The impact of PSS autocorrelation on cell search based on robust maximum likelihood scheme in LTE system


The respected Comrade Kim Jong Un said:

"The sector of science and technology should conduct a vigorous campaign to surpass the cutting edge."

Robust Maximum Likelihood scheme for primary synchronization signal (PSS) detection and integer frequency offset (IFO) in Long Term Evolution (LTE) system has the best performance of cell search. This scheme provides good synchronization performance based on correlation properties of Zadoff-Chu (ZC) sequences and reduced-rank representation of channel frequency response in multi-path propagation fading environment. The three ZC sequences used for PSS in LTE system cannot be said to have sufficient correlation properties in considering impact of symbol timing error. In this paper, we evaluate impact of the autocorrelation of PSS on PSS detection and IFO estimation and raise an issue of using other ZC sequence with better correlation properties. The simulation results show that PSS detection and IFO estimation performances are improved by using proposed ZC sequence.

Also, in LTE system, typical cell search procedure in cellular systems such as Wideband Code Division Multiple Access (WCDMA) or Global System for Mobile Communications (GSM) system must be performed, through which user equipment (UE) recovers the cell-ID of the serving base station and acquires correct timing and frequency synchronization. The cell search procedure must be firstly performed after UE is switched on and it is performed in presence of carrier frequency offset (CFO) and symbol timing offset. In LTE system the cell search is accomplished by exploiting a dedicated synchronization channel (SCH) periodically inserted in DL radio frame to convey Primary Synchronization Signal (PSS) and Secondary Synchronization Signal (SSS). LTE cell search procedure consists of three steps. Firstly, the receiver utilizes the cyclic prefix to estimate fractional frequency offset (FFO) and coarse symbol timing information. Then, integer frequency offset (IFO) and sector identity are jointly detected by using PSS. Last, the fine timing information and cell group identity are acquired by using SSS.

Since the performance of cell search decisively depends on correct detection of PSS, the literatures have focused on PSS detection and IFO estimation in the last few years. Time domain detection schemes evaluate peak output of matched filter to identify position of PSS and sector identity. Frequency domain schemes can jointly estimate sector identity and IFO based on correlation between PSS and SCH subcarriers of the receive discrete frequency transform (DFT).

Robust maximum likelihood scheme using AMMSE reduced-rank representation provides high estimation performance because it effectively compensates the channel response to improve detection performance of the PSS. However, for AMMSE scheme to provide higher performance, correlation properties of the PSS must be perfect.

In this paper, we evaluate the correlation properties of Zadoff-Chu (ZC) sequences included in PSS in the presence of CFO and timing error and investigate impact of correlation of PSS on the performance of AMMSE scheme. ZC sequences a25(n), a29(n) and a34(n) used as PSS have good correlation properties but autocorrelation of a25(n) does not show single peak characteristics in case of OFDM symbol timing error. Also, in viewpoint of correlation chart it is not sufficiently satisfied and hence the value of likelihood function based on AMMSE reduced-rank representation of CFR can be negative results.

In LTE system the root indices of ZC sequences used for PSS are u∈{25, 29, 34}. These indices are selected to give a good autocorrelation and cross-correlation properties. For the PSS, the basis is the two sequences corresponding to root indices u0=25 and u1=29. The third index is derived by subtracting u2=NZC – u1 = 34. The sequences based on indices u0 and u1 show the best auto-correlation and cross correlation properties. The reason for selecting u1 and u2 depends on the very good conjugate complex symmetry of these two sequences in the time domain. This reduces the processing effort required for synchronization.

However, the cross-correlation between a25(n) and a34(n) does not exhibit sufficiently low correlation. AMMSE scheme that use uniform delay spread profile channel model is affected by autocorrelation properties of ZC sequence with root index 25. When ZC sequence with root index 23 that has better autocorrelation and cross-correlation properties is used the IFO estimation performance of robust ML scheme is substantially improved.

The results was published in the Journal "Digital Signal Processing" under the title of "The Impact of PSS Autocorrelation on Cell Search Based on Robust Maximum Likelihood Scheme in LTE System" ( j.dsp.2020.102958).