Cell Search Techniques in LTE

Cell Search Techniques in LTE

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55

Abstract
LTE is a part of evolved UMTS beyond HSPA and the air interface differs from WCDMA in three major areas the downlink waveform is based on OFDM, the system supports a variable system bandwidth ranging from 1.25 MHz to 20 MHz and finally, there is maximum commonality between FDD and TDD designs of LTE.. The LTE cell search approach has been carefully designed keeping the overall search space in consideration, while minimizing the complexity of the terminal. This paper presents an overview of cell search techniques which have been standardized in LTE. The topics include the baseline cell search procedure, synchronization signal design and interference management across synchronization signals in different cells. Further, the paper highlights the design options for two-step cell search procedure and provides the performance comparison between two-step cell search procedure and threestep cell search procedure which has been standardized in WCDMA systems.
Index TermsCell search, OFDM, LTE, E-UTRA, WCDMA, UMB, SFN

1. Introduction
Cell search techniques for LTE have been standardized over the past one year and contain the following aspects: Cell Identity (ID) identification procedure including symbol timing and radio frame boundary timing identification Synchronization signal design Synchronization interference management across different cells. The cell search procedure in WCDMA systems is divided in three steps [17]: Symbol timing detection using Primary synchronization signal (PSC) Radio frame boundary and cell ID group detection using Secondary synchronization signal (SSC) Cell identity (ID) identification using primary common pilot channel (P-CPICH). The key difference between LTE and WCDMA is that in the former, the cell ID identification finishes in two steps as shown in Figure 1: PSC and SSC uniquely provide the cell ID identification [1][2][4][5] [6]. In Figure 1, PBCH is the physical broadcast channel and RS is the reference signal in LTE while P-CCPCH is the primary common control physical channel in WCDMA.

56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80

Figure 1: Cell search procedure comparison: LTE vs WCDMA In this paper, we provide an overview of cell search techniques in LTE applicable to both initial acquisition and neighboring cell search. Considering the significance of adopting a two-step cell search procedure in LTE, we highlight several design options and compare the performance of two-step cell search procedure and three-step cell search procedure.

2. LTE Synchronization Overview

2.1 Synchronization signals
This section briefly describes the main features of the synchronization signals in LTE systems. 2.1.1 Synchronization bandwidth The synchronization signals (PSC and SSC) in LTE occupy the center 1.08MHz 1 bandwidth Figure 2 irrespective of the system bandwidth.
20-MHz bandwidth SCH 10-MHz bandwidth

5-MHz bandwidth

2.5-MHz bandwidth

1.25-MHz bandwidth

81 82 Figure 2: PSC/SSC placement in frequency 83 domain 84 85 2.1.2 Radio frame structure in LTE and PSC/SSC 86 placement in time domain 87 Downlink and uplink transmissions in LTE are 88 organized in radio frames of 10ms duration. Two 89 frame structures are supported by LTE:

Only 0.96MHz bandwidth is used for transmission.

Cell Search Techniques in LTE

1 2 3 4 5 6 7 8 9

Frame structure type 1 (FS1), applicable to both FDD and TDD, Frame structure type 2 (FS2), applicable to TDD only

51 52 53 54 55 56

measurement periods in GSM are approximately 5ms long. In order to enable radio-frame boundary detection, the SSC in slot 0 and in slot 10 are different. The combination of the PSC and SSC provide the LTE cell ID.

FS2 numerology is designed to support backward compatibility with TD-SCDMA. In the remaining of this paper, we will only consider FS1 numerology, which is depicted in Figure 3.

10 11 Figure 3: Downlink frame structure in LTE 12 13 The time durations of interest are: 14 15 Radio frame: The radio frame length in 16 LTE is 10ms. 17 Subframe: A radio frame consists of 10 18 subframes of 1ms duration each. 19 Slot: Each subframe consists of two slots of 20 0.5ms duration each. Depending on the 21 cyclic prefix (CP) configuration of a LTE 22 system, the slot structure varies accordingly. 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 o For the normal CP configuration (short CP), each slot consists of 7 OFDM symbols. The CP duration is approximately 4.7us and the subcarrier spacing is 15 KHz. Note that the CP duration in a slot is not the same for all the OFDM symbols: the first OFDM symbol has a slightly longer CP duration compared with the others. For the extended CP configuration (long CP), each slot consists of 6 OFDM symbols. The CP duration is 16.66us and the subcarrier spacing is 15 KHz. For the MBSFN dedicated carrier configuration (DL stand-alone SFN broadcast carrier), each slot consists of 3 OFDM symbols. The CP duration is 33.33us and the subcarrier spacing is 7.5 KHz.

