the correlation between them is low enough, then the signals?can be separated by the receiver.

It may be that the signals must be transmitted with a more robust modulation and/or?coding scheme, but in many cases, such as when the channel is sufficiently diverse, the?conditions will be right and transmitting two streams of information at a lower rate to a user?will result in a higher overall data rate than transmitting a single stream of information at a higher data rate. This mechanism of transmitting two or more independent signals,?or layers, to the same receiver, is called spatial multiplexing or single-user MIMO (SU-?MIMO) as illustrated in Figure 17. It can also be referred to as multi-layer MIMO as there?are multiple layers of data transmitted on the same physical resources. We now have?MIMO because there are multiple inputs to the channel and multiple outputs. A SU-MIMO?mechanism can increase the data rate that a user can achieve.

SU-MIMO does not always give again. If the correlation between the combination of the?different layers sampled in each receive antenna is high, then the equalization process will?not be able to accurately separate and decode the symbols from each layer. Additionally,?the data throughput that can be achieved overall will either drop or, in serious cases, the?data may not be able to be decoded at all. To control this, the receiver must feedback to the transmitter on the state of decorrelation between each of the channels. This is?typically done by the UE with a metric referred to as the “channel rank.” This indicates the?order of decorrelation between the channels and thus how many unique transmissions can?be sent between the transmitter and the receiver. Further improvement can be achieved if?the receiver provides additional feedback on the overall quality of the channel. Even if they?are decor-related, the signal-to-noise ratio may be poor and the capacity of the channel?limited, despite the use of SU-MIMO.

Figure 17. Single user-multiple input single output (SU-MIMO) transmission

A further piece of feedback can be provided by the receiver. Using especially known signals sent by the transmitter, the receiver can provide an estimate of how the transmitted signals could be modified to maximize the separation between them. These are known as precoder weights. As part of the transmission process, usually in the digital domain, these weights can be applied to each layer in the transmission to optionally change the?phase and gain of each layer. This is known as precoding and can deliver substantial performance gains.

SU-MIMO can be generalized from communication between one transmitter and one receiver to multiple transmitters or multiple receivers. This is multi-user MIMO (MU- MIMO), which is illustrated in Figure 18. This allows the multiplexing of different RF streams onto the same spectral resources using different antennas. Again, multiple antennas at the receiver receive these signals and under the right channel, conditions can separate them.

Channel feedback can deliver better utilization of the channel and better orthogonality of the channels through precoding.

Figure 18. Multiuser-multiple input single output (MU-MIMO) transmission

An increased benefit of MIMO is achieved if the coupling between antenna elements is?reduced. This is the phenomenon where power transmitted by one element in the array?is absorbed by other antennas in the same array, impairing the transmitted signal. This requires the antenna elements to be spaced sufficiently far apart. A minimum spacing of?the antenna elements of half a wavelength is generally accepted although larger spacings?can increase the benefit up to a point. Increasing the spacing between elements in the?array also lowers the correlation between the signals received at each, as the increased?spatial separation allows different sampling of the channel between antennas.

These factors and constraints lead to the engineering contradiction that large arrays are?desired to maximize the benefit of MIMO, but small form factors are also valued for the?arrays to minimize cost, weight, wind shear, and other undesirable characteristics. The good?news is that the ideal size of even complex antenna arrays decreases as the frequency?increases, making massive MIMO viable and acceptable for FR2 bands in particular.

Although the underlying physics is the same, there are differences between the reality for MU-MIMO at sub-6 GHz and in the mmWave bands. At mmWave, the distances over which signals decorrelate are very short. The success of MIMO depends on the ability to provide feedback rapidly enough to update the precoding matrix to maintain good decorrelation between the channels most effectively. At mmWave frequencies, the precoding required at one location can vary dramatically from the precoding required at another, even millimeters away. Even for deployments serving static UEs, such as fixed wireless access with static antennas attached to the outside of the house, the dynamics in the environment such as people, vehicles, and other scattering objects have a similar effect as motion. This makes the job of maintaining the decorrelation between the channels hard. High dimension MU- MIMO is, therefore, less effective in the mmWave spectrum.

