Sub-band full-duplex, a compromise for the 5G dream?

​Hello everyone, I am Mayfly.

In this issue, let’s talk about one of the most basic topics in the field of wireless communications: duplex.

Duplex refers to the bidirectional data transmission between two communication devices. Specifically, duplex technology includes full-duplex and half-duplex modes.

Full-duplex means that two-way data transmission can be carried out simultaneously. In other words, both parties in communication can send data while receiving data from the other party, and both sending and receiving can be carried out simultaneously.

Half-duplex is much simpler. Sending and receiving cannot be done at the same time, but can only be done in turns: you can't receive when sending, and you can't send when receiving. Our common walkie-talkies are in this mode.

Full duplex, are we using it?

After five generations of development, mobile communication technology has reached perfection. So, when our base stations and mobile phones interact, they must use full-duplex, right?

At this time, the two terms we use most often are: FDD (Frequency Division Duplex) and TDD (Time Division Duplex). So, what are they, full duplex and half duplex?

For FDD, we use two spectrums, one for base stations to send signals to mobile phones, also called downlink; the other for mobile phones to send signals to base stations, also called uplink. In order to prevent interference between downlink and uplink, a certain isolation band must be left between the two spectrums, which is called "duplex bandwidth".

It can be seen that the downlink and uplink links of FDD are both half-duplex. When they are combined, a "pseudo full-duplex" system is formed at the cost of doubling the spectrum resource usage. This is like the lanes on the road. Each lane can only be one-way, but when lanes in different directions are combined, two-way traffic can be achieved.

For TDD, the spectrum only needs to be occupied for a certain period, but the uplink and downlink can only be used in turns. In other words, when the base station is sending data, the mobile phone can only receive it quietly, and if it wants to send data, it can only hold it in until the time allocated to it is up.

Isn't this the real half-duplex? The 5G frequency bands we commonly use are all TDD modes, but the switching time between uplink and downlink is extremely short, in milliseconds, and we can't feel it at all. Therefore, TDD uses the fast switching of half-duplex at the micro level to achieve the "pseudo full-duplex" at the macro level.

Can't we transmit and receive at the same time on the same spectrum to achieve true "simultaneous and full-duplex"? In this way, the spectrum efficiency will be doubled! Isn't it wonderful that the country is safe and the people are happy?

However, problems that everyone can see have remained unresolved for so many years, and there must be huge difficulties that are extremely difficult to solve.

To achieve full-duplex is equivalent to having two trains running in opposite directions at high speed on the same track. The result is self-evident.

The reason for this is that when transmitting and receiving at the same time in the same frequency band, huge interference will be generated. This includes not only the self-interference of the base station's own transmission to its own reception, but also the interference between base stations, between base stations and mobile phones, and between mobile phones. These cross-link interferences are extremely difficult to handle.

Therefore, everyone can only focus their main efforts on increasing the number of lanes, continuously pushing up the frequency bands used, continuously widening the carrier bandwidth, and continuously doubling the number of transmitting and receiving channels.

For example, from 2G to 5G, the frequency bands used have expanded from low frequency (less than 1GHz) to medium frequency (less than 6GHz), and then to millimeter wave and even terahertz. The channel bandwidth has also expanded from a few megabytes to tens of megabytes, hundreds of megabytes and even gigahertz; the number of transmitting and receiving channels has also increased from single channel to dual channel, 4 channels, 8 channels, 32 channels, 64 channels and even 128 channels.

As for full-duplex technology, although it was widely discussed in the early stages of 5G standardization and was considered one of the key technologies of 5G, it was ultimately shelved due to difficulties in implementation.

Making wireless communications truly full-duplex has become the biggest dream buried deep in the hearts of communications professionals.

Realistic compromise: sub-band full-duplex

Time has passed. 5G has been commercially used for several years. The technical standards for the second half of 5G: 5G-Advanced standards are being formulated in full swing.

Full-duplex has once again entered everyone's field of vision.

This is because, with the penetration of 5G industry applications into industrial site networks, the demand for the network to simultaneously support ultra-large uplink bandwidth and ultra-low latency has become prominent, and the current FDD and TDD modes are unable to cope with it.

For example, applications such as video surveillance, electronic fences, and machine vision in factories are mainly large uplink services, and the bandwidth requirements of multiple terminals range from hundreds of Mbps to even Gbps; industrial AR requires a latency of less than 10 milliseconds, AGV collaborative handling requires a latency of less than 5 milliseconds, and machine motion control requires a latency of less than 4 milliseconds.