57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92

Figure 4: PSC/SSC placement in time domain Although LTE systems support three different CP lengths, only short CP and long CP can be used in the subframe where PSC/SSC are sent. The CP length needs to be detected blindly during initial acquisition. Note that the time difference (5ms) between 2 PSC symbols is independent of the cyclic prefix (CP) length used in the system: this enables PSC combining across 5ms periods before detecting the CP length during initial acquisition.

2.1.3 Downlink reference (DL-RS) structure Figure 5 shows the placement and multiplexing of the downlink reference signal (RS), PBCH, physical downlink shared channel (PDSCH), physical downlink control channel (PDCCH), and ACK channel in subframe 0 for the short CP case. Up to 4 Tx antennas at the eNode-B are supported in LTE. The RS from Tx antenna 3 and 4 is sent in OFDM symbol 1 while RS from Tx antenna 1 and 2 is sent in both OFDM symbol 0 and 4 of the slot. RS is sent only in three OFDM symbols in each slot RS from different Tx antennas are time and frequency domain multiplexed RS spans the entire system bandwidth RS density is 1/6 in one OFDM symbol: one subcarrier every 6 This is different from WCDMA systems where PCPICH is transmitted continuously in time. During initial acquisition, the UE is not aware of the system bandwidth nor the number of Tx antennas until decoding of PBCH.

Figure 4 shows the details of the first and the 10th slots in one radio frame for the short CP case. Note that this structure enables detection of the LTE cell ID and radio-frame boundary from a single observation period of 5ms. This feature is designed to facilitate an easy mechanism for GSM-to-LTE handover since the available continuous

93 94 Figure 5: RS placement in time domain: PDCCH 95 spans three OFDM symbols in this illustration 96

Cell Search Techniques in LTE

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59

3. Synchronization design issues

3.1 SFN effects
One of the design goals for LTE is maximum commonality between the FDD and TDD modes of operation. TDD systems require synchronous operation while FDD systems could be synchronous or asynchronous. Also, as it was discussed before, the design should enable LTE cell ID identification from a single 5ms observation in order to facilitate the handover from GSM, where the measurement window is approximately 5ms in duration. It has been shown that coherent detection of SSC improves cell search performance [2]. Note that during initial acquisition, PSC is the only phase reference available for SSC. In synchronous systems, if only one PSC sequence is used in the system and it is located in the same position across different cells, there would be a large phase mismatch between the detected PSC (recall that PSC occupies the same tones in all cells) and the SSC (which is unique to the cell group). This is called SFN effect of the PSC. The SFN effect may happen more often at the cell edge where the cell search performance is of most interest, especially for neighbor cell search. Thus, the synchronization signal design should avoid or minimize this SFN effect. One way to achieve this is by using multiple PSC sequences in the system. Alternatively, we could use a single PSC and randomize the location of PSC across different cells. Knowing that the computational complexity of symbol timing detection is very high, even for one PSC sequence [1], the design should balance the UE complexity and the system performance. Note that while the SFN effect causes channel phase mismatch between the PSC and the SSC, it does increase the received energy of PSC. A different design choice is to exploit this SFN effect instead of avoiding them [2].

3.2 Synchronization design issues

The main design issues in LTE for synchronization are: Hypothesis partitioning among PSC, SSC and RS Coherent vs. non-coherent SSC detection Avoid vs. exploit SFN effect? PSC signal structure: repeated vs. nonrepeated structure Exact bandwidth occupancy of PSC/SSC SSC signal structure

60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116

The main goal is to characterize the cell search time for initial and neighbor cell search.

