In addition to speed, what are the key technologies of 5G?

[[285321]]

5G is not just about faster internet speeds, but more about disrupting lifestyles and catalyzing all walks of life. 5G is not just about high bandwidth, but also about many points of integration with B-side users (enterprises). Next, let's learn about the key technologies of 5G with Zishuo from Alibaba Entertainment.

1. Key technologies of 5G

5G has many core technologies, including many technology sets. Those who have a little knowledge of 5G should know that 5G has actually defined three major scenarios:

• eMBB: Enhanced Mobile Broadband, as the name implies, is aimed at high-traffic mobile broadband services;
• URLLC: Ultra-high reliability and ultra-low latency communication (3G response is 500ms, 4G is 50ms, and 5G requires 1ms), which will be used in autonomous driving, telemedicine, etc.;
• mMTC: Massively Connected Internet of Things, targeting large-scale Internet of Things services.

1.1 eMBB

4G is already so fast, so how does 5G continue to increase its capacity?

Capacity = bandwidth * spectrum efficiency * number of cells

According to this formula, there are only three ways to increase capacity: increasing spectrum bandwidth, improving spectrum efficiency, and increasing the number of cells. Increasing the number of cells means building more base stations, which is too costly.

As for spectrum bandwidth, resources in the mid- and low-frequency bands are very scarce, so 5G has expanded its vision to the millimeter wave field. As will be introduced later, the millimeter wave band has a high frequency band and rich resources, making it a key spectrum development area. In addition to expanding more spectrum resources, another effective way is to make better use of the existing spectrum. Cognitive radio has also made some progress after years of development. Cognitive radio can be used to improve the utilization rate of radio and television white spectrum.

White spectrum refers to the spectrum that can be used by wireless communication devices or systems at a specific time and in a specific area without interfering with higher-level services. The so-called radio and television white spectrum refers to the white spectrum in the broadcast and television band. Because the frequency band where the broadcast and television signals are located is a very high-quality band and is very suitable for wide-area coverage, the application of cognitive radio in this band deserves attention.

Operators prefer to increase capacity by improving spectrum efficiency. They use error correction and coding methods to approach the Shannon limit rate. Compared with 4G's Tubor code, 5G's channel coding is more efficient.

The modulation technology currently used by 4G and WiFi is mainly OFDM. The capability of this modulation method has been greatly improved compared with previous CDMA, but OFDMA requires that each resource block is orthogonal, which will limit the use of resources. Therefore, if the signal can be demodulated normally even if it is not orthogonal, it will greatly improve the system capacity. Therefore, NOMA (non-orthogonal multiple-access) technology came into being. After the improvement of modulation technology has reached its limit, another more effective method is multi-antenna technology, which can achieve a significant increase in capacity through Massive MIMO.

★ 1.1.1 Channel Coding Technology

There are three main data coding schemes: LDPC code was proposed by Americans, Polar code was proposed by a university professor in Türkiye, and there is also Europe's Turbo2.0 code.

In October 2016, 3GPP held the RAN1#86bis meeting (hereinafter referred to as the 86th meeting) in Lisbon, Portugal. At this international conference, Turbo, which was previously dominated by 3G and 4G, had almost no supporters, and the protagonists of the debate were LDPC and Polar. In this meeting, the three factions criticized the technical shortcomings of the solutions proposed by other camps. However, LDPC had the upper hand due to its technical advantages and gained a large number of supporters, such as Samsung, Qualcomm, Nokia, Intel, Lenovo, Ericsson, Sony, Sharp, Fujitsu, Motorola Mobility, etc. At this time, only Huawei was still insisting on Polar code, and even if Lenovo voted for Polar code, it would be of no avail. At this meeting, LDPC had a clear advantage and became the solution adopted by 5G mobile broadband in the data transmission part.

In November 2016, 3GPP held the RAN1#87 meeting in the United States, which mainly discussed the 5G data channel short code solution and the 5G control channel solution. The final result of the vote was that in the channel coding technology solution for the 5G eMBB scenario, the long code coding and the uplink and downlink short code solutions for the data channel adopted the LDPC code promoted by Qualcomm; the control channel coding adopted the Polar solution promoted by Huawei.

