Part II Physical Layer for Downlink 121
12.5 Interference Coordination and Frequency Reuse
One limiting aspect for system throughput performance in cellular networks is inter-cell interference, especially for cell-edge users. Careful management of inter-cell interference is particularly important in systems such as LTE which are designed to operate with a frequency reuse factor of one.
The scheduling strategy of the eNodeB may therefore include an element of Inter-Cell Interference Coordination (ICIC), whereby interference from and to the adjacent cells is taken into account in order to increase the data rates which can be provided for users at the cell edge.
This implies for example imposing restrictions on what resources in time and/or frequency are available to the scheduler, or what transmit power may be used in certain time/frequency resources.
The impact of interference on the achievable data rate for a given user can be expressed analytically. If a userkis experiencing no interference, then its achievable rate in an RBmof subframe f can be expressed as
Rk,no-Int(m,f)=Wlog
1+Ps(m,f)|Hks(m,f)|2 PN
(12.5) whereHks(m, f) is the channel gain from the serving cellsto userk,Ps(m,f) is the transmit power from cells,PNis the noise power and W is the bandwidth of one RB (i.e. 180 kHz). If neighbouring cells are transmitting in the same time-frequency resources, then the achievable rate for userkreduces to
Rk,Int(m,f)=Wlog
1+ Ps(m,f)|Hks(m,f)|2 PN+(
isPi(m, f)|Hki(m, f)|2
(12.6) where the indicesidenote interfering cells.
The rate loss for userkcan then be expressed as
Rk,loss(m, f)=Rk,no-Int(m,f)−Rk,Int(m, f)
=Wlog
⎛⎜⎜⎜⎜⎜
⎜⎜⎜⎜⎝ 1+SNR
1+3
1
SNR+(isPs(m,f)|HPi(m,f)|ksH(m,ki(mf)|,f)|2 2
4−1
⎞⎟⎟⎟⎟⎟
⎟⎟⎟⎟⎠ (12.7)
Figure 12.4 plots the rate loss for userkas a function of the total inter-cell interference to signal ratioα=#(
isPi(m, f)|Hki(m,f)|2$ /#
Ps(m, f)|Hks(m,f)|2$
, with SNR=0 dB. It can
−10 −8 −6 −4 −2 0 2 4 6 8 10 0
10 20 30 40 50 60 70 80 90
α (dB)
rate loss (in %)
Figure 12.4: User’s rate loss due to interference.
easily be seen that for a level of interference equal to the desired signal level (i.e.α≈0 dB), userkexperiences a rate loss of approximately 40%.
In order to demonstrate further the significance of interference and power allocation depending on the system configuration we consider two examples of a cellular system with two cells (s1ands2) and one active user per cell (k1andk2respectively). Each user receives the wanted signal from its serving cell, while the inter-cell interference comes from the other cell.
In the first example, each user is located near its respective eNodeB (see Figure 12.5(a)).
The channel gain from the interfering cell is small compared to the channel gain from the serving cell (|Hks1
1(m,f)| |Hks21(m, f)|and|Hks22(m,f)| |Hks12(m,f)|). In the second example (see Figure 12.5(b)), we consider the same scenario but with the users now located close to the edge of their respective cells. In this case the channel gain from the serving cell and the interfering cell are comparable (|Hks1
1(m, f)| ≈ |Hks2
1(m, f)|and|Hks2
2(m,f)| ≈ |Hks1
2(m,f)|).
The capacity of the system with two eNodeBs and two users can be written as RTot=W
log
1+ Ps1|Hks11(m, f)|2 PN+Ps2|Hks12(m,f)|2
+log
1+ Ps2|Hks22(m,f)|2 PN+Ps1|Hks21(m,f)|2
(12.8) From this equation, it can be noted that the optimal transmit power operating point in terms of maximum achievable throughput is different for the two considered cases. In the first scenario, the maximum throughput is achieved when both eNodeBs transmit at maximum power, while in the second the maximum capacity is reached by allowing only one eNodeB to transmit. It can in fact be shown that the optimal power allocation for maximum capacity for this situation with two base stations is binary in the general case; this means that either both base stations should be operating at maximum power in a given RB, or one of them should be turned offcompletely in that RB [17].
From a practical point of view, this result can be exploited in the eNodeB scheduler by treating users in different ways depending on whether they are cell-centre or cell-edge users.
Each cell can then be divided into two parts – inner and outer. In the inner part, where users experience a low level of interference and also require less power to communicate with the
H1,1 H1,2 H2,2 H2,1
H1,1
H1,2 H2,2 H2,1
(a)
(b)
Figure 12.5: System configuration: (a) users close to eNodeBs; (b) users at the cell edge.
serving cell, a frequency reuse factor of 1 can be adopted. For the outer part, scheduling restrictions are applied: when the cell schedules a user in a given part of band, the system capacity is optimized if the neighbouring cells do not transmit at all; alternatively, they may transmit only at low power (probably to users in the inner parts of the neighbouring cells) to avoid creating strong interference to the scheduled user in the first cell. This effectively results in a higher frequency reuse factor at the cell-edge; it is often referred to as ‘partial frequency reuse’ or ‘soft frequency reuse’, and is illustrated in Figure 12.6.