3.2.1 Cell ID identification procedure

To ease the deployment, approximately 510 cell IDs are needed in LTE systems. This is similar to WCDMA systems, where 512 cell IDs are defined. The key differences between LTE and WCDMA in this respect are as follows. PSC/SSC are present twice per 10ms in LTE o 15 times per 10ms in WCDMA PSC and SSC are time multiplexed in LTE o Code multiplexed in WCDMA DL-RS is time multiplexed with PSC/SSC in LTE o PSC/SSC are non-orthogonal to PCPICH in WCDMA DL-RS is not transmitted continuously in time in LTE o P-CPICH is continuously transmitted in WCDMA The differences above justify redesigning the cell ID identification procedure for LTE. During initial acquisition (before decoding of PBCH) UE is not aware of system bandwidth o Only DL-RS in the center 1.08MHz can be used UE does not know whether the system is TDD or FDD o Only subframe #0 and #5 are known to be downlink subframes o Thus, only RS in those two subframes may be used for cell ID identification UE does not know the number of transmit antennas o Only the RS from the first Tx antenna may be used for cell ID identification In one radio frame, some slots may be configured with short CP while other slots may be configured with long CP. The only assumption that the UE can make is that CP length in subframe #0 and #5 is the same. o Only RS signal in those two subframes may be used for cell ID identification In summary, only 48 RS tones are usable in each 5ms observation interval for initial cell search. This corresponds to the RS subcarriers in the center 1.08MHz, from the first Tx antenna and only in subframe 0 or subframe 5.

Cell Search Techniques in LTE

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52

10 ms radio frame

53 54
PSC SSC Reference Symbol

This reduces the second stage hypotheses from 510 to 170

1 ms
1 2 3 4 5 6 7 1 2 3 4 5 6 7

55 56 57

SSC carries 170 hypotheses for cell ID identification

-0.5 ms

Figure 6: RS signal placement with respect to PSC/SSC in time domain

Figure 6 shows an example of RS placement for the short CP case. PSC/SSC and 4 RS symbols (highlighted in green) within the same TTI are separated by Two reference symbols in the first slot o o 5 OFDM symbols from SSC and 6 from PSC 1 OFDM symbols from SSC and 2 from PSC

58 59

Figure 7: PSC/SSC placement for short CP

Two reference symbols in the second slot o o 1 OFDM symbols from PSC and 2 from SSC

60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96

Figure 8: PSC/SSC placement for long CP

3.2.3 PSC signal design

3.2.3.1 Sequence consideration
PSC sequences should be designed in a way such that they Have good auto-correlation and crosscorrelation properties Enable low complexity implementation. In [7][14], a general framework of designing three different PSC sequences based on one good PSC sequence has been proposed: Choose a good base sequence Apply permutation or scrambling to get three different PSC sequences The resulting benefit is to maximize reusing the same efficient correlator for the base sequence at the receiver. The PSC sequences adopted in LTE requires two correlators, i.e., one correlator is reused.

5 OFDM symbols from PSC and 6 from SSC Thus, if coherent detection of RS is required using PSC/SSC as the phase reference, there may be a large channel estimation loss due to the large separation in time between PSC/SSC and RS symbols. Moreover, during initial acquisition, there may exist a large frequency offset between the incoming signal and the UEs local oscillator. While PSC may be used for frequency offset estimation after symbol timing has been detected, the estimate may be very noisy especially for low SNR. Due to the large separation in time domain between PSC/SSC and RS signal, only non-coherent accumulation of 4 RS symbols in one TTI may be feasible due to the large residual frequency offset [3]. Thus, using DL-RS to carry cell ID information is not a desired structure in LTE [1][2].

3.2.3.2 Number of tones used for PSC

The number of tones used for PSC is another important consideration. While up to 72 tones can be used for PSC, to enable low complexity PSC detection, up to 64 tones are used for PSC to allow for a sampling rate at 0.96MHz rather than 1.92MHz [8].

3.2.2 Two-step cell search procedure

Figure 7 and Figure 8 show the adopted PSC/SSC placement for short CP and long CP respectively. Note that three different PSC sequences carry group ID information. Three groups ( , and represented

) can be

3.2.3.3 PSC time domain signal property

To enable low complexity implementation of the PSC detector, a popular design is to use repetitions of one sequence in time domain. While sequence

Cell Search Techniques in LTE

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55

repetition in time domain enables auto-correlation based time synchronization which is popular in high SNR applications, it is not adequate for cellular systems where low SNRs are of most interest. Furthermore, with neighbor cell search, the two extra peaks (3dB below the peak at the desired timing) of the stronger cells due to 2X repetition may hide the peaks from other weaker cells.

