Tom Marzetta and other researchers have been very skeptical about the performance of Massive MIMO using FDD (Frequency Division Duplex) rather than TDD (Time Division.) Massive MIMO requires regular and robust reporting from the cell phone, which they feared would be impractical with FDD. Huawei, ZTE, and now Ericsson have done field trials they believe prove otherwise.
In trial with T-Mobile. Nishant Batra is confident many antenna Massive MIMO will add to performance in FDD frequencies as well as the TDD frequencies beginning to widely deploy. While the performance gain may not be exponential, directing the beam significantly increases capacity even in FDD systems.
Batra expects 3 carrier, 60 MHz TDD Massive MIMO in early 2018. He didn't provide performance estimates, but that class of technology should provide double or better the effective capacity of the "Gigabit LTE" now deploying.
Just as Gig LTE is most often 75-300 megabits, the soon come Massive MIMO will often be far less than the 2-3 gigabits achieved under ideal test circumstances. Batra expects 100 MHz Massive MIMO by yearend 2018. That will provide exceptional performance for networks like Sprint that have enough spectrum.
Ericcson's Kathy Egan writes about FDD, "Capacity increase as compared to 2 transmit antennas is between 1,5x – 3x on average and 1,9x – 5x for cell edge." This is far less than the hope of Massive MIMO proponents. Simply going from two to four antennas would often yield 1.5x and sometimes more. THis may be an inherent limit of the FD technology or might just reflect the early
Simply going from two to four antennas would often yield 1.5x and sometimes more. This may be an inherent limit of the FDD technology or might just reflect the early stage of the research. Maximizing Massive MIMO requires sophisticated algorithms and I expect much better results as the software improves.
The superior results on the cell edge correspond to what Softbank told me last fall. Softbank went from a significant number of cell edge phones not having enough capacity for video to nearly universal performance at an adequate level. Rarely will a mobile phone customer notice a difference if her speed goes from, say, 50 to 75 megabits down. But sputtering video leaves unhappy customers.
Sprint's spectrum at 2500 MHz is generally used for TDD, as is the spectrum from 3400-4200 MHz. Most spectrum below 2100 is allocated to FDD. That applies to almost all the spectrum by many of the major carriers. Blue Danube, working with AT&T, is testing an analog system for Massive MIMO that may also provide robust performance in FDD bands.
I've asked all the companies involved for the actual test data from the field and will share what they send me. The results so far have been somewhat disappointing, perhaps 2X to 3X what's achieved with 4x4 MIMO. That is likely to improve significantly as the systems mature, especially from better beaming-forming algorithms.
Note. An earlier version of this article confused FD (Full Dimension) MIMO with FDD.