Antenna Selection Based Initial Ranging Method for IEEE m MIMO-OFDMA Systems

Transcription

1 Antenna Selection Based Initial anging Method for IEEE 8.6m MIMO-OFDMA Systems Department of Physics & Electronics Information Luoyang ormal Uniersity o.7, Longmen oad, Luoyang, Henan, 47 CHIA Abstract: - An antenna selection based initial ranging method is proposed in orthogonal frequency diision multiple access (OFDMA networks with multiple-input multiple-output (MIMO for IEEE 8.6m standard. In the proposed scheme, the initial ranging signal processing is selected by the antenna selection at the base station, where the receie antenna with the largest receied signal-to-interference-and-noise ratio (SI is chosen. The actie initial ranging users and the timing offset estimation can easily obtain by means of the adaptie threshold. Simulation results show that the proposed initial ranging scheme can achiee a much better initial ranging performance than the noncoherent antenna combining method, and dramatically improe the initial ranging efficiency for a large number of actie SSs simultaneously to access into networks. Key-Words: - orthogonal frequency diision multiple access (OFDMA; initial ranging; antenna selection (AS; multiple-input multiple-output (MIMO; IEEE 8.6 Introduction Orthogonal Frequency Diision Multiple Access (OFDMA is a promising multiple access scheme and has been adopted by seeral wireless communication system standards to proide efficient broadband wireless access to subscribers. Multiple subscriber stations with different timing transmit simultaneously in the uplink channel of OFDMA systems, synchronization can be achieed by a random access process referred to as initial ranging (I in the IEEE 8.6m standard []. In the I process, a new ranging subscriber station (SS, transmits a randomly chosen frequencydomain ranging code on ranging subchannel in specific ranging time-slot (one or seeral OFDMA symbol interals. Howeer, the initial ranging performance degrades with the ranging channel frequency selectiity and the magnitude of SSs since the correlation properties of the ranging codes are affected by the channel frequency selectiity resulting in large multiuser access interference (MAI which is amplified by the larger number of actie SSs. The I methods in OFDMA systems generally fall into two categories. The first category []-[3] is discussed for single receie antenna at the base station (BS. The I methods []-[5] are based on the I codes correlations in either the frequency or the time domain, but the performance of these methods seerely deteriorate because of treating the MAI as noise. In order to combat the harmful impact of the ranging channel frequency selectiity, the methods in [6]-[3] are proposed to suppress the MAI effect. The works [6]-[9] are proposed by using a new ranging code structures and associated ranging subchannel allocation, but these schemes are different from that of the IEEE 8.6m standard. Following the IEEE 8.6m standard, the literatures []-[3] present a successie multiuser detection and interference cancellation methods in the frequency selectie channels. Howeer, the complexity of those methods is ery high, and the accumulated residual MAI degrades the I performance and limits the number of SSs simultaneously accessing into the network. The second category on the I scheme [4]-[6] is presented for the multiple receie antennas equipped at the BS. The works [4] and [5] discuss the multiuser diersity and multiantenna diersity gains by utilizing the I subchannel allocation. Each of the I subchannel is allocated with a little number of adjacent subcarriers, and most of the SSs are expected to transmit on different I subchannels. Those methods work well by exploiting the channel state information (CSI at the transmitter side in time diision duplex (TDD model. Howeer, the adjacent subcarriers allocation increases the sensitiity to residual carrier frequency E-ISS: Volume 4, 5

2 offsets (CFOs. In addition, the subchannel allocation scheme is different from the IEEE 8.6m standard for the I. The literature [6] presents a noncoherent antenna combining (AC initial synchronization scheme in Long Term Eolution (LTE systems, Howeer, when the receie antennas experience fading independent, the AC method degrades the I performance duo to the MAI of each receie antenna. In this work, we proposed an antenna selection based I scheme for the IEEE8.6m MIMO- OFDMA systems. In the proposed method, the I signal processing is selected by the antenna selection, and the selection is based on the receied signal-to-interference-and-noise ratio (SI, where the receie antenna with the largest receied SI is chosen. The actie