From MIMO to FD-MIMO to massive MIMO – what’s it all about, Part 1
Everyone in the cellular industry will be familiar with the term multiple input and multiple output (MIMO), writes Moray Rumney, the lead technologist at Keysight in the first of a two-part article.
It took root from the start of LTE and later evolved into higher order MIMO and then full dimension MIMO. The new kid on the block is clearly massive MIMO as it shows up so frequently in the 5G narrative. But what do all these MIMO forms mean? Is it better to beamform or beamsteer? Would you recognize an Eigenmode if you bumped into one, and why should you care? What performance benefits should we expect in the traditional frequency bands and new mmWave bands?
Exploit multi-path propagation
The first MIMO specifications showed up in 3GPP at the tail end of the 3G UMTS era but never quite caught on. It was only with the introduction of LTE in 2008 that MIMO started to be taken seriously. The basic principle of MIMO is that when there are multiple antennas at each end of the radio channel, it is possible to cheat the Shannon-Hartley channel capacity theorem by exploiting multi-path propagation.
The goal is to increase data rates by sending multiple data streams at the same time in the same frequency, known as spatial multiplexing. In a single antenna system, sending multiple streams of data just results in interference, but with MIMO, the signals transmitted from each antenna take different paths to the receivers. By applying the right mix of each data stream to each transmit antenna, the signals received at each receiving antenna only see one of the original data streams, which have become unscrambled by the propagation characteristics of the channel.
For example, the simplest channel that can support spatial multiplexing is a 2×2 channel where the path from transmitters 1 and 2 towards receiver 1 are identical, the path from transmitter 2 to receiver 1 is also the same but the path from transmitter 2 to receiver 2 has a 180-degree phase inversion. In such a channel if we have two data streams A and B, and we transmit A+B from transmitter 1 and A-B from transmitter 2, then receiver 1 will see (A+B) + (A-B) = 2A, and receiver 2 will see (A+B) – (A-B) = 2B.
This may appear simple but in reality, such a perfect MIMO channel rarely exists and in any case for MIMO to work, it is necessary to know the actual channel propagation conditions which will vary between each transmitter and each receiver in time, frequency, phase and interference level. In frequency division duplexing (FDD) systems, the channel conditions are estimated by the user equipment (UE) and fed back using channel state information (CSI) messages (consuming uplink bandwidth!), and this process is never exact. In time division duplexing (TDD) systems the base station (BS) can estimate the channel very accurately with no assistance from the UE.
Provided the channel between the BS and UE is not identical for each receiver, it is mathematically possible to precode the transmitted signals with the inverse of the channel as in the simple example above. However, the inability to precisely know the channel conditions coupled with the existence of noise or interference means the practical ability to mathematically recover both streams becomes a function of how well the channel is conditioned to support orthogonal paths and the signal to interference and noise (SINR) level.
The 2×2 principle
For any given signal to interference plus noise ratio (SINR), if 2×2 MIMO is used, the signal power has to be shared between the transmitters meaning the SINR reduces by 3 dB. This illustrates a key principle of MIMO in that it only provides meaningful gain over single input-single output (SISO) when the SINR of the channel gets higher than is necessary to support the maximum SISO data rate. These high SINR conditions occur when the user is near the cell centre, or when interference from adjacent cells is low due to less traffic. It is estimated in a typical urban macro environment that 2×2 MIMO provides around 20% gain over SISO, and not the 2x theoretical gain had the channel conditions been precisely known and the SINR infinite. MIMO gains increase towards the 2x limit as the channel conditions improve.
The 2×2 principle described thus far can be increased to higher orders by adding more antennas at each end of the link. In the original 3GPP Release 8 LTE standard in 2008, 2x and 4x operation was specified, and 8×8 was added later in Release 10. As the number of antennas increases, so it becomes less likely that the channel will support orthogonal transmission paths. These orthogonal paths are known as Eigenmodes. Another MIMO principle is that for UE (smartphones etc.) it becomes increasingly difficult to support higher order MIMO as there is insufficient space to add the required number of receive antennas.
This is evident in that it was eight years after Release 8 specified 4x SU-MIMO that UE with four receivers started appearing on the market. And to take full advantage of that, networks would have to upgrade their BS with 4 Tx/Rx antennas. This is a point of contention in the industry since the practical gains from 4x MIMO, in loaded networks where the gains are most needed, are not yet known.
However, this UE limitation on receiver count does not mean the end of MIMO, rather it leads us to an alternative form of MIMO where instead of transmitting multiple streams of data to one user (known as single-user MIMO – SU-MIMO) to increase peak data rates, the same number of streams can be transmitted towards multiple users, each getting one stream. This is known a multi-user-MIMO (MU-MIMO) and has the effect of increasing cell capacity, but not increasing peak data rates to any one user over the SISO case.