5G data channels pursue transmission speed, mainly for large packets, and in this regard, LDPC has obvious advantages, which is why LDPC can successfully take over the long code of the data channel. Regarding the 5G control channel, due to the small amount of data transmitted, reliability is more important than speed. In this regard, Polar code has important advantages. In addition, with the broad support of Chinese manufacturers (including Lenovo's vote in favor), Polar code has become the international coding standard for 5G mobile broadband control channels.

From the performance of different channel coding under large information block length, it can be seen that the transmission efficiency of LDPC is still significantly higher than the other two.

★ 1.1.2 Non-orthogonal multiple access technology

4G network uses Orthogonal Frequency Division Multiple Access (OFDM) technology. OFDM can not only overcome the multipath interference problem, but also cooperate with MIMO technology to greatly improve the data rate. Due to the orthogonality of multiple users, there is no far-near problem between the mobile phone and the cell, and fast power control is abandoned. Instead, the AMC (adaptive coding) method is used to achieve link adaptation.

From 2G, 3G to 4G, multi-user multiplexing technology is nothing more than working on the time domain, frequency domain, and code domain, while NOMA adds a dimension based on OFDM - the power domain.

The purpose of adding this power domain is to realize multi-user multiplexing by utilizing the different path losses of each user.

• What NOMA hopes to achieve is to revive the non-orthogonal multi-user multiplexing principle of the 3G era and integrate it into the current 4G OFDM technology.
• NOMA can use the difference in path loss to superimpose multi-path transmission signals to improve signal gain. It enables all mobile devices in the same cell coverage area to obtain the maximum accessible bandwidth and can solve the network challenges caused by large-scale connections.

★ 1.1.3 Millimeter Wave

As early as 2015, the Federal Communications Commission of the United States took the lead in planning the four frequency bands of 28 GHz, 37 GHz, 39 GHz and 64-71 GHz as the recommended frequency bands for 5G millimeter waves in the United States. The FCC of the United States held a 28GHz spectrum auction, and the total transaction amount of 2,965 spectrum licenses was nearly US$703 million. (PS: Foreign spectrum is publicly auctioned, while domestic spectrum is allocated by the Radio Management Committee).

The great advantage of millimeter waves is that they have high frequency bands, abundant spectrum resources, and wide bandwidth. In addition, the high spectrum and short wavelength mean that the antennas are correspondingly shorter, making it easier to build multi-antenna applications on small devices such as mobile phones. According to the formula of speed of light = wavelength * frequency, the wavelength of a 28GHz frequency is about 10.7mm, which is millimeter waves. Generally speaking, the antenna length is proportional to the wavelength. Basically, the antenna is optimal when it is one-quarter or one-half of the wavelength. Therefore, the shorter wavelength of millimeter waves also makes the antenna shorter.

In the Massive MIMO system, a large-scale antenna array can be designed at the base station end of the system, so that the millimeter wave application can be combined with the beamforming technology, which can effectively improve the antenna gain. However, due to the short wavelength of the millimeter wave, in millimeter wave communication, the transmission signal is easily interfered by external noise and attenuated to varying degrees when the millimeter wave is used as a carrier. The signal is not easy to pass through buildings or obstacles, and can be absorbed by leaves and rain.

★ 1.1.4 Massive MIMO and Beamforming

MIMO (Multiple-Input Multiple-Output) can be translated as multiple input multiple output, that is, the transmission and reception of multiple antennas. MIMO is not a new technology and has been used in the LTE (4G) era. Through higher-order MIMO technology, combined with carrier aggregation and high-order modulation, the industry has been able to make LTE reach gigabit speeds (1Gbps and above), which is ten times the speed of the initial LTE.

MIMO technology breaks through the limitations of Shannon's theorem, jumps out of the point-to-point single-user framework, and transforms a single point-to-point channel into multiple parallel channels for processing, so that the spectrum efficiency mainly depends on the number of parallel channels, thereby improving system capacity and spectrum efficiency.

As shown in the figure below, LTE and LTE-A base stations and mobile phones use a small number of antennas. The small number of antennas used on mobile phones is mainly limited by the size of the mobile phone. In the current low and medium frequency bands, the corresponding antenna size is still large, and it is impossible to integrate too many antennas in the mobile phone. After 5G uses millimeter waves, the size of the antenna becomes very small, and a large number of antennas can be easily integrated. Massive MIMO can support up to 256 antennas.