Although the terminology is somewhat subjective, and the distinction arbitrary, low order MIMO involves up to around eight antennas, or channels, where the phase and amplitude can be controlled independently as illustrated in Figure 19. These can be dedicated to one user (SU-MIMO) or a lower number of devices simultaneously. In contrast, massive MIMO supports more antenna elements, generally accepted to be significantly more than eight, and thus can transmit to more devices simultaneously.

Figure 19. Massive MIMO

The LTE standards have supported high order MIMO since Release 13 which saw 16TX MIMO and Release 14 with 32TX MIMO. Support for at least 4x4 MIMO is essential for 5G NSA NR but Release 15 can support up to 32TX MIMO. As the number of antenna elements increases, the effective isotropic radiated power (EIRP) that can be achieved will increase.

Thus, M-MIMO can increase the gain by two orders of magnitude for hundreds of antenna?elements. The 5G revolution that Release 15 brings is the flexible and dynamic combination?of MIMO together with beamforming.

?

2.6.6 Introduction to Beamforming

Up to now, we have discussed the use of multiple antennas to increase the capacity and/?or reliability of the channel to transfer more data between the transmitter and the receiver.?Multiple antennas can also be utilized to shape how the energy is radiated from the antenna. A signal can be transmitted from each element in an antenna array. Each of these?signals will tend to constructively or destructively interfere with the others with the exact?result of this interference depends on the location of the receiver.

Given a fixed physical arrangement of antenna elements, by changing the phase and?amplitude of the transmission at each element, energy can be directed in one or more?specific directions. This is known as beamforming and although it depends on multiple?antenna elements, it is a concept distinct from MU-MIMO described above.

A two-dimensional panel antenna array presents the ability to shape and direct the beam?in the vertical axis as well as the horizontal axis, facilitating fine control of beams so that coverage can be surgically directed to where it is required. This could be at different?horizontal and vertical angles from the antenna boresight and at different distances,?assuming the antenna is at an elevated location.

To control the beam direction effectively, a spacing of antenna elements of around?half a wavelength is ideal. However, larger spacings between antenna elements above half?a wavelength increase the occurrence of grating lobes where instances of higher power?occur away from the intended beam. This will not only reduce the power in the direction?of the intended beam; it will also increase interference and reduce the spectral efficiency?and capacity. Having larger numbers of antenna elements results in the ability to focus the?energy in narrower beams and reduce instances and effects of sidelobes directing energy in?unwanted directions.

Beamforming can be operated in a passive mode, which is also known as switched?beamforming. This means that the beams are static and a user in motion will move?between beams. Which beam or beams provide service changes over time. Another mode?of operation is active beamforming where the direction in which the energy is focused?changes over time to track the user.

An important characteristic of beamforming is how tightly the energy can be focused.?A system with a few wide beams will generally have less capacity than one with?many narrower beams as the spectral resources can be fully reused in each beam. In a beamforming antenna system, one factor affecting the width of a beam is the number of?antenna elements. As this number increases, the beam can be focused more tightly. The?individual elements should ideally be around half a wavelength apart to allow the energy?to be focused effectively.

At 1 GHz, this spacing is around 15 cm while at 30 GHz it is around 5 mm. This means that?beamforming antennas at mmWave can be smaller, cheaper, lighter, and less subject to?factors such as wind shear, which can disrupt communications. However, if smaller antenna?sizes are used, the aperture is also smaller, and less energy can be collected.

?2.6.7 Pure Digital Beamforming and Hybrid Analog/Digital Beamforming

Modification of the phase and amplitude of the transmitted signal is required for both?MIMO and beamforming. In the case of MIMO, the amplitude and phase of the baseband?streams are modified by a precoding matrix chosen for that receiver in its current?conditions. In the case of beamforming, the phase of each signal must be aligned to?achieve the desired direction for that beam.

In principle, both types of manipulation could take place in the digital domain where the?digital signal processing could apply both the MIMO precoding and the phase adjustments?to achieve the desired beam direction. In general, there are advantages to signal processing?as much as possible in the digital domain with signal processing microprocessors, rather?than converting to the RF domain and performing direct manipulation of the phase and?amplitudes of the signals. As soon as the signal is converted to the RF domain, conveying?the signal and processing it will incur a loss of power. The approach of performing digital-to-analog conversion close to the antenna unit with no further manipulation, therefore,?holds a lot of appeals. However, there are issues with this pure digital beamforming. The?complexity of the signal processing for large antenna systems with hundreds of antenna?elements places high demands on the signal processing stage. In some deployments, there?may be constraints on the availability of real estate for signal processing and the associated?HVAC along with the bandwidth requirements between the baseband and RF components?of the system.

