VIEWS: 536 PAGES: 6 CATEGORY: Emerging Technologies POSTED ON: 12/4/2010 Public Domain
(IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 8, November 2010 Channel Estimation Algorithms, Complexities and LTE Implementation Challenges Md. Masud Rana Department of Electronics and Communication Engineering Khulna University of Engineering and Technology Khunla, Bangladesh Email: mrana928@yahoo.com Abstract—The main purposes of the long term evolution Many CE techniques have already been proposed for (LTE) are substantially improved enduser throughputs, the LTE OFDMA systems. The simple least square (LS) low latency, reduced user equipment (UE) complexity, algorithm, which is independent of the channel model, high data rate, and significantly improved user experience is commonly used in CE [10-14]. But the radio channel with full mobility. LTE uses single carrier-frequency is time-variant; hence a method has to be found in order division multiple access (SC-FDMA) for uplink transmission and orthogonal frequency division multiple to perform estimation in a time-varying channel. The access (OFDMA) for downlink transmission. The major minimum mean-squared error (MMSE) estimate has challenges for LTE terminal implementation are efficient been shown to be better than the LS estimate for CE in channel estimation (CE) method as well as equalization. wireless communication systems [15]. The important This paper discusses the basic CE techniques and future problem of the MMSE estimate is its high direction for research in CE fields. Simulation results computational complexity, which grows exponentially demonstraters that the linear mean square error (LMMSE) with inspection samples [16]. In [17], a low rank CE method outperforms the least square (LS) CE method approximation is applied to a linear MMSE (LMMSE) in term of mean square error (MSE) by more than around estimator that employs the correlations of the channel. 3dB. Hence, based on a given LTE systems resources and specifications, a appropriate method among the presented To further improve the system performance, Wiener methods can be applied for OFDMA systems. estimation has been investigated [18]. Although it exhibits the best performance among the existing linear Keywords— LS, LMMSE, LTE, OFDMA. algorithms, it requires accurate knowledge of second order channel statistics, which is not always feasible at a mobile receiver. Also, this scheme requires higher I. INTRODUCTION complexity. This paper outlines the developments of the LTE OFDMA systems, and highlights some upcoming The wireless evolution has been stimulated by an challenges, where advanced signal processing could explosive growing demand for a wide variety of high play a important role in resolving them. Specifically, we quality of services in voice, video, and data. This investigates various types of CE techniques such as LS, rigorous demand has made an impact on current and and LMMSE CE methods and find out which is the future wireless applications, such as digital audio/video more efficient one. The performance is measured in broadcasting, wireless local area networks (WLANs), terms computational complexity, and mean square error worldwide interoperability for microwave access (MSE). Simulation results shows that the LMMSE CE (WiMAX), wireless fidelity (WiFi), cognitive radio, and algorithm outperforms the existing LS CE in term of 3rd generation partnership project (3GPP) long term MSE by more than around 3dB. Hence, based on a evolution (LTE) [1], [2]. LTE uses single carrier- given LTE systems resources and specifications, a frequency division multiple access (SC-FDMA) for appropriate method among the presented methods can uplink transmission and orthogonal frequency division be applied. multiple access (OFDMA) for downlink transmission The rest of the paper is organized as follows. We give [3], [4]. SC-FDMA utilizes single carrier modulation a brief overview of the wireless communication systems and frequency domain equalization, and has similar in section II. The classification of CE is described in performance and essentially the same overall section III. The LS and LMMSE CE methods are complexity as those of OFDMA system. These describes in section IV and its performance are analyzed advanced applications in which the transmitted signal in section V. In section VI, we highlight the challenges disperses over the time and the frequency domains, for LTE terminal implementation. Finally, some show the need for highlydeveloped signal processing conclusions are made in section VII. algorithms. In particular, one of the main challenges in The following notations are used in this paper: bold the mobile communication is a wireless channel that face lower and upper case letters are used to represent suffers from numerous physical impairments due to multipath propagation, interference from other users or vectors and matrices respectively. Superscripts XT layers, and the time selectivity of a channel [5-9]. 