23.3 Interference Issues in Unpaired Spectrum
23.3.1 Adjacent Carrier Interference Scenarios
For an FDD cellular system, adjacent channel frequency separation of an interfering transmitter and a victim receiver naturally implies that the interferer and victim are of differing equipment types (i.e. one is a mobile terminal whilst the other is a base station).
Transmitter–receiver interference between one User Equipment (UE) and another, or between one eNodeB and another is avoided by virtue of the duplex spacing.
The same is also generally true in TDD systems if they are time-synchronized so that overlap between uplink and downlink transmission periods is avoided. However, when synchronization is not or cannot be provided, or when TDD systems operate on carriers adjacent to an FDD system, the possibility arises for interferer and victim to be of the same device type. Figure 23.3 depicts a relatively common scenario in which an unpaired spectral allocation is located in a region between an FDD downlink band and an FDD uplink band (as is the case for the 2.5–2.6 GHz UMTS extension band, for example).
Unpaired (TDD) FDD Downlink FDD Uplink
sync FDD FDD/TDD
border FDD non-sync
FDD/TDD border
carrier frequency
uplink TDD TDD downlink
Figure 23.3: Possibilities for Adjacent Carrier Interference (ACI) between FDD and TDD systems.
The vertical arrows in Figure 23.3 represent the desired communication between the base station and the mobile terminal, unidirectional on a per-carrier basis for FDD and bi- directional for TDD. The diagonal dotted lines represent base-station-to-mobile and mobile- to-base-station adjacent channel interference that results from imperfect ACIR.
Two TDD scenarios are shown, synchronized and non-synchronized – the latter encom- passing any general possibility for partial or full uplink-downlink overlap in time between two systems. For the synchronized case, the interference scenarios between the base station and mobile terminal (and vice versa) are the same as for the corresponding FDD-to-FDD adjacent channel cases. However, for the non-synchronized case additional interference paths exist between TDD mobiles and between TDD base stations, represented by the horizontal bidirectional arrows. At the FDD/TDD border regions, these same ‘horizontal’ interference paths exist but are unidirectional in nature.
There are many facets of a deployment which affect the severity of these various interference paths. For example, the locations of the interfering transmitter and victim receiver, as well as the characteristics of the propagation between them, clearly influence the overall coupling that exists. Macrocellular deployments typically use base station transmit antennas mounted on masts located above roof-top level, thereby resulting in an increased likelihood of Line-Of-Sight (LOS) propagation between base stations, with a correspondingly low path-loss exponent. The common use of macrocell base station antennas with vertical directivity and hence high gain can further worsen this situation. These aspects are less problematic for microcellular and dense-urban deployments in which LOS propagation between base stations is less likely due to their antennas being located below rooftop level.
Coupling between mobile terminals is often mitigated by the surrounding local clutter, and due to the lower antenna gain in the terminals. One typical scenario is depicted in Figure 23.4 in which non-co-located macro base stations have some potential to exhibit stronger mutual coupling than between the mobiles which they respectively serve. However, the figure also shows that it is not always possible to rely on local clutter to provide the necessary isolation between terminals, due to the fact that when terminals are closely spaced (for example, in the same office or café), LOS propagation again becomes more likely and the potential for interference is increased – as a result of both the lower path-loss exponent between the terminals and the small physical separation between them.
The base-to-base and UE-to-UE interference scenarios that are particular to unsyn- chronized TDD deployments and to TDD deployments adjacent to FDD deployments are reviewed in more detail in the following two subsections.
23.3.1.1 Base-Station to Base-Station Interference
Base stations of relevance to a particular base-to-base interference scenario may be either co-located (i.e. antennas mounted at the same cell-site) or non-co-located. Nevertheless, base-to-base interference is generally deterministic. This is because the locations of the base stations are fixed, and furthermore the link adaptation strategy typically employed for the LTE downlink usually results in all available transmit power being used to maximize the throughput of the link. It is therefore reasonable to analyse the interference assuming full transmit power from each base station.
Strong coupling potential, low path-loss exponent e.g. LOS
Weaker coupling potential, high path-loss exponent Local
clutter Local
clutter
spaced mobiles (e.g. same office) Low path-loss exponent for LOS closely
Figure 23.4: Typical RF interference scenario for a TDD system.
