Interference is currently the largest impediment to increasing spectral efficiency in cellular networks.

A typical LTE site, with 4 receive antennas and without any interference, could achieve 15.1 bps/Hz spectral efficiency on the uplink, leading to 301.5 Mbps data rate across a 20 MHz bandwidth. Unfortunately, real deployments operate at only 1.5 to 2.0 bps/Hz [1, 2, 3], meaning they only achieve 13% of their potential capacity.

The maximum achievable uplink spectral efficiency is easily derived from the 3GPP LTE specification, and is calculated from a combination of:

• Maximum transport block size given highest MCS in 20 MHz bandwidth: 75,376 bits
• Maximum number of MIMO layers in 4 receive antenna site: 4
• Subframes per second: 1,000

15.1 bps/Hz ≈ 75,376 bits * 4 layers * 1,000 subframes-per-second / 20 MHz 

Could networks really be underperforming by that much?

Yes! For a typical urban site where the transmit power level of the UE is not the limiting factor to performance, near 15.1 bps/Hz would be attainable IF there were no interference. However, the interference from UEs in adjacent sites and the residual interference from the MIMO layers drives down the number of allowable layers, modulation and coding selection by the scheduler, leaving our networks at 13% peak performance.

Can we avoid interference through sharing the bandwidth in time and frequency?

There are many techniques that avoid interference in modern cellular networks, such as fractional frequency reuse, intercell interference coordination, almost blank subframes, etc. Although sometimes effective at providing basic service to the UEs at the edge of the cell or addressing specific issues with heterogenous networks, avoiding interference typically results in a lower overall spectral efficiency.

What are alternative solutions?

Collision has demonstrated that better receivers can suppress significant interference and result in closing the gap to the theoretical LTE. For example, for a stadium C-RAN scenario that utilizes 2 receive antennas per site, Collision demonstrates 6.15 bps/Hz/site (avg.) — that is 81.6% of the theoretical LTE uplink maximum rate (for 2-antenna sites). These receivers are applicable across various cellular architectures from traditional macrocells, small cells, massive-MIMO sites and C-RAN, although the interference suppression capability does vary depending on the architecture.

[1] 3GPP TR.36.912, Feasibility study for Further Advancements for E-UTRA (LTE-Advanced).

[2] 3GPP TR.36.814, Further advancements for E-UTRA physical layer aspects.

[3] Chen, Wen-Tzu. “Spectral efficiency analysis for LTE networks.” 2014 IEEE Fourth International Conference on Consumer Electronics Berlin (ICCE-Berlin). IEEE, 2014.