Six Differences Between MU-MIMO and Massive MIMO

Multi-user MIMO (MU-MIMO) is not a new technology, but the basic concept of using multi-antenna base stations (BSs) to serve a multitude of users has been around since the late 1980s.

An example of how MU-MIMO was illustrated prior to Massive MIMO.

I sometimes get the question “Isn’t Massive MIMO just MU-MIMO with more antennas?” My answer is no, because the key benefit of Massive MIMO over conventional MU-MIMO is not only about the number of antennas. Marzetta’s Massive MIMO concept is the way to deliver the theoretical gains of MU-MIMO under practical circumstances. To achieve this goal, we need to acquire accurate channel state information, which in general can only be done by exploiting uplink pilots and channel reciprocity in TDD mode. Thanks to the channel hardening and favorable propagation phenomena, one can also simplify the system operation in Massive MIMO.

This is how Massive MIMO is often illustrated for line-of-sight operation.

Six key differences between conventional MU-MIMO and Massive MIMO are provided below.

Conventional MU-MIMO Massive MIMO
Relation between number of BS antennas (M) and users (K) MK and both are small (e.g., below 10) K and both can be large (e.g., M=100 and K=20).
Duplexing mode Designed to work with both TDD and FDD operation Designed for TDD operation to exploit channel reciprocity
Channel acquisition Mainly based on codebooks with set of predefined angular beams Based on sending uplink pilots and exploiting channel reciprocity
Link quality after precoding/combining Varies over time and frequency, due to frequency-selective and small-scale fading Almost no variations over time and frequency, thanks to channel hardening
Resource allocation The allocation must change rapidly to account for channel quality variations The allocation can be planned in advance since the channel quality varies slowly
Cell-edge performance Only good if the BSs cooperate Cell-edge SNR increases proportionally to the number of antennas, without causing more inter-cell interference

Footnote: TDD stands for time-division duplex and FDD stands for frequency-division duplex.

22 thoughts on “Six Differences Between MU-MIMO and Massive MIMO”

  1. So, with the above difference can we say that Massive MIMO is not supporting FDD? Please suggest.

    1. Massive MIMO of course works both in TDD (its canonical form) and FDD, but it works much better (=more reliably, higher capacity) in TDD because then the uplink-downlink channel reciprocity can be leveraged.

      You may also be interested in this comparison, https://arxiv.org/pdf/1704.00623.pdf

  2. In FDD Massive MIMO system, some experts have proposed a method called ‘Joint Spatial Division and Multiplexing’ to reduce the pilot overhead. What do you think is the shortcoming of this method?

    1. JSDM is a very nice concept, but it is not clear to what extent it can be applied in practice. As said in the original paper “The main idea of JSDM consists of partitioning the user population into groups with approximately the same channel covariance eigenspace”. Basically, we need to divide the world into regions where the covariance matrices are approximately equal, while they are substantially different between regions. The users in a given region will be one group. I’m not sure if that type of partitioning is possible in practice. I think that the covariance eigenspaces change gradually as you move around instead of changing more abruptly when you move from one region to another.

      That said, it is definitely important to take spatial channel correlation (channel eigenspaces) into account in the resource allocation in Massive MIMO.

  3. As a basic principle, what is the difference between Massive MIMO, MU-MIMO and SU-MIMO. Please in a simple manner, my confusion is: SU-MIMO is like 2×2 MIMO today where a single beam covers a particular sector in telecom site and as per scheduler, resource blocks are shooted from antenn in form of analogue signals with a particular time lap (0.5ms for 1RB). Now this shooting of RBs is divided as per application demand by the users. Similarly MU-MIMO is applied as short focused signal beams towards potential users but it is continuous, while 2×2 MIMO is having a time lap between shooting of RBs. Massive MIMO is approach of having transmission of signals from multiple antenna arrays but again here we can get max 8×8 MIMO parallel streams not like for 64×64 streams as it looks for massive MIMO, but these 8x8MIMO will be shooted from one set of antenna array panel in antenna and then 08 different antenna array panels to make total 64×64 antenna arrays so these 8×8 MIMO is shooted from 64×64 antenna array system periodically not parallel. Please elaborate if my concept is wrong. thanks really

    1. To answer these things in detail, I would have to write a book…

      To answer your question briefly:

      SU-MIMO: Transmission from a multi-antenna transmitter to a multi-antenna receiver. One can either send one or multiple simultaneously beams (also called layers or streams) to the receiver. The number of beams is limited by the minimum of the number of transmit antennas and receive antennas.

      MU-MIMO: Transmission from a multi-antenna transmitter to multiple receivers, each having either one or multiple antennas. The transmitter sends at least one simultaneous beam to each receiver. The opposite is also included: Simultaneous transmission from multiple transmitter to a single multi-antenna receiver.

      Massive MIMO: This is a name for MU-MIMO with very many antennas, typically at least 64 antennas at the base station. This base station is capable of sending 64 different beams simultaneously, but in typical operation it will probably use more like 10 beams since a core insight in the design of Massive MIMO is that is preferable to have many more antennas than beams. This makes the beams relatively narrow so there is empty space between them, leading to less interference.