71 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 8, November 2010 X+ denote the transpose and congugate transpose of the The 3GPP members started a feasibility study on the enhancement of the universal terrestrial radio access X , and I is the identity matrix. (UTRA) in December 2004, to improve the mobile phone standard to cope with future requirements. This project was called LTE [5], [8]. 3GPP LTE uses SC- II. COMMUNICATION SYSTEMS FDMA for uplink transmission and OFDMA for Nowadays, cellular mobile phones have become an downlink transmission [9]. Fig. 2 summarizes the cellular mobile communication systems and its access important tool and part of daily life. In the last decade, cellular systems have experienced fast development and schemes [10]. there are currently about two billion users over the world [6]. Mobile penetration is based on population, pay TV and broadband is households. 60 Mobile Total pay TV 50 Broadband P e n e t ra t io n p e rc e n t a g e 40 30 20 10 Fig.2 (a): Evolution path in mobile communication systems. 0 2004 2005 2006 2007 2008 2009 2010 2011 User Fig. 1 Mobile is the key growth platform. The idea of cellular mobile communications is to divide large zones into small cells, and it can provide radio coverage over a wider area than the area of one cell. This concept was developed by researchers at AT & T Bell laboratories during the 1950s and 1960s. The initial cellular system was created by Nippon telephone & telegraph (NTT) in Japan, 1979. From then on, the cellular mobile communication has evolved. The mobile communication systems are frequently classified as different generations depending of the service offered. The first generation (1G) comprises the analog communication techniques, and it was mainly built on frequency modulation (FM) and frequency Fig. 2 (b) Multiple access schemes. division multiple accesses (FDMA). Digital communication techniques appeared in the second From the beginning wireless communications there is generation (2G) systems, and main access schemes are a high demand for realistic mobile fading channels. The time division multiple access (TDMA) and code motive for this significance is that efficient channel division multiple access (CDMA). The two most models are necessary for the investigation, design, and commonly accepted 2G systems are global system for deployment of wireless communication system for mobile (GSM) and interim standard-95 (IS-95). These reliable transfer of information between two parties. systems mostly offer speech communication, but also Correct channel models are also important for testing, data communication limited to rather low transmission parameter optimization, and performance evolution of rates [7]. The concept of the third generation (3G) wireless communication systems. The performance and system started operations on October, 2002 in Japan. complexity of signal processing algorithms, transceiver 72 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 8, November 2010 designs, and smart antennas, employed in future transmitted at certain positions of the OFDMA wireless communication systems, are highly dependent frequency time pattern, in its place of data as shown in on design methods used to model mobile fading Fig. 4. An amount of training sequences is raise the channels. The effect of the channel on the transmitted accuracy of CE, but it is reduces the system efficiency, information must be estimated in order to recover the because there isn’t any new information in the training transmitted information correctly [11]. symbols. III. CLASSIFICATION OF CE Channel can be described everything from the source to the destination of a radio signal. This includes the physical medium between the transmitter and the receiver through which the radio signal propagates. On the other hand, CE is the process of characterizing the effect of the physical channel on the input sequence. It can be employed for the purpose of detecting received signal, improve signal to noise ratio (SNR), channel equalization, reduced ISI, mobile localization, and improved system performance [8], [9]. Fig. 4 Positions of data and pilot symbols. In general, both iterative and noniterative CE techniques can be divided into three categories such as the training CE, blind CE, and semi-blind CE [10], B. Blind CE [21]. A blind CE method requires no training sequences [13]. They exploit certain underlying mathematical information regarding the type of data being transmitted. These CE methods might be bandwidth efficient but still have their own downfalls. These methods are enormously computationally intensive and convergence is slow [21]. A popular category of blind CE method is decision directed algorithms. These methods rely upon the demodulated and detected signal at the receiver to reform the transmitted signal. The drawback of these CE algorithm is that a bit error at the receiver is cause the construction of an erroneous transmitted sequence. Fig. 3 Outline of the