nb-iot system and method for radio cell synchronization

NB-IoT System and Method for Radio Cell Synchronization White Paper Omri Isaacs, Algorithm Engineer Zeev Kaplan, Algorithm Team Leader Tal Shalev, Senior System Architect

Rev. 1.0 February 2018

NB-IoT System and Method for Radio Cell Synchronization White Paper

Documentation Control History Table

Version Date Description Remarks

1.0 27 February 2018 Initial release

Rev. 1.0 Copyright © 2018 – CEVA®, Inc. i


Disclaimer and Proprietary Information Notice The information contained in this document is subject to change without notice and does not represent a commitment on any part of CEVA®, Inc. CEVA®, Inc. and its subsidiaries make no warranty of any kind with regard to this material, including, but not limited to implied warranties of merchantability and fitness for a particular purpose whether arising out of law, custom, conduct or otherwise.

While the information contained herein is assumed to be accurate, CEVA®, Inc. assumes no responsibility for any errors or omissions contained herein, and assumes no liability for special, direct, indirect or consequential damage, losses, costs, charges, claims, demands, fees or expenses, of any nature or kind, which are incurred in connection with the furnishing, performance or use of this material.

This document contains proprietary information, which is protected by U.S. and international copyright laws. All rights reserved. No part of this document may be reproduced, photocopied, or translated into another language without the prior written consent of CEVA®, Inc.

CEVA®, CEVA-XC™, CEVA-XC5™, CEVA-XC321™, CEVA-XC323™, CEVA-Xtend™, CEVA-XC4000™, CEVA-XC4100™, CEVA-XC4200™, CEVA-XC4210™, CEVA-XC4400™, CEVA-XC4410™, CEVA-XC4500™, CEVA-XC12™, CEVA-TeakLite™, CEVA-TeakLite-II™, CEVA-TeakLite-III™, CEVA-TL3210™, CEVA-TL3211™, CEVA-TeakLite-4™, CEVA-TL410™, CEVA-TL411™, CEVA-TL420™, CEVA-TL421™, CEVA-Quark™, CEVA-Teak™, CEVA-X™, CEVA-X1620™, CEVA-X1622™, CEVA-X1641™, CEVA-X1643™, Xpert-TeakLite-II™, Xpert-Teak™, CEVA-XS1100A™, CEVA-XS1200™, CEVA-XS1200A™, CEVA-TLS100™, Mobile-Media™, CEVA-MM1000™, CEVA-MM2000™, CEVA-SP™, CEVA-VP™, CEVA-MM3000™, CEVA-MM3100™, CEVA-MM3101™, CEVA-XM™, CEVA-XM4™, CEVA-XM6™, CEVA-X1™, CEVA-X2™, CEVA-NeuPro™, CEVA-Audio™, CEVA-HD-Audio™, CEVA-VoP™, CEVA-Bluetooth™, CEVA-SATA™, CEVA-SAS™, CEVA-Toolbox™, CEVA-Dragonfly™, SmartNcode™, PentaG™, and ClearVox™ are trademarks of CEVA, Inc.

All other product names are trademarks or registered trademarks of their respective owners.

ii Copyright © 2018 – CEVA®, Inc. Rev. 1.0


Support CEVA® makes great efforts to provide a user-friendly software and hardware development environment. Along with this, CEVA provides comprehensive documentation, enabling users to learn and develop applications on their own. Due to the complexities involved in the development of DSP applications that might be beyond the scope of the documentation, an online Technical Support Service has been established. This service includes useful tips and provides fast and efficient help, assisting users to quickly resolve development problems.

How to Get Technical Support:

● FAQs: Visit our website http://www.ceva-dsp.com or your company's protected page on the CEVA website for the latest answers to frequently asked questions.

● Application Notes: Visit our website http://www.ceva-dsp.com or your company's protected page on the CEVA website for the latest application notes.

● Email: Use the CEVA central support email address [email protected]. Your email will be forwarded automatically to the relevant support engineers and tools developers who will provide you with the most professional support to help you resolve any problem.

● License Keys: Refer any license key requests or problems to [email protected]. For SDT license keys installation information, see the SDT Installation and Licensing Scheme Guide.

Email: [email protected]

Visit us at: www.ceva-dsp.com

Rev. 1.0 Copyright © 2018 – CEVA®, Inc. iii

http://www.ceva-dsp.com/


mailto:[email protected]






List of Sales and Support Centers Israel USA Ireland Sweden

2 Maskit Street P.O. Box 4047 Herzelia 46120 Israel

1174 Castro Street Suite 210 Mountain View, CA 94040 USA

18/19 South William Street 2nd Floor Dublin 2 Ireland

Klarabergsviadukten 70 Box 70396 107 24 Stockholm Sweden

Tel: +972 9 961 3700 Fax: +972 9 961 3800

Tel: +1-650-417-7923 Fax: +1-650-417-7924

Tel: +353 1 237 3900 Fax: +353 1 237 3923

Tel: +46(0)8 506 362 24 Fax: +46(0)8 506 362 20

China (Shanghai)

China (Beijing)

China (Shenzhen)

Hong Kong

Unit 1203, Building E Chamtime Plaza Office Lane 2889, Jinke Road Pudong New District Shanghai, 201203 China