To achieve Massive MIMO, the base station must accurately grasp the channel information and terminal location, which is not a big problem for the time-division multiplexing TDD system, but it is troublesome for the frequency-division multiplexing FDD system. Since the TDD system uses the same frequency band for uplink and downlink, the downlink channel can be estimated unilaterally based on the uplink channel status, that is, the reciprocity of the uplink and downlink channels is used to infer the downlink link from the base station to the terminal. In the FDD system, since the uplink and downlink are not in the same frequency band, the uplink channel status cannot be directly used to estimate the downlink channel status. In order to achieve channel estimation, CSI feedback needs to be introduced. With a large amount of CSI feedback, as the number of antennas increases, not only the overhead increases, but the accuracy and timeliness of the feedback information may also decrease. Therefore, the industry has always believed that Massive MIMO is more difficult to deploy on FDD.

In fact, when China was developing 3G, the domestic TD-SCDMA mentioned smart antennas. The base station system used digital signal processing technology and adaptive algorithms to enable smart antennas to dynamically form directional beams for specific users in the coverage space. Although TD-SCDMA did not take off, it is undeniable that it allowed major Chinese manufacturers to accumulate more experience in MIMO antennas and beamforming. Foreign countries have been promoting FDD, and it seems that TDD has an indispensable advantage in Massive MIMO.

China Mobile conducted field tests in Hangzhou, using Huawei's 5G solution from chip to core network end-to-end. The network side uses Huawei's 2.6GHz NR wireless equipment that supports 160MHz large bandwidth and 64T64R MassiveMIMO, connecting to the core network that supports 5G SA architecture centrally deployed in Beijing, and the terminal side uses a test terminal based on Huawei's Balong 5000 chip. It can be seen that the base station side uses 64T64R, that is, 64 transmitting antennas and 64 receiving antennas, a total of 128 antennas.

MIMO technology has gone through a development process from SU-MIMO (single-user MIMO) to MU-MIMO (multi-user MIMO). SU-MIMO is characterized by serving only a single terminal, and the terminal is limited by the number of terminals and the complexity of design, which limits further development. MU-MIMO combines multiple terminals for spatial multiplexing, and the antennas of multiple terminals are used at the same time. In this way, a large number of base station antennas and terminal antennas form a large-scale virtual MIMO channel system. This is a more macro-thinking about improving system capacity from the perspective of the entire network. However, the introduction of so many antennas and the cross-signaling will inevitably lead to interference, which requires preprocessing and beamforming technology.

This spatial multiplexing technology changes the signal coverage from omnidirectional to precise directional service. The beams will not interfere with each other, providing more communication links in the same space and greatly improving the service capacity of the base station.

Suppose there is an omnidirectional base station (red dot) on the edge of a square surrounded by dense buildings, and three terminals (red, green, and blue X) are distributed in different directions around it. In the Massive MIMO scenario, after introducing precise beamforming, the situation magically becomes as follows. Doesn't it look very high-end? It can already accurately control the direction of electromagnetic waves. It's easier said than done, and there are countless high-techs involved.

Image source: https://www.cnblogs.com/myourdream/p/10409985.html

★ 1.1.5 Cognitive Radio

Why cognitive radio exists? The main reason is that low-frequency spectrum resources are very scarce and have been allocated to some systems before, but it is found that these systems do not use the spectrum very effectively. Therefore, cognitive radio technology is considered to make use of these spectrums without affecting the main communication system.

Cognitive radio can be understood as a radio that gains awareness of its surroundings and adjusts its behavior accordingly. For example, a cognitive radio can identify an unused frequency band and use it for transmission before jumping to another unused frequency band. The term cognitive radio was coined by Joseph Medora and refers to an intelligent radio that can sense the external environment, learn from history, and make intelligent decisions to adjust its transmission parameters based on the current environmental situation.

Cognitive radio is a combination of SDR (software defined radio) and MIND (artificial intelligence). We can imagine that radio gives humans some functions, perceives the outside world through observation, and then decides whether to send and how to send. There will be a lot of research and applications related to cognitive radio in 5G.