Also, digital artifacts and intermodulation products can arise from purely digital signal?processing. An alternative is to perform some aspects of the signal processing in the RF?domain. For example, beam steering can be performed in the RF domain by applying?analog phase offsets in the antenna array itself. This reduces the digital signal processing?requirement but involves more manipulation in the RF domain. In this case, controlling the?phasing and directivity of beams will be slower and more complex.

?However, the ease with which these phase shifts can be performed in the RF domain?means that this is an ideal solution, especially when dynamic beams are not required, and?static or semi-static beams are sufficient. The system design must consider the cost of?manufacture, HVAC demands, suitability for the use case, and availability of space for the?various disaggregated components (CU, DU, and RU) and the distances between them, as?well as the overall achieved performance of the resulting communication.

Beamforming is good for coverage-limited systems. This is important for mmWave which?suffers from higher propagation losses. It is also important for allowing the reuse of existing?base stations designed for 2 GHz bands. Focusing the energy of beamforming means that?these same sites can be used for 5G NR even with higher frequency carriers in FR1.

2.6.8 Support for MIMO and Beamforming in 5G NR

The 5G NR standard introduces various features to support multi-layer MIMO and?beamforming. Fundamental to this is the concept of the beam, which is defined centrally?within the standards. As in earlier technologies, the cell still exists as a logical entity. The?cell has a primary synchronization signal (PSS), which indicates the start of the frame, along?with the secondary synchronization signal (SSS), which indicates the locations of the start?of the subframe.

The PSS takes one of three values, while the SSS takes one of 336 values. Together these?determine the physical cell identity (PCI) for that cell for which there are 1,008 unique?values. In addition to the cell logical entity, the beam is also a first-class logical entity in 5G NR, which eclipses the cell in terms of importance. Each beam has its own unique?Synchronization Signal Block (SSB) which comprises the PSS and SSS, and thus the PCI from?the parent cell, along with the Physical Broadcast Channel (PBCH) and its Demodulation?Reference Signal (DM-RS). Although the beam exists as a logical entity and is named in a way that could be suggestive of some type of shaping, there is no requirement for the?beam to be formed or shaped in any way.

The 5G NR standard defines the concept of antenna ports (See TS 38.211). These correspond?to layers of transmissions of different data spatially multiplexed on the same physical?resources. The standards define the use of 1, 2, 4, 8, 12, 16, 24, or 32 ports for flexible?multiplexing dependent on the channel conditions (38.211). This is highly flexible and means?that even where there is some combination of highly correlated channels, high interference,?low signal strength, or other harsh conditions, the communication can be successful. This?happens by using fewer ports, or low order MIMO, yet when the conditions are good the capacity can be maximized by increasing the number of ports. The CSI-RS symbols?(defined below) are mapped onto physical resource elements in such a way that they can?be decoded independently by the UE for each channel.

The standards define mechanisms for channel estimation so that the optimal use of?multi-layer MIMO and beamforming can be achieved. The mechanism is flexible; how it?is configured in practice will depend on factors such as the frequency band and whether the channel can be assumed to be sufficiently similar in the UL and the DL, a characteristic?known as channel reciprocity.

The Channel-State Information Reference Signal (CSI-RS) is transmitted on each beam by?the gNB. This known signal allows the UE to estimate the degree to which the channels?are decorrelated and thus the degree of spatial multiplexing, or the rank, that can be?supported. Once the rank is known, the UE may then have to calculate the precoding?required for its transmissions from the gNB, although this is not always the case.

The precoding must be applied by the transmitter in the digital domain by applying a?precoding matrix. The matrix will be larger for higher-order MIMO with many ports. The?matrix itself does not need to be transmitted. That would require a lot of data transmission?and would therefore impact capacity and add latency to the ability to respond to channel?conditions. Instead of transmitting the raw precoding matrix, an index into a codebook of pre-defined precoding matrices is signaled. The number of codebooks available places?the upper limit of ports that can be multiplexed onto the same physical resources. More?codebooks would allow for higher and higher multiplexing, and it is the number of defined?codebooks that places the limit of thirty-two ports.

