CRS interference cancellation algorithm for heterogeneous network H. Luo✉, W. Li, Y. Zhang, L.-K. Huang, J. Cosmas and Q. Ni Heterogeneous network is introduced to improve the network capacity in LTE Release 9 and system beyond. The potential traffic congestion due to increased users can be alleviated by the cooperation between macro-cell and pico-cell. However, the inter-cell interference caused by RF signal from macro-cell will reduce the performance severely. Enhanced inter-cell interference coordination is proposed in Rel. 10 to solve this problem using almost blank subframe (ABS). Yet, the cell specific reference signal in ABS can still cause interference to the data resource element (RE) from the pico-cell inevitably for noncolliding scenario. In this Letter, a novel interference cancellation (IC) algorithm is proposed to mitigate the interference. First, the timing and carrier frequency offset of interference signal is estimated and compensated. Second, the interfering channel response is estimated by utilising the channel statistics. Third, the interference signal is reconstructed based on the channel estimation and cancelled in the received signal in time domain. The experiment results show that the performance of proposed IC algorithm is robust.
Introduction: Heterogeneous network (HetNet), first introduced in LTE Release 9, is a promising network topology for achieving high special efficiency. With the macro-cell providing basic coverage and the picocell serving as a complementary cell, pico-cell can increase the off-load data rate and network coverage of macro-cell. However, the user equipments (UEs), served by pico-cell, will suffer from the RF signal named inter-cell interference caused by the neighbour high power macro-cells. This issue will become even severer if the UEs are within the coverage of macro-cells. To solve this problem, enhanced inter-cell interference coordination (eICIC) [1, 2] was introduced in 3GPP Rel. 10 with two techniques. On the one hand, the signal strength is biased to pico-cell which can reduce the interference power. On the other hand, macro-cell keeps silent for certain periods called almost blank subframe (ABS). In ABS, the interference is alleviated because UEs will not receive the physical downlink shared channel from macrocell. However, the cell specific reference signal (CRS), paging channel, physical broadcast channel and synchronisation channels (PSS/SSS) can still be received and the performance will still be degraded. Therefore, in Release 11, further eICIC was proposed to cancel the CRS interference problem. There are few literatures about CRS interference cancellation (IC). In [3, 4], the authors studied a traditional CRS IC which is realised by first estimating the interference channel and then cancelling the interference. A log-likelihood ratio muting/puncturing method was investigated in [5, 6]. A receiver algorithm that combines IC with direct decision channel estimation was proposed for non-colliding CRS in [7]. According to Priyanto et al. [8], a robust equalisation technique was studied which is with similar performance to the traditional CRS IC but has lower complexity and latency. Nevertheless, timing and frequency offset will decrease the system performance severely for non-colliding CRS scenario. In this Letter, a CRS IC algorithm, with timing offset (TO) and carrier frequency offset (CFO) taken into account, is studied for the noncolliding scenario. The interference signal is reconstructed and then mitigated by using TO and CFO obtained and the estimated interference channel response. The simulation results show that our algorithm can achieve significant performance in different channel conditions when the signal to interference and noise ratio is within −3 dB to 9 dB.
yi(n)
FFT
TO/CFO estimation PSS/SSS
interfering CE CRS
interference y˜ (p)(n) reconstruction and cancellation CRS
interfering CRS modelling
Fig. 1 Proposed IC algorithm
The LTE receiver IC algorithm is briefly shown in Fig. 1. For LTE downlink, the received signal yi(n) for the ith symbol can be modelled as: (m) yi (n) = y(p) i (n) + yi (n) + ni
(1)
(m) where y(p) i (n) and yi (n) denote the desired and interference signal respectively; ni is the additive Gaussian noise. The fast Fourier transform is first done in order to transfer the signal into OFDM symbols. After N-point FFT, the OFDM symbols with TO/ CFO can be written as: (p) (m) + Yi,k + Ni,k Yi,k = Yi,k (p) (p) = Hi,k Xi,k +
N /2 n=−N /2
(m) (m) e(2pnDn/N ) Hi,k Xi,k Fi + Ni,k
(2)
(p) (m) and Xi,k represent the ith symbol at kth subcarrier for desired where Xi,k (p) (m) and Hi,k are the channel coefsignal and interference, respectively; Hi,k ficients of the serving and interfering channel at kth subcarrier, respectively; Φi stands for the inter-carrier interference, which arises from CFO; and Δn stands for the relative TO between the interfering cell and serving cell.