Rm 503, Tower C Raycom InfoTech Park No. 2, Kexueyuan South Road Haidian District Beijing 100190 China

Rm 709, Tower A SCC Financial Centre No. 88 First Haide Avenue Nanshan District Shenzhen 518064 China

Level 43, AIA Tower 183 Electric Road North Point Hong Kong

Tel: +86-21-20577000 Fax: +86-21-20577111

Tel: +86-10 5982 2285 Fax: +86-10 5982 2284

Tel: +86-755-84356027 Tel: +852-39751264

South Korea Taiwan Japan France

#478, Hyundai Arion 147, Gumgok-Dong Bundang-Gu Sungnam-Si Kyunggi-Do, 463-853 South Korea

Room 621 No. 1, Industry E, 2nd Rd Hsinchu, Science Park Hsinchu 300 Taiwan R.O.C

1-6-5 Shibuya SK Aoyama Bldg. 3F Shibuya-ku, Tokyo 150-0002 Japan

RivieraWaves S.A.S 400, avenue Roumanille Les Bureaux Green Side 5, Bât 6 06410 Biot - Sophia Antipolis France

Tel: +82-31-704-4471 Fax:+82-31-704-4479

Tel: +886 3 5798750 Fax: +886 3 5798750

Tel: +81-3-5774-8250 Tel: +33 4 83 76 06 00 Fax: +33 4 83 76 06 01

iv Copyright © 2018 – CEVA®, Inc. Rev. 1.0


Table of Contents

1. INTRODUCTION .................................................................................................. 1 1.1 Scope .............................................................................................................1 1.2 Audience .........................................................................................................1

2. OVERVIEW .......................................................................................................... 3 2.1 Benefits ..........................................................................................................3 2.2 Deployment .....................................................................................................3 2.3 Challenges ......................................................................................................4

3. NB-IOT INITIAL SYNCHRONIZATION PROCESS .................................................. 7 3.1 Synchronization Flow ........................................................................................7

4. NPSS AND NSSS .................................................................................................. 9 4.1 Synchronization Signal in Radio Frames ..............................................................9 4.2 NSSS Structure .............................................................................................. 12

5. SYNCHRONIZATION METHODS .......................................................................... 15 5.1 Overview ...................................................................................................... 15

5.1.1 Coarse Stage ........................................................................................ 16 5.1.2 Fine Stage ............................................................................................ 17

5.2 Coarse Stage Detection Methods ...................................................................... 17 5.2.1 Cross-Correlation Synchronization Method ................................................ 17

5.3 Auto-Correlation Synchronization Method .......................................................... 19 5.4 Two-Phase Synchronization Method .................................................................. 21

5.4.1 Step 1: Segment-Length Cross-Correlation ............................................... 22 5.4.2 Step 2: Segmented-Wise Correlation ....................................................... 23 5.4.3 Step 3: Cost Function Formulation ........................................................... 23 5.4.4 Step 4: Coherent Combining ................................................................... 24 5.4.5 Step 5: Threshold for NPSS Detection ...................................................... 24

6. PERFORMANCE .................................................................................................. 25

7. SUMMARY ......................................................................................................... 29

8. REFERENCES ..................................................................................................... 31

9. GLOSSARY ......................................................................................................... 33

Rev. 1.0 Copyright © 2018 – CEVA®, Inc. v


List of Figures Figure 4-1: Radio Frame Synchronization Signals ..............................................................9 Figure 4-2: NPSS Structure (Frequency Domain) ............................................................. 10 Figure 4-3: NSSS Structure (Frequency Domain) ............................................................ 12 Figure 5-1: Synchronization Process .............................................................................. 16 Figure 5-2: Two-Phase Coarse Synchronization Method Flow ............................................ 22 Figure 6-1: Comparison of Different Synchronization Methods (ETU-1) .............................. 25 Figure 6-2: Comparison of Different Synchronization Methods (EVA-5) .............................. 26 Figure 6-3: Comparison of Different Synchronization Methods (EPA-0) ............................... 26

List of Tables Table 4-1: NPSS Structure (Time Domain) ..................................................................... 11 Table 9-1: Acronyms ................................................................................................... 33

vi Copyright © 2018 – CEVA®, Inc. Rev. 1.0


1. Introduction 1.1 Scope

This document describes the narrowband IoT (NB-IoT) radio access technology, and focuses mainly on the primary synchronization process of the NB-IoT in the downlink, which is activated when the user equipment (UE) attempts to receive data from the base station (BS) when it leaves its dormant state and wakes up.

When the UE wakes up, it must first synchronize with the time, frequency, and base station parameters (for example, cell ID or radio network temporary identifier (RNTI)) to be able to receive data. In this document, we mainly focus on the UE's initial cell search, in which the range of the carrier frequency uncertainty is very high. This process is one of the most power-consuming processes in NB-IoT because it requires the UE to continuously activate an RX radio frequency (RF) chain and to constantly process input samples until the synchronization stage is complete.

This document presents a synchronization procedure developed by CEVA to enhance the performance of the NB-IoT primary synchronization procedure when the UE wakes up. This new procedure enables the timing and frequency of the transmitted NB-IoT signal to be estimated by the UE with relatively low detection latency even in very difficult urban environments. This procedure was developed to be efficiently executed on the CEVA-X core family, and is integrated into the CEVA-Dragonfly NB1 product. To learn more about the CEVA-Dragonfly NB1 product, go to https://www.ceva-dsp.com/product/dragonfly-nb1.