The overall quality of the channel is important, as this will determine how robust a?modulation and coding scheme should be used. Thus, the following pieces of information?are transmitted by the UE to the gNB. The rank indicator (RI) is the number of independent?layers that the UE can use. The precoding matrix indicator (PMI) is the index into the?codebook of precoding matrices corresponding to the rank and what is most appropriate?for the channel conditions. The channel quality indicator (CQI) indicates which combination?of modulation scheme and channel coding is appropriate.

Another mechanism for channel estimation is possible. This can be used in conditions where?there is channel reciprocity, e.g. when the TDD mode is used, such that UL transmission is?on the same carrier as the DL. In this case, the sounding reference signal (SRS) can be used.?This is transmitted by the UE to the gNB and allows the precoding matrix for the UE to be?calculated directly by the gNB. In this case, it becomes unnecessary for the UE to transmit?the PMI to the gNB.

This has the advantage of the precoding weights being calculated where they are needed,?and so are available with much lower latency and with a saving of UL bandwidth,?compared to cases where the PMI must be calculated and then reported by the UE.

This has the further advantage that the choice of precoding weights is not limited to a codebook of choices, but rather the weights can be fine-tuned more precisely to the?channel conditions. Thus, this method of channel estimation is suitable for cases where?the required precoding can change more rapidly, such as in FR2 carriers. The use of SRS is insufficient on its own in this case. The UE must still transmit the RI and the CQI to indicate?the degree of MIMO that it can support and to allow the modulation scheme and channel?coding appropriate for the channels to be calculated.

There is a further feedback mechanism beyond that of the channel. The same logical cell?can have more than one beam. A cell with a single PCI can have multiple beams, each with?a unique SSB. Each beam will also have its own CSI-RS, allowing the UE to establish the?quality of the channels between each beam and itself. This feedback is transmitted from?the UE to the gNB using the CSI-RS resource indicator (CRI), allowing the gNB to select the best beam for communication.

Now that we have established the way that channel conditions can be quantified,?communicated, and transmission layers appropriately weighted, how does this relate to?MIMO and beamforming? This is where some flexibility in MIMO and beamforming is?established. Let’s consider an antenna panel that has four rows of four cross-polarized?pairs, making thirty-two discrete antennas in total. Initially, we consider the case where?these are not beamformed. A UE close enough to the transmitter will be able to measure?the CSI-RS on at least some of these independent transmissions and measure the channel?state. If the conditions conspire in favor of the UE, it may be able to separate all thirty-two?transmissions. So, the UE can report the channel state to the gNB using the CQI, the RI, and?the PMI. This situation is illustrated on the left-hand side of Figure 20.

Now consider the situation where some beamforming is performed so that the thirty-two?ports are separated into four beams of eight ports each. Now the CSI-RS is used by each?mobile to calculate the relative strength of each beam and uses the CRI to report the?information back, along with the CQI and RI as before. This situation is illustrated on the?right-hand side of Figure 20. In contrast, consider these scenarios when there is channel?reciprocity, with and without beamforming. This situation is illustrated in Figure 21. Note?the PMI has been replaced by the SRS.

The 5G NR standards are thus very flexible in how beams can be formed and how spatial?channel multiplexing can be configured.

Figure 20. Channel and beam estimation and feedback for cases without channel reciprocity

Figure 21. Channel and beam estimation and feedback for cases with channel reciprocity (e.g., TDD)

2.6.9 Comparison of MIMO and Beamforming

A comparison of MIMO and beamforming is provided in Table 5.

Table 5: Comparison of MIMO and beamforming

2.7 Energy Saving Support in 5G NR

The energy required for running a 5G gNB is reduced. In LTE, the eNB is required to transmit?a reference signal every 0.25 ms and a synchronization signal every 5 ms. Even in cases of low?utilization were few or no UEs are actively using the cell, these must still be broadcast.

In contrast, a 5G gNB is not required to transmit with the same frequency. 5G NR has no?cell-specific reference signal and the synchronization signal needs only be broadcast every?20 ms by default with the broadcast data. This periodicity is configurable and can be?extended up to 160 ms.

Somewhat linked to energy saving is power saving in the UE. It can be costly for the UE to?have to continuously monitor the physical downlink control channel (PDCCH) for scheduled?DL data. This is inconsistent with the need for battery-powered devices that must last for many years without replacement. This is addressed in 5G NR with the extended?discontinuous reception cycle (eDRX) feature. When operating under eDRX the UE only?needs to monitor the PDCCH during specific periods and can be dormant at other times as?long as it has no data it needs to send in the meantime.