In general, systems operating on adjacent carriers may use differing deployment topolo- gies and cell sizes, giving rise to varying distances between the base stations on the adjacent carriers. We therefore consider the worst-case base-to-base distance only, and assume that at some point in the network the transmit and receive antenna patterns are aligned to provide maximum gain at this worst-case distance. The co-channel Power Spectral Density (PSD) of the received interference at the victim base station antenna connector can then be written as:
PSDRx=PSDTx+2GBS−ρBS-BS(x0) (23.1)
where PSDTxis the transmitted PSD at each interfering base station antenna connector,GBSis the antenna gain at each base site,ρBS-BS(x) is the path-loss between base sites as a function of distance xin metres and x0 is the worst-case (smallest) distance between base sites of different carriers.
In order for the inter-system interference to have only a minor effect, one can assume that the PSD of the interference after benefiting from any available ACIR should be of the order of PSDN+NFBS−6 dB or less if it is to produce no more than a 1 dB desensitization of the base station receiver (where NFBSdenotes the base station noise figure and PSDNis the PSD of thermal noise, e.g.−174 dBm/Hz at typical temperatures):
(PSDRx−ACIR)≤(PSDN+NFBS−6) dB (23.2) Note that in the case of an SC-FDMA3uplink victim receiver, consideration needs to be paid not only to the ACIR averaged over the system bandwidth, but, more challengingly, to the localized frequency resource blocks located closest to the interfering carrier, especially if the important uplink control signalling on the Physical Uplink Control CHannel (PUCCH) is to be protected (see Section 16.3.1). In general, however, the ACIR requirement varies directly with PSDTx:
ACIR≥PSDTx+2GBS−ρBS-BS(x0)−PSDN−NFBS+6 dB (23.3)
3Single Carrier Frequency Division Multiple Access.
In order to arrive at an ACIR requirement, we must therefore know the transmitted PSD and the intervening path-loss. To do so, it is reasonable to assume that the base station transmit power capabilities are dimensioned in order that each of the two systems are interference-limited on the downlink (at least this can apply for small- and medium-sized cells without exceeding the eNodeB output power capabilities). We therefore assume here that the spectral density of the downlink signal for 95% of the total area isγDLdB larger than the spectral density of the thermal noise in the UE receivers. Thus
PSDTx=PSDN+NFUE+γDL+ρBS-UE−GBS (23.4) where NFUEis the noise figure of the UE receiver,γDLis the received downlink signal-to- noise ratio in dB which is exceeded at 95% of the UE receivers, andρBS-UEis the 95-percentile path-loss between UEs and their serving eNodeBs in the interfering network.
For ease of representation, an empirical approximation toρBS-UEis applied here specific to this particular example deployment: let σ be the standard deviation of the log-normal shadow fading between eNodeB and UE andLbrepresent the additional building penetration loss (assuming indoor coverage); then withρBS-UE(x) denoting the path-loss between base stations and UEs separated by distancexmetres, we can write
ρBS-UE≈ρBS-UE(0.58s)+0.7σ+Lb (23.5) Substituting Equations (23.5) and (23.4) into (23.3) we obtain an approximate expression for the necessary ACIR to maintain an acceptable adjacent channel interference level at a victim base station:
ACIR≥NFUE−NFBS+γDL+ρBS-UE(0.58s)+0.7σ+Lb+GBS−ρBS-BS(x0)+6 dB (23.6) This is plotted in Figure 23.5 as a function of the smallest inter-carrier eNodeB–eNodeB separationx0for several selected values of the inter-site spacingsbetween co-channel base- stations on the interfering carrier under the following assumptions:
• uniform base station deployment on the interfering carrier;
• free-space propagation between eNodeBsρBS-BS(x);
• ρBS-UE(x)=128.1+37.6 dB, from [2];
• γDL=6 dB, NFUE=9 dB, NFBS=5 dB,GBS=14 dBi,Lb=20 dB.
The transmit PSD of the eNodeB (PeNB) for 95% coverage is also listed in the legend for each Inter-Site Distance (ISD).
For a given ISD, the required ACIR naturally decreases as the worst-case separation between interfering eNodeBs is increased. Notice, however, that the base-to-base problem worsens significantly as the ISD in the interfering network increases, due to the fact that the path-loss exponent from eNodeB to UE is higher than the path-loss exponent between eNodeBs. The transmit power needed by the interfering eNodeB to reach UEs at its cell edge increases at a faster rate than can be compensated by the path-loss to a victim eNodeB receiver at the same cell-edge location. It should be remembered, however, that this is representative of a macrocellular scenario with eNodeB antennas mounted above rooftop level; for the smaller cell sizes (characteristic of microcells) the eNodeB-to-eNodeB situation
0 100 200 300 400 500 600 700 800 60
70 80 90 100 110 120 130
Worst-case base-to-base separation x0 (m)
Required ACIR (dB)
s=250m (PeNB=16.2 dBm/5MHz) s=500m (P
eNB=27.5 dBm/5MHz) s=750m (PeNB=34.1 dBm/5MHz) s=1000m (P
eNB=38.8 dBm/5MHz) s=1250m (P
eNB=42.5 dBm/5MHz)
Figure 23.5: Required ACIR for 1 dB desensitization in an eNodeB-to-eNodeB interference scenario.
will be significantly improved by the higher propagation exponent between eNodeBs whose antennas are mounted below rooftop level.