The following sections describe different known synchronization procedures, discuss their advantages and disadvantages, and propose a new synchronization procedure that better utilizes the structure of the narrowband primary synchronization signal (NPSS) for DSP-based implementation.

1.2 Audience This document is intended for communication and signal processing engineers who want to expand their knowledge about new innovative ways to perform synchronization, as well as engineers who are interested in the NB-IoT communization system.

Rev. 1.0 Copyright © 2018 – CEVA®, Inc. 1

https://www.ceva-dsp.com/product/dragonfly-nb1


2. Overview The internet of things (IoT) is a fast-growing market [1, 2] based on machine-type communication that requires minimal human supervision. Ericsson predicts there will be around 28 billion IoT-enabled devices by 2021, of which more than 15 billion will be machine-to-machine (M2M) type and consumer electronic devices. Cisco estimates a 38% annual growth rate from 2015 to 2020 in M2M connections (from 604 million to 3.1 billion).

These recent rapid technology developments have created a pressing need to supply simple and robust connectivity solutions for machine-type communication in a wide range of areas, such as sensors, meters, smart grids, monitoring, and security, to name just a few. However, because the IoT market has different needs than the cellular market does (for example, low complexity and robustness), a new radio access technology was created, called narrowband IoT (NB-IoT). This technology was recently introduced by the Third-Generation Partnership Project (3GPP) in its 13th release [3].

2.1 Benefits NB-IoT technology has many benefits for the IoT market, for example, reduced device complexity, extended coverage, increased battery life, and low cost [4-7]. In addition, NB-IoT uses only a small portion of the available spectrum, which enables it to be transmitted alongside other communication signals such as Long-Term Evolution (LTE) and UMTS. NB-IoT is designed mainly for ultra-low-end IoT applications.

2.2 Deployment NB-IoT requires a minimum bandwidth of 180 kHz, which is equal to the smallest size of the LTE physical resource block (PRB). It can be deployed in the following scenarios (depending on the available spectrum):

● As a standalone deployment ● In the guard-carriers of the existing LTE/UMTS spectrum (guard-

band deployment) ● Within an LTE carrier, replacing one of its PRBs (in-band

deployment)

A conflict of resources between an LTE system and the NB-IoT requires the downlink control channel to be used at the beginning of each subframe for all of these possible deployments. For an in-band deployment, additional cell-specific reference signals (CRSs) must be inserted into every subframe before allocating the resources of the NB-IoT system.



Because NB-IoT reuses significant parts of the LTE designs, such as orthogonal frequency division modulation (OFDM) in its downlink, and single-carrier frequency division multiple access (SC-FDMA) in its uplink, as well as channel coding, rate-matching, and interleaving, it can be deployed alongside the LTE spectrum. However, because NB-IoT also includes changes from the legacy LTE, such as synchronization signals, broadcast channels, and control channels, its design enables good performance while co-existing with the legacy LTE.

2.3 Challenges The NB-IoT is specifically designed to support M2M communications between IoT devices. As a result, this communication technology must address the following challenges:

● Low Signal-to-Noise Ratios (SNRs) and Difficult Channels: IoT devices might be placed in deep indoor locations (for example, cellars and basements) or in urban environments (for example, smart cities), which will result in low SNRs and channels with multipath fading propagations. To handle this type of environment and/or the SNR issues, the NB-IoT communication channel must contain very robust signal-processing methods and be designed to withstand very low SNRs (for example, -12.6 dB).

● Frequency Offset: NB-IoT is intended for use in ultra-low-cost UE devices, which receive information from base stations in their surrounding areas. These UE are often equipped with low-cost oscillators, which might cause an initial carrier frequency offset (CFO) due to limited stability of up to 20 parts per million (ppm) and an additional ±7.5 kHz from the frequency raster (100 kHz raster).

4 Copyright © 2018 – CEVA®, Inc. Rev. 1.0


● Low Detection Latency: The time duration that the UE RF chain is on and active is called RF-ON, and includes amplifiers, the analog-to-digital converter (ADC), and the digital front end (DFE). The RF-ON stage is very energy consuming; to increase the battery lifetime, we aim to minimize it by reducing the detection time of the NPSS, which requires constant process input.

● Complexity: IoT devices might be stationed in places without reliable power supplies or where it is not affordable to change their batteries. As a result, the device lifespan is often determined by its battery consumption. Because the processing power of ultra-low-cost UE devices is limited, the NB-IoT signal processing algorithm must be simple to implement with a low sampling rate.



3. NB-IoT Initial Synchronization Process

3.1 Synchronization Flow IoT UE devices are expected to be dormant most of the time. When powered on, the UE must do the following:

Align its carrier frequency with that of the base station for it to receive 1.data at a low bit error probability.

Because the base station transmits information in a repeated structure (called a radio frame) every 10 msec, the UE must detect the beginning of the radio frame to extract data at the correct time.

This means that the UE must perform time and frequency synchronization with the base station (eNodeB) to be able to transmit and receive data. In this document, this synchronization process is called the primary synchronization process.