Bandwidth adaptation (BA) allows the UE to be configured to receive the PDCCH on only?the active BWP. This also contributes to saving UE energy use.

?

2.8 Standards Evolution in 5G NR

The 3GPP Release 15 version of the standards was the first release referred to as 5G. However,?the work will not be complete until the standards are sufficient to satisfy the requirements?laid out by the ITU in IMT-2020. The focus of Release 15 was to deliver the requirements of?the eMBB class use cases. There has been less focus on the URLLC, and very little attention?has been given to the mMTC classes of use case. These other classes are receiving more?attention in Release 16 and will continue to receive further attention beyond that release.

To understand the direction that the 3GPP standards will take towards delivering IMT-2020?capability, we can look at the features and studies that are taking place in 3GPP as part of?Release 16. Of course, if the 5G network is transcending a pure telecommunications industry?focus, 3GPP must expand to include new industry verticals. As 3GPP developed?Release 15, the ranks of 3GPP membership have swelled to include companies from industry?verticals that previously had not taken part in 3GPP. These include the agriculture, satellite,?automotive, aeronautical, and rail sectors. This is encouraging and will ensure that the evolution of the standard will be tailored to the new verticals for which 5G is developed. 3GPP?plans to complete Release 16 in June 2020 and at that point is expected to meet the ITU?requirements for IMT-2020 across all the service enablers; eMBB, mMTC, and URLLC.

2.8.1 Evolution of URLLC

With the introduction of 5G into 3GPP with Release 15 came a drive to deliver the eMBB?service enabler. The URLLC received less attention and was predominantly focused on?TTI for low latency scheduling. To achieve the full IMT-2020 requirements for 5G,?much attention has been focused on the URLLC service enabler in Release 16. Studies are being carried out in 3GPP to build on the ability of NR to deliver a wider range of URLLC use cases. For example, the study on physical layer enhancements for NR URLLC (TR?38.824) sets out to expand the modest aspirations of Release 15 for URLLC, whose scope?was limited to entertainment-type applications such as virtual reality. The goal for Release?16 is to address use cases with stricter requirements, such as factory automation, remote driving along with other vehicular applications, and power distribution. Within scope are?enhancements to the physical layer including for example; PDCCH, PUSCH, UCI, MIMO?feedback, and scheduling. It also includes UL prioritization and multiplexing of inter-UE?transmissions and enhanced grant-free UL transmissions.

The PDCCH contains the Downlink Control Information (DCI) which is information that?describes where in the OFDMA frame the downlink and uplink data channels are allocated.?Until this is received a UE does not know where to look for the data it is to receive or when?to transmit its data to send. Awaiting reception of this will naturally delay transmissions?and add to latency. An enhanced monitoring capability studied in Release 16 means that?this latency can be reduced in some cases.

The study on non-orthogonal multiple access (NOMA) for NR (TR 38.812) aims to reduce?latency by removing the need for the uplink transmissions to be granted. There can be?a significant delay if a device needs to send some data. In some circumstances, if it does?not have a sufficiently imminent opportunity for UL transmission, it will have to use the contention-based random-access procedure to establish a connection with the network, before requesting resources for its transmission. NOMA aims to strip away this delay and cut?the latency to an absolute minimum.

Other enhancements to support ultra-reliability are being considered. The Release 15 study?on NR to support non-terrestrial networks (TR 38.811) envisages connectivity delivered by?airborne or satellite access.

?

2.8.2 Evolution of mMTC

Delivering massive Machine Type Communications will place further demands on the?standards. Use cases in this class will be expected to deliver a service in extraordinarily?diverse subscriber densities. Densities of millions of devices per square mile are expected?in some scenarios such as smart cities, campuses, factories, and seaport environments.

This will require enhancements to the NR to enable, for example, the ability to access the?network without congesting the random-access channels.