Proposed IC algorithm: The IC algorithm proposed mainly consists the following three steps. First, the relative timing and frequency offset between the interfering cell and serving cell are estimated by using PSS/SSS generated in the interfering CRS modelling. Second, the interfering channel estimation is done based on the signal after TO/CFO compensation. Third, the interfering signal is reconstructed according to previous interfering CE and then cancelled from the received LTE signal. The details of these three steps are shown below from step A to step C. TO/CFO estimation: Most of the timing and frequency synchronisation algorithms existed utilise the periodic nature of time domain signal by using cyclic prefix (CP) [9] or pilot data [10]. However, in ABS, the data resource elements (REs) of the macro-cell are zero which reduces the power of CP significantly. The low SNR of CP makes it hard for timing and frequency synchronisation. Yet, the PSS/SSS signal, located at the last and second-last symbol in slot 0 and slot 10, can be used for synchronisation as well. The TO/CFO can be estimate by utilising the cross-correlation of PSS/SSS symbol in time domain [11]: (3) {D˜n, Df˜ } = arg max C Dn, Df Dn,Df
where M s∗i (m)r(Dn + m) e(−2pDfn/N) C Dn, Df =
(4)
m=1
r(n) and si (n) are the received and locally generated symbol that contain PSS/SSS, respectively, and the correlation length M could be multiple times of PSS/SSS symbols when SNR is low. After TO/CFO estimation, the interference signal can be synchronised and the interference channel response can be estimated. Interfering CE: To reconstruct interfering signal, it is important to estimate the interfering channel response at first. For the sake of decreasing compute complexity, the least square channel estimation algorithm is chosen. According to (2), the interfering channel estimation can be written as: (m) (m) (m) (p) (p) (m) (m) = Hi,k + Hi,k Di,k /Pi,k + Ni,k /Pi,k H˜ i,k = Yi,k /Pi,k
(5)
(m) Xi,k
(m) is replaced by the interfering CRS Pi,k Here, the interference signal because we only care for interfering CRS signal from the macro-cell. According to (5), the data REs of serving cell D(p) i,k become interference with relative higher power. Therefore, the estimation in (5) is not accurate. Many studies, such as [12], show that the distribution of the interference signal is close to Gaussian for large size RB and non-Gaussian for small size RB. The mean of the distribution converges to 0. Hence the expectation of (5) can be derived as: (m) (m) (p) (p) (m) (m) Di,k /Pi,k + E Hi,k + E Ni,k /Pi,k E H˜ i,k = E Hi,k (m) (6) ≈ E Hi,k
The mean value of the interfering channel can be estimated by utilising (m) moving average in time domain. Then Hi,k can be approxi (m)window mated by E H˜ i,k when the moving average window length M is within the coherence time of the channel.
ELECTRONICS LETTERS 8th January 2016 Vol. 52 No. 1 pp. 77–79
1.0
serving cell. The signal with desired data and interference pass through a fading channel with a delay spread smaller than CP duration. In this simulation, the AWGN channel model is used. In addition, Different CFO and arriving time are applied to the interfering signal to evaluate its effect. The effect of CFO and TO on BLER performance is shown in Figs. 2 and 3, respectively. The performance without IC is shown as a comparison. It can be seen that our proposed IC achieves good performance both for CFO and TO. The BLER performance of the proposed IC algorithm for different Doppler frequency and SNR is shown in Fig. 4. The system performance is significantly improved and therefore the robustness of this method can be proved.
0.9 BLER without IC BLER with IC BLER without interference
0.8 0.7 BLER
0.6 0.5 0.4 0.3 0.2 0.1 0 0
1000 2000 3000 4000 5000 6000 7000 8000 CFO, Hz
Conclusion: In this Letter, a novel IC algorithm is proposed for HetNet receiver based on the interference signal reconstruction. First, the TO/ CFO compensation and interfering channel estimation are applied for reconstructing the interfering CRS. Finally, the interference is cancelled in time domain after reconstruction by subtracting it from the received signal. The performance shown in the simulation results indicate that the IC algorithm can achieve very good performance in different channel conditions.
Fig. 2 BLER performance against frequency offset 0.9 0.8 0.7
BLER without IC BLER with IC BLER without Interference
BLER
0.6 0.5 0.4
© The Institution of Engineering and Technology 2016 Submitted: 31 July 2015 E-first: 17 November 2015 doi: 10.1049/el.2015.2669 One or more of the Figures in this Letter are available in colour online.