CE. C. Semi-blind CE A. Taining CE Semi-blind CE methods are used a combination of data aided and blind methods [11]. Since, there are a The training CE algorithm requires probe sequences; large number of channel coefficients, a large number of the receiver can use this probe sequence to reconstruct pilot symbols may be required. It would result in a the transmitted waveform [10]. It has the advantage of decrease of data throughput. To avoid it, the semi-blind being used in any radio communications system quite CE methods with fewer pilot symbols can be used. As a easily. Even if this is the most popular CE method, it result, improve system performance in compared with still has its drawbacks. The drawback of training using equal pilots in LS method. Moreover, there is a sequence methods is that the probe sequence occupies trend to use superimposition of pilot and data symbols. valuable bandwidth, reducing the throughput of the In fact, these methods by superimposing pilot and data communication system. This scheme also suffers due to symbols in the same time economize the system the fact that most communication systems send bandwidth. But in superimposed training sequence information lumped frames. It is only after the receipt of scheme, there is disadvantage due to the interference of the entire frame that the channel estimate can be information data. So, an accurate CE has been one of reconstructed from the embedded probe sequence. Since, the most important issues for reliable mobile the coherence time of the channel might be smaller than communication systems. So, CE can be performed by the frame time, for rapid fading channels this CE might many ways inserting pilot tones into each OFDMA not be sufficient. Training symbols can be placed either symbol with a specific period or blind CE. at the beginning of each burst as a preamble or regularly through the burst [21]. Training sequences are 73 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 8, November 2010 IV. LS and LMMSE CE ALGORITHMS w LS = (ρ I + SS ) -1S + r , (4) Pilot estimators are often achieved by multiplexing where ρ is regularization parameter and has to be training sequence into the data sequence. These pilot chosen such that the resulting eigenvalues are all -1 symbols allow the receiver to extract channel defined and the matrix (ρI + SS ) is least perturbed. attenuations and phase rotation estimates for each Here the channel is considered as a deterministic received symbol, facilitating the compensation of parameter and no knowledge on its noise statistics is channel fading envelope and phase. A general CE needed. The LS estimator is computationally simple but procedure for communication system is shown in Fig. 5. problem is that the inversion of the square matrix turns out to be ill-conditioned (sometime). So, it will need to regularize the eigenvalues of the matrix to be inverted by adding a small constant term to the diagonal [14]. MMSE CE method proposes at the minimization of the MSE between the actual and estimated channel impulse response (CIR). The most important problem of the MMSE estimate is its high computational complexity, which grows exponentially with inspection samples [15], [16]. In [17], a low rank approximation is applied to a linear MMSE (LMMSE) estimator that Fig. 5 General CE procedure. employs the correlations of the channel. The general expression of LMMSE is described as The signal S is transmitted via a unknown time- varying channel w, and corrupted by an additive white Gaussian noise (AWGN) z, before being detected in a w est = R ww (R ww + ΓI /SNR) -1w LS , (5) receiver. The channel coefficient w est , is estimated where R ww is the auto-covariance matrix of w, w LS is using any kind of CE method. In the channel estimator, the channel response in LS estimation, and is a constant transmitted signal S is convolved with an estimate of depending on the modulation constellation the channel w est . The error between the received 2 2 signal and its estimate is Γ = E[ S k ] E[ 1/S k ]. (6) For QPSK modulation, is 1[17]. Here, w LS is not e = r - rest (1) very important issue in the matrix computation, the The aim of most CE algorithms is to minimize the MSE, while utilizing as little computational resources as inversion of R ww does not require to be estimated every possible in the estimation process. time the transmitted sybmols in w LS varies. Also, if The idea behind LS CE method is to fit a model to signal to noise ratio (SNR) and R ww are identified measurements in such a way that weighted errors earlier or are set to fixed nominal values, the matrix between the estimation and the true model are minimized [14]. The received signal can be written as (R ww + ΓI /SNR)-1 needs to be computed at once. vector notation as Under these situation, the estimation requires L multiplications per tone. r = Sw + z , (2) In order to calculate computational complexity, we T where r = [r1 , r2 ......rL ] is the received signal, assume that