When this step is completed, the UE must extract additional data, such 2.as the base station cell ID and the radio-frame 80 msec boundary. This synchronization process is called the secondary synchronization process.

After detecting the necessary system configurations, the cell 3.synchronization process is completed, and the UE begins decoding the control channels (NPBCH, NPDCCH) and the data channel (NPDSCH).

The NB-IoT radio frame is comprised of 10 subframes, each with a time duration of 1 msec. A subframe contains 14 OFDM symbols; the first three symbols are for the control channel of the legacy LTE, and are assumed to be unused for NB-IoT transmissions.

The purpose of the synchronization procedure is to estimate the system parameters based on the narrowband primary and secondary signals (NPSS and NSSS), which are repeated every one and two radio frames, respectively. The primary synchronization procedure attempts to detect the NPSS in a radio frame window, which requires the UE to receive data until this signal is located.



4. NPSS and NSSS The NB-IoT physical channels are mostly based on the legacy LTE; however, the synchronization signals have been completely redesigned to enable different deployments inside and adjacent to LTE carriers:

● In a legacy LTE system, the synchronization signals are as follows: ○ The primary synchronization signal (PSS) is used for frame timing

and frequency estimation, as well as for part of the cell identity detection (NID2).

○ The secondary synchronization signal (SSS) holds the complementary part of the cell identity (NID1), and is used for CP detection and duplexing mode determination (FDD/TDD) ([ 3]).

● In an NB-IoT system, however, narrowband synchronization signals are constructed from several OFDM signals that occupy a different bandwidth than they do in the LTE (180 kHz as opposed to 1.08 MHz): ○ The narrowband PSS (NPSS) obtains the estimation of the frame

timing and the carrier frequency. ○ The narrowband SSS (NSSS) obtains the cell identity and the

frame boundary in an 80 msec window, and enables cell tracking.

4.1 Synchronization Signal in Radio Frames The synchronization signals occupy a bandwidth of 180 kHz. The NPSS is located in the sixth subframe of every radio frame, and the NSSS is located in the tenth subframe of every even radio frame, as shown in Figure 4-1.

Figure 4-1: Radio Frame Synchronization Signals



The NPSS is transmitted every 10 msec in the sixth subframe. It is comprised of the last 11 OFDM symbols in the subframe (the first three are reserved for the legacy LTE control region), and uses the first 11 subcarriers (out of 12) of the allocated PRB.

Note: For an in-band deployment, the NPSS is punctured by a CRS.

Figure 4-2 shows the structure of the NPSS.

Figure 4-2: NPSS Structure (Frequency Domain)

0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9

Frame nf

Nf mod2 = 0

NPSS

NSSS

NPSS

0 1 2 3 4 5 6 7 8 9 10 11 12 13

0123456789

1011



The NPSS is comprised of OFDM symbols that are generated in the frequency domain based on length-11 Zadoff-Chu (ZC) sequences with a root index of 5 [3].

To enable efficient implementation and good synchronization performance, NB-IoT uses a predetermined pseudo-random code for the sign of its OFDM symbols, as listed in Table 4-1.

Table 4-1: NPSS Structure (Time Domain)

Symbol Index 0 1 2 3 4 5 6 7 8 9 10 11 12 13

Cover Code 1 1 1 1 -1 -1 1 1 1 -1 1

OFDM Symbol Vector X X X X -X -X X X X -X X

The NPSS is predetermined and known to the UE, which enables it to be used for the synchronization process. Because of the high uncertainty in the timing and frequency of the transmitted signal, NPSS detection is one of the most computationally consuming operations performed by the UE.



4.2 NSSS Structure The NSSS is transmitted every 20 msec in the tenth subframe in the even-numbered radio frame indexes. It is comprised of the last 11 OFDM symbols in the subframe (the first three are reserved for the legacy LTE control region), and uses all 12 subcarriers of the allocated PRB.

Note: For an in-band deployment, the NSSS is punctured by a CRS.

Figure 4-3 shows the structure of the NSSS.

Figure 4-3: NSSS Structure (Frequency Domain)

The NSSS is comprised of a total of 132 frequency domain elements (that is, 12 subcarriers in each of the 11 OFDM symbols) whose values are based on length-131 ZC sequences. The root of the ZC sequence is based on the cell ID of the base station. The NSSS is generated by element-wise multiplication between the ZC sequence and a binary code that is based on the system frame number (SFN) in a boundary of 80 msec.

0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9

Frame nf

Nf mod2 = 0

NPSS

NSSS

NSSS

0 1 2 3 4 5 6 7 8 9 10 11 12 13

0123456789

1011



The NSSS is known to the UE given the cell ID and radio frame index, that is, it is used to determine the cell ID and radio frame index. These parameters can be detected by the UE only after the time and frequency synchronization is completed; otherwise, a four-dimensional estimation problem must be solved. The cell ID and SFN are required for the UE's operation and data extraction.

After the time and frequency are estimated, and the cell ID and radio frame boundary are detected, the UE can start decoding the control channels (NPBCH, NPDCCH) and the data channel (NPDSCH).



5. Synchronization Methods 5.1 Overview

The synchronization procedure can be viewed as a multi-dimensional estimation problem with a few unknown parameters, such as time, frequency, phase, and so on. These types of problems have been extensively investigated in various fields, for example, in communications, GPS systems, radar, and sonar.