1.2uRLLC

The theoretical latency of 5G is 1ms, which is a few tenths of the latency of 4G, and basically reaches the level of quasi-real time. This will naturally give rise to many application scenarios. In fact, the full name of uRLLC is ultra-reliable, low-latency communication, so it is not just low latency but also high reliability. With low latency and reliability, some technologies such as industrial automation control, telemedicine, and autonomous driving can be gradually built. The changes brought about in this regard may be earth-shaking. Things that seemed impossible before are slowly becoming possible. Let's see what has been done to make these a reality.

★ 1.2.1 5GNR frame structure

First, let me explain what 5GNR is. It is actually the 5G air interface standard. 3gpp gave it a name, called 5GNR (New Radio). In the 4G era, the air interface is generally named LTE (Long Term Evolution). Like LTE, a 5GNR wireless frame is 10ms long. Each wireless frame is divided into 10 subframes, and the subframe length is 1ms; each wireless frame can be divided into two half-frames (half-frame). The first half-frame is 5ms long and contains subframes #0~#4, and the second half-frame is 5ms long and contains subframes #5~#9; the structure of this part is fixed.

The subcarrier spacing of 5G NR is no longer fixed at 15Khz like LTE, but is variable and can support five configurations: 15kHz, 30kHz, 60kHz, 120kHz, and 240kHz. Why can't it be less than 15KHz or greater than 240KHz?

Phase noise and Doppler effect determine the minimum subcarrier spacing, while cyclic prefix CP determines the maximum subcarrier spacing. We certainly hope that the subcarrier spacing is as small as possible, so that more data can be transmitted under the same bandwidth. However, if the subcarrier spacing is too small, phase noise will produce excessive signal errors, and eliminating this phase noise will place too high requirements on the local crystal oscillator.

If the subcarrier spacing is too small, the physical layer performance is also susceptible to interference from Doppler frequency deviation; if the subcarrier spacing is set too large, the duration of the CP in the OFDM symbol will be shorter. The purpose of designing the CP is to eliminate delay spread as much as possible, thereby overcoming the negative impact of multipath interference. The duration of the CP must be greater than the delay spread of the channel, otherwise it will not play a role in overcoming multipath interference. Therefore, the choice of 15KHz~240KHz is a compromise result of a series of comprehensive considerations such as technology and implementation cost.

As shown in the following figure, the larger the subcarrier spacing, the shorter the time slot (the minimum subcarrier spacing of 15KHz corresponds to a time slot length of 1ms, and the maximum subcarrier spacing of 240KHz corresponds to a time slot length of 0.0625ms). For the uRLLC scenario, low transmission latency is required. In this case, the network can meet the latency requirement by configuring a relatively large subcarrier spacing.

The flexible framework design of 5G NR can expand TTI upward or downward (i.e., use longer or shorter TTI), depending on specific needs. In addition, 5G NR also supports multiple transmissions with different TTIs at the same frequency. For example, mobile broadband services with high QoS (quality of service) requirements can choose to use a 500 µs TTI instead of the standard TTI in the LTE era. At the same time, another service that is very sensitive to latency can use a shorter TTI, such as 140 µs, instead of having to wait until the next subframe arrives, which is 500 µs later. That is to say, after the last transmission ends, both can start at the same time, thus saving waiting time.

★ 1.2.2 Multi-carrier technology improvements

In the OFDM system, each subcarrier is orthogonal in the time domain, and their spectra overlap, so it has a higher spectrum utilization rate. OFDM technology is generally used in data transmission in wireless systems. In the OFDM system, due to the multipath effect of the wireless channel, interference occurs between symbols.

In order to eliminate inter-symbol interference (ISI), a guard interval is inserted between symbols. The general method of inserting a guard interval is to set the symbols to zero, that is, to stay for a period of time after sending the first symbol (not sending any information), and then send the second symbol. In the OFDM system, although this weakens or eliminates the inter-symbol interference, it destroys the orthogonality between subcarriers, thus causing inter-subcarrier interference (ICI). Therefore, this method cannot be used in the OFDM system. In the OFDM system, in order to eliminate both ISI and ICI, the guard interval is usually played by CP (Cycle Prefix). CP is a system overhead and does not transmit valid data, thereby reducing the spectrum efficiency.