the evaluation of the scalar addition or subtraction needs L addition and multiplying the scalar S = diag[s1 , s 2 ......s L ] is the transmitted signal, by the vector requires L multiplications, and multiplying w = [w1 , w2 ......wL ]T is the unknown channel two matrix need 4L multiplications and 4L-1 additions. T Table I summarizes the computational complexity of the coefficients, and z = [z1 , z2 ......z L ] is additive white different CE methods. Gaussian noise (AWGN). The LS estimate of such a system is obtained by minimizing square distance Table I Complexity of the CE methods between the received signal and its estimate as [14] Operation LS CE LMMSE CE Matrix inversion 1 2 j = (r - Sw ) (r - Sw )† . (3) Multiplication 11L 17L Now differentiate this with respect to w and set the Addition 11L - 3 17L - 5 results equal to zero to produce [14]: 74 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 8, November 2010 We calculate the number of complex addition and on some preferred requirements such as cell-edge multiplications which are needed to implement the spectral efficiency in the uplink transmission [19]. LTE algorithm. It shows that the LS CE algorithm has lower implementation poses the following signal processing complexity than LMMSE method. For this LMMSE challenges in terms of performance, cost and power estimator, the main contribution to the complexity consumption: comes from the term R ww (R ww + ΓI /SNR) . -1 • Regrettably, the development of data rates is not matched by advanced in semiconductor structures, terminal power consumption improvements. Therefore, advanced signal processing architectures V. ANALYTICAL RESULT as well as algorithms are needed to cope with these data rates [19]. In this simulations, we consider a system operating • High performance multiple input multiput output with a bandwidth of 1.25MHz, with a total symbol (MIMO) receivers such as sphere decoders, period of 520µs, of which 10 µs is a cyclic prefix. The maximum likelihood receivers offer substantial entire channel bandwidth is divided into 128 sub- system performance gains but enforce an carriers, implemented by 128-point IDFT. Sampling is implementation challenge, especially when the high performed with a 1.92MHz. The data symbol is based peak data rates are targeted [19]. on BPSK. In practice, the ideal channel coefficient is • LTE utilizes precoding, which requires accurate CE. unavailable, so estimated channel coefficient must be used instead. The more accurate estimated channel Advanced methods like iterative decision directed CE and pilot based CE offer system performance coefficient is, the better MSE performance of the CE improvements, but pose again a computational will achieve. The performance is measured using MSE complexity challenge [19]. between the actual and the estimated channel response. Fig. 6 shows the MSE versus SNR for the different • LTE has a large ”toolkit” of MIMO methods and channel estimators. We can see that LMMSE CE can adaptive methods. The choice and combination of always achieve better performance than LS CE. The the accurate technique in a cell with heterogeneous main reason is, LMMSE CE method uses channel devices, channel conditions and bursty data correlation as well as SNR but the LS CE method does services is a challenge [19]. not uses channel correlation. Finally, it concludes that • It is very difficult to implement many antennas in a the LMMSE CE method has higher computational small hand portable unit. In near future, we need to complexity and around 3dB better performance use wearable antenna on head. compared with the LS CE method. • LTE roll-out will be gradual in most cases- interworking with other standards such as GSM or 0 10 HSPA is required for a long time. This imposes not only a cost and computational complexity issue. One of the reasons many early 3G terminals had LMMSE CE poor power consumption was the need for second -1 10 LS CE generation (2G) cell search and handover in addition to normal 3G operation. Reduced talk-time for dual-mode devices is not suitable [19]. Fig. 7 -2 shows the estimated complexity based on the 10 baseline receiver. Note that the complexity of the LTE receiver grows linearly with respect to system MSE bandwidth and the corresponding maximum -3 10 nominal throughput. Interestingly, MIMO mode requires less than double the SIMO mode complexity. -4 10 -5 10 0 5 10 15 20 25 30 35 SNR [dB] Fig. 6 MSE of the LS and LMMSE CE methods. VI. IMPLEMENTATION CHALLENGES LTE meets the important obligations of next generation mobile communications, but still falls short 75 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No. 8, November 2010 Proc. Personal, Indoor and Mobile Radio Commun., Sept. 2009. 