The synchronization procedure is the first step before receiving data or estimating parameters of interest. A predetermined signal (also called a pilot) is either explicitly or partially known to the receiver, as is the general construction of the pilot signal (for example, duration and repetition pattern). Based on its knowledge about the pilot, the UE aligns its system configuration to improve its information processing capabilities.

In the NB-IoT primary synchronization stage example here, we will focus mainly on the case where the receiver knows the pilot signal, and can therefore estimate the system parameters.

The NPSS is used by the NB-IoT UE to perform the primary part of the cell synchronization, that is, to estimate the essential system parameters (for example, the timing of the radio frame and the frequency of the transmitted signal). The estimation of the beginning of the radio frame is straightforward based on the estimation of the beginning of the pilot (that is, the NPSS), without ambiguity.

For synchronization purposes, we assume that the signal can be processed using either the native sampling frequency of 240 kHz (low rate) or at a higher sampling frequency of 1.92 MHz (high rate).

The primary synchronization procedure is made up of the following stages:

The first stage is a coarse estimation of the unknown parameters, as 1.described in Section 5.1.1.

The second stage is a refinement of the time and frequency estimations, 2.as described in Section 5.1.2.

After the primary synchronization stage, the UE will continue to the secondary synchronization stage to detect essential data, such as the cell ID and the 80 msec radio boundary.



Figure 5-1 shows the synchronization process, followed by data extraction.

Figure 5-1: Synchronization Process

5.1.1 Coarse Stage The coarse stage of the primary synchronization procedure is done at a lower rate of 240 kHz to reduce computational complexity. The coarse estimation stage is essential because of the high range of uncertainty about the timing and frequency in a multi-dimensional estimation problem.

The coarse stage estimates coarse estimations of the timing and frequency of the received signal.

This document focuses on the coarse detection stage in the primary synchronization procedure, and elaborates on known estimation methods that could be utilized for the coarse stage. Finally, a new estimation method is proposed, followed by simulation performances.

Primary Synchronization Stage - Coarse Stage

Primary Synchonization Stage - Fine Stage

Secondary Synchonization Stage

NPBCH Processing

NPDSCH/NPDCCH Processing



5.1.2 Fine Stage To improve the UE's data extraction capabilities after the synchronization procedure is completed, a time and frequency refinement is implemented after the coarse stage. This stage is known as the fine stage. The coarse estimations are used to reduce the complexity of the fine stage by decreasing the uncertainty of the time and frequency.

In this stage, the following occurs:

The algorithm opens a "short signal window" around the coarse time 1.estimation at a sampling rate of 1.92 MHz.

The frequency of the received signal is shifted to baseband based on the 2.coarse frequency estimation.

The fine stage outputs are timing and frequency estimations, with lower errors than the coarse estimations. After the timing and frequency are estimated, the UE continues to detect the cell ID and frame boundary based on the NSSS at a low rate of 240 kHz, for example, by applying frequency domain cross-correlation and cyclic prefix removal.

5.2 Coarse Stage Detection Methods

5.2.1 Cross-Correlation Synchronization Method The most straightforward synchronization algorithm is the cross-correlation method, referred to here as full-length cross-correlation.

This method relies on knowledge of the synchronization signal at the receiver's side. When applied to NB-IoT systems, full-length cross-correlation methods can be implemented by correlation of the received signal at time 𝜏𝜏 denoted by 𝑥𝑥[𝜏𝜏] in an 11 msec window with the conjugate of the NPSS given a frequency hypothesis 𝑓𝑓𝑛𝑛, for example:

𝝆𝝆(𝝉𝝉) = � 𝒙𝒙[𝒏𝒏 + 𝝉𝝉]𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵∗[𝒏𝒏]𝐞𝐞𝐞𝐞𝐞𝐞 (𝒋𝒋𝟐𝟐𝝅𝝅𝒇𝒇𝒏𝒏𝒏𝒏) 𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵−𝟏𝟏

𝒏𝒏=𝟎𝟎

where:

● 𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁 is the length of the NPSS. ● 𝜏𝜏 is a candidate synchronization point in the radio frame (for example,

a point in time or a sample number within the radio frame).



The full-length cross-correlation synchronization method produces the most gain because it uses complete knowledge of the system and the transmitted signal [8] (that is, it checks different candidates for the timing offset and frequency offset to formulate the explicit NPSS). The cross-correlation output is used as the "cost function" to evaluate the synchronization performance, where the peak of the cost function is the time offset candidate in the radio-frame window in which the NPSS begins. Based on the peak-to-average ratio (PAPR) of the cost function, the reliability of the synchronization procedure can be evaluated.

Because of the difficult outdoor conditions in the NB-IoT system and the urban environment, the UE cannot estimate the timing of the NPSS using an 11 msec window while maintaining a low false alarm probability. To get around this issue, the NPSS is repeated every 10 msec, enabling the cost function to be summed incoherently until a predetermined condition is met, (that is, a high PAPR of the cost function).

Note: The full-length cross-correlation method produces the most gain out of a single frame, which means that the number of cost function accumulations needed for a low false alarm probability is relatively low compared to other methods.

A low number of accumulations results in a short synchronization time, which results in a shorter RF-ON period. Reducing this time considerably reduces the power consumption of the UE and increases battery lifetime.