initial ranging users and the timing offset estimation can easily obtain by means of the adaptie threshold. The performance of the I by the proposed scheme is highly improed compared to AC method. The rest of this paper is organized as follows. We introduce the I description and signal models oer multipath channel in Section. The I scheme is presented in detail in Section 3. The I performance and simulation results are gien in Section 4 and 5. Finally, we conclude paper in Section 6. System Description and Signal Models. I procedure I process is the first step of wireless initial access for each actie SS, and it contains I signal transmission, I detection, and resource allocation. If some SSs transmit their I signals at the same time, the BS will recognize each SS by its specific ranging code. As a result of different locations and mobility, each SS has its specific transmission time delay (TTD. The BS must execute uplink time synchronization by the I process to compensate TTD of each SS. The I procedure based on the IEEE 8.6 standard can be described as follows: Step each SS first acquires downlink synchronization and uplink transmission parameters from downlink control frames. Step each SS randomly chooses an I timeslot and an I code and then transmits it on the I channel. Step 3 BS detects I codes and estimates its timing and power from the receied I signal. BS broadcasts an I response message which adertises the receied I code and the I time-slot where the I code has been identified. The I response message also contains all the adjustment information (e.g., timing and power adjustment and a status notification (e.g., success, continue. Step 4 if an SS receies the success notification and its I process is completed; otherwise, the SS repeats its I process at the following I attempts.. I signal models We consider a multiantenna wireless communication system, and each SS transmits the I signal with single antenna and the BS receies the I signals with x receie antennas. According to the IEEE 8.6m standard [], a typical OFDMA uplink has subcarriers, and an I timeslot consists of ranging subcarriers with index { Jk ; k=,, } oer two OFDMA symbols. Each SS selects an aailable I code which is randomly chosen from a predefined code set { C,, C,, C }, where is the total number of u c I codes. Assume that all SSs perform frequency synchronization based on downlink control channel before initiating the ranging process and the conflict of more than one SS using the same I code is not considered in this paper. Suppose that the u th SS transmits C ={ C ( l, l=,,, } on the ranging u u channel. The ector of Cu c is mapped onto an OFDMA symbol Xu = [ Xu (,, Xu ( k,, Xu ( ] and its corresponding element X ( k is gien by Cu ( k, k = J k, k =,..., X u ( k = (, otherwise After the inerse discrete Fourier transform (IDFT and cyclic prefix (CP insertion, the samples of the u th SS in the time domain are transmitted oer two consecutie OFDMA symbols, and they are expressed by j π Jk n/ xu ( n = Cu ( k e, g n + g k= ( where is the length of CP. g Let K be the number of SSs that are simultaneously actie on the one I time slot. The fading channel between the transmit/receie antenna pair is assumed to be frequency selectie, and multipath ayleigh fading channels for different SSs are independent. The receied I signal at the BS is gien by u E-ISS: Volume 4, 5

3 ( ( ( [ (,, (,, x T y n = y n y n y ( n] (3 ( m where y ( n is the receied sample at the m th antenna and can be expressed as K L ( m u, l u( u + (4 y ( n= h x n l d z ( n where u, l u= l= h is the fading coefficient of the uth SS for the l th path at the m th receie antenna and L is the channel length (normalized by the sampling period T s. du represents the round-trip delay and which is related to the different distances between the SS and the BS. The maximum alue for a SS located at the cell boundary is gien by d,max = / ( ct with being the cell radius and c denoting the speed of light. { z ( n } are independent and identical distributed complex Gaussian random ariables with ( ariance δ [ m z = E z ( n ] at the m th receie antenna. At the BS, remoing the CP and after fast Fourier transform (FFT for the receied symbol, the I signal of the m th receie antenna on the k th I subcarrier is T Y = [ Y (,, Y ( k,, Y ( ] K L (5 h Λ C + Z = u, l du+ l u u= l= where du+ l = diag Λdu+ l, Λdu+ l, k Λdu+ l, Λ {,,,, } is a diagonal matrix with diagonal element j π Jk ( du + l/ Λ =. du+ l, k e ( ( ( ( Z m = [ m (,, m (,, m ( ] T is a Z Z k Z complex Gaussian ector with coariance j π Jk n/ matrix δ z I, and Z ( k = z ( n e. n= Define M = d + L as the maximum number t,max of resoled path for each SS, equation (5 can be considered as a linear combination of M t possible path signals for eery actie SS. Equation (5 can be further rewritten as K M t = = hu, α α u + u= α= Y Y Λ C Z (6 When the path { u, α } is present in the obseration ector Y, the path is alid and h α ; otherwise, the path is inalid and h α =. u, u, s 3. Proposed I Algorithm To derie coneniently, { τ, } denotes the path of the I codes with index ( [, c ] and the timing offset τ ( τ [, M t ]. For correlation-based detection, the BS correlates the receied signal in the frequency domain with the signal of path { τ, }, separately in each receie antenna output. For the m th receie antenna, the correlation output of path { τ, } is H H ( τ = C Λτ Y (7 Depending on whether the path { τ, } is present in the obseration ector Y, equation in (7 can be expressed as ( τ = W (8 for m=,, x, for the MAI-plus-noise hypothesis H, where K M t - H H H H W = hu, α C Λ τ Λ α Cu + C Λ τ Z is ( u= α= u the MAI-plus-noise on the m th receie antenna, and ( τ = h ( τ + W (9 for m=,, x, for the signal-plus-noise hypothesis ( H. The problem of the I detection is to test the MAI-plus-noise hypothesis against a signal-plusnoise hypothesis, gien x obserations in (7 at the BS. The binary hypothesis test can be formulated as : ( = H τ W ( : ( = ( H τ h τ + W ( To alleiate the fading impairment, the I signal processing is selected by the antenna selection criteria, and the selection is based on the receie SI, where the receie antenna with the largest receied SI is chosen. Thus, een if some of the receied ersions are deeply faded, it is probable that not all copies are faded. The block of the proposed I scheme is depicted in Fig.. 3 Proposed I Scheme In this section, we propose a noel I scheme with multiple receie antennas and derie the threshold setting in detail. Fig.. Block of the Proposed I scheme The MAI can be considered as an independent Gaussian random ariable with zero mean and E-ISS: Volume 4, 5

4 unknown power ( σ MAI, and is independent for ( different antennas. Hence, W m is a Gaussian ariable with power ( σ = ( σ MAI + δ z. From (6, for gien H, the power of the MAI plus noise on the m th receie antenna is obtained as H H ( σ = E[ ( W W = ( Y Y ( From (7, the signal power of path { τ, } on the m th receie antenna under H is estimated by yˆ ( τ ( τ, τ,, M t = = (3 The estimated signal-to-interference-and-noise ratio (SI of path { τ, } on the mth receie antenna is obtained as ( ˆ ( / ˆ y σ m, ( τ / σ γ τ = τ = (4 The most possible timing offset for the reference code on the m th receie antenna is estimated by ˆ τ = arg max γ ( τ (5 τ M t A receie antenna is selected by mˆ = arg max γ ( ˆ τ (6 m x If γ ( ˆ τ λ where λ is a predefined threshold which is set according to the false alarm probability acceptable at the BS, the path { τ, ˆ } is alid, so the ( ˆ SS with code index is declared present and ˆ τ m is the estimated timing offset for the actie code. Otherwise, the path { τ, ˆ } is inalid and the th SS is assumed to be absent. 3. Threshold Setting ( m Let q ( τ = ( τ / σ, for gien H, q ( τ is a complex Gaussian random ariable with zero mean and unit ariance. Equation (4 can be written as γ ( τ = q ( τ (7 { γ ( τ, τ =, M t, m=,, x} follow central Chi-square distribution with degree of freedom, and the probability density function (PDF of ( γ m ( τ is ( p, ( m η = e η η (8 γ ( τ In the proposed I method, the SS will pick the antenna with the largest SI defined in (6. Define Zi = max{ γ ( τ, τ =, M t, m=,, x} (9 The cumulatie distribution function (CDF of Z i is λ Mt x P[ Zi λ H ] = [ p ( η dη ] γ ( τ ( λ Mt x = ( e A false alarm could happen if one or more inalid paths reach the threshold λ for all x receie antennas and c referenced I codes at the BS. The probability of false alarm P fa acceptable at the BS can be determined as λ P ( e Mt x c λ = M e ( fa t x c From (, the threshold λ is determined as λ = ln( M / P ( t x c fa 4 Performance analysis In this section, the probability of correct detection is analyzed for the proposed I scheme, and the computational complexity is compared with the AC I method. 4. Probability of Correct Detection For H, when { τ, ˆ } is the alid path for the th actie SS where ˆm is the index of selected antenna, the decision ariable is gien by s= γ ˆ ˆ ( τ = ( τ / σ (3 = h ( ˆ τ / σ + W / σ W / Due to σ is a complex Gaussian random ariable with zero mean and unit ariance, so s is a non-central Chi-square random ariable with degree of freedom. The PDF of s is gien by ( s+ β β p( s = e I ( 4 s (4 where β = h ( ˆ τ / σ, and I ( the zeroorder modified Bessel function of the first kind. The probability of the correct detection is gien by λ P[ s λ H] = p( s ds= Q ( β, s (5 = b. ( where (, x + a Q a b xe I ( ax dx 4. Computational Complexity Consider that a complex multiplication (CM is equialent to four real multiplications (Ms plus two real additions (As, a complex addition (CA is translated into two As, and represents two Ms and one A. Because a M takes much more hardware resources than that of a A, computational complexity of the proposed scheme is ealuated in terms of the number of the Ms. E-ISS: Volume 4, 5