The CP-OFDM technology currently used in LTE can solve the problem of multipath delay very well, but it is sensitive to frequency deviation and time deviation between adjacent sub-bands. This is mainly due to the large spectrum leakage of the system, which easily leads to interference between sub-bands. At present, the LTE system uses a guard interval in the frequency domain, but this reduces the spectrum efficiency and also increases the delay to a certain extent. Therefore, 5G needs to consider some new waveform technologies. The current CP-OFDM will encounter challenges in MTC and short access scenarios, polar delay services; burst, short frame transmission; low-cost terminals have large frequency deviations, which are not conducive to orthogonality. In multi-point collaborative communication scenarios, it is difficult to transmit and receive signals at multiple points.

There are some candidate improvement technologies. The new waveform candidates proposed by various companies at the 3gpp conference include: Windowed Orthogonal Frequency Division Multiplexing (CP-OFDM with WOLA), Shifted Filter Bank Multi-Carrier (FBMC-OQAM), Filter Bank Orthogonal Frequency Division Multiplexing (FB-OFDM), Universal Filter Multi-Carrier (UFMC), Filter Orthogonal Frequency Division Multiplexing (F-OFDM) and Generalized Frequency Division Multiplexing (GFDM). These technologies are too professional to be introduced here. Interested students can use keywords to search for them. I haven't worked on this for many years, so it's a bit difficult to understand them, but it doesn't matter. It's good to know what problems they solve.

★ 1.2.3 Network Slicing

As a crucial technology in 5G, network slicing technology has greatly liberated operators and is deeply loved by operators. Traditional routers are all hard-switched, and rules and other things need to be configured in advance by connecting to the network cable. Modification is very inconvenient. Of course, if there is no need to process data packets on demand, this is actually quite good, fast and stable. However, with the increasing demand for differentiated services, how to manage the network more quickly and efficiently has become a headache. The emergence of SDN just solves this problem. Software Defined Network (SDN) is a new type of network innovation architecture proposed by the Clear State research group at Stanford University in the United States. It is a way to implement network virtualization.

After SDN transformation, there is no need to repeatedly configure the routers of each node in the network. The devices in the network are automatically connected. You only need to define simple network rules when using them.

What SDN does is to separate the control rights on network devices and manage them by a centralized controller, without relying on the underlying network devices (routers, switches, firewalls), shielding the differences from the underlying network devices. The control rights are completely open, and users can customize any network routing and transmission rules and policies they want to implement, making it more flexible and intelligent. The control plane and data plane are separated, and different forwarding rules can be configured for different data packet types/sources, thereby distinguishing different service levels for data packets, which in turn leads to differences in service quality.

Someone made a vivid metaphor for SDN, which helps to better understand SDN.

1.3 mMTC

Let's first look at the KPIs of mMTC. The connection density is 1,000,000/km2. The battery life is 10 to 15 years when the MCL (maximum coupling loss) is 164dB, which means that it can still work for 10 to 15 years even when the signal is very poor (the worse the signal, the greater the transmission power and the more power it consumes). Coverage enhancement requires a rate of 160bps when MCL=164dB. The complexity and cost requirements of the UE are very low.

LTE-M, or LTE-Machine-to-Machine, is an IoT technology based on the evolution of LTE. It is called LTE enhanced MTC (eMTC) in 3GPP R13 and is designed to meet the needs of IoT devices based on existing LTE carriers. NB-IoT is a fusion of NB-CIoT and NB-LTE. NB-IoT technology is mainly promoted in China.

What is the difference between the two technologies? As shown in the figure below, the two technologies have their own advantages. If you have high requirements for voice, mobility, speed, etc., choose eMTC technology. On the contrary, if you do not have high requirements for these aspects, but have higher requirements for cost, coverage, etc., you can choose NB-IoT.

Although the cell capacity is 50,000 connections, they basically use PSM and eDRX mechanisms, so the devices are in sleep mode most of the time, reducing the signaling interaction with the base station and indirectly increasing the cell capacity. This capacity increase is mainly due to the long-term sleep of the device. It can be seen that the eDRX cycle time of NB-IoT is longer than that of eMTC, so the response speed to downlink data will be slower.