0.3 0.2 0.1
H. Luo, W. Li and Y. Zhang (Institute for Research in Applicable Computing, University of Bedfordshire, Luton LU1 3JU, UK)
0 0
20
40
60 80 100 timing offset, samples
120
140
160
✉ E-mail:
[email protected] L.-K. Huang (Cobham Wireless, Stevenage, SG1 2AN, UK)
Fig. 3 BLER performance against timing offset
J. Cosmas (Brunel University, London UB8 3PH, UK) Q. Ni (Lancaster University, Lancaster LA1 4YW, UK)
MCS 20; SIMO 100
W. Li: Also with the Cobham Wireless, Stevenage, SG1 2AN, UK References
BLER
10–1
10–2
BLER without IC BLER with IC BLER without Interference
10–3
10–4
7
8
9
10 SNR, dB
11
12
13
Fig. 4 BLER performance against different Doppler frequency and SNR
Interfering signal reconstruction and cancellation: After TO/CFO and interfering CE estimation, the interference signal can be reconstructed based on the local CRS in time domain. As the relative TO Δn is larger than the duration of CP potentially which will cause intersymbol interference within the OFDM window of desired signal, the interference signal is reconstructed in time domain and subtracted from the received signal directly: ˜
y˜ (p) (n) = y(n) − x˜ (m) (n + D˜n) e(j2pDf n/N ) ⊗ h˜ l
(7)
where ⊗ is circular convolution and x˜ (m) (n) is the interference CRS reconstructed and (m) (8) h˜ l = IFFT H˜ i,k Results: The performance of the proposed IC algorithm is evaluated via Monte Carlo simulation. With different modulation and coding schemes to deliver the service, the serving cell is set to work with 10 MHz bandwidth. The neighbour interfering cell transmits ABS with bandwidth of 5 MHz. The CRS from the interfering cell overlaps the data REs of the
1 Heli, Z., Shanzhi, C., Xi, L., Hong, J., and Xiaojiang, D.: ‘Interference management for heterogeneous networks with spectral efficiency improvement’, IEEE Wirel. Commun. Mag., 2015, 22, pp. 101–107 2 Deb, S., Monogioudis, P., Miernik, J., and Seymour, J.P.: ‘Algorithms for enhanced inter-cell interference coordination (eICIC) in LTE HetNets’, IEEE/ACM Trans. Netw., 2014, 22, pp. 137–150 3 Tyrrell, A., and Schoeneich, H.: ‘Advanced receiver signal processing techniques: Evaluation and characterization’, ARTIST4G D2.2 Deliverable, June 2011 4 Yoo, T., et al.: ‘Dynamic switching between common reference signal interference cancellation & resource element puncturing in a co-channel heteregoneous network’. Patent US2012/0087261 A1, 2012 5 Qualcomm Inc.: ‘Enabling communication in harsh interference scenarios’, R4-102350, 3GPP-RAN WG4 AH#10-03, Bratislava, 2010 6 Qualcomm Inc: ‘Link level simulations for FeICIC with 9 dB cell range expansion’, R4-123313, 3GPP-RAN WG4 #63, Prague, May 2012 7 Huang, M., and Xu, W.: ‘Macro-femto inter-cell interference mitigation for 3GPP LTE-A downlink’. IEEE Wireless Communication and Network Conf, April 2012, pp. 75–80 8 Priyanto, B.E., Kant, S., Rusek, F., Hu, S., Chen, J., and Wugenshi, C.: ‘Robust UE receiver with interference cancellation in LTE advanced heterogeneous network’. IEEE Vehicular Technology Conf., Las Vegas, September 2013, pp. 1–7 9 Astely, D., Dahlman, E., Furuskar, A., Jading, Y., Lindstrom, M., and Parkvall, S.: ‘LTE: the evolution of mobile broadband’, IEEE Commun. Mag., 2009, 47, pp. 44–51 10 Filippi, A., and Serbetli, S.: ‘OFDM symbol synchronization using frequency domain pilots in time domain’, IEEE Trans. Wirel. Commun.., 2009, 8, (6), pp. 3240–3248 11 Tsai, Y.-H., and Sang, T.-H.: ‘A new timing synchronization and cell search procedure resistant to carrier frequency offsets for 3GPP-LTE downlink’. IEEE Int. Conf. on Communication in China (ICCC), Beijing, August 2012, pp. 334–338 12 Feng, C., Cui, H., Ma, M., and Jiao, B.: ‘On statistical properties of co-channel interference in OFDM systems’, IEEE Commun. Lett., 2013, 17, pp. 2328–2331
ELECTRONICS LETTERS
8th January 2016 Vol. 52 No. 1 pp. 77–79