Why is it difficult for both the current FDD and TDD modes to simultaneously meet the requirements for large bandwidth and low latency? Let's talk about this below.

Since the division of spectrum usage has been determined historically and cannot be changed without taking all currently used systems offline, different frequency bands are actually strongly bound to FDD or TDD duplex modes.

Binding relationship between frequency band and duplex mode

The characteristics of the FDD frequency band are low frequency band, small available bandwidth, and limited speed. For example, 900M uplink and downlink have 35M bandwidth each, and 1800M uplink and downlink have 75M bandwidth each. These few precious resources have to be distributed to multiple operators, and the bandwidth available to each operator is even more limited. Although the coverage is good, the network speed is not high.

Although the rate is limited, the FDD mode has a prominent advantage, that is, the uplink and downlink data are sent on their own independent spectrums. Basically, as long as there is data, it can be sent without the time limit like TDD, so FDD can achieve a relatively short delay.

TDD spectrum is the opposite, with generally higher frequency bands and larger available bandwidth. For example, at 3.5GHz, China Unicom and China Telecom each have 100M bandwidth; at 2.6GHz, China Mobile has an exclusive 160M bandwidth.

These TDD large-bandwidth carriers can achieve high uplink or downlink rates by setting different uplink and downlink time slot ratios. However, due to the half-duplex characteristics of TDD itself, it is difficult to reduce the latency.

Although we cannot feel the impact of latency, the communication between machines in the factory is extremely sensitive to it. And such stringent latency requirements are rigid, and it will not work if they are not met.

If the advantages of TDD and FDD can be integrated in the same frequency band, wouldn’t it be possible to support large bandwidth and low latency at the same time?

So someone came up with the idea: Isn't the bandwidth of the TDD spectrum large? I can cut the TDD carrier into two sub-bands (called sub-bands). Both sub-bands are still TDD mode, but the uplink and downlink time configurations are opposite. In this way, when you send, I receive, and when you receive, I send. Doesn't this have the qualities of FDD?

In this way, we can integrate FDD technology into the TDD carrier at a relatively low cost through sub-band division and time slot configuration, thus realizing pseudo "full-duplex" within the TDD carrier.

TDD and sub-band full-duplex

However, such pseudo "full-duplex" is essentially a stitching of TDD and FDD technologies, and does not actually achieve an improvement in spectrum efficiency. It is just a small step in the long march to achieve full-duplex. Therefore, it is called "sub-band full-duplex", abbreviated as SBFD (Subband Full Duplex).

How to divide sub-bands?

From a purely technical perspective, the uplink and downlink can be divided in any way. They can each occupy half, so that the uplink and downlink rates are balanced; more downlink sub-bands can be allocated, so that a higher downlink rate can be achieved; more uplink sub-bands can be allocated, so that a higher uplink rate can be achieved.

From the demand perspective, what we ordinary people need when watching videos is a high downlink rate, but the latency requirement is not high, and there is no demand for sub-band full-duplex. In factories, applications such as data reporting, surveillance cameras, and machine vision require a large uplink, and a large number of control applications require low latency. Therefore, sub-band full-duplex is useful in industrial scenarios and needs to be configured to be mainly based on the uplink sub-band.

As for how many sub-bands need to be divided, two are enough from the perspective of usage, but in reality this mainly depends on the interference situation.

If the spectrum of an operator that wants to deploy sub-band full-duplex is adjacent to that of other operators, it is best to keep the adjacent spectrum as is, with the uplink sub-band as the main one. In this case, the uplink sub-band should be placed as far away as possible, so as to minimize interference.

Specifically, if there are spectrums of adjacent operators on both sides of the spectrum, it is recommended to divide it into two downlink sub-bands and one uplink sub-band, and put the uplink sub-band in the middle, and configure it in a sandwich manner of downlink + uplink + downlink; if there are other operators on only one side, then it is sufficient to divide it into one downlink sub-band and one uplink sub-band, which will have a better effect.

"Sandwich" division of uplink and downlink sub-bands

In terms of the frame structure, in order to be compatible with existing terminals, the traditional frame structure of DFFFU can be maintained. The first time slot is all downlink, the middle three time slots are configured with subbands and flexible uplink and downlink scheduling is performed as needed, and the third time slot is all uplink.

Two configurations of upper and lower subbands

How to eliminate interference?

Self-interference within the system is the core problem that must be solved in sub-band full-duplex.

Since the uplink and downlink sub-bands of the full-duplex sub-band are closely adjacent, unlike the traditional FDD uplink and downlink frequency bands with a duplex interval of tens of MHz, this will cause serious interference between the transmitter and the receiver.