8 [5] L. Sanguinetti, M. Morelli, and H. V. Poor, “Frame detection and timing acquisition for OFDM transmissions with 7 unknown interference,” IEEE Trans. on Wireless Commun., LTE SIMO 16-QAM vol. 9, no. 3, pp. 1226–1236, Mar. 2010. LTE MIMO 16-QAM [6] M. Morelli, G. Imbarlina, and M. Moretti, “Estimation of 6 residual carrier and sampling frequency offsets in OFDM- HSDPA 16-QAM SDMA uplink transmissions,” IEEE Trans. on Wireless Commun., vol. 9, no. 2, pp. 734–744, Feb. 2010. C o m p le x i t y [ G ip s ] 5 [7] Z. Wang, Y. Xin, G. Mathew, and X. Wang, “Multipath parameter estimation for radio propagation channel measurements with a vector network analyzer,” IEEE Trans. 4 on Vehicular Technology, vol. 59, no. 1, pp. 48–52, Mar. 2010. [8] J. I. Montojo and L. B. Milstein, “Channel estimation for 3 non-ideal OFDM systems,” IEEE Trans. on Commun., vol. 58, no. 1, pp. 146– 156, Jan. 2010. 2 [9] M. H. Hsieh and C. H. We, “Channel estimation for OFDM systems based on comb-type pilot arrangement in frequency selective fading channels,” IEEE Trans. on Consumer 1 Electronics, vol. 44, no. 1, pp. 217–225, Feb. 2004. [10] F. Wan, W. P. Zhu, and M. N. S. Swamy, “A semiblind channel estimation approach for MIMO-OFDM systems,” 0 IEEE Trans. on Signal Process., vol. 56, no. 7, pp. 2821– 0 2 4 6 8 10 12 14 16 18 20 2834, July 2008. System bandwiudth [MHz] [11] H. Wu, X. Dai, and H. Zhang, “Semi-blind channel estimation for the uplink of multi-carrier code-division multiple access systems with timing offset,” IET Commun., Fig. 7 Complexity of LTE receiver. vol. 3, no. 12, pp. 1832–1842, 2009. [12] B. Karakaya, H. Arslan, and H. A. Curpan, “An adaptive channel interpolator based on Kalman filter for LTE uplink in high Doppler spread environments,” EURASIP Journal VII. CONCLUSION on Wireless Commun. And Networking, vol. 2009, Article ID 893751, 2009. An accurate CE is one of the most important issues for [13] T. Fusco and M. Tanda, “Blind synchronization for OFDM systems in multipath channels,” IEEE Trans. on Wireless reliable future wireless communication systems such as Commun., vol. 8, no. 3, pp. 1340–1348, Mar. 2009. LTE. In this paper, we briefly insvesteget LS and [14] A. Ancora, C. B. Meili, and D. T. Slock, “Down sampled LMMSE CE techniques for LTE terminal implemtation. impulse response least-squares channel estimation for LTE Simulations demonstrated that the MSE performance of OFDMA,” In Proc. International Conference on Acoustics, Speech, and Signal Processing, pp. III–293–III–296, Apr. the LMMSE CE algorithm is at least 3dB better than 2007. existing LS estimator. Even though, the LMMSE CE [15] J. J. V. D. Beek, O. Edfors, and M. Sandell, “On channel technique requires a little high computational complexity, estimation in OFDM systems,” In Proc. Vehicular the advantage in the MSE and convergence towards true Technology Conference, vol. 2, pp. 815–819, Sept. 1995. [16] M. H. Hsieh and C. H. We, “Channel estimation for OFDM channel coefficient may be significantly useful for future systems based on comb-type pilot arrangement in frequency mobile communications which allow broadband selective fading channels,” IEEE Trans. on Consumer multimedia Internet access and wireless connection Electronics, vol. 44, no. 1, pp. 217–225, 1998. anywhere, and any time. This paper also discusses the [17] O. Edfors, M. Sandell, J.J. V. D. Beek, S. K. Wilson, and P. O. Borjesson, “OFDM channel estimation by singular value challenges imposed by developments in the LTE terminal decomposition,” IEEE Trans. on Commun., vol. 46, no. 7, implementation. Hence, based on a given LTE systems pp. 931– 939, 1998. resources and specifications, a appropriate method [18] L. A. M. R. D. Temino, C. N. I. Manchon, C. Rom, T. B. among the presented methods can be applied. Sorensen, and P. Mogensen, “Iterative channel estimation with robust Wiener filtering in LTE downlink,” In Proc. Vehicular Technology Conference, pp. 1–5, Sept. 2008. [19] R. Irmer and S. Chia, “Signal processing challenges for REFERENCES future wireless communications,” In Proc. International Conference on Acoustics, Speech, and Signal Processing, [1] Q. Li, G. Li, W. Lee, M. I. Lee, D. Mazzarese, B. Clerckx, pp. 3625–3628, Apr. 2009. and Z. Li, “MIMO techniques in WiMAX and LTE: a feature [20] M. M. Rana, “Channel Estimation Techniques and LTE overview,” IEEE Commun. Magazine, vol. 48, no. 5, pp. 86– Terminal Implementation Challenges,” in proc. Int. Con. on 92, May 2010. Computer Sciences and Convergence Information [2] M. Alasti, B. Neekzad, C. J. Hui, and R. Vannithamby, Technology, Bangladesh, December 22, 2010. “Quality of service in WiMAX and LTE networks,” IEEE [21] X. G. Doukopoulos and G. V. Moustakides, “Blind adaptive Commun. Magazine, vol. 48, no. 5, pp. 104–111, May 2010. channel estimation in OFDM systems,” IEEE Trans. on [3] Z. Lin, P. Xiao, B. Vucetic, and M. Sellathurai, “Analysis of Wireless Commun., vol. 5, no. 7, pp. 1716–1725, 2006. receiver algorithms for LTE SC-FDMA based uplink MIMO systems,” IEEE Trans. on Wireless Commun., vol. 9, no. 1, pp. 60–65, Jan. 2010. [4] M. Zhou, B. Jiang, T. Li, W. Zhong, and X. Gao, “DCT- based channel estimation techniques for LTE uplink,” In 76 http://sites.google.com/site/ijcsis/ ISSN 1947-5500