What exactly is Massive MIMO?

The next generation of 5G networks are expected to deliver on demands which are substantially higher than today's demands. 5G will be built upon the transformation that is happening in the current releases of 4G Long Term Evolution (LTE). Predominantly, 5G will be a unified network, scalable to meet new requirements, which will thereby connect new industries and devices and will be deployed across all types of bands and technologies. 5G will have to deliver extremely high data rates of multi-gigabits per second, consume extremely low energy, offer very low latency, and superior reliability and be ultra-dense in deployment. This article deals with such a technology which enables us to offer these kinds of advantages - Massive Multiple Input Multiple Output or Massive MIMO. Massive MIMO with is superior spectrum efficiency and energy efficiency can significantly enhance the wireless access for 5G.

Massive MIMO is one of the main drivers of 5G wireless communications. Massive MIMO allows us to scale the benefits of MIMO by several orders of magnitude by employing potentially hundreds of antennas at the base station and the User Equipments (UEs). This is why it is called Massive MIMO. A large number of antenna elements at the base station will help us in achieving beamforming gains, throughput gains, reliability gains and interference management. Massive MIMO ensures that several users are serviced simultaneously using Time Division Duplex (TDD) method while still maintaining a robust link with each terminal. Basically, the more the antennas on either the base station or the user equipment, the more links in between those two and hence better data rate and link reliability. Now let's look at some of the key aspects of Massive MIMO followed by the benefits and limitations of using this technology as an enabler of 5G.

• Channel Estimation in Massive MIMO: 

Channel estimation strategy becomes a key aspect, especially in Massive MIMO systems as the number of antennas is much higher than the conventional MIMO systems. In a Frequency Division Duplex (FDD) scheme, where uplink and downlink have separate frequencies, though the uplink channel estimation remains fairly the same, the downlink estimation becomes a resource-consuming process. This is due to the significant increase in the number of antennas at the BS side which translates into more number of unique pilots being utilized. Hence, this process consumes a large amount of time-frequency resources [1]. A possible solution to this problem lies in using TDD as the strategy which eliminates the need for downlink channel estimation and relies on channel reciprocity [1], [3]. This way TDD transmit training also takes much less time as compared to FDD, which involves more than one step, especially in the downlink channel estimation [3]. Nevertheless, as of today, TDD schemes for channel estimation seem much more promising as far as Massive MIMO is considered.  

• Signal Detection Techniques: 

Signal detection is a critical part of Massive MIMO communication as a large number of users transmit information on the uplink simultaneously. These signals can easily interfere with each other and cause multi-user interference. To avoid this, powerful detection algorithms and techniques must be used in Massive MIMO to be able to recover each user’s signal. This involves recovering channel coefficients and then demodulating the signals.  

Linear as well as non-linear detection methods are considered as strong candidates for Massive MIMO. Some of the linear techniques include the Matched Filter (MF), Zero Forcing (ZF) and minimum mean square error(MMSE) methods [1]. Nonlinear methods can also be used to detect signals in a Massive MIMO system but these methods have a high order of complexity and due to this linear decoding techniques will most likely be preferred. More discussion about these techniques can be found in “An Overview of Massive MIMO: Benefits and Challenges” by L.Lu, et al. (IEEE Journal of Selected Topics in Signal Processing, Oct. 2014, Volume: 52, No. 5, Page: 742-758). 

• Precoding Techniques: 

Precoding, in a nutshell, is a way of estimating the channel condition and minimizing the errors in detection at the receiver. By doing so, we reduce the complexity of mobile phones and make the transmitter more complicated. In Massive MIMO, this is still the fundamental reason to use efficient precoding. As can be expected, non-linear, more complicated precoding techniques offer better performance as compared to linear techniques. But as we incorporate more and more antennas at the BS, it is only logical to use linear techniques like MF and ZF which still give us acceptable performance and keep complexity low [1]. 

