An Analysis on 5G Implementation Feasibility

Release Date:2017-01-09 By Xiang Jiying Click:

 

This article analyzes a couple of 5G candidate technologies, particularly focusing on their implementation feasibility. It comes up with a conclusion that most of the 5G technologies depend on analytic methods rather than statistical methods to achieve performance improvement at the cost of increased complexity. For example, more antennas in massive MIMO systems may not necessarily bring linearly better performance. Although TDD massive MIMO has technological advantages, precise calibration is required in implementation. Multi-user shared access (MUSA) and filter bank-orthogonal frequency division multiplexing (FB-OFDM) offer a better trade-off between the performance and complexity.

 

Fundamental Methods for 5G Technologies

There are two problem solving methodologies: analytic method and statistical method. Statistical method is used to predict future events based on a large amount of historical data (big data) as well as a series of extensive regularities. Analytical method accurately measures relevant input parameters and calculates unknown events through a precise regularity or an analytic expression.


Both statistical and analytical methods are kinds of mathematical modeling of events. The statistical method is a sub-optimal solution based on simple approximate modeling with less input, while the analytical method is the optimal solution with a large amount of input. Both methods have their theoretical limitation; however, the analytical method has higher theoretical limits and thus achieves better performance. This is determined by its own principle.


The single-channel capacity has almost reached the Shannon limit. If more independent, non-interfering and equivalent channels (or SDMA channels) are built on the same band, in the same time frame and within the same radio propagation space, the capacity can be increased significantly. Theoretically, the statistical method cannot implement SDMA, while the analytical method can do. The amplitude and phase of signals are described in the form of complex numbers, and a separate logical space can be built by using the orthogonality of instantaneous phase.

 

The first issue that the analytical method needs to solve is un-measurable random signal. Randomness is not absolute but relative. An event is an un-measurable random event for the method A, but may become a certain measurable parsing event for the method B. For example, Rayleigh fading was considered as random, un-measurable sensitive events. However, if a coherent method (multi-antenna joint equalization is essentially a spatial coherence) is used, Rayleigh fading becomes parsing events that can be accurately measured. Therefore, the avoidance of Rayleigh fading by using the statistical method has evolved into man-made Rayleigh fading for null steering so as to multiply the capacity.


In the existing radio communications systems still operating upon the classical wave theory, most of the events can be parsed at the cost of complex implementation. 5G is seeking a trade-off between the complexity and performance. In other words, events are parsed at the appropriate time so that maximum performance can be achieved at appropriate costs.

 

As the potential for communications has been gradually exhausted and the theoretical limits have been gradually approached, more analytical methods should be used to achieve optimal performance at the cost of higher complexity. 5G SDMA and anti-interference are such technologies that increase capacity through complex analytical calculation. Non-orthogonal telecom technologies in 5G massive IoT also improve near-far effects and increase the number of accessed users through the complex calculation. However, the filter bank multi-carrier (FBMC) method introduces inter-symbol interference and then eliminates the interference to reduce frequency-domain leakage. The above-mentioned calculations are more complicated than linear calculations. Achieving even small capacity enhancement through the huge, disproportionate analytic calculations will become quite normal in driving 5G technologies.

 

Non-Linear Growth of Massive MIMO Capacity

The massive MIMO technology can achieve SDMA and multiply the capacity. Although in principle the maximum capacity of a single site is directly proportional to the number of antennas. For example, 64 antennas should provide 64-fold capacity, but actual networking capacity is far below the theoretical linear value due to many factors such as noise, non-ideal spatial channels, and measurement errors. Especially when too many antennas are used, non-ideal factors may offset the gain brought by additional antennas. Therefore, the number of massive MIMO antennas is not the more the better. Actual implementation conditions should be considered for the best price-performance ratio.


TDD provides symmetric uplink and downlink, so massive MIMO can be implemented with TDD and even offer better performance by keeping standards transparent (with the same air-interface on mobile phones).


However, there are still some problems with TDD massive MIMO. To leverage its symmetry, TDD massive MIMO has extremely high requirements on accurate calibration. For example, a 2.6 GHz carrier with a phase difference of around 10 degrees accounts for about 1/300 ns calibration error, and no device can directly meet this requirement (GPS can only reach up to several nanoseconds). In this case, a two-level calibration mechanism is needed. In the future, when evolving towards 5G, devices are faced with two challenges: the more antennas the higher requirement for calibration precision, and the higher frequency the higher requirement for absolute calibration precision.

 

Another problem to be addressed for TDD massive MIMO is the number of transmit channels at the terminal side. Since terminals are sensitive to cost and power consumption, more antennas are used for reception while fewer antennas or even a single antenna is used for transmission. This damages symmetry to some extent. Currently, switching transmit is used or the amount of single-user streams is reduced to avoid this damage.

 

Reducing Receiver Complexity for Non-Orthogonal Technology

MUSA is a non-orthogonal 5G technology proposed by ZTE. Traditional telecom technologies such as 4G use the orthogonal method to differentiate users, in which two users cannot share the same degree of freedom (DOF). However, MUSA allocates a code sequence to each user and assigns the same DOF (time, sub-carrier, or space) to these users. A successive interference cancellation (SIC) receiver is needed to identify users and solve near-far effect problems. In this way, the number of accessed users can be increased.


MUSA can use the SIC receiver, which is relatively simple but can significantly improve performance. With its code energy evenly distributed, MUSA provides better coverage under the same circumstances. Furthermore, there are abundant available MUSA code resources, and even if the extended code length is short (as short as 4), there are still thousands of code resources available (complex number value is 9). Therefore, it is easier to eliminate scheduling operations, and a base station can even identify users through blind detection, with its complexity being controllable. This is of great significance to the IoT, for it increases the standby time and makes it possible for a button cell to work for several years.

 

FBMC: A New Filter-Based 5G Waveform Technology

Traditional orthogonal frequency division multiplexing (OFDM) has solved the orthogonal issue between the time domain and frequency domain, but in principle a serious out-band leakage problem still exists. Therefore, a large-enough protective band must be reserved between the OFDM system and other systems. In the time domain, each OFDM symbol is transmitted at a constant power within the symbol period and is orthogonalized. As the spectrum of rectangular waves is a SINC function that features oscillation, it is difficult to control the interference between adjacent systems.

 

5G proposes to change the shape of filters to improve frequency domain performance, which is called generalized frequency bank multicarrier (FBMC). Because the time domain and frequency domain are two mathematical descriptions of the same signal, the two descriptions are eventually equivalent. Following the principle of energy conservation, the more compact the time domain response, the looser the frequency domain (serious out-band leakage). Therefore, theoretically it is impossible to find a method that meets the infinite compactness in both time and frequency domains.


Researchers have proposed a variety of new filtering methods. Their differences are as follows:
● Time-domain windowing, cross-signal convolution, and narrowing each sub-carrier in the frequency domain;
● Time-domain convolution and frequency-domain filter over the entire bandwidth.
In addition to the differences in the time domain, the filter is also divided into complex-domain filtering and real-domain filtering according to the filter function.

 

From the complexity perspective, since all filtering methods introduce inter-symbol interference (ISI) in the time domain, their complexity is higher than that in OFDM. Under the same circumstances, the complexity of real-domain filtering is higher than that of complex-domain filtering.

 

In addition to suppression effects, receiver complexity and its related delay should also be considered. In short, a trade-off is needed between the time domain and frequency domain. FB-OFDM proposed by ZTE uses time domain windowing and complex-domain filtering, which is applicable to all bandwidths regardless of the number of sub-carriers. With good near-end suppression effects and simple receivers, FB-OFDM is fairly a balanced filtering method.