The use of multiple antennas (MIMO) relies on two different principles:
- Improving the SNR (Signal to Noise Ratio).
- Sharing the SNR.
This means that in scenarios where the SNR is low, improving the SNR is the way to go. By using beamforming, Tx diversity and/or RX diversity, the SINR improves greatly.
Beamforming concentrates the transmitted (and/or received) energy in desired direction(s).
Tx diversity achieves diversity against the channel fading by transmitting the information at different times and/or from different antenna locations. Open Loop tx diversity does not exploit any channel information (no feedback from receiver) while closed loop tx diversity uses feedback from the UE in order to maximize the performance.
When the SNR is high, the data rate and spectrum efficiency improve by e.g. increasing the modulation order (e.g. going from 16-QAM to 64-QAM). This gives us more bits/s/Hz.
However, the improvement in throughput and spectrum efficiency as a function of the SNR is logarithmic. This means that the throughput saturates at high SNRs, resulting in an excessive need for power/link budget in order to reach high data rates. In that case, it is better to share the SNR (or energy) between different dimensions (layers). This is achieved by spatial multiplexing, where several data streams (layers) are transmitted on the same physical radio resources (simultaneously at the same frequency and code). The streams are separated in the spatial domain instead.
The data rate only increases logarithmically as a function of the SNR (or SINR – Signal to Interference and Noise Ratio), at a high SNR. This is according to the Shannon theorem
rdata = BW x log2(1+SNR)
(max data rate rdata is equal to the bandwidth, BW, multiplied by the base-2 logarithm of the SNR plus 1).
On the other hand, at a low SNR, the max data rate increases almost linearly. Therefore, it is not efficient aiming only to obtain a high SNR. It is more efficient to try to create several “data pipes” with lower SNR (sharing SNR), which will lead to a multiplication of the maximum achievable data rate with up to the channel rank rmax.
A traditional way of sharing the SNR is actually by spread spectrum techniques, e.g. CDMA, where the transmission is multiplexed over a wider bandwidth. Different configurations of multiple antennas are shown in below picture. These include SISO (Single Input Single Output), MISO (Multiple Input Single Output), SIMO (Single Input Single Output) and of course, MIMO (Multiple Input Multiple Output). The naming convention refers to input/output of the radio channel. This means that the transmitter antenna(s) correspond to input to the radio channel and the receiver antenna(s) reception correspond to the output of the radio channel.
With multiple antennas at the transmitter and only single antenna at the receiver (referred to as MISO) it is possible to obtain so called beamforming. With this method the transmission signal is steered in a beneficial direction (typically towards the UE). This is accomplished by adjusting the phase (and sometimes amplitude) of the different antenna elements by multiplying the signal with complex weights. This method increases the SNR (Signal to Noise Ratio) and then the capacity.
With this configuration it is also possible to achieve Transmit Diversity. This is done by transmitting time-shifted copies of the signal and thus achieving diversity in the time-domain. This method also increases the SNR.
With multiple antennas at the receiver (SIMO or MIMO), it is possible to use receive diversity. A combining method (typically MRC – Maximum Ratio Combining) is applied to increase the SNR of the received signal.
With multiple antennas at both transmitter and receiver, it is possible to use all of the above mentioned methods.
However, with multiple antennas at both transmitter and receiver, it is also possible to achieve spatial multiplexing, also referred to as MIMO. This method creates several layers, or “data pipes” in the radio interface. The maximum number of layers created depends on: The radio channel characteristics and the number of tx and rx antennas.
The channel rank equals the maximum number of layers that the radio channel can support. Effectively, the maximum number of layers that can be used is equal or less than the minimum number of antenna elements at the TX or RX side or the channel rank. The number of layers used for transmission is referred to as the transmission rank.
At optimal circumstances, The data rate increases by the number of layers.
In a simple beamforming example, the complex weights adjusts in order to align the carrier phase in such a way that the individual antenna elements’ signals add constructively at the receiver side. With two antenna elements we need two complex weights, here denoted w1 and w2. These weights adjust with the support of some kind of feedback from the receiver (in this case the UE), in order to maximize the reception.
With multiple antennas at both tx and rx side and multiple layers, the complex weights will form a matrix. In case of 4×4 MIMO and four layers, a 4×4 matrix will be needed. This means that each layer will have its own complex weight vector of length equal to the number of antenna elements.
