Further Enhanced ICIC (FeICIC)

Guest post by Faris Alfarhan*

In an earlier post, R10-LTE enhanced inter-cell interference coordination (eICIC) techniques for heterogeneous networks were discussed, along with the concept of small cell range expansion. The purpose of cell range expansion is to offload more traffic from macro cells to small cells and hence achieve larger cell splitting gains. By adding a cell selection bias, the service area of small cells increases and more users are offloaded to small cells. The need for heterogeneous networks interference management schemes stems from the fact that users in the small cell range expansion area are vulnerable to stronger interference signals than useful signals from the associated serving small cell. In the previous post, it was explained how time domain partitioning based eICIC schemes – known as Almost Blank Subframes (ABS) – could be used to control the interference on the data channels in the range expansion region. Further, carrier aggregation based techniques – known as Cross Carrier Scheduling – could be used to control interference on the control channels (such as the PDCCH, PCFICH, and PHICH channels). However, R10 eICIC schemes did not address interference control on cell-specific reference signals (CRS), which cannot be blanked in order to ensure backward compatibility with R8 and R9 UEs. In this post, R11 improvements to eICIC schemes are discussed, along with the shortcomings of R10 eICIC schemes. First, the concept of Reduced Power Almost Blank Subframes (RP-ABS) is explained along with its advantages over ABS. I then discuss the R11 techniques of Further enhanced ICIC (FeICIC) to control the interference on CRS resources.

Reduced Power Almost Blank Subframes – ABS requires the macro cell to completely blank the transmit power on PDSCH resource elements in almost blank subframes. This effectively translates into resource loss that is a result of the TDM split of the LTE radio frame into eICIC subframes and non-eICIC subframes. While the small cell could use eICIC and non-eICIC subframes, the macro cells could only use non-eICIC subframes to schedule its users. With reduced power almost blank subframes on the other hand, the macro cell doesn’t completely blank the power on eICIC subframes, as it could use these subframes with reduced power to serve cell center users in the macro cell. This of course requires intelligent scheduling and coordination between the macro cell and the coordinated small cells. The amount of power reduction could be static, for example the power reduction in eICIC subframes could be equal to the range expansion bias in dB, or dynamic, which is more optimal but requires better coordination. The capacity gain from using reduced power subframes over almost blank subframes depends on the ratio of eICIC subframes to non-eICIC subframes within a radio frame and the intelligence of the scheduling and coordination. With static RP-ABS, the split of eICIC subframes to non-eICIC subframes is usually fixed and the amount of power reduction is fixed as well for all eICIC subframes. With dynamic RP-ABS, the small cell scheduler must communicate with the macro cell scheduler to exchange the amount of eICIC resource blocks and the corresponding required power reduction on each eICIC resource block. The macro cell scheduler then allocates the RP-ABS resource blocks to cell centers, where the quality of RP-ABS resource blocks are ranked in terms of fading and power reduction then allocated to cell center to achieve the desired quality of experience determined by the scheduling goal. A typical choice for the power reduction per user is the difference in received reference powers between the interfering macro cell and the small cell. In order to have effective coordination, backhaul X2 links between the macro cell and the coordinated small cells must have relatively low latencies. The good news is that only coordination of scheduling decisions is the required, which doesn’t require a lot of backhaul bandwidth but requires low latency. As the degree of coordination of scheduling decisions nears perfection, the described dynamic RP-ABS scheme becomes almost identical to the Coordinated Scheduling feature in Coordinated Multipoint (CoMP). It is worth noting that achieving perfect coordination between macro and small cell nodes requires a centralized coordination structure (e.g. cloud RAN). The RP-ABS implementation details are not standardized by 3GPP, as they are vendor specific.

An illustration of the difference between ABS and RP-ABS.

CRS Signal Interference – While R10 eICIC schemes handle the macro cell interference on the data and control channels in the range expansion region, they do not address interference handling of CRS signals (Cell-specific Reference Signals). In eICIC subframes, the CRS must still be transmitted to support UE measurements and reporting of legacy UEs. CRS interference could be detrimental for the demodulation of various data and control channels. With reference signals frequency hopping, the CRS symbols of neighbouring cells are classified as “non-colliding” and CRS symbols interfere with data and control symbols. Without reference signals frequency hopping, CRS symbols of neighbouring cells are classified as “colliding” and CRS symbols interfere with CRS symbols of neighbouring cells, which leads to poor channel estimation. With R11 FeICIC schemes, both cases of CRS interference can be either avoided by the transmitter or removed by the receiver.

The receiver based FeICIC approach relies on interference cancellation at the UE to eliminate the dominant CRS interference, which requires estimating the dominant interfering CRS at the receiver. This should not be problematic for the UE in the case of “non-colliding CRS”, since the UE is able to distinguish the strong interferers then select the dominant interferer. Once the UE estimates the CRS of the strongest interferer, it subtracts it from the received signal. It is worth noting that the UE could perform this process for multiple interferers, but it would require an iterative process to subtract each interfering signal. However, it is not advisable to cancel the interference from weak interferers, as it would difficult to estimate such signals; weak CRS interference does not significantly contribute to the overall interference, so there is little harm in neglecting it at the receiver. Therefore, cancelling the dominant interfering CRS brings the bulk of the benefit. In the case of “colliding CRS”, the UE may find some trouble estimating the interfering reference signals. Transmitter based assistance is important in such case, as the UE requires information from the network regarding which cells are suitable for CRS cancellation at the UE.

The transmit based FeICIC approach rely on muting PDSCH resource elements within a small cell ABS that experiences strong interference from macro CRS. In addition, rate matching is then applied to compensate for the minor PDSCH resource loss by changing the rate of coding and rate matching. The approach is illustrated in Figure 2 for the non-colliding CRS case. In such case, the transmitter mutes PDSCH resource elements vulnerable to strong CRS interference. However, in order to support legacy UEs, this type of muting is not currently feasible for the downlink control channels, as PDCCH resource could not be muted or rate matched.

Transmit-based FeICIC CRS interference avoidance in almost blank subframes.

In a 3GPP contribution, Ericsson illustrates the throughput capacity gains achieved from using reduced power subframes over almost blank subframes. The study shows that ABS can only provide small capacity gains in highly loaded non-full buffer traffic profiles, whereas RP-ABS could provide substantial gains in system performance over ABS. The study also shows the adverse effect of not handling CRS interference in the range expansion region. The simulations show that suppressing CRS interference from strongest neighboring cell can significantly improve the cell edge throughput.

* Faris is wireless systems engineer in the research and specifications team at InfoVista. His domain of interest and expertise include radio access network design and optimization, performance simulations, and advanced technologies.