For co-sited eNodeBs (i.e. very small x0), a Minimum Coupling Loss (MCL) value of 30 dB has been used to replace 2GBS−ρBS-BS(x0) in Equation (23.1), resulting in
ACIRco-siting≥PSDTx−MCL−PSDN−NFBS+6 dB (23.7) For wide area base stations, the ACIR required for co-siting can rise towards a challenging 120 dB or so. This problem, however, is not new or specific to LTE and has been encountered previously for FDD/TDD coexistence in WCDMA. Some practical solutions to this problem have been documented in [3], in which RF bandpass cavity resonator filters were used to improve greatly the ACLR and ACS of base station transmitters and receivers respectively either side of a TDD/FDD boundary. These significantly exceed the standardized minimum requirements which were not intended to cope unaided with the case of co-sited base stations.
Similar techniques also apply to LTE yet remain significantly challenging. With careful design, however, adjacent channel deployment of FDD and TDD LTE base stations, or of two non-synchronized LTE TDD base stations, should be feasible, even for co-sited arrangements, provided that appropriate measures are adopted in both the interfering and victim base stations.
23.3.1.2 Mobile-to-Mobile Interference
The UE-to-UE interference scenario requires a more probabilistic approach than for base-to- base interference for the following reasons:
• the locations of the interferers and victims are variable and dynamic;
• the physical resources assigned by the scheduler to the UEs are variable;
• the transmit powers of the interfering UEs are a function of their channel conditions and of the power control policy implemented;
• the received levels of the wanted base station signals at the victim UEs are also variable as a function of the UEs’ channel conditions.
It is difficult, therefore, to formulate a definitive analysis of UE-to-UE interference for LTE. Nonetheless, a basic analysis is presented here together with some discussion of the attributes of the LTE system which have some bearing on the magnitude of the interference effects.
We consider two similar overlaid tri-sectored LTE deployments on adjacent 5 MHz carriers. One deployment contains the interfering UEs while the other contains the victim UEs. Both carriers have regions in which uplink transmissions overlap in time with downlink transmissions. The base stations of the deployments are either co-located or maximally spaced non-co-located. Cells in the interfering network each schedule groups of four contiguous uplink Resource Blocks (RBs) to a number of randomly selected UEs, resulting in six simultaneously scheduled interferers per subframe.
The impacts of the scheduled interferers’ transmissions on a UE receiver in the victim network are calculated for the case in which a randomly selected victim UE is scheduled a downlink transmission resource in the one RB next to the band edge separating the two carriers. The impact caused by the interfering adjacent-carrier UEs is analysed in terms of the mean percentage reduction in victim UE downlink throughput,Rloss, caused by the presence of the interfering UEs.
As assumed in [2], the ACIR increases by an additional 13 dB for localized SC-FDMA interferer transmissions located anywhere other than the four RBs next to the band edge of the interferer network. Additionally, for the purposes of this analysis an uplink power control strategy is employed whereby the transmit PSD of each UE is set such that it is not received at any co-channel non-serving eNodeB receiver any higher than 6 dB above the eNodeB receiver’s thermal noise floor.
The transmit PSD of the eNodeB in the victim network is set via Equations (23.4) and (23.5) in the same manner as for the eNodeB-to-eNodeB analysis, such that the downlink is in an interference-limited region of operation (but is not excessively ‘over-powered’). The same path-loss model between eNodeBs and UEs is assumed. The path-loss between UEs is assumed to be given by Equation (23.8) (with a carrier frequency fc of 2000 MHz), based upon a simple two-slope microcellular model from [4] with break point at xb=45 m to reflect the likelihood of free-space-like propagation (with exponent 2) for low separation distances (i.e.x≤xb), and increased attenuation exponentz=6.7 due to local clutter at higher
distances (x>xb).
