An Effective Approach to Detect Random Access ... - ScienceDirect.com

Available online at www.sciencedirect.com Available online at www.sciencedirect.com

Procedia Engineering

ProcediaProcedia Engineering 00 (2011) Engineering 15 000–000 (2011) 2339 – 2343 www.elsevier.com/locate/procedia

Advanced in Control Engineering and Information Science

An Effective Approach to Detect Random Access Preamble in LTE Systems in Low SNR Ping’an Lia, Bin Wua,a* School of Information Engineering,Wu Han University of Technology,Wu Han 430070,China

Abstract In the long time evolution systems, random access is a very important procedure for user equipment to request resources from base stations. In the paper, a receiver detection scheme is presented. A down sampling technique is proposed to lower the computational complexity. Multiple degree of freedom Chi-square distribution properties of the additive whiten Gaussian noises are used to set detection threshold for multi-antenna non-coherent combination. A detection algorithm for low signal-to-noise ratio scenarios (SNR) is proposed. The simulation results show that the scheme proposed can improve the performance of detection in the low SNR scenarios.

© 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license. Selection and/or peer-review under responsibility of [CEIS 2011] Keywords: Random access; preamble detection; low SNR;

1. Introduction To ensure competitiveness for the next 10 years and beyond, the 3rd generation partnership projiect (3GPP) started the long time evolution (LTE) in November 2004. LTE, sometimes referred to as 3.9G or Super-3G, will enable performance enhancement over the existing 3G technology through improved coverage and system capacity as well as increased data rates and reduced latency. Assuming a 20 MHz spectrum allocation, 2 receive antennas in the base station (BS) and 1 transmit antenna in the user

* Corresponding author. Tel.:+86 13307183193. E-mail address: [email protected].

1877-7058 © 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license. doi:10.1016/j.proeng.2011.08.438

22340

Ping’an Wu / Procedia Engineering 15 000–000 (2011) 2339 – 2343 Pingan Li, Li ，and et al/Bin Procedia Engineering 00 (2011)

equipment (UE), the system should support an instantaneous peak data rate of 100 Mbps in the downlink and 50 Mbps in the uplink [1]. Random Access Channel (RACH) in LTE systems is mainly used for UE to request resources from BS [2]. In the physical random access channel (PRACH), the Zadoff-Chu (ZC) sequences are used as preambles. This is based on the fact that the ZC sequences have the ideal auto-correlation and crosscorrelation properties [3]. Different ZC sequences or the same ZC sequence with different cycle shifts are used for generating preambles by different UEs [4] [5]. In high SNR scenarios, the receiver can detect the access preamble easily, while in low SNR scenarios, however, the performance of detection will become poor. In this paper, a simple and effective receiving scheme in base-band for detecting the preamble signal in the random access of LTE frequency division duplex systems is presented. An effective signal pre-processing approach is proposed to improve performance of detection in low SNR scenarios by applying smoothing operation on more than one preamble burst. 2. PRACH transmitter structure In the transmitter, a ZC sequence with length 839 is generated for Format0, see details in [4], before mapping to subcarriers. An IFFT module is then used to convert the signal from the frequency domain to the time domain. If the preamble format is Format2 or Format3, the data samples after IFFT should be repeated, and the length will be twice as that in Format0. After the cyclic prefix (CP) insertion, the signal will be send to the radio frequency processing module. The base band signal in the time domain for PRACH is defined by Nzc −1 Nzc −1

s(t) = βPRACH ∑

∑x

k =0 n=0

u,v

exp{− j2πnk / Nzc}exp{j2π[k + ϕ + K(k0 +1/ 2)]Δf RA(t − Tcp )}, 0 ≤ t < TSEQ + TCP ,

(1)

xu , v is a cyclic shifted ZC sequence, see [4] for details, N ZC is the length of the ZC sequence. The variable ϕ is a fixed frequency offset [4] where Δf RA is the subcarrier spacing for the random access preamble,

and K = Δf / Δf RA accounts for the difference in subcarrier spacing between the random access preamble RA UL RA being a parameter and the uplink data transmission. In [4], k 0 = n PRB N scRB − N RB N scRB / 2 with nPRB UL derived in 5.7.1 in [4] for controlling the location of the preamble signal in the frequency-domain, N RB

and N scRB denote the uplink bandwidth expressed as a number of resource blocks and the resource block size expressed as a number of subcarriers, respectively. ZC sequence of odd length is expressed as ⎡ n ⋅ (n + 1) / 2 + l ⋅ n ⎤ , (2) aq = exp⎢− j ⋅ 2 ⋅ π ⋅ u ⋅ ⎥,0 ≤ n ≤ N ZC − 1 N ZC ⎣ ⎦

where u ∈ {1,...N ZC − 1} is the root ZC sequence index, l can be any integer. In the LTE, l is set to 0 in general. The ZC sequence for PRACH in LTE is defined by

xu (n ) = exp(− j

π ⋅ u ⋅ n ⋅ (n + 1) N ZC

), 0 ≤ n ≤ N ZC − 1 .

