Is it possible to increase the capacity of a mobile network with antenna arrays? This is the question we will answer in this video. We have already seen beamforming and antenna arrays. Let's take a set of antennas lambda/2 apart. By controlling the phase of the signal from each antenna, we can focus the power in a particular direction. The advantage is that we increase the power received in this direction compare to a standard antenna. We reduce the power radiated in other directions. If we consider the neighboring base station, this means that we will reduce the power towards the terminal in question so there will be less interference. The background noise is unchanged since we are not changing the terminals receiver. In short, we have increased the signal to noise and interference ratio SINR. When we apply a formula like Shannon's capacity theorem, we see that the higher the SINR, the higher the real output we can achieve. But can we do even better? Let's suppose we have two users, A and B, who are in quite different areas of the cell. We can imagine a first data transmission to A, focusing the energy in the direction of A and canceling it in the direction of B. We can also imagine transmission to B canceling, transmission in the direction of A and superimpose the two. We can demonstrate that with n antennas at the base station and one antenna for example, on each UE, we can transmit up to n simultaneous streams. This is called MU-MIMO for Multi-user Multiple Input, Multiple Output. Now, let's look at a little vocabulary equation. The radio propagation channel is like a black box with inputs and outputs. The number of inputs corresponds to the transmission part, namely the number of transmitting antennas. The output corresponds to the receiving port. The number of receiving antennas, M for multiple, S for single. We can imagine all the different possible combinations. SISO, Single Input, Single Output, the simplest, MISO, SIMO or MIMO. When we have several transmitting antennas and several receiving antennas. What we have just seen is the case of Multi-user MIMO. Here, the global propagation channel corresponds to a base station with several antennas and several different terminals. This is for downlink transmission. But be careful, multi-user is not always possible. Here, we have a case where it works well, because A and B are in clearly different directions. But here, if A and B are in the same direction, we necessarily have the same signal radiated to A and B. We cannot separate them because we are dealing with propagation in free space. But this is not always true. What we have assumed so far is line-of-sight propagation. There is no obstacle between the transmitter and the receiver, or there is no obstacle between the transmitter and the receiver, or in the general vicinity. This is the case when we are completely in a vacuum or when we are in a flat rural area. But it's rarely true, quite often, there are buildings, for example, in urban areas, which causes reflections, diffraction, reflection off the walls of the building. Sometimes the direct path is not obstructed, there is no obstacle between the transmitter and the receiver. The received power along this path is often dominant, so it's easy to determine the angle of view for the beam. But sometimes there is no direct path, which is represented here. It's clear that the signal is received at the receiver side from very different directions. It's difficult to know in which direction to direct the beam, this is a case of non-line-of-sight propagation or NLOS. Actually, it's a very subtile question: Are multipaths a friend or an enemy. This is a complex question. Why an enemy? Well because we need to identify the different phase shifts, and to do that, we need to send reference symbols or symbols with known values. This makes it possible to characterize the channel. Since the terminal can move, we can get rapid fluctuations in this propagation channel. But multipath propagation is also a friend, because we will have very different propagation conditions for a very small shift, and this will make it possible to have simultaneous transmissions between a pair of users. Let's look at a very simple example to get a feel for what's behind this. We'll look at a transmitter and three points A, B, and C. When we have only a direct path, we see that the signal at A and C have the same phase. Why that? Because A and C are on the same wavefront. If we add diffraction indicated in red, the overall received signal is the sum of the signal from the direct path and the signal subject to diffraction. If we look between A and C, the wave will travel Lambda divided by two more. We assume that E_1 and E_2 have the same phase on A. For C, E_1 has the same phase, but E_2 undergoes an extra Lambda-over-2 paths, so each will arrive in phase opposition. If we have a diffraction loss, we are going to have a fairly large signal with one phase in A and a significant different phase at C. At B, the phase will also be different. In any case, in multipath propagation, we have different phases at A, B, and C. This is not a rigorous demonstration, but we want to get the idea accross that these different phases make it possible to process a signal that allows us to transmit different streams simultaneously. We assume that with n transmitting antennas and p receiving antennas, it is possible to have up to minimum of n and p simultaneous parallel streams. This is not always possible. It's really depends on the propagation conditions and the way the multipath fit together. We can say, for example, that in free space, we will only be able to transmit a single stream. To conclude, MIMO consists of n antennas on the transmission side, and P on the receiving side, where n and p are greater than one. When considering MU-MIMO, we have a set of antennas at the base station, and it's the set of UEs that form the set of antennas on the receiver side. In single user MIMO or SU-MIMO, we have a set of antennas at the base station and the UE side. If we consider the up link, both modes can co-exist. The maximum number of parallel flows is mean of n and p. But careful, this is a maximum number. There may be propagation conditions where there are fewer (and far fewer sometimes) parallel streams. With four antennas at the base station and two at each terminal, we can expect to add up to two streams per UE and two UE set simultaneously. There is a strong dependency on the propagation channel. A need to periodically estimate the channel, and this is only possible by frequently transmitting reference symbols, symbols that are already known. Massive MIMO involves taking a large number of antenna elements. For example, up to 256, we will arrange them in a square of 16 times 16 or in a cylinder. This is especially feasible above six gigahertz, since there are often size constraints. By taking a very large number of antennas, the beams will be extremely narrow, which means that we will generate very little in interference in the other directions. Cellular system will therefore no longer be limited by interference which was up and including the fourth generation, the main capacity limit of a cellular system. This is a major paradigmatic shift. On the other end, in massive MIMO, even if we have 256 antenna elements, we transmit a relatively limited number of parallel streams, typically eight. The major difficulties are that a huge amount of computing power is needed. Lastly, we have to consider health requirements. By concentrating the flow, we will increase the power in a particular direction, and the operators must ensure that the power of the electromagnetic field does not exceed existing safety standards.