5 Let be the number of subcarriers for an OFDMA symbol, c, M t, and x are the number of the reference I codes, the maximum number of resoled path for each SS, and the number of receie antennas, respectiely. For the proposed I method, the correlation detection can be performed by the -point FFT operation, so it needs about c x log + c M t x Ms, while the AC scheme is 4 c x M t + c M t x. In general, the number of is large, thus, computational complexity of the proposed method is lower than that of the AC scheme. two methods drops, but the proposed method still has a good correct detection performance compared to the AC method. As expected, with x increasing, the probability of correct detection of the proposed method is dramatically superior to that of the AC method. 5 Simulation esults We consider an I system with multiple antennas at the BS, and the system parameters are specified in the IEEE 8.6m []. The number of subcarriers in OFDMA system is 4, and the length of CP is 8 samples. The sample rate is.mhz and the carrier frequency is.5ghz. The cell radius is set to 3km, and the maximum round-trip delay is 4 samples. In the I procedure, an I time-slot consists of 44 non-contiguous subcarriers oer two OFDMA symbols and the number of codes resered for I is 3. The main tasks of the I process at the BS are multiuser I code detection and multi-user timing estimation. The timing requirement according to [] is that all uplink OFDMA symbols for each SS should arrie at the BS within an accuracy of /8 of the length of CP, and which indicates that the timing offset should be within 6 samples. If a SS is correct detected with timing offset estimation within 6 samples, the SS receies the success response and its I process is completed. If a SS is detected with timing offset estimate beyond the range of 6 samples, the SS receies the retransmission response and repeats its I process at the following I attempts. When a SS is not detected by the BS, a missed detection eent occurs. In the simulation, the fading channel is simulated by the WIEII-B channel model and the mobile speed is 6km/hr. Comparisons are made between the proposed method and the AC scheme under the false alarm probability is limited within -3. Fig. shows the probability of correct detection ersus number of actie SSs in one I time-slot, where the S for each SS is equal to 9 db. It is shown that the correct detection probability of the proposed method is much better than that of the AC method under the same receie antennas x condition. With the number of actie SSs increasing, the probability of correct detection by Fig.. Probability of correct detection s. number of actie SSs Fig.3 shows the standard deiation of timing offset estimation ersus number of actie SSs in one I time-slot at S equal to 9 db. Clearly, the proposed method exhibits a superior timing estimation performance compared to the AC method. With the number of actie SSs increasing, the timing performance of the proposed slightly declines while that of the AC method seerely degrades. It also reealed that the timing performance of the two methods is improed with x increasing. Fig. 3. Standard deiation of timing s. number of actie SSs Fig.4 shows the probability of successful detection of the two methods ersus number of actie SSs in one I time-slot at S equal to 9 db. It can be seen that the successful detection E-ISS: Volume 4, 5

6 performance of the proposed method is distinctly superior to that of the AC method. Duo to the worse correct detection performance and timing accuracy of the AC method, the successful detection performance of the AC method seerely degrades, een if x increases. It also reealed that the proposed method can dramatically improe the I efficiency for each SS to access into network with a small number of retransmissions compare with the AC method. offset estimation ersus Ss with x equal to 8. It can be seen that the timing accuracy of the two methods is improed with the S increasing. As K increases, the timing performance of the two methods declined. Howeer, the timing accuracy of the proposed method is much higher than that of the AC method, een if the number of actie SSs in one I time-slot is great. Fig. 6 Standard deiation of timing s. Ss Fig. 