Generally, the base station uses a common antenna for signal transmission and reception, and the strong signal transmitted will be directly received, causing the weak signal from the mobile phone to be blocked. In addition, when processing inside the base station, coupling interference will also occur between the RF transmission and reception links.

Self-interference suppression can be achieved through means such as spatial domain, RF domain, and digital domain, using a multi-pronged approach and multi-level elimination.

Self-interference cancellation in spatial, RF and digital domains

The simplest means of suppressing spatial self-interference is to separate the transmit and receive antennas. By using separate antennas for transmission and reception and adding multiple isolation barriers between the two antennas, the transmit signal can be effectively prevented from entering the receive antenna. In addition, the beam nulling technology of the transmit antenna in the direction of the receive antenna can further reduce interference.

High isolation antenna for transmitting and receiving

There are two ways to suppress interference in the RF domain: sub-band filters and RF interference elimination.

By adding a subband filter in the base station, the downlink subband can filter out the uplink subband signal through the filter, and the uplink subband can filter out the downlink subband signal through the filter. This method is relatively simple, but the adjustment of the filter bandwidth is not flexible and will increase insertion loss.

Radio frequency interference cancellation is achieved by collecting a copy of the known downlink transmission signal and transmitting it to the uplink receiving end, and then canceling it by constructing an opposite signal. This method is relatively complex and costly.

RF Interference Cancellation

When implementing RF domain interference suppression, the required interference suppression capability can be evaluated and one or a combination of two methods can be selected for implementation.

The idea of ​​digital domain interference suppression is similar to the second type of interference suppression in the RF domain. An auxiliary RF channel is introduced in the RF domain and converted into a digital signal, and then an opposite signal is constructed in the digital domain to cancel it out, thereby further reducing the residual interference.

By eliminating self-interference at three levels: spatial domain, RF domain, and digital domain, self-interference can be suppressed to an acceptable level with slightly reduced sensitivity.

Self-interference has been resolved, that is, a single base station can work normally by itself, but in actual deployment, it is impossible to have only one base station and one terminal, but multiple base stations must form a network and serve multiple different terminals at the same time. This involves a more difficult problem: cross-link interference.

To eliminate cross-link interference, it is necessary to design a corresponding interference measurement mechanism to know both the self and the enemy, and to transmit known interference characteristics, and then eliminate interference through beam nulling, interference suppression merging and other technologies. This process is more complicated than eliminating sub-interference within a single base station, and the industry is still studying it.

In order to take the first step smoothly, we should proceed step by step from easy to difficult. First, we can deploy sub-band full-duplex micro base stations in smart factories, which have low power and are relatively easy to isolate from outdoor macro base stations.

Later, we will consider networking between multiple sub-band full-duplex base stations, and finally we will try to solve the networking between sub-band full-duplex macro base stations and existing large downlink macro base stations. With the progress of solving the networking interference problem, the industrial ecology will naturally mature.

The road to standardization

Sub-band full-duplex has been launched in 3GPP R18 and is currently in the SI (Study Item) stage. Theoretical and engineering technology research has been fully launched.

China Mobile took the lead in standardizing sub-band full-duplex technology and packaged it into the UDD (Unified Division Duplex) series of technologies. Among them, S-UDD (Single carrier UDD) refers to sub-band full-duplex. Samsung also packaged the technology into XDD (cross division duplex).

Although the current research has made some progress, the technology is still far from being officially commercialized. According to the pace of R18 research and R19 standardization, the relevant protocols are expected to be frozen in 2025, and commercial use is expected to be after 2026.

In 2026, there are only three years left until 6G. Therefore, to smoothly promote the commercialization of sub-band full-duplex technology, it is necessary to focus on compatibility with existing terminals. This is because the upgrade and transformation of the base station side is usually easier to promote, while the popularization of the terminal industry chain is more lagging behind.

After achieving the above-mentioned sub-band full-duplex without overlapping sub-bands, we can go a step further and allow the sub-bands to overlap and study how to make the small amount of spectrum in the overlapping area capable of simultaneous and same-frequency full-duplex. The next step is to push the entire carrier towards simultaneous and same-frequency full-duplex. This is a step-by-step process.

In any case, sub-band full-duplex will serve as an important milestone on the road to simultaneous and same-frequency full-duplex, and will play a valuable role in connecting the past and the future at the turn of the 5G and 6G eras.

The difference between FDD and TDD will eventually become a thing of the past.