In order to coordinate the scheduling in different cells in such a way, communication between neighbouring cells is required. If the neighbouring cells are managed by the
PowerPowerPowerPower
Frequency
Frequency Frequency Frequency
Figure 12.6: Partial frequency reuse.
same eNodeB, a coordinated scheduling strategy can be followed without the need for standardized signalling. However, where neighbouring cells are controlled by different eNodeBs, standardized signalling is important, especially in multivendor networks. The main mechanism for ICIC in LTE Releases 8 and 9 is normally assumed to be frequency- domain-based, at least for the data channels, and the Release 8/9 inter-eNodeB ICIC signalling explained in the following two sections is designed to support this. In Release 10, additional time-domain mechanisms are introduced, aiming particularly to support ICIC for the PDCCH and for heterogeneous networks comprising both macrocells and small cells;
these mechanisms are explained in Section 31.2.
12.5.1 Inter-eNodeB Signalling to Support Downlink Frequency-Domain ICIC in LTE
In relation to the downlink transmissions, a bitmap termed the Relative Narrowband Transmit Power (RNTP6) indicator can be exchanged between eNodeBs over the X2 interface. Each bit of the RNTP indicator corresponds to one RB in the frequency domain and is used to inform the neighbouring eNodeBs if a cell is planning to keep the transmit power for the RB below a certain upper limit or not. The value of this upper limit, and the period for which the indicator is valid into the future, are configurable. This enables the neighbouring cells to take into account the expected level of interference in each RB when scheduling UEs in their own cells. The reaction of the eNodeB in case of receiving an indication of high transmit power in an RB in a neighbouring cell is not standardized (thus allowing some freedom of implementation for the scheduling algorithm); however, a typical response could be to avoid scheduling cell-edge UEs in such RBs. In the definition of the RNTP indicator, the transmit power per antenna port is normalized by the maximum output power of a base station or cell.
The reason for this is that a cell with a smaller maximum output power, corresponding to smaller cell size, can create as much interference as a cell with a larger maximum output power corresponding to a larger cell size.
12.5.2 Inter-eNodeB Signalling to Support Uplink Frequency-Domain ICIC in LTE
For the uplink transmissions, two messages may be exchanged between eNodeBs to facilitate some coordination of their transmit powers and scheduling of users:
A reactive indicator, known as the ‘Overload Indicator’ (OI), can be exchanged over the X2 interface to indicate physical layer measurements of the average uplink interference plus thermal noise for each RB. The OI can take three values, expressing low, medium, and high levels of interference plus noise. In order to avoid excessive signalling load, it cannot be updated more often than every 20 ms.
A proactive indicator, known as the ‘High Interference Indicator’ (HII), can also be sent by an eNodeB to its neighbouring eNodeBs to inform them that it will, in the near future, schedule uplink transmissions by one or more cell-edge UEs in certain parts of the bandwidth, and therefore that high interference might occur in those frequency regions. Neighbouring cells may then take this information into consideration in scheduling their own users to limit the interference impact. This can be achieved either by deciding not to schedule their own
6RNTP is defined in [18, Section 5.2.1].
cell-edge UEs in that part of the bandwidth and only considering the allocation of those resources for cell-centre users requiring less transmission power, or by not scheduling any user at all in the relevant RBs. The HII is comprised of a bitmap with one bit per RB and, like the OI, is not sent more often than every 20 ms. The HII bitmap is addressed to specific neighbour eNodeBs.
In addition to frequency-domain scheduling in the uplink, the eNodeB also controls the degree to which each UE compensates for the path-loss when setting its uplink transmission power. This enables the eNodeB to trade off fairness for cell-edge UEs against inter-cell interference generated towards other cells, and can also be used to maximize system capacity.
This is discussed in more detail in Section 18.3.2.
12.5.3 Static versus Semi-Static ICIC
In general, ICIC may be static or semi-static, with different levels of associated communica- tion required between eNodeBs.
Forstatic interference coordination, the coordination is associated with cell planning, and reconfigurations are rare. This largely avoids signalling on the X2 interface, but it may result in some performance limitation since it cannot adaptively take into account variations in cell loading and user distributions.
Semi-static interference coordinationtypically refers to reconfigurations carried out on a time-scale of the order of seconds or longer. The inter-eNodeB communication methods over the X2 interface can be used as discussed above. Other types of information such as traffic load information may also be used, as discussed in Section 2.6.4. Semi-static interference coordination may be more beneficial in cases of non-uniform load distributions in eNodeBs and varying cell sizes across the network.