The LTE specifications support the use of multiple antennas at both transmitter (tx) and receiver (rx). MIMO (Multiple Input Multiple Output) uses this antenna configuration.
In the first release of LTE it is likely that the UE only has one tx antenna, even if it uses two rx antennas. This leads to that so called Single User MIMO (SU-MIMO) will be supported only in DL (and maximum 2×2 configuration).
There are seven different transmission modes in LTE. Switching between the modes is done by RRC signaling.
- Mode 1 (”Single antenna port, port 1”)
o One antenna
o Used for classical beamforming without precoding feedback
- Mode 2 (”Transmit Diversity”)
o SFBC
o 2 or 4 tx antennas
- Mode 3 (”Open loop spatial multiplexing”)
o 2 or 4 tx antennas
o CQI & RI feedback
o Tx schemes:
– Tx diversity
– Large delay CDD
- Mode 4 (”Closed Loop spatial multiplexing”)
o 2 or 4 tx antennas
o CQI, PMI & RI feedback
o Tx schemes:
– Tx diversity
– CL SM
- Mode 5 (”Multi User MIMO”)
o Two UEs can be scheduled in the same RB
o Tx schemes:
– Tx diversity
– MU-MIMO
- Mode 6 (”Closed loop spatial multiplexing, single layer”)
o As mode 4, but with RI hardcoded to 1
o Tx schemes:
– Tx diversity
– CL SM
- Mode 7 (”Single antenna port, port 5”)
o Used for classical beamforming without feedback.
Every transmission mode can use one or more transmission schemes. Typically, the transmission mode is set-up at session establishment and not changed during the session, while the transmission scheme is dynamically decided every TTI.
To achieve high SINR, less likely that cell edge UEs use multiple layers, go for beamforming or diversity instead. Also less likely that cell edge UEs use MU-MIMO.
Multiple layers means that the time- and frequency resources (Resource Blocks) can be reused in the different layers up to a number of times corresponding to the channel rank. This means that the same resource allocation is made on all transmitted layers.
In MU-MIMO, the different UEs that are spatially multiplexed are allocated the same resource blocks, but spatially separated by the antenna arrangement and associated processing. This means that each UE does not experience a data rate multiplication as with SUMIMO. Instead the cell benefits from the reuse of the resources.
LTE uses precoder based MIMO operation. A precoder, which is a complex matrix of size equal to the number of layers times the number of transmit antennas, is used to form the “beams” for the different layers separately.
For Spatial Multiplexing:
- Multiple parallel data streams (Higher data rates).
- One or two code words.
- Number of layers ≤ Number of antenna ports.
- Pre-coding weights based on UE feedback (beamforming).
For Transmit Diversity (”open-loop”):
- Transmission of same information from multiple antenna ports (Diversity).
- One code word.
When spatial multiplexing is used, maximum two codewords (transport blocks) per TTI will be used. When no spatial multiplexing is used, only one codeword per TTI will be used.
Given the minimum number of transmit or receive antennas, the transmission rank can at maximum be equal to that minimum number. For example, if four transmit antennas and two receive antennas are used, and then the transmission rank cans one or two. If the transmission rank is lower than the channel rank (i.e. there are more antennas than layers), then the remaining antenna ports can be used for beamforming at the same time as the spatial multiplexing.
The number of useful layers is here denoted r, and the maximum number of layers rmax.
In the DL, the UE estimates the spatial properties of the radio channel by measuring the DL reference symbols from the different antenna ports. This estimation is used to give feedback to the eNodeB so that the eNodeB can allocate an appropriate amount of resources make an optimal antenna mapping.
The feedback is the CQI (Channel Quality Indicator), the PMI (Precoder Matrix Indicator) and the RI (Rank Indicator). The CQI reflects the channel quality and is used whether or not spatial multiplexing is used. The RI indicates the number of useful layers, as estimated by the UE and the PMI indicates the precoder matrix that the UE considers as the best (gives the highest estimated SINR).
The UE will measure the different reference signals, which are transmitted separately on the different antenna ports in DL, and estimate the optimal precoder weights. The UE will choose a precoder matrix from a finite codebook, that consists of a number of precoder matrixes. Only the index of the chosen precoder matrix (Precoder Matrix Index – PMI) and Rank Indicator (RI) will be signaled to the eNB, together with the CQI.
The eNB does not have to follow the UE “recommendation” of precoder matrix, since the eNB might consider another matrix more appropriate when taking other UEs in the cell into account.