56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76

The maximum false frequency estimation error (the difference between the true frequency offset and the estimated frequency offset) is over all possible timing offset errors (chipx32, up to one Tc where Tc corresponds to one sample at 1.92MHz sampling rate) and the estimator uses the phase-differential based method. The assumptions for those plots are: Single path static channel model Noise free Frequency offset ranging from 100Hz to 10kHz Sinc interpolation is used to generate the ZC based over sampled time domain waveform For length 64 ZC sequence, those 64 samples are mapped to 64 tones including the DC tone o o DC tone is not set to zero (original ZC sequence in the plot) DC tone is set to zero (DC tone set to zero ZC Sequence in the plot)

3.2.3.4 PSC frequency domain property

It is well-known that a flat spectrum minimizes the frequency domain channel estimate mean-square error (MSE). For coherent detection of SSC, a flat spectrum of PSC is desirable though not essential. The adopted PSC sequences in LTE are based on frequency domain defined Zadoff-Chu (ZC) sequences. Some useful properties of ZC sequence are [8]: There exists duality for ZC sequence defined in time or frequency domain. For example, for length 64 ZC sequence, frequency defined ZC sequence with root index 31 is equivalent to time domain defined ZC sequence with index 33. ZC sequence is defined only for a discrete set of samples. After band-limiting or timelimiting operation at the transmitter, the interpolated samples (the samples between the original ZC sequences) are no longer ZC sequences.

While ZC based PSC sequence has flat spectrum, there may be time and frequency ambiguity due to the phase ramping property of ZC sequences. To achieve better SSC detection, the UE may use PSC to estimate the frequency offset before detecting SSC: this is regardless of coherent or non-coherent detection of SSC. For ZC based PSC sequences, a simple method to estimate the frequency error is the phase-differential based method which is summarized as follows: 1. Divide the received signal into two parts 2. 3. 4. De-spread each part by the detected PSC sequence and sum the results Get the differential product of the two summed results Estimate the frequency offset based on the phase of the differential product

77 78 79 80 81

Figure 9: Frequency offset estimation capability of ZC based sequence: sequence length 64

Since there always exists residual timing offset (depending on sampling rate) between the true timing and the detected timing, in Figure 9, we plot the maximum false frequency estimation error as a function of ZC root index defined in frequency domain

82 83 84 85

Figure 10: Frequency offset estimation capability of ZC based sequence: sequence length 63

Cell Search Techniques in LTE

1 2 From these plots, we can observe 3 The frequency offset estimation capability 4 of ZC sequence is a function of 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 o o ZC root index In general, ZC root indexes in the middle range are better

51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94

Figure 11: PSC/SSC sequence mapping in frequency domain (63 tones are used including DC tone)

In LTE, length 63 ZC sequence with DC puncturing is used. Three different ZC sequences are chosen such that two of them are complex conjugated of each other in time domain. Note that those three sequences in frequency domain are derived from one base sequence with complex conjugate or modulated by a periodic signal.

3.2.5 Interference management for SSC

PSC/SSC pairs can be assigned arbitrarily to each cell by the operator. In that sense, any form of planning is not mandated in LTE specifications. Figure 12 shows one example for bad PSC/SSC cell planning: different PSCs are assigned to three cells in one eNode B but the same SSC is used for all three cells. This deployment results in large channel mismatch between PSC and SSC and makes coherent detection of SSC difficult. Note that SSC has other applications: can be used as an additional phase reference for PBCH demodulation for single transmit antenna systems can be used for time tracking or frequency tracking Since SSC contains two short sequences, there are many cases where two different SSCs share one short sequence. Thus it is necessary to minimize the channel mismatch between PSC and SSC: randomizing the interference across different SSC sequences by using scrambling on top of SSC is a good design choice [13]. The scrambling sequences can not be eNode-B specific since the UE does not know the eNode B cell ID before SSC detection. Also the scrambling sequences can not be common across cells since it does not help in synchronous systems. [13] proposed to use three different scrambling sequences on top of SSC. Each scrambling sequences is linked to one of three PSC sequences such that once PSC sequence is identified, UE knows the scrambling sequence for its SSC. The advantage of this method is that it reduces the interference across different cells/sectors regardless of synchronous or asynchronous systems. The same scheme was also proposed in [16].