Conversely, some industry verticals will require very low densities of devices with very?long-range communication. This will place a different set of challenges on the networks,?such as the need for link budgets exceeding 160 dB. This might be achieved through some?combination of robust channel coding and repetition, perhaps supported by beamforming.?Many use cases will be viable only if the cost of devices can be driven down. This means?that support for extremely simple devices with a single antenna will have to be delivered.?Some devices will be battery-powered and have decades-long service intervals; this will?require extreme battery lives along with enhancements in the standards to support the?devices’ long sleep times.

Another aspect of mMTC is the connected car, which is the focus of a Release 16 study on?vehicle-to-everything (V2X) (TR 38.885). 3GPP envisages a rich set of functionalities in 5G?to support vehicular communications. Four use case groups are targeted for support in the?standards. The first use case group is platooning. Vehicles in the platoon travel close together?and can be regarded as a single entity while the platoon persists. This offers various?advantages. For example, not all vehicles in a platoon may need an active human driver.

One scenario is that only the lead vehicle needs a human driver. Another advantage of?platooning is that the vehicles can drive closer together. This can increase road capacity and?ease congestion as well as improving driving efficiency by reducing losses to air resistance.

Another use case group is extended sensors. It is common for vehicles to have numerous?sensors for velocity, position, proximity, hazards and nearby objects, temperature, etc.

Traditionally these sensors reside in the vehicle. Additional benefits can potentially come?from sensors that are not attached to the vehicle but rather are in the vicinity of the current?location of the vehicle. These can include video or photographic sensors that are part of the?road infrastructure, sensors carried by pedestrians or nearby vehicles for example. The richer?sensor data can provide a human or autonomous driver with a more complete view of the?surroundings for safer decisions to be made. As some types of sensors are characterized by?high data rates, the various 5GNR service enablers are ideal for satisfying this use case group.

The third use case group is advanced driving. This envisages multiple vehicles cooperating?by sharing information, sensor data, and intent. This goes beyond the extended sensors?use case group in that it facilitates coordinated vehicle maneuvers in addition to sharing of?sensor data.

The final use case group is remote driving. This includes use cases such as relieving a?human driver, for example when they want to perform other activities, or where the?driving conditions are challenging. It also envisages the guidance of vehicles that lack an?onboard driver, either human or computer.

Many use cases within these use case groups have highly ambitious target latencies. Rather?than relying on all communications going via the RAN infrastructure, these use cases are?facilitated by a feature that allows radio connection directly between vehicles. This is?known as the PC5 side link. The physical layer resembles that of regular NR; that is, it is?based on CP-OFDM and supports several subcarrier spacings; 15, 30, and 60 kHz in?FR1. As well as FR1, it is envisaged that FR2 carriers will be used and support 60 and 120?kHz subcarrier spacing. The proximity of the vehicles in a platoon or advanced driving?group for example will be ideally suited to these higher frequency carriers. Bandwidth?parts are compatible with the side link. This will facilitate the allocation of smaller portions of a?larger 5G carrier to V2X activity.

Sideline can operate in unicast, multicast, and broadcast modes where multicast will be useful for supporting various use case groups such as platooning and advanced?driving. In communication between a UE and a gNB, the gNB is in control of who can transmit at what times and in what subcarriers. In contrast, there is no inherent hierarchy?between vehicles communicating via side link and therefore no obvious way of managing?to schedule to achieve efficient use of spectral resources. The standards identify two modes of coordination. In the first, a base station controls the scheduling of side link?resources. This of course requires there to be coverage by the network and so on its own?would rule out autonomous of side link away from network coverage. To manage this?scenario a second mode is identified. This encompasses an autonomous selection of side?link resources by UEs as well as coordinated scheduling. The coordinated schemes include?UE scheduling other UEs and the assistance inside link resource selection of one UE by another.

?

2.8.3 Evolution of eMBB

Although the eMBB set of use cases was the central focus of Release 15, there are still?active efforts to enhance this further. For example, the study on requirements for NR?beyond 52.6 GHz (TR 38.807) is laying the groundwork for this and aspires to open even?higher spectrum bands. This addresses challenges such as the increased phase noise?encountered in these frequency bands, not to mention the increased propagation loss.?This may even mean that new waveforms and spread spectrum utilization mechanisms?are needed. Addressing the challenge of designing efficient power amplifiers for these?bands is also being considered as part of the study.