• Key Benefits of Massive MIMO: 

1. Robustness: A Massive MIMO antenna system can be incredibly robust since it relies on a large number of antennas using precise beamforming. These antennas can make sure that the signals reach the maximum number of terminals and that the streams interfere constructively at the receiver and destructively at the other terminals.  
2. Fading Resiliency: Massive MIMO systems can combat fading much better than conventional MIMO systems. As the number of the antenna increases considerably in Massive MIMO, the number of multipath reaching the terminal also increases and creates a richly scattering environment. Using Orthogonal Frequency Division Multiplexing (OFDM), the gain can be extracted from these multipath propagation. Hence Massive MIMO relies heavily on NLOS communication where good signal diversity is obtained. 
3. Spectral and Energy Efficiency:  By transmitting beams accurately in directions where terminals are located Massive MIMO can be extremely energy efficient. At the same time, serving tens of terminals simultaneously improves the spectral efficiency [2].
4. Economical: Normally we would expect a dramatic increase in the cost of a Massive MIMO BS because it uses hundreds of antennas. But, Massive MIMO systems may not be a burden on a carrier’s monetary budget. Since Massive MIMO relies on a large number of antennas, each antenna need not be a high power antenna. In fact, according to the power scaling law of Massive MIMO systems, the power of each antenna can be reduced as we increase the number of antennas. Consequently, there is a reduction in the cost of each antenna, but as a whole, the antennas perform better than the conventional MIMO systems [2]. 

• Key Challenges in a Massive MIMO System: 

1. Pilot Contamination Effect:  As we already know that a Massive MIMO system serves tens of users simultaneously, we approach the limits of the number of orthogonal sequences available to be used [1], [2]. Orthogonal sequences like Walsh Codes and Pseudo Noise (PN) codes are used by the pilot signals of each terminal on uplink as well as downlink so that the BS can estimate the channel in both the UL and DL. But these codes are limited since the channel conditions can change quickly. In such a case, every user is not able to get an orthogonal code for its pilot [1]. This results in either non-orthogonal codes being assigned to users in different cells or fewer users being served [1]. When users in neighboring cells have the same pilot codes, the channel estimation for their channels can be corrupted since the BS now receives two codes and uses them to estimate the channel. This results in errors in channel estimation, obviously, and ultimately leads to strong interference [1]. This effect is called the pilot contamination effect and is a prime source of concern among several researchers working on Massive MIMO. One solution to this includes serving only those users who can get orthogonal codes, which directly impacts the capacity of the system. Other solutions include protocol-based methods, better precoding techniques and blind methods which are described more in detail in [1].  
2. Challenges in TDD and FDD modes: TDD has been widely considered as a good option for channel estimation even though FDD gives a better channel estimation. This is mainly due to less efficient precoding and signal detection algorithms which are fairly complicated to implement for FDD. Hence more work needs to be done on precoding algorithms which work efficiently even with FDD mode and can predict the channel conditions more accurately [1]. TDD relies mostly on channel reciprocity, but there might be situations where the channel conditions change in ways that are unknown to the network [2]. In this case, TDD may not be the best way of channel estimation. 
3. Implementing Massive MIMO in 5G Networks: 5G networks are expected to be heterogeneous and implementing Massive MIMO BS in these networks could be a challenging task. These heterogeneous networks will consist of macro, pico, femto and small cells co-existing with each other, while Massive MIMO BS will be communicating with several users at a time. Therefore, interference management must be dealt with greater care. A Massive MIMO BS is expected to have more antennas than any terminal in its serving cell giving it more degrees of freedom and with beamforming, these antennas can be used to focus beams to any terminal and combat interference [4].  

Massive MIMO technology has significant advantages in critical aspects of today’s 4G networks thus ascertaining they are indeed the building blocks for 5G. The article discussed about key aspects of this technology, its benefits as well as challenges establishing that these this technology is successive alternatives to today’s 4G networks. I think this technology can solve the critical issues we face today in terms of developing 5G networks. But on the other hand, a lot of work and research needs to be done to address the challenges in Massive MIMO and make it more feasible. 

References:

[1] L. Lu, G. Y. Li, A. L. Swindlehurst, A. Ashikhmin and R. Zhang, "An Overview of Massive MIMO: Benefits and Challenges," in IEEE Journal of Selected Topics in Signal Processing, vol. 8, no. 5, pp. 742-758, Oct. 2014. 

[2] E. Larsson, O. Edfors, F. Tufvesson and T. Marzetta, "Massive MIMO for next generation wireless systems," in IEEE Communications Magazine, vol. 52, no. 2, pp. 186-195, February 2014. 

[3] T. L. Marzetta, "Massive MIMO: An Introduction," in Bell Labs Technical Journal, vol. 20, no. , pp. 11-22, 2015. 

[4] T. E. Bogale and L. B. Le, "Massive MIMO and mmWave for 5G Wireless HetNet: Potential Benefits and Challenges," in IEEE Vehicular Technology Magazine, vol. 11, no. 1, pp. 64-75, March 2016.