These two technologies have their own advantages for different types of IoT technologies. Therefore, some people say that the two technologies are complementary and each is suitable for different IoT usage scenarios.

• Category 1 services: water meters, electricity meters, gas meters, street lights, manhole covers, garbage cans and other industries/scenario, which are characterized by static, small data volume, and low latency requirements, but have strict requirements on working hours, equipment costs, network coverage, etc. For such services, NB-IoT is more suitable technically.
• The second type of business: elevators, smart wearables, logistics tracking and other industries/scenarios have certain requirements for data volume, mobility and latency. For this type of business, eMTC is technically superior.

The following figure shows that 3GPP will address the massive and critical issues in the 5G era. Massive refers to the problem of large-capacity IoT communications, and critical includes high reliability and low latency. The standards are still evolving. Currently, China Telecom's NB-IoT network construction speed is the fastest in China. From our online use, it can basically cover our business areas.

2. 5G networking and coverage

2.1 Domestic spectrum allocation

The results of the domestic 5G spectrum allocation have been released, which should be the result of an assessment of the current status of operators. The green part in the figure below is the spectrum allocated this time, with China Telecom and China Unicom each getting 100MHz, and China Mobile getting 260MHz.

1) China Unicom and China Telecom have obtained the international mainstream 5G frequency band near 3.5GHz, which has the following characteristics:

• The industry chain is relatively mature, the research and development is relatively complete, and it has the greatest global applicability;
• The development progress is relatively fast and commercialization is relatively early;
• Lower frequency, more economical, lower base station density required, and relatively smaller capital expenditure.

2) China Mobile obtains the 2.6+4.9 GHz combined spectrum, which has the following characteristics:

• The 100M bandwidth of 4.9GHz can support more users and more traffic, but the density of base stations required is higher, which puts some pressure on capital expenditures;
• The 2.6GHz spectrum industry chain is less mature, and China Mobile needs to take the initiative to promote the cultivation and layout of the industry chain. However, it has a wide coverage range and low capital expenditure, and can also provide dual-band insurance for 5G commercial use.

3) It can be seen that the total frequency band allocated to China Mobile is 260MHz. However, since the 2515-2675MHz allocated this time includes the range of 4G in this frequency band before, excluding the previous 4G allocation, the actual new spectrum allocated this time is 200MHz.

China Mobile already has a large number of TD-LTE equipment at 2.6GHz (2575MHz~2635MHz), which will have a speed advantage in 5G construction. In particular, the upgrade and transformation of existing equipment can increase 5G coverage, but 2.6GHz is not the mainstream 5G spectrum at present, so China Mobile will need to spend more effort to cultivate the industry chain.

3.5GHz is the mainstream spectrum in China. The industrial chain in this frequency band is relatively mature, so it is also the focus of competition among operators.

2.2 Hotspot coverage or continuous coverage?

2.6GHz has the feasibility of outdoor continuous coverage, but its uplink coverage is limited by terminal capabilities and power, and the uplink coverage capability is relatively weak. The uplink coverage is 4dB lower than that of 1800MHz, and more than 10dB lower than that of 800MHz. The propagation loss of wireless signals in free space follows certain rules. The higher the spectrum, the greater the propagation loss and the shorter the propagation distance. In fact, continuous coverage or hotspot coverage mainly involves the issue of investment cost and return on investment, because the higher the propagation loss means that the base stations have to be built more densely, and the cost will increase greatly. China Mobile is allocated a lower frequency band, which has a greater possibility of continuous coverage.

According to conservative estimates, the number of 5G base stations (macro base stations) will be 1.2 to 1.5 times the number of existing 4G base stations. Since 5G networks operate at higher frequency bands, the penetration ability of traditional macro base stations is weakened, so small base stations or indoor distributed system base stations will become a great supplement, such as deploying distributed systems in some hot spots, shopping malls, stadiums, underground parking lots, etc.

2.3 SA or NSA?

First, let's explain SA and NSA. Non-Standalone (NSA), Standalone (SA).