3.2.4 SSC signal design

SSC should provide sufficient number of hypotheses and preferably have the following features. Good cross-correlation properties Enable low complexity implementation. Robust against frequency selective channel Two concatenated short sequences have been chosen for SSC in LTE in order to provide sufficient number of hypotheses. There were two main candidates for the short sequence: Hadamard based SSC [12] M-sequence based SSC [9][15] M-seq based SSC has been adopted in LTE based on [9]. The detailed signal is as follows. Let S(0,n) represent one M-sequence generated from the primitive polynomial x5+x2+1 over GF(2), i.e. S(0,n)= {1,-1,-1,1,1,1,1,1,-1,-1,-1,1,1,-1,1,1,1,-1,1,1,1,-1,-1,-1,-1,1,-1,-1,1,-1,1}. Define S(k,n)=S(0,(n+k)mod N) where N=31 k=1,2,30 S(k,n) is a cyclic shift (k shifts) of S(0,n)

38 Thus, we have N sequences in total including S(0,n) S(0,n), S(1,n).S(30,n) 39 40 If we stack those N sequences, we get an M-sequence 41 matrix. 42 The properties of using M-sequence based SSC are: The cross-correlation between any two rows 43 44 of M-sequence matrix is -1. 45 SSC is based on the interleaving of S(m,n) and S(v,n) 46 in frequency domain where m, v =0,1,2,.30, i.e., 47 48 49 50 [S(m,0), S(v,0).,S(v,14),S(m,15), 0, S(v,15),S(m,16), S(v,16),, S(m,30), S(v,30)] They are mapped to the centre 63 tones

Cell Search Techniques in LTE

A2: Joint time dithering 45 approach with uniform 46 distance between 2 PSCs 47 Multiple PSCs are used in the system and 48 they are located in the same position 49 o The sequences themselves carry 50 partial cell ID information 51 o This option was adopted in LTE 52 53 54 4.1 Time dithering approach: non-uniform 55 dithering [2] 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 Figure 13 and Figure 14 show one possible setting of PSC/SSC placement for A1. Note that the relative distance between the two PSCs carries group ID information as well as CP length information. At least three groups ( , and be represented o

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44

Figure 12: PSC/SSC channel mismatch example

3.2.6 Transmit diversity techniques for PSC/SSC

Different transmit diversity techniques have been proposed for PSC/SSC. PVS (precoding vector switch): This scheme consists of transmitting the PSC/SSC pair on different Tx (virtual) antennas at different OFDM symbols/slots/radio frames. To fully utilize transmit power amplifier at different Tx antennas, a pre-coding vector selected from a unitary matrix is applied before PSC/SSC transmission and the pre-coding vector may change from different slots/radio frames. FSTD (frequency switch transmit diversity): in this scheme, within one PSC/SSC symbol, different tones may transmit from different Tx antennas. PVS was adopted in LTE.

) can

This reduces the second stage hypotheses from 510 to 170

CP length information is also included o Long CP and short CP

PSC in the 9th slot benefits from SFN gains in synchronous systems, i.e., increasing symbol timing detection performance PSC in the 0th slot does not suffer SFN effects and is used as the phase reference for SSC detection

4 Two-step cell search procedure: alternative designs

Several other options for two-step cell search design have been evaluated [1][2]. The key components of those proposals are: Cell ID identification finishes in two steps: using only PSC and SSC Multiple PSC sequences are used to represent partial cell ID information There are different ways to achieve multiple PSC sequences: Only one PSC sequence is used in the system o The location of PSCs or relative distances between two PSCs carry partial cell ID information and/or CP length information [1][2] A1: Time dithering approach with nonuniform distances between 2 PSCs 73 74 75 76 77

Figure 13: PSC/SSC relative position for short CP: non-uniform dithering

78 79 80 81 82

Figure 14: PSC/SSC relative position for long CP: non-uniform dithering

Cell Search Techniques in LTE

1 2 3 4

In Table 1, we list six different relative timings between the 2 PSCs for A1. By comparing different combinations of PSC correlator outputs, UE can uniquely identify one group with short or long CP.

29 30

5 Note that this design does not allow for GSM-to-LTE 6 measurement if GSM window is limited to 5ms and 7 the window is fixed over time. 8
Relative timing Between 2 PSCs 4.5 ms + 286 us 4.5 ms 4.5 ms 4.5 ms 4.5 ms 4.5 ms + + + + + 214 us 0 us 333 us 250 us 83 us

: short CP Group : short CP Group : short CP Group : long CP Group : long CP Group : long CP
Group

31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

Figure 16: PSC/SSC relative position for long CP: joint dithering

4.3 Performance Comparison

In this section, we provide some numerical results to compare two approaches: Approach 1: three-step cell search, i.e., PSC does not carry cell ID information and SSC and DL-RS together carry 510 cell ID information. Approach 2: two-step cell search with nonuniform dithering.