      I recommend my video series about multiple antenna communications:
      https://www.youtube.com/watch?v=OA4viERrlzA&list=PLTv48TzNRhaKz0C-dCAwimXSypV_5UTxg

      And also this video: https://youtu.be/xGkyZw98Tug

  4. Hi, Dr. Bjornson
    In the last part of the table, you claim that by utilizing more antenna elements at the BS, we can enhance the performance of cell-edge users without increasing the inter-cell interference.
    I think if we increase the BS antennas, the power received by users will be increased without increasing the transmit power and for users in the edge of the cell this increased power may cause inter-cell interference too.

    1. If you have a look at the effective SINR expressions in Table 4.1 in Fundamentals of Massive MIMO, you can see how the numerator grows with the number of antennas, while the interference terms are constant. This is because you beamform towards the desired user, while the signal received at other users is generally non-coherently combined.

      What is the reason for you believing that it will work in a different way?

      1. I just thought that by increasing the power because of more antennas at BS both the power received by desired users and the power of other users should be increased the same but if i rightly understood from your answer only the desired users receive the power by beamforming and so the interfrerence is not increased by utilizing more antennas at BS.

  5. I have a doubt how the point to point MIMO system works. What I have understood as Point to Point MIMO system is that the Base station has multiple antennas and user equipment also has multiple antennas. But how these operate in a cell? I was wondering is it like that a few antennas are dedicated to particular users equipment?

    1. The general rule for MIMO is that one always use all the antennas to transmit to all the users. In a point-to-point MIMO system, the user has multiple antennas and there is only one user. If the channel is sufficiently rich, one can send and receive multiple layers of data at the same time by sending them in different directions. All these layers are transmitted from all antennas and received on all antennas. However, one uses transmit precoding and receive combining to extract the different layers in the baseband processing. You can view it as directing each layer in a different spatial direction and direct the listening in different directions as well.

      Here are two videos about point-to-point MIMO:
      https://youtu.be/CXDr-glqzx8
      https://youtu.be/Yr9r-vupmeo

  6. Dear Sir
    Resource scheduling is necessary for multiple access.
    Spatial multiplexing (SM) and SDMA use same subcarriers for different stream or different users respectively.
    1) When do we use different subcarrier for different users and when use same subcarriers for different stream/users (SM/SDMA)?
    2) Massive MIMO achieves sharp beamforming. But multiple statistically indepentdent path is exploited for diversity, and rich scattering (different signal direction) is exploited in SM/SDMA. Doesn’t sharp beam make these channel conditions worse?

    Regards
    Micael

    1. 1) If one does SDMA, then there are by definition multiple users that are scheduled at the same time on the same subcarrier. To make use of this, we need a sufficiently large number of antennas to separate the users spatially. We typically have more antennas at the base station than users that are spatially multiplexed, to make sure that they can be well separated in space.

      2) No, more antennas always improve the performance, if they are used properly, compared to a case with having fewer antennas.

  7. In figure 2 that shows the beams transmitted to users, does each beam have its own MIMO configuration (e.g., one beam is using 2×2 MIMO, another is using 4×4 MIMO)?

    1. The figure only illustrates one beam per user, which means one layer of data. However, there are situations when one can transmit multiple layers per user, if the channel supports that (e.g., by using dual polarizations).

  8. Hello, Dr. Björnson.

    Why does mMIMO consider that the number of BS antennas is much greater than the number of users?

    1. To protect the users from interfering with each other. More antennas leads to narrower beams, thus the interference leakage between the users will be smaller (and the loss in signal power from applying interference suppression schemes such as zero forcing becomes smaller).

  9. Hi Sir,

    When narrow beams are used for each user with help of beamforming, what will happen if two users are very near/adjacent to each other? How effective will beamforming work? Is there any limit for distance between users? Kindly explain.

    1. This is a good question and the answer depends on the propagation environment. If both users are in line-of-sight to the base station, then the beam width determines how far apart the users need to be. With current base stations, I would guess that the users need to be 10 degrees apart (as seen from the base station). The number of meters that corresponds to depends on the propagation distance. In a non-line-of-sight scenario, it is enough that the users are around five wavelengths apart to get rather different channel realizations. One can then deal with the interference very effectively using zero forcing.

  10. Dear Sir.
    Channel correlation (rank of the channel) limits the capacity of SU-MIMO. MU/massive mimo is the same as SU-MIMO?
    Rank of LoS SU-MIMO channel is close to 1. How about LoS MU/massive mimo channel?
    Assuming rich scattering, any advantages of MU-MIMO (K users) compared with SU-MIMO (K antennas)?

    1. Spatial channel correlation can be either good or bad for the MU-MIMO capacity. It depends on whether the users have distinctly different dominant subspaces of their correlation matrices, or overlapping. I recommend you to read Section 2.4 in Massive MIMO Networks: massivemimobook.com

      A MU-MIMO channel normally has full rank. Even if each user has a LOS channel, the spatial directivity will be different since users are spread out.

      Regarding rich scattering, the SU-MIMO capacity is higher than the MU-MIMO capacity, if the channel matrices have the same distribution and the total transmit power is the same. The key reason is that the MU-MIMO case has restrictions on which antennas are allowed to help each other out.

Leave a Reply

Your email address will not be published. Required fields are marked *