4 Probability of successful detection s. number of actie SSs Fig.5 shows the probability of correct detection as a function of Ss with different number of SSs in one I time-slot, where the receie antennas x at the BS is equal to 8. As the S increases, the correct detection performance of the two methods is improed, but the proposed method has a much better correct detection performance than that of the AC method. When the number of SSs K is 8 and the S is equal to 8dB, the probability of correct detection of the proposed method approaches the alue of.95, and which is increased by % compared with the AC method. Fig.7 shows the probability of successful detection as a function of Ss with x equal to 8. Obiously, the successful detection performance of the proposed scheme is much better than that of the AC method with the S increasing. As K increases, the probability of successful detection of the proposed method slightly decreases while that of AC method seerely degrades. This indicated that the proposed scheme can accommodate more actie SSs simultaneously to access into networks than the AC method. Fig. 7 Probability of successful detection s. Ss Fig. 5 Probability of correct detection s. Ss Fig.6 shows the standard deiation of timing 6 Conclusion E-ISS: Volume 4, 5

7 A noel I algorithm is proposed in MIMO- OFDMA uplink systems. By selecting the receie antenna with the largest receied SI, the proposed scheme achiees significant gains in I signal detection with great number of SSs. The simulation results show that the proposed method enhances the I performance and increases the capacity of actie SSs compared with the AC scheme. The proposed method not only can be used to enhance I performance for MIMO-OFDMA systems based on the IEEE 8.6 standard, but also can be applied into the random access for the Long Term Eolution (LTE systems. Acknowledgement This work was supported by the Key Scientific and Technological Project of He'nan Proince of China ( 389 and by the Applied esearch Progra ms of Science and Technology of Luoyang ormal Uniersity of China (4-YYJJ-. eferences: [] IEEE 8.6m, IEEE standard for local and metropolitan area networks, part 6: air interface for broadband wireless access systems amendment 3: adanced air interface, May. [] X. Fu, H. Minn, Initial uplink synchronization and power control (ranging process for OFDMA systems, IEEE Global Telecommunication Conference, Dallas Texas: IEEE, 4, pp [3] D. H. Lee, OFDMA uplink ranging for IEEE 8.6e using modified generalized chirp-like polyphase sequences, in Proc. International Con-ference in Central Asia on Internet (5, Bishkek, Kyrgyz epublic, Sept. 5, pp. 5. [4] Y. Zhou, Z. Zhang, and X. Zhou, OFDMA initial ranging for IEEE 8.6e based on timedomain and frequency-domain approaches, in Proc. 6 Int. Conf. on Commun. Techn., pp. -5. [5] H. A. Mahmoud, H. Arslan, and M. K. Ozdemir, An efficient initial ranging algorithm for WiMAX (8.6e OFDMA, Computer Communications, ol. 3, no., Jan. 9, pp [6] X. Fu, Y. Li and H. Minn, A new ranging method for OFDMA systems, IEEE Trans. wireless Commun., ol. 6, no.,, Feb. 7, pp [7] J. Zeng and H. Minn, A noel OFDMA ranging method exploiting multiuser diersity, IEEE Transactions on Communications, ol. 58, no. 3, Mar., pp [8] M. Morelli, L. Sanguinetti and H. V. Poor, A robust ranging scheme for OFDMA-based networks, IEEE Trans. Commun., ol. 57, no.8, Aug. 9, pp [9] L. Sanguinetti, M. Morelli, and H. V. Poor, An ESPIT-based approach for initial ranging in OFDMA systems, IEEE Trans. Commun., ol. 57, no., o. 9, pp [] M. uan, M. C. eed, and Z. Shi, Successie multiuser detection and interference cancelation for contention based OFDMA ranging channel, IEEE Trans. Wireless Commun., ol. 9, no., Feb., pp [] C. Lin and S. Su, A robust ranging detection with MAI cancellation for OFDMA systems, in Proc. Int. Conf. on Ad. Commun. Technol., pp [] C. Lin and S. Su, A differential successie multiuser ranging detection for OFDMA systems, in Proc. Int. Conf. on Wireless Communications and Signal Processing,, pp. -5. [3] L. Sanguinetti and M. Morelli, An initial ranging scheme for the IEEE 8.6 OFDMA uplink, IEEE Trans. Wireless Commun., ol., no. 9, Sep., pp [4] J. Zeng, H. Minn and C. Chong, anging signal designs for MIMO-OFDMA systems, IEEE Global Telecommunication Conference, ew Orleans: IEEE, 8: -6. [5] J. Zeng, and H. Minn, Diersity exploiting MIMO-OFDMA ranging, IEEE International Conference on Information, Communications & Signal Processing, Zhengzhou: IEEE, 7: -5. [6] Sanguinetti L, Morelli M and Marchetti L, A random access algorithm for LTE systems, Transactions on Emerging Telecommunications Technologies, ol. 4, no., Jan. 3, pp E-ISS: Volume 4, 5