This contribution of access to the higher bands to eMBB is complemented by work on?NR-based access to the unlicensed spectrum. This is particularly challenging as radio?access must co-exist with other transmissions. This leads to lower spectral efficiency and?less reliability unless it is well managed. However, as a complement to licensed spectrum,?unlicensed bands can bring significant benefits. This has implications for most aspects of?the radio physical layer and associated procedures.

The use of NR operating in unlicensed spectrum, NR-U, is being developed in Release 16.?This is targeted at 5 and 6 GHz. A design principle that is being followed by 3GPP is that?of fair coexistence. That is, the addition of a 5G NR network to an unlicensed carrier should?have an impact no greater than the addition of a Wi-Fi network on that carrier.

Several scenarios for using unlicensed spectrum bands are envisaged across carrier?aggregation, dual connectivity, and stand-alone. The carrier aggregation scenario between?licensed and unlicensed bands would see a licensed band NR primary cell and an NR-U?secondary cell. Two dual connectivity scenarios are under consideration where NR-U is in dual connectivity either with LTE or NR licensed. The stand-alone scenario NR-U,?connected to the 5GC, would be appropriate for applications not requiring high reliability?or guaranteed data rates.

2.9 Navigating the 3GPP Standards for 5G NR

This section provides a quick guide to the key specifications for 5G NR.

TS 38.201: Physical layer; General description; very high-level description of NR physical?layer in terms of the relationship to other layers, the choices for multiple access,?physical channels, modulation, and channel coding, along with physical layer procedures?and measurements.

TS 38.202: Services provided by the physical layer; high-level description of the layer 1?functions along with the various uplink and downlink channels.

TS 38.211: Physical channels and modulation; covers physical channels and modulation,?including physical channels, frame structure, modulation, random access, and?synchronization symbols. Introduces antenna ports, resource grid, resource elements,?resource blocks, and bandwidth parts. Introduces the UL and DL physical channels and signals.

TS 38.212: Multiplexing and channel coding; covers how logical channels are mapped?to physical channels, how data are encoded, protected, and multiplexed onto the?physical channels.

TS 38.213: Physical layer procedures for control; describes synchronization, radio link?monitoring and recovery procedures, power control, random access, and procedures for?exchanging control information.

TS 38.214: Physical layer procedures for data; describes the procedures concerning the?physical layer that the UE must perform to maintain the connection. Includes receiving and?sending transmissions and reporting the channel state.

TS 38.215: Physical layer measurements; describes the measurements that the UE and gNB?are expected to make including received strength, quality, and signal-to-noise ratios of?various signals, timing measurements, and measurements of LTE.

TS 38.300: NR and NG-RAN overall description; provides a comprehensive overview of 5G.?Describes the interfaces within the 5G RAN and into the 5G core. Introduces the physical?layer, medium access control (MAC), radio link control (RLC), packet data convergence?protocol (PDCP), and radio resource control (RRC). Covers mobility in idle, inactive, and?connected mode along with mobility between radio access technologies (RATs). Covers?scheduling, QoS, security, self-configuration, and self-optimization.

TS 38.304: User Equipment (UE) procedures in idle mode and RRC Inactive state; describes?the various RRC states when the mobile is not connected and procedures for the selection of?the radio network, selection and reselection of NR cells, broadcast information, and paging.

?

2.10 5G NR and Network Slicing

Network slicing is key to delivering mixes of use cases in the same RAN. The vision of?many industry verticals being enriched by new services with different mixes of low latency,?ultra-reliability, massive connectivity, and enhanced Mobile Broadband is most valuable if these can be delivered simultaneously in the same network. There are many design?decisions in 5G NR that have been made with this desire to address numerous use cases?together simultaneously.

?

TS 38.306: User Equipment (UE) radio access capabilities; describes the various categories?of UE in terms of the capabilities they possess including maximum data rates and buffer?sizes along with parameters of the PDCP, RRC, RLC, MAC, RF, and physical layers.

TS 38.321: Medium Access Control (MAC) protocol specification; defines the MAC. Includes?coverage of random access, maintenance of timing alignment, scheduling of transmissions,?retransmissions, MAC protocol data units, and control elements.

TS 38.322: Radio Link Control (RLC) protocol specification; defines the RLC along with the?acknowledged-, unacknowledged-, and transparent-modes for RLC entities. Also covers?procedures including those for management of RLC entities, data transfer, and automatic?repeat request (ARQ). Defines how RLC protocol data units (PDUs) are constructed along?with the RLC parameters.