ρUE-UE(x)=⎧⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪
⎩
−27.56+10 log10(fc2x2) dB forx≤xb
−27.56+10 log10 fc2xz
xzb−2
dB forx>xb (23.8) A microcellular model is considered applicable to UE-to-UE interference as it reflects the case where both the transmitting and receiving antennas are below rooftop level. Other system parameters assumed for this analysis are generally in line with those of [4] for macrocell simulation. With these assumptions, and forLb=20 dB, the results of Figure 23.6 are obtained, displaying the relationship between the band-edge ACIR and the throughput lossRlossfor various values of ISDs.
0 10 20 30 40 50
0 10 20 30 40 50 60 70 80 90 100
ACIR for interferer allocations within 4 RBs from the band edge (dB) R loss (%)
s = 250m, L
b = 20dB s = 750m, L
b = 20dB s = 1250m, L
b = 20dB s = 250m, L
b = 0dB s = 750m, L
b = 0dB s = 1250m, L
b = 0dB
Figure 23.6: Throughput reduction due to UE-to-UE interference.
The results in Figure 23.6 are given for the case of co-located base stations. Those for the non-co-located case are very similar showing only a small further degradation for low ACIR values; the absence of a significant difference is a result of the uplink power control strategy employed as described above (whereby the mobile transmit power is correlated with the strongest non-serving cell path-loss rather than with the path-loss of the serving cell).
The LTE specifications are based upon a 30 dB ACLR which would provide an ACIR of 28 dB assuming an ACS of 33 dB. In this case, it can be seen from Figure 23.6 that the
worst-case throughput loss for the band-edge downlink resource block would be between 7%
and 16% depending on cell size.
This analysis is, however, rather sensitive to certain assumptions, especially the value of the in-building penetration lossLb. When the penetration loss is increased, the serving eNodeB may instruct the UE to increase its transmit power by the same amount in order to maintain its received Signal-to-Interference-plus-Noise Ratio (SINR) at the serving eNodeB (subject to maximum UE output power constraints) without causing additional interference to non-serving eNodeBs which are also protected by the same building penetration loss.
However, the path-loss to a worst-case nearby victim UE (e.g. with LOS propagation between the UEs) is not affected by the increased building penetration loss. Thus, increased building penetration loss can have the effect of increasing worst-case UE-to-UE interference levels.
This effect is clearly evident from the fact that the curves of Figure 23.6 for the case of Lb=20 dB show a greaterRlossthan the dotted curves forLb=0 dB.
This suggests that the susceptibility of the system to UE-to-UE interference is lowered considerably in an outdoor scenario. In these cases, the system’s need to control the uplink inter-cell interference between UEs and neighbouring eNodeBs in a frequency-reuse-1 network constrains the quantity of interference power that is injected by those UEs into adjacent carriers. System throughput loss is then minimal (∼2%) for commonly expected ACIR levels. The fact that a reduction in penetration loss mitigates UE-to-UE interference also points towards possible cell-planning solutions to alleviate the problem, for example using picocells or home base stations rather than macrocells to provide in-building coverage.
The statistical nature of UE-to-UE interference is also of relevance when assessing its impact. LTE allows for randomization of the allocated radio resources for both the interferer and the victim in both the time and frequency domains. Uplink frequency hopping is able to provide the necessary randomization in frequency, and in the time domain a degree of randomization can be provided by different resource scheduling strategies, as well as the possibility for differing retransmission delays due to the fact that the downlink retransmissions to the victim UE are dynamically scheduled rather than synchronous. The use of Hybrid Automatic Repeat reQuest (HARQ) also provides robustness against those instantaneous events in which high interference levels are experienced. Thus, in the case of frequency-adjacent LTE systems, UE-to-UE interference may be heavily randomized.
Its effects can therefore be ‘smoothed’ and shared amongst all users of the system on a probabilistic basis, helping to avoid persistent effects on specific pairs of users with close RF coupling.
UE-to-UE interference may also be alleviated by receive processing at the UE. As discussed in Section 26.2.6, Interference Rejection Combining (IRC) receivers [5] can be used to maximize the received SINR, taking into account the instantaneous direction of arrival of the wanted and interfering signals. Forms of the IRC receiver that make use of averaged correlation (e.g. in time or in frequency) of the received signals across antennas (e.g. [6]) are particularly applicable to the adjacent-channel UE-to-UE interference scenario in which explicit channel estimation of interferers is likely to be impractical. A nearby closely coupled UE on an adjacent channel would typically present a single dominant interference source, enabling an IRC receiver to provide a gain when the instantaneous interference-to- thermal-noise ratio is relatively high. Thus, one could anticipate that use of the IRC receiver may provide some additional robustness against UE-to-UE interference. Figure 23.7 confirms that this is the case (here shown for a 500 m ISD). This figure shows the benefits that an IRC