9 Table 1: Six different relative timings between 2 10 PSCs for Approach 2 11 12 13 14 4.2 Time dithering approach: joint dithering 15 The relative timing distance between any two 16 consecutive PSCs is fixed as shown in Figure 15 and 17 Figure 16. 18 19 20 21 22 23 24
P

4.3.1 Single cell performance

We show the average cell search time results in Figure 17 and Figure 18.

The distance is 5ms regardless of long CP or short CP The phase of SSC carries cell group ID information (3-PSK modulation) o This requires coherent detection of SSC
10 ms Radio Frame Short CP
Shared Data M Shared Data

51
6

0.5 ms
0 1 2 SSC1

4.5 ms
PSC 5 6

0.5 ms
0 1 2 SSC2

4.5 ms
PSC 5

Short CP Group

52 53

Figure 17: TU3 channel model and 10 kHz frequency offset

5 ms Short CP Group

SSC1

PSC

SSC2

PSC

5 ms
0 SSC1 PSC 3 4 5 6 0 SSC2 PSC 3 4 5 6

Short CP Group

25 26 27 28

5 ms

Figure 15: PSC/SSC relative position for short CP: joint dithering

Cell Search Techniques in LTE

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Figure 18: TU120 channel model and 10 kHz frequency offset

21 Figure 20: 2GHz carrier, site-to-site distance 22 1.732 Km, TU3 channel model with 20 dB 23 penetration loss 24 25 26 From the results presented, we observe that 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 Approach 2 (two-step cell search) outperforms Approach 1 (three-step cell search) in both single cell and multi-cell cases

4.3.2 Multi-cell performance

Note that all the results presented in this document are generated with a multi-cell link level simulator. Since the performance of cell edge users is of most interest, we show the average cell search time results for users at geometry lower than -3 dB in Figure 19 and Figure 20 for two different deployment scenarios [18].

5. Summary
This paper has presented an overview of the cell search design issues in LTE systems. Major features and components of the cell search design have been reviewed and some performance results comparing two-step and three-step cell search schemes have been presented.

REFERENCES
[1] R1-062691, Link analysis of initial acquisition, Qualcomm Europe, RAN1#46bis, Seoul, Korea, October 2006. [2] R1-063431, Initial cell search: analysis and simulations, Qualcomm Europe, RAN1#47, Riga, Latvia, November 2006 [3] R1-070428, Further analysis of initial cell search for approach 1 and 2-singel cell scenario, Qualcomm Europe, RAN1#47bis, Sorrento, Italy, January 2007 [4] R1-070429, Further analysis of initial cell search for approach 1 and 2-multi cell scenario, Qualcomm Europe, RAN1#47bis, Sorrento, Italy, January 2007 [5] R1-070650, Hypothesis partition for initial cell search: link performance, Qualcomm Europe, RAN1#48, St Louis, USA, February 2007 [6] R1-071277, Placement of FH flag and number of transmit antennas for cell search, Qualcomm Europe, RAN1#48bis, St Julians, Malta, March 200 [7] R1-071278, Complexity efficient PSC design, Qualcomm Europe, RAN1#48bis, St Julians, Malta, March 2007 [8] R1-072009, Details on PSC sequence design, Qualcomm Europe, RAN1#49, Kobe, Japan, May 2007 [9] R1-072727, Details on SSC sequence design, Qualcomm Europe, RAN1#49bis, Orlando, USA, June 2007

16 17 18 19 20 Figure 19: 2GHz carrier, site-to-site distance 500 m, TU3 channel model with 20 dB penetration loss

Cell Search Techniques in LTE

10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

[10] R1-072726, Details on PSC sequence design, Qualcomm Europe, RAN1#49bis, Orlando, USA, June 2007. [11] R1-071584, Secondary synchronization signal design, Ericsson. [12] R1-072663, Cell search email-reflector summary, Motorola, Ran1#49bis, Orlando, USA, June2007 [13] Tao Luo, Scrambling techniques for synchronization channels in E-UTRAN, May 24, 2007, CR&D Espresso Internal memo [14] Tao Luo and Peter Gaal, Permutation based synchronization codes in E-UTRA, February 6, 2007, CR&D Espresso Internal memo [15] Tao Luo and Peter Gaal, Coding schemes for synchronization codes in LTE, April 26, 2007, CR&D Espresso Internal memo [16] R1-072940, Scrambling method for S-SCH in EUTRA downlink, NTTDoCoMo [17] 3GPP TS 25.214, Physical layer procedure (FDD) [18] 3GPP TR 25.814, Physical layer aspects for evolved Universal Terrestrial Radio Access (UTRA)