The full-length cross-correlation method requires solving a two-dimensional estimation problem, that is, estimating the time and frequency simultaneously.

The frequency dimension is tested using a grid for the frequency hypothesis, where the pilot's frequency is determined based on the frequency values in the grid. The grid of the frequency candidates must be minimal in size to reduce complexity, but if the maximal distance between the true frequency and its closest frequency hypothesis is too high, the attenuation from the cross-correlation will deteriorate the system performances considerably because of the long duration of the NPSS.

In the case of initial cell search, the range of the CFO error is very high due to a large frequency offset of ±20 ppm ±7.5 kHz and continuous time drift. Considering the range of the CFO and the long duration of the NPSS signal (around 0.78 msec), the number of frequency hypotheses needed for the full-function cross-correlation method is very high (around 50 frequency hypotheses).



In an NB-IoT system, a high number of hypotheses will result in high complexity and high memory consumption (for the cost function of each frequency hypothesis), which means that synchronization using the full-length cross-correlation method in an NB-IoT system seems very infeasible. In [7], it was proposed that the cross-correlation in the frequency domain be computed to reduce the complexity of the cross-correlation methods. However, even after implementing dedicated hardware for the cross-correlation methods, this method still had approximately 10 times more complexity than other known methods like the auto-correlation method (which is described in Section 5.3).

To summarize, the advantage of the full-length cross-correlation method is that it provides very low synchronization latencies, especially at low SNR. However, this method has the following disadvantages in an NB-IoT system:

● High complexity ● High memory requirements

5.3 Auto-Correlation Synchronization Method Another well-known synchronization method is called the auto-correlation method [4, 5].

Because the NPSS is comprised of 11 identical OFDM symbols (which are multiplied by a predetermined 11-length pseudo-random binary cover code), an auto-correlation synchronization solution could be a suitable choice for an estimation algorithm in an NB-IoT system.

For example, consider computing the auto-correlation lags of up to four NPSS symbols (𝑘𝑘 = 1,2,3,4):

𝑨𝑨𝒌𝒌(𝝉𝝉) =𝟏𝟏

𝟏𝟏𝟏𝟏 − 𝒌𝒌� 𝐞𝐞𝑯𝑯(𝒍𝒍+ 𝒌𝒌, 𝝉𝝉)𝐞𝐞(𝒍𝒍, 𝝉𝝉)𝒔𝒔(𝒌𝒌)𝒔𝒔(𝒌𝒌+ 𝒍𝒍)𝟏𝟏𝟏𝟏−𝒌𝒌

𝒍𝒍=𝟏𝟏

where:

● 𝐞𝐞(𝑙𝑙, 𝜏𝜏) is the 𝑙𝑙th symbol starting from synchronization candidate point 𝜏𝜏.

● 𝑠𝑠(𝑘𝑘) is the cover code.

Note: Aside from the cover-code, the autocorrelation does not use the NPSS in its synchronization algorithm.



At the true candidate time offset, the expectation of the phase rotation caused by the CFO between two adjacent NPSS symbols will be proportional to the fractional CFO and the distance between the symbols, that is:

𝑬𝑬[𝑨𝑨𝒌𝒌(𝝉𝝉)]~𝒆𝒆𝒋𝒋𝟐𝟐𝝅𝝅𝒌𝒌𝒇𝒇 The fractional CFO is the modulo of the true CFO and the maximal CFO value, which can be estimated. This results in ambiguity in the frequency estimation, which can be resolved in the fine stages of the synchronization.

As opposed to the full-function cross-correlation method described in Section 5.2, the cost function in the auto-correlation method can be coherently accumulated. As a result, the cost function is formed from different auto-correlation lags, in which the phase of the cost function at the true time offset is equal to the phase of the auto-correlation with a single lag. This cost function can be expressed as:

𝝆𝝆(𝝉𝝉) = 𝑨𝑨𝟏𝟏(𝝉𝝉) + �𝑨𝑨𝒒𝒒∗ (𝝉𝝉)𝑨𝑨𝒒𝒒+𝟏𝟏(𝝉𝝉)𝟒𝟒

𝒒𝒒=𝟏𝟏

The cost function is coherently accumulated every radio frame, which reduces the number of accumulations needed because the noise in the system is uncorrelated in time.

The phase of the expectation at the true timing of the NPSS is equal to:

[𝝆𝝆(𝝉𝝉)]~𝒆𝒆𝒋𝒋𝟐𝟐𝝅𝝅𝒇𝒇 This means that the frequency of the transmitted signal can be estimated; however, due to the high range of the CFO, a number of frequency solutions can be considered (that is, there is high ambiguity in the frequency estimation process).

To summarize, the advantages of the auto-correlation method are:

● The cost function enables the starting point of the NPSS and the frequency of the signal to be estimated.

● The cost function can be coherent accumulated in time. ● It is simple to implement efficiently in a recursive fashion.

However, this method has the following disadvantages in an NB-IoT system:

● The performance is inferior to that of the full-function cross-correlation method, especially at low SNR [3].

● The frequency estimation has ambiguities.



Note: The low performance of the auto-correlation method at a low SNR results in a high number of accumulations needed to estimate the received signal at high probability.

Because the auto-correlation method uses knowledge of the repetitions of the NPSS (as opposed to the full-function cross-correlation method, which uses full knowledge of the NPSS ([3]), this method has a lower output SNR. This means that the power consumption in the auto-correlation method deteriorates due to the high RF-ON time of the system.

5.4 Two-Phase Synchronization Method The following sections describe a new synchronization procedure, called the two-phase cell synchronization method, which is designed specifically for NB-IoT systems and provides a compromise between detection performance and complexity.

The method is made up of a number of steps, as described in the following sections. It detects the time offset of the NPSS in a radio frame, as well as the frequency of the received signal, by processing the signal post-decimation at a low sampling rate of 240 kHz.



Figure 5-2 shows a detailed flow of the two-phase synchronization signal method, where the flow begins after the signal is received from the DFE.

Figure 5-2: Two-Phase Coarse Synchronization Method Flow

5.4.1 Step 1: Segment-Length Cross-Correlation The first step in the two-phase synchronization signal method is performing segment-length cross-correlation between a predetermined length portion of the received signal and a segment of the NPSS.

By reducing the length of the cross-correlation from a full-length NPSS signal to the shorter length segment, the number of frequency hypotheses is reduced in an inverse proportion.

Because the NPSS has a repetitive structure, this short-length cross-correlation method has a complexity advantage – the number of computations required to implement it can be reduced by using a recursive computation algorithm.

DIVIDE THE SIGNAL INTO SEGMENTS

CALCULATE CROSS-CORRELATION AT SEGMENT LENGTH

CALCULATE SEGMENT-WISE CORRELATIONS BETWEEN THE CROSS-CORRELATION OUTPUTS

CALCULATE COST FUNCTION

ACCULMULATE COST FUNCTION

DETERMINE A COARSE STARTING POINT

ANALYZE THE THRESHOLD

ESTIMATE THE FREQUENCY OFFSET



5.4.2 Step 2: Segmented-Wise Correlation After the segment-length cross-correlation described in the previous step, the SNR of the cross-correlation outputs increases dramatically, because the power of the noise decreases proportionally to the length of the segment.

We propose to compute a segment-wise auto-correlation using the cross-correlation outputs, which have a much higher SNR than the received signal. Considering this, the SNR output of auto-correlation computed in the equation below is much higher than the auto-correlation of the received signal. The computation of the auto-correlation can be expressed as:

𝑵𝑵𝒉𝒉(𝝉𝝉,𝒎𝒎) = � 𝒚𝒚𝒉𝒉∗ (𝝉𝝉,𝒌𝒌)𝒚𝒚𝒉𝒉(𝝉𝝉,𝒌𝒌 + 𝒎𝒎)

𝑵𝑵𝒔𝒔𝒆𝒆𝒔𝒔𝒎𝒎𝒆𝒆𝒏𝒏𝒔𝒔𝒔𝒔

𝒌𝒌=𝟏𝟏

where 𝐍𝐍𝐬𝐬𝐞𝐞𝐬𝐬𝐬𝐬𝐞𝐞𝐬𝐬𝐬𝐬𝐬𝐬 is the number of segments.

The auto-correlation can be computed recursively in a simple fashion via very low memory allocations.

Because this method uses the structure and values of the known NPSS, it is expected to achieve better performance than the auto-correlation method, which only uses the repetitive nature of the NPSS.

5.4.3 Step 3: Cost Function Formulation The cost function can be expressed by a weighted combination of 𝑆𝑆ℎ(𝜏𝜏,𝑚𝑚) as:

𝝆𝝆𝒉𝒉(𝝉𝝉) = 𝒘𝒘𝟏𝟏𝑵𝑵𝒉𝒉(𝝉𝝉,𝟏𝟏) + 𝒘𝒘𝟐𝟐𝑵𝑵𝒉𝒉(𝝉𝝉,𝟐𝟐)𝑵𝑵𝒉𝒉∗ (𝝉𝝉,𝟏𝟏) + ⋯

where 𝑤𝑤𝑚𝑚 denotes the weight for coherent combining.

The time offset can be estimated based on the maximum value of the cost function:

𝝉𝝉𝟎𝟎 = 𝒂𝒂𝒂𝒂𝒔𝒔𝐬𝐬𝐦𝐦𝐞𝐞𝝉𝝉,𝒉𝒉

|𝝆𝝆𝒉𝒉(𝝉𝝉)|

It can be easily shown that the frequency offset can be estimated based on the following relation:

𝑬𝑬[𝝆𝝆𝒉𝒉 (𝝉𝝉𝟎𝟎 )]∝ 𝒆𝒆𝒋𝒋(𝒇𝒇−𝒇𝒇𝒉𝒉)



5.4.4 Step 4: Coherent Combining To enable the system to withstand difficult environments and low SNR, the NPSS is transmitted repeatedly every 10 msec. The cost function can be coherently accumulated to enhance estimation performances. The accumulation is performed via the following simple IIR filter:

𝝆𝝆𝚺𝚺,𝒉𝒉(𝝉𝝉) = 𝜶𝜶𝝆𝝆𝚺𝚺,𝒉𝒉(𝝉𝝉) + (𝟏𝟏 − 𝜶𝜶)𝝆𝝆𝒉𝒉(𝝉𝝉), 𝟎𝟎 < 𝜶𝜶 < 𝟏𝟏

where 𝜌𝜌Σ,ℎ(𝜏𝜏),𝛼𝛼 are the accumulated cost function and the decay factor, respectively.

5.4.5 Step 5: Threshold for NPSS Detection The presence of the NPSS is found by comparing the peak of the cost function with a dynamic threshold.

When the threshold condition is passed, the location of the NPSS and the carrier frequency offset of the signal are estimated based on the location of the peak of the cost function and its phase, respectively. The time offset and the carrier frequency estimated during the coarse stage will require refinement, which is done in the fine stage.



6. Performance The latency performance for the primary stage of the initial cell search was simulated using the three synchronization methods described in Section 5. The simulations were conducted through 500 independent trials in the following channel scenarios:

● ETU with Doppler 1 Hz, as shown in Figure 6-1 ● EVA with Doppler 5 Hz, as shown in Figure 6-2 ● EPA with Doppler 0 Hz, as shown in Figure 6-3

In these tests, the detection rate is greater than 99%. In the following figures, the performance of the latency is based on the residual latency at the 90th percentile, that is, 90% of the tests have lower or equal latency at a given SNR and channel.

Because the full-function cross-correlation method has the highest gain from the received signal (and therefore the minimum latency), the latency of the other methods was normalized by it. The performance latency is shown for the case of guard-band deployment with an SNR of down to -12.6 dB.

Figure 6-1: Comparison of Different Synchronization Methods (ETU-1)



Figure 6-2: Comparison of Different Synchronization Methods (EVA-5)

Figure 6-3: Comparison of Different Synchronization Methods (EPA-0)



These figures show that, as SNR decreases, the residual latency performance of the auto-correlation method increases dramatically compared to the latency of the cross-correlation method; however, the performance of our two-phase synchronization method remains very close to the performance of the cross-correlation synchronization method even at a very low SNR.



7. Summary In this paper, a new synchronization method is proposed for the NB-IoT primary synchronization coarse stage.

This method obtains coarse time and frequency estimations based on the decimated received signal. It is explained that the proposed method can be efficiently implemented in a feasible manner using recursive computations.

Simulations show that this method surpasses the latency performances of another well-known method, the auto-correlation method, while having lower complexity and memory allocations than full-length cross-correlation.



8. References 1. Cisco, "Visual networking index: Global mobile data traffic forecast

update, 2015-2020", White Paper, 2016.

2. Ericsson, "Cellular networks for massive IoT", white paper, Jan. 2016.

3. 3GGP TS 36.211: 'Evolved Universal Terrestrial Radio Access (E-UTRA); Physical Channels and Modulation '3GGP TS 36.211: 'Evolved Universal Terrestrial Radio Access (E-UTRA); Physical Channels and Modulation'.

4. 3GPP TSG RAN WG1 Meeting #84, Qualcomm Incorporated 3GPP TSG RAN WG1 Meeting #84, Qualcomm Incorporated.

5. Y. P. E. Wang et al., "A Primer on 3GPP Narrowband Internet of Things," in IEEE Communications Magazine, vol. 55, no. 3, pp. 117-123, March 201

6. A. Adhikary and X. Lin and Y. P. E. Wang, "Performance Evaluation of NB-IoT Coverage", 2016 IEEE 84th Vehicular Technology Conference (VTC-Fall), Sept. 2016

7. H. Kroll, M. Korb, B. Weber, S. Willi and Q. Huang, "Maximum-Likelihood Detection for Energy-Efficient Timing Acquisition in NB-IoT," 2017 IEEE Wireless Communications and Networking Conference Workshops (WCNCW), San Francisco, CA, 2017, pp. 1-5

8. Lee, Yuk Wing, T. P. Cheatham, and Jerome B. Wiesner, "The application of correlation functions in the detection of small signals in noise", Research Laboratory of Electronics, Massachusetts Institute of Technology, 1949



9. Glossary Table 9-1 defines the acronyms used in this document.

Table 9-1: Acronyms

Term Definition

ADC Analog-to-Digital Converter

CFO Carrier Frequency Offset

CRS Cell-Specific Reference Signals

DFE Digital Front End

EPA Extended Pedestrian A

ETU Extended Typical Urban

EVA Extended Vehicular A

FDD Frequency Division Duplex

IIR Infinite Impulse Respond

LTE Long-Term Evolution

M2M Machine-to-Machine

NB-IoT Narrowband Internet of Things

NPBCH Narrowband Physical Broadcast Channel

NPDCCH Narrowband Physical Downlink Control Channel

NPDSCH Narrowband Physical Downlink Shared Channel

NPSS Narrowband Primary Synchronization Signal

NSSS Narrowband Secondary Synchronization Signal

OFDM Orthogonal Frequency Division Modulation

PAPR Peak-to-Average Ratio

PRB Physical Resource Block

PSS Primary Synchronization Signal

SC-FDMA Single-Carrier Frequency Division Multiple Access

SFN System Frame Number

SNR Signal-to-Noise Ratio

SSS Secondary Synchronization Signal

TDD Time Division Duplex

UE User Equipment

UMTS Universal Mobile Telecommunication System

UTRA Universal Terrestrial Radio Access


nb-iot system and method for radio cell synchronization

Documents