TS 38.323: Packet Data Convergence Protocol (PDCP) specification; defines the PDCP in?terms of its architecture, services, and functions. Also covers procedures including those?for management of PDCP entities, data transfer, discarding service data units (SDUs) and?data recovery, reporting on transmit and receive operation, header compression, ciphering,?integrity protection, and PDCP duplication. Defines how data and control PDCP protocol?data units (PDUs) are constructed along with the PDCP parameters.

TS 38.331: Radio Resource Control (RRC); Protocol specification; defines the RRC in terms?of its architecture, services, and functions. Also covers procedures including those for?management of PDCP entities, data transfer, discarding service data units (SDUs) and?data recovery, reporting on transmit and receive operation, header compression, ciphering,?integrity protection, and PDCP duplication. Defines how data and control PDCP protocol?data units (PDUs) are constructed along with the PDCP parameters.

TR 38.912: Study on New Radio (NR) access technology; good overview of the deployment?scenarios, forward compatibility, and protocol architectures of the user and control plane of the radio interface. Also includes the physical layer modulation, multiplexing, data and?control channels, waveforms, multiple access, channel coding, multiple-antenna schemes,?physical layer procedures, scheduling, power control, and random access. Also covers the?protocol stack including MAC, RLC, PDCP, AS, and RRC. Also covers the architecture including?the responsibility of core and RAN, the NG and Xn interfaces, quality of service, dual?connectivity with LTE, and network slicing. Includes procedures for initial access, mobility,?dual connectivity, and session management along with radio transmission and reception.

Flexible numerologies and bandwidth parts are key to enabling network slicing. As?described previously, the higher numerologies have larger subcarrier spacing, and hence?shorter symbols and slots can be scheduled more rapidly. As the slots are self-contained?in terms of how data are scheduled on them, this is key to supporting low-latency applications. For the numerology with 120 kHz subcarrier spacing (the highest numerology?that supports data), the slot lasts a mere 125 μs. This gives extraordinarily frequent?opportunities for scheduling low-latency data in conditions where the channel conditions?support it and can deliver the low latency required.

?

But it would be wasteful if an entire carrier had to be given up to numerology with?the shortest slots. Splitting a block of the spectrum into two carriers could overcome this?but would reduce the spectral efficiency with the need for guard bands. Thus, another?key part of delivering network slicing is the use of bandwidth parts. The ability to use?different numerologies within the same carrier means that services with vastly different?characteristics can be supported while preserving spectral efficiency.

While the use of high subcarrier spacing and short slots is consistent with low latency, it?is also subject to the issue of inter-symbol interference. This will reduce the reliability of?the transmission and could have an impact on Ultra-Reliable Low Latency Communication?(URLLC). But even when high-order numerologies are used where the delay spread is significant and indicates against the use of such short symbols, there are mitigations in the?5G standard.

First, the most robust modulation and coding schemes may be sufficient to overcome?the inter-symbol interference. But packet duplication can also assist in this regard. This?is a mechanism that creates a second RLC entity along with a second logical channel on?the radio bearer when more reliability is required. The data is then transmitted twice—?once in each RLC entity. This significantly raises the chance that each packet will be?delivered successfully.

Mini slots described earlier, are also a key feature for low latency. Resource blocks in the?resource grid can be reserved for mini-slots and not be used for transmissions scheduled?as part of normal slot scheduling. Mini slots can start at any time in the period and don’t need to be aligned to other slot boundaries. This is ideal for low-latency applications as the?transmissions can start as soon as they are needed by the low-latency application with no?constraints on delay. Mini slots can be used in the DL and UL and can be as short as one?symbol in length.

The 5G standard makes it possible for data to be grouped into different logical channels.?Logical channel prioritization (LCP) allows these logical channels to be assigned different?prioritizations. This supports network slices that can respect the relative importance?and latency requirements of the slices. Restrictions can be placed on the logical channels?so that they must be restricted to being scheduled on specific combinations of configured?cells, numerologies, or PUSCH transmission durations. Logical channels thus have a well-defined hierarchy of priority consistent with their importance, and the latency can be?controlled by arranging for the latency to below for slices that need it.

These various features, in combination with the other flexible uses of the 5G infrastructure,?facilitate network slicing for delivery of the rich mix of heterogeneous services that 5G will provide.

?

?