(3)

3. Detection algorithm The preamble receiver structure in base band for random access preamble is depicted in Fig.1. The receiver is based on the frequency domain correlation property. The computational complexity of this structure is significantly smaller than that based on the correlation in time domain. This is based on the

Ping’an Li Li, andBin BinWu/ Wu Procedia / ProcediaEngineering Engineering0015(2011) (2011)000–000 2339 – 2343 Pingan

fact that the correlation computation is simpler in the frequency domain than in the time domain. Diversity reception in the base station is assumed with two antennas. After the CP removal in the receiver, A DFT operation should be used to transmit signal from the time domain to the frequency domain. For Format 0, a 24576 point DFT is required. In order to reduce the complexity and delay, down sampling filtering [DSF] is used. In the proposed scheme, a 3 order DSF is employed before DFT. That is, 11 taps symmetric filters, 7 taps symmetric filters, and 14 taps symmetric filters are used in series to shorten the samples to 1536, After the DFT operation, 839 points ZC sequence with fading and noises will be extracted after subcarrier demapping. Furthermore, frequency correlation is carried out before IDFT. The length of input of IDFT module should be padded to 840 due to hardware implementation. At last, peak detection will be carried out in the time domain after non-correlation combination. The threshold setting presented in this section is based on the false alarm probability in additive whiten Gaussian noise (AWGN) environments. When the transmitter does not transmit signal, the receiver signal is complex AWGN only, and both the real part and the imaginary part of a complex AWGN variable satisfy the standard normal distribution, i.e., the Gaussian distribution with mean ‘0’ and variance ‘1’. It is well known that the power envelope of a Gaussian random variable follows the central chi-square distribution [6]. The degrees of freedom k is determined by how many Gaussian random variables are summed together. For complex signals, k = 2 ⋅ N RX , N RX is the number of diversity antennas combined non-coherently. The threshold factor can be determined after antennas combining by the following equation (4) T = G −1 ( PFA ) , where PFA is the desired false alarm probability and G is the function defined as (5) G ( x) = 1 − Fk ( x) , where Fk (x) is cumulative distribution function (CDF) with degrees of freedom k. So if we want to keep the false alarm below 10−3 , and the number of receiver antennae is 2, the threshold factor T should be set to 9.2 approximately, as shown in Fig.2. Notice that T is based on the assumption that the variance of the AWGN is “1”. For a received signal in practice, a modifying factor should be used and a modified threshold can be set as (6) Tdet = T ⋅ m N , A simple approach to evaluate the modifying factor is to calculate mean value of all samples, i.e. 1 NCV (7) mCV = ∑CV(i) , NCV i=1 where N CV is the number of all samples, vector CV holds the correlation samples after non-coherent combination. Consider the fact that the signal component will destroy the whiten Gaussian noise hypothesis. A better approximation of the actual noise level is to discard peaks higher than a predefined level before the mean calculation and, thus, a little more accurate threshold value can be achieved. That is, the noise level estimate is achieved by re-calculating mean value from those samples that are smaller than N ⋅ mCV

1 Ns (8) ∑(CV (i) < N ⋅ mCV), NS i=1 where N S is the number of samples that fulfil the condition given in (8). N can be theoretically derived for AWGN based on the false alarm probability. For example, if we want to cut off 10% of samples (highest peaks) , N can be set to 4 as shown in Fig.2, when the number of receiver antennas is 2. mN =

2341 3

42342

Ping’an Wu / Procedia Engineering 15 000–000 (2011) 2339 – 2343 Pingan Li, Li ，and et al/Bin Procedia Engineering 00 (2011)

In low SNR scenarios with higher energy of noise, the correlation characteristics of the ZC sequence become poor. It is unable to detect the correlation peaks when the SNR decrease to -18dB, as shown in Fig.3a. An effective approach to increase the performance of detection in low SNR is to smooth the noise before correlation. The smoothing process is illustrated by Fig.4. First, compute the mean value of received preamble1 and preamble2, measured as m1. Then, compute the mean value of m1 and preamble3, measured as m2, and go on. Of course, the more the preambles used, the better the performance achieved. The cost of the approach is that the receiver needs to store the last previous preamble average value. The domain correlation peak becomes obvious after noise smoothing, as shown in Fig.3b. 4. Numerical Results

The simulations are executed on the condition that the system bandwidth is 20MHz with carrier frequency 2.6GHz. The ideal AWGN channel is assumed. Preamble Format 0 [4] is used. A potential detection for the random access of an UE is defined by finding a correlation peak within the search window which exceeds the threshold. The probability of a missed detection PM is specified simply as N (9) PM = 1 − D , N PRE where N D is the number of preambles detected and N PRE is the total number of preambles sent. In the simulations, 10 preambles is used in smoothing processing for suppress the noise. The results are given in Tab.1. It can be seen from the table that in low SNR scenarios, the probability of missing detection decreases obviously by using noise smoothing pre-processing. Table 1 Probability of missing detection SNR(dB)

-17

-18

-19

PM ( without smoothing)

88.59%

96.10%

98.20%

PM ( with smoothing)

2.80%

13.71%

34.53%

Acknowledgements

This work is supported by the National Special and Important Project of China under Grant: 2009 ZX03002-009. References [1] 3GPP TSG TR 25.913 v8.0.0, Requirements for UTRA and UTRAN. Dec. 2008. [2] 3GPP TS36.213 v8.4.0. Physical layer procedures (Release 8). September 2009. [3] Mansour, M.M., “Optimized architecture for computing Zadoff-Chu sequences with application to LTE,” GLOBECOM 2009. IEEE , Nov. 30 2009-Dec. 4, pp. 1-6, 2009.s [4] 3GPP TS36.211 v8.4.0. Physical Channels and Modulation (Release 8). September 2009. [5] 3GPP TSG RAN WG1 Meeting #44bis, R1-060998.Ericsson.E-UTRA Random Access Preamble Design. March, 2006. [6] Wilson, E. B, Hilbert, M.M, “The distribution of chi-square,” Proceedings of the National Academy of Sciences, Washington, 17, 684–688, 1931.

Ping’an Li Li, andBin BinWu/ Wu Procedia / ProcediaEngineering Engineering0015(2011) (2011)000–000 2339 – 2343 Pingan

*

2

*

2

Fig.1 Receiver structure

Fig.2 Derivation of threshold factor

Fig.3a Correlation before smoothing

Fig.3b Correlation after smoothing

m1

m2

Fig.4 Approach for noise smoothing

2343 5