Massive MIMO Rocks      Back to

Massive MIMO Feb 2018

Dark blue: Building actively: China Mobile, Softbank Japan, Bharti India, 
Jio India, Vodafone India, Singtel, Globe Phillippines

Dark green: Announced: DT, FT/Orange, BT, Sprint USA, Qatar, Verizon USA, T-Mobile Netherlands, 

Light green: Talking: Vodafone England, Vodafone Turkey,
Safaricom Kenya, Telekom South Africa   


multi user channel estimation 230

Large gains in wireless performance are possible with the right software.

Interference between cells and multiple signals seriously limits the performance of wireless networks. The problem becomes worse with today's very dense networks. Massive MIMO addresses that challenge with beamforming and intelligent design. Many of the theoretical problems remain open. The available solutions require enormous computing power.

Perfect solutions would require orders of magnitude more computing power than available today, but significant results are practical today.

Joe Farkas of Collision Communications believes they have particularly effective software that yields results approaching the ideal.

They can optimize multiple sites with multiple users with reasonable computing power.

Under certain circumstances, Collision claims they improve performance dramatically, delivering "98.8% on the Uplink and 84.8% on the Downlink." Farkas believes current industry techniques only deliver 66.6% and 54.2%." 

I've no doubt these are serious people with valuable results. However, dozens of first-rate engineers at all the vendors are working on these problems. I'd be surprised if anyone is as far ahead as Farkas believes Collision is. I'd be delighted to be proven wrong, of course.

The best results are coming from TDD systems with 64 or more antennas. Improvements are most significant in 5G mmWave. The results in FDD systems are presumably more limited.  

Farkas' answer to doubters is to show their software suite to the equipment manufacturers and it in their scenarios alongside their systems. "We use their model and show results."

Today, most of that work is done in simulations and controlled lab conditions. There aren't that many systems in the field yet.

Both Huawei and now Nokia have designed custom network processors dedicated to tasks like this. The Nokia device is part of their recently announced Reef Shark chipset. (Get specs here.)

Stanford's Andrea Goldsmith has told me finding practical solutions with many signals and transmitters is a crucial problem in communications theory.

Here's how they describe their tools:

Direct Modeling of Interference
By recognizing that interference is of a known structure, we are able to model that interference, providing significantly more interference suppression than traditional approaches.

Multi-User Channel Estimation

Sophisticated multi-user channel estimation approaches obtain near perfect channel estimation for many simultaneous signals, enabling more sophisticated multi-user detection algorithms and beamforming to thrive.

Multi-User Detection

Sophisticated multi-user detection “separates” interfering signals through innovative receiver architectures and algorithms.

Cross-Layer Optimization

MAC layer optimizations, which are particularly valuable given very advanced signal processing, provide for the opportunity to take advantage of new capabilities and to further realize spectral efficiency improvements.