Massive MIMO, which depends on using a large array of antennas, is the keystone technology for realizing the improvement necessary to justify the evolution from 4G to 5G wireless networks.

Fifth generation (5G) wireless access networks are being defined to meet the perpetual growth in demand for capacity and address new use cases and applications in 2020 and beyond. 5G New Radio (NR) targets up to 10Gbps peak data rates per user to offer enhanced mobile broadband (eMBB) services, which represents roughly 100× improvement over 4G networks.

Massive MIMO is a primary means of accomplishing this. The technology is particularly well suited for underutilized TDD (time division duplex) bands below 6 GHz, such as band 40 (2.3 GHz), band 41 (2.5 GHz), band 42 (3.5 GHz), and band 43 (3.7 GHz), many of the unlicensed bands set aside for wireless communications, along with other bands that will be newly allocated for commercial wireless networks.

Massive MIMO is important for enabling dynamic digital beamforming to implement user-by-user beams to theoretically offer full cell capacity to each user, which otherwise is shared amongst users on a time and frequency basis. There is no change required in the existing user equipment to benefit from massive-MIMO-enabled cell towers.

The promise of massive MIMO is so appealing that many of the operators do not want to wait for the completion of 5G NR standards and are considering its deployment on 4G equipment. However, these benefits come with a set of challenges.

The larger footprint and higher power and cost due to multifold increase in system complexity in implementing massive MIMO radios is a major hurdle. Overcoming the challenges will require the integration of the analog signal chain with digital front end (DFE) devices in the radio, along with substantial increase in the signal processing compute power.

Massive MIMO and beamforming
Beamforming is not a new concept and has been around in the cellular market as active antenna systems (AAS) that use static beamforming in the radio as a tradeoff to contain the system cost and complexity. AAS are applicable in coverage limited networks, but today’s congested networks need dynamic digital beamforming to get the maximum possible spectral efficiency improvements.

Massive MIMO with full digital beamforming adds a spatial dimension to frequency and time dimensions to significantly boost spectral efficiency. The resulting SNR (signal to noise ratio) improvements brought about by the array gain and orthogonality of multiple beams means the same time and frequency allocations can be reused by multiple users.

Figure 1 Active antenna systems and massive MIMO

Massive MIMO system: Base station dis-aggregation and functional partitioning
The complexity associated with the massive MIMO architecture mandates a significant design alteration: the disaggregation of the base station to support new functional partitions to manage in-system connectivity bandwidth.

For example, in a 100 MHz 64T64R (transmit/receive) antenna array system, the bandwidth between baseband and radio functions is 230 Gbps, assuming that the baseband and radio functions are implemented using one device each. In reality, systems use multiple devices to implement either 8T8R or 16T16R array DFE radio function modules resulting in more than doubling in-system connectivity bandwidth requirements.

Figure 2 provides a conceptual diagram of a massive MIMO radio system. Digital radio processing blocks implement 8T8R or 16T16R DFE functions with integrated analog to digital converters (ADCs) and digital to analog converters (DACs). This is a must-have to eliminate the JESD204B connectivity links ordinarily required for interfacing digital and analog domains; their elimination will help to lower system footprint, power, and cost.

Figure 2
Conceptual massive MIMO architecture

The beamforming device brings Layer 1 baseband functionality to radio to substantially reduce connectivity bandwidth requirements with the higher-layer baseband functions that can now potentially be virtualized in the mobile edge. Integration, flexibility, and higher compute power are three critical requirements to optimally implement massive MIMO systems and evolve associated beamforming and DFE algorithms to continually improve performance, cost, and power.

Programmable RFSoCs

A 5G NR massive MIMO implementation requires a large number of active signal chains in the radio to connect to each antenna or subset of antennas in the array. These active signal chains, which traditionally comprise data converters, filters, mixers, power amplifier, and low noise amplifier, can lead to significant increase in power, form factor, and cost of the system. The large number of active signal chains in massive MIMO systems increase in system power and footprint, making it difficult to realize commercially viable systems. The costs associated with moving data between the RF front-ends (RFFE) and the DFE is one of the key challenges that must be resolved in 5G; the issue must be addressed at the software, hardware, and system levels.

In order to address this challenge, Xilinx has replaced multiple ADCs and DACs along with many other RF components on the board by integrating direct RF-sampling data converters into the existing 16nm FinFET multi-processing SoC (MPSoC) family of products designed and deployed for radio applications. This new SoC device family, called “All Programmable RFSoC,” monolithically integrates RF sampling data converter technology to provide a fully hardware and software programmable wide bandwidth platform for radio systems.

Based on an ARM-class processing subsystem merged with FPGA programmable logic, the architecture features 12-bit, 4GSPS (gigasamples per second) RF-sampling ADCs, and 14-bit, 6.4GSPS direct RF DACs, along with optimized digital down-conversion and up-conversion signal processing.

Moving RF into the digital domain by integrating RF-sampling data converter technology not only overcomes power, space, and cost disadvantages but enables implementation of wide bandwidth and multi-band systems.

Analog RF in existing radio systems is typically designed to relax discrete data converter specifications. In addition, discrete data converters and analog RF components use older process nodes and are typically optimized for narrow bandwidths. This results in analog RF solutions that are expensive in size, power, and cost for wide bandwidth MIMO and massive MIMO radio systems. Integrating high speed data converters, 6.4GSPS direct RF DACs and 4GSPS RF-sampling ADCs, allows digital RF to be flexible, low power, and wide bandwidth, ideally suited for building MIMO and massive MIMO systems with lower footprint, power, and cost.

[Continue reading on EDN US: 16nm FinFET technology]

Paul Newson is a System Architect for wireless communication systems, Hemang Parekh is a Senior Engineering Manager, and Harpinder Matharu is the Director of the Communications Business, all at Xilinx.

Related articles: