Over The Air Bandwidth Aggregation Challenge At Baseband Processing

One of the main challenge of today’s Massive MIMO testbeds resides in the need to aggregate the whole set of RF channel baseband digitized samples within a single and common baseband processing engine, a key requirement for any Massive MIMO algorithms implementation.

The maximum data that can be managed by the baseband processing is defined as the system throughput, and contributes to defining the maximum RF over-the-air (OTA) bandwidth that the testbed would be able to process for real-time scenarios.

Existing testbeds have been constrained by aggregation issues due to fixed, switch backplanes or network infrastructure with limited system throughput….thereby limiting the amount of "real world" Massive MIMO waveform propagation scenarios that could be investigated, while being limited to a few OTA MHz of real-time bandwidth.

To better understand the throughput demands which must be met by today’s Massive MIMO testbeds, Table 1 illustrates the over-the-air RF-to-baseband processing for three scenarios: the throughput required for 1) an 8×8 antenna 4G/WLAN MIMO system, 2) a 100×100 Massive MIMO testbed using 4G/WLAN radios (20 MHz BW), and 3) a 100×100 Massive MIMO testbed using the 5G/802.11AC radios (100 MHz BW) of tomorrow.

Wireless Generation

Over The Air

RF BW (MHz)

Min. Baseband Data Resolution (Bits)

# Antennas

Minimum Required System Throughput (Gbps)

4G/WLAN

20

12

8

3.8

4G/WLAN

20

12

100

48

5G/802.11AC

≥100

12

100

≥240

Other advanced

≥250

12

100

≥600

Table 1: Massive MIMO Testbed Throughput Requirements
 

To better understand the limitation imposed by existing switch PCIe or Network solutions (which enforce an N:1 aggregation topology to the central baseband processing unit), the following table compares them to Nutaq's TitanMIMO multi point-to-point Aurora-4x based interfaces (which uses an N:M aggregation topology to the central baseband processing unit).

Minimum Required System Throughput (Gbps)

10 GigE Throughput (Gbps)

PCIe 4x Gen2 Throughput (Gbps)

7x Aurora-4x Throughput (Gbps)

16x Aurora-4x Throughput (Gbps)

48x Aurora-4x Throughput (Gbps)

3.8

10

20

112

256

768

48

10

20

112

256

768

≥240

10

20

112

256

768

≥600

10

20

112

256

768

Table 2: Throughput Capabilities of PCIe, 10 GigE, and Aurora-4x Point to Point
 
From these tables we can see Nutaq’s TitanMIMO series is the only Massive MIMO testbed that allows real-time baseband processing of more than 20 MHz OTA bandwidth on 100×100 systems, providing a baseband processing engine which can scale to support OTA bandwidths up to 250 MHz. 
 

The shortage in computing resources

As the physical layers which would benefit from the Massive MIMO technology (such as 5G) aren’t yet standardized nor implemented, we cannot know the exact computation resources needed for a 100×100 testbed implementation.

To ensure the future of a Massive MIMO testbed, and maximize the return on investment, it is important to avoid having the baseband processing capabilities be limited i.e. having a central baseband processing engine which is limited to a single processor which is incapable of expansion or upgrades. More importantly, expansion of the central baseband processing engine can’t be performed simply by adding more processor cards in a backplane infrastructure. The same rule of “system throughput” must apply for the inter-processor communication on top of the channel aggregation issue. This is crucial for proper Massive MIMO algorithm implementation, since each added processor within the central baseband processing unit must access information from all RF channels.

To address the need for "unlimited" computing resources, Nutaq's TitanMIMO testbed provides seamless baseband processing upgrades through the addition of FPGA clusters (8x FPGAs) that are interconnected via a mesh topology using multiple high speed point multigigabit transceivers. These clusters can be stacked together on a mesh backplane infrastructure, ensuring future 5G algorithms will have the necessary processing power.

 

Cost effective system upgrades

Designed according to the latest industry modular telecom standards (AMC / MTCA / ATCA / FMC), the TitanMIMO series allows seamless system upgrades through either easy card addition or replacement.

Additionally, the baseband processing cards for the TitanMIMO testbed can be reused throughout the available model series, enabling 80% system cost reuse as the FMC radio front ends are upgraded to future 5G requirements.