In fact, this concept is easy to understand, as shown in the figure below. There are two options for upgrading from 4G to 5G. Those with deep pockets can choose to build a completely independent 5G core network and 5G base stations. Some less powerful companies can consider transitioning and reuse the existing 4G core network to enjoy the new air interface features brought by 5G base stations. The air interface rate will be improved, but some new features of the 5G core network, such as network slicing, cannot be used.

Because many foreign operators are not financially strong, and the cost of 4G has not been recovered, and they have to lay such a large network, which is really powerless. Therefore, in order to let everyone play happily with 5G, 3GPP also provides various NSA upgrade packages for everyone to choose from. Because after the air interface rate of 5G is increased, the original 4G base station may not be able to support such a high rate, and may face some transformation.

Since NSA's 5G air interface carrier only carries user data, system-level service control still relies on the 4G network, and it is to expand capacity by adding new carriers to the existing 4G network. Because it still relies on the core network and control plane of the 4G system, the non-independent networking architecture cannot give full play to the technical characteristics of the 5G system's low latency, nor can it flexibly support diversified business needs through network slicing, mobile edge computing and other features.

Globally, most operators chose NSA in the early stages, which is faster to deploy. However, this can only meet the enhanced mobile broadband part of the three major 5G scenarios, and cannot meet the low latency, high reliability and massive connection scenarios. In addition, the 5G NSA standard was closed relatively early, and the SA standard is still in progress. Therefore, some existing 5G terminal chips only support NSA. If we only consider the bandwidth, it is not a big problem for mobile phones to only support NSA.

2.4 Ultra-Dense Networking (UDN)

5G has high-density networking capabilities in some hot spots, such as large-scale stadium performances and events that are close to Damai's business, which will be an ultra-dense networking scenario. In the high-capacity and dense hotspot scenario, the wireless environment is complex and the interference is changeable. The ultra-dense networking of base stations can improve the spectrum efficiency of the system to a certain extent, and can quickly allocate wireless resources through fast resource scheduling, thereby improving the utilization rate of wireless resources and spectrum efficiency of the system, but it also brings many problems.

The coexistence of high-density wireless access sites may bring serious system interference problems; high-density sites will make switching between cells more frequent, which will cause a sharp increase in signaling consumption and a decline in user service quality; in order to achieve rapid and flexible deployment of low-power small base stations, small base stations are required to have plug-and-play capabilities, including autonomous backhaul, automatic configuration and management functions.

The key technologies to solve these problems are:

1) Multi-connection technology: The main purpose of multi-connection technology is to achieve simultaneous connection between UE (user terminal) and multiple macro and micro wireless network nodes. In dual-connection mode, the macro base station serves as the main base station of the dual-connection mode, providing a centralized and unified control plane; the micro base station serves as the secondary base station of the dual-connection mode, only providing data bearing for the user plane. The secondary base station does not provide control plane connection with the UE, and only the RRC (radio resource control) entity corresponding to the UE exists in the main base station.

2) Wireless backhaul technology: In the existing network architecture, it is difficult to achieve fast, efficient, and low-latency horizontal communication between base stations, and base stations cannot achieve ideal plug-and-play. In order to improve the flexibility of node deployment and reduce deployment costs, using the same spectrum and technology as the access link for wireless backhaul transmission can solve this problem. In the wireless backhaul mode, wireless resources not only serve the terminal, but also provide relay services for the node.

3) Dynamic adjustment of small cells to maximize spectrum utilization. For sudden gatherings and events such as exhibitions or football matches, the traffic fluctuation characteristics are more obvious, and the network sharing behavior of user groups is more common, so the uplink capacity requirements are higher. For relatively closed indoor venue areas, it is necessary to implement dynamic UL/DL subframe ratio adjustment according to the real-time traffic situation, such as adjusting to an uplink-dominated configuration to meet the uplink video backhaul requirements. Specifically, for scenarios with high demand for downlink resources such as big data downloads such as movies and music, more downlink resources need to be expanded for transmission, such as adjusting D/U from 3:1 to 8:1; large-scale conference live broadcasts, video or audio content uploads, there is a great demand for uplink resources, such as adjusting D/U from 3:1 to 1:3. In addition, user groups with similar service types are usually clustered, or even exist in cell units, that is, in the deployment area, when the user service demand statistics over a period of time show a stable and obvious feature, such as an increase in the demand for uplink services, then it is necessary to make unified time slot adjustments to the cells in this area.

The communication experience requirements in complex and diverse scenarios are getting higher and higher. In order to meet the needs of users to obtain a consistent service experience in ultra-dense scenarios such as large gatherings, open-air gatherings, and concerts, 5G wireless networks need to support a 1,000-fold capacity gain and 100 billion. For such future hotspot high-capacity scenarios, UDN (ultra-dense networking) can achieve a huge improvement in system frequency reuse efficiency and network capacity by increasing the density of base station deployment, and will become a key solution for hotspot high-capacity scenarios. In the near future, the popularity of ultra-high-definition, 3D, and immersive videos will greatly increase data rates. A large amount of personal data and office data will be stored in the cloud, and massive real-time data interaction requires a transmission rate comparable to that of optical fiber.

3. Conclusion

To sum up, in this article, we can learn about the key technologies of 5G.

1) The peak rate of a single base station must reach 20Gbps, and the spectrum efficiency must reach 3 to 5 times that of 4G. These are indicators of eMBB ultra-wideband. The main technologies used include new coding technologies such as LDPC/Polar codes to improve capacity, use millimeter waves to expand more spectrum, use beamforming to bring spatial division multiple access gain, use NOMA technology to achieve PDMA power domain gain, and use Massive MIMO technology to obtain greater capacity. Millimeter waves make the wavelength shorter and the antenna shorter, so more antennas can be installed on the mobile phone. The base station side can support 64T64R antenna arrays with a total of 128 antennas.

2) The latency reaches 1 millisecond. This is a scenario related to uRLLC. The new air interface standard 5GNR defines a more flexible frame structure and a more flexible subcarrier spacing configuration. The maximum subcarrier spacing of 240KHz corresponds to a time slot length of 0.0625ms, which makes ultra-low latency applications possible. The new multi-carrier technology solves the resource waste such as the guard interval in the current CP-OFDM, reduces latency and increases utilization. In addition, there is network slicing technology, which makes the network more flexible, can better support ultra-low latency applications, and establish an end-to-end high-speed power. Network slicing technology is mainly the application of SDN and NFV in the core network.

3) The connection density reaches 1 million per square kilometer. This is the scenario of mMTC. At present, the standard is mainly based on the evolution of eMTC and NB-IoT. The two standards have their own advantages and disadvantages. eMTC is more suitable for scenarios with certain requirements for data volume, mobility, and latency. It has the characteristics of static, small data volume, and low latency requirements. However, NB-IoT is more suitable for scenarios with strict requirements for working hours, equipment costs, network coverage, etc. NB-IoT is currently the main coverage in China. The number of connections here is actually a relatively elastic or ideal value, because the increase in the number of connections is mainly brought about by the terminal's sleep through PSM or eDRX technology. In the future, more concurrent capabilities, smaller network signaling consumption, more burst data packets and other scenarios need to be considered. This part of the evolution still has a long way to go.

4. Postscript

Today's AI is very prosperous and popular, and is more concentrated in the field of image recognition. It is undeniable that CNN and deep neural networks have brought about tremendous changes in this field, but AI is not equal to DNN, image recognition, or even face recognition. To achieve a smarter world, AI technology needs to make breakthroughs in more aspects.

AI has made breakthroughs in the field of images, which is equivalent to the eyes of the intelligent world becoming brighter. Images that computers cannot understand are gradually becoming structured and understandable. Image recognition, image tracking, image segmentation, etc. have made the front-end smarter. The progress in speech recognition is equivalent to the ears of the intelligent world becoming heard and understandable. The advancement of various sensing technologies will gradually approach the perception of the physical world of human touch, smell, etc. Finally, it will gather to the brain to complete intelligent decisions and upload and release instructions, and 5G networks are gradually becoming a neural network connecting various parts of the intelligent world. The future is worth looking forward to, and we also hope that our Alibaba's urban brain can become an important part of the future intelligent world.

The 5G eMBB scenario will definitely develop earlier because this area has a relatively clear demand and high user perception. The other two scenarios of 5G are estimated to need to be combined with the scenarios more, and are more industrial applications. Operators are currently actively involved in B-side applications.