Extending the precision time protocol to a metropolitan area...

116
Degree project in Communication Systems Second level, 30.0 HEC Stockholm, Sweden MOZHDEH KAMEL Synchronizing radio base stations Extending the precision time protocol to a metropolitan area network KTH Information and Communication Technology

Transcript of Extending the precision time protocol to a metropolitan area...

  • Degree project inCommunication Systems

    Second level, 30.0 HECStockholm, Sweden

    M O Z H D E H K A M E L

    Synchronizing radio base stations

    Extending the precision time protocolto a metropolitan area network

    K T H I n f o r m a t i o n a n d

    C o m m u n i c a t i o n T e c h n o l o g y

  • Extending the precision time protocol to a metropolitan area network

    Synchronizing radio base stations

    Mozhdeh Kamel

    2014.02.06

    Master thesis

    Examiner: Gerald Q. Maguire Jr. Supervisors: Jiang Wu

    School of Information and Communication Technologies KTH Royal Institute of Technology

    Stockholm, Sweden

  • © Mozhdeh Kamel, February 2014

  • i

    Abstract

    When building various types of wide area cellular radio networks there is a need to synchronize all of the base stations within a given system. Today this is typically done by attaching a highly accurate clock to each radio base station. A GPS radio receiver is commonly used as such a clock. This thesis explores the use of the Precision Time Protocol (PTP) to provide synchronization of radio base stations, rather than the current practice of using GPS radio receivers.

    Advantages of utilizing PTP rather than a GPS radio receiver include the ability to easily locate radio base stations (without the need for connecting the GPS radio receiver to an antenna that has line of sight to a sufficient number of GPS satellites); the system is not vulnerable to interference with or jamming of GPS radio signals; the system is not vulnerable to spoofing of GPS radio signals, and because the new generations of radio base stations are connected to a packet based backhaul link – the system can potentially utilize the existing packet network interface (thus avoiding the need for a serial interface to the GPS receiver and a pulse per second input).

    At the start of this thesis project it was not known what the limits of PTP are (in terms of utilizing PTP together with radio base stations). Thus it was not clear whether PTP could be extended to much longer distances than it had originally been designed for.

    This thesis shows that PTP can be used as an accurate timing source to synchronize base stations in networks with up to four switches between the PTP grandmaster and any PTP slave.

    This project was performed in the Common Transport Feature department at Ericsson.

    Keywords: precision time protocol, radio base station, synchronization

  • iii

    Sammanfattning

    Vid konstruktion av wide area cellular radio networks finns det behov av att synkronisera samtliga basstationer inom ett givet system. Detta görs idag typiskt genom att ansluta en klocka med stor tillförlitlighet till varje basstation. En GPS radiomottagare används vanligen som klocka för detta syfte. Detta examensarbete undersöker användandet av Precisions Tid Protokoll (PTP) för att synkronisera radiobasstationer, istället för att som nu typiskt använda GPS radiomottagare.

    Fördelar med att använda PTP istället för GPS radiomottagare är att en radiobasstation lätt kan lokaliseras (utan att ansluta en GPS-mottagare till en antenn vilken har mottagning mot flera GPS-satelliter); systemet är inte sårbart mot interferens eller störningar av GPS radio signaler; systemet är inte sårbart mot spoofing av GPS radio signaler och på grund av att den nya generationens radiobasstationer är anslutna till ett paketförmedlande backhaul nätverk kan systemet potentiellt använda sig av det redan existerande paketförmedlande nätverksgränssnittet (och på sätt undvika ett seriellt gränssnitt mot en GPS-mottagare och en puls per sekund ingång).

    När detta examensarbete startades var det inte känt var gränserna för PTP låg när det gäller att använda PTP tillsammans med radiobasstationer. Det var således inte klart ifall räckvidden för PTP kunde utvidgas till mycket längre avstånd än det ursprungligen var ämnat för. Detta examensarbete syftar till att visa att PTP kan användas som tillräckligt noggrann synkroniseringskälla för basstationer i nätverk med upp till fyra nätverksswitchar mellan PTP Grand Master och PTP slav.

    Examensarbetet har utförts vid avdelning Common Transport Feature på Ericsson.

    Nyckelord: precisions tid protokoll, radiobasstation, synkronisering

  • v

    Acknowledgement

    I would like to express my gratitude to my academic supervisor and examiner at KTH, Professor Gerald Q. Maguire Jr. for guiding me during this thesis project, providing helpful feedback and ideas during this thesis project.

    I also would like to thank my industrial supervisor and my manager in Ericsson Jiang Wu and Sheng Chu Li and who gave the opportunity to do this thesis at Ericsson AB. I would also like to thank Mikael Johansson and Mikael Olofsson in the Network Synchronization department for explanation about PTP protocol and current synchronization technologies. Moreover, I like to thank to Seth Norgren, Thomas Bäverlid and Goran Matovic in CTF lab for lending me necessary equipment for this thesis process.

    Furthermore, I like to special thanks to my friend Pooya Moazzeni Bikani for supporting me spiritually throughout my thesis.

    Last but not least, many thank to my parents and my sisters for their unconditional support and love throughout my life.

  • Tab

    AbstractSammanAcknowlTable of List of FList of TList of A

    11.1 1.2 1.3 1.4

    1.4.1.4.1.4.1.4.

    1.5

    22.1

    2.1.2.1.

    2.2 N2.2.2.2.2.2.

    2.3 2.3.2.3.2.3.2.3.2.3.2.3.2.3.2.3.2.3.2.3.2.3.

    2.4 A2.4.2.4.2.4.2.4.2.4.

    2.5 2.5.2.5.2.5.2.5.

    ble of C

    t ....................nfattning ......ledgement ....

    f Contents .....Figures ..........Tables ............Acronyms and

    IntroductionOverview .....Goals ............Limitations ...Methodology

    Literat.1 Experi.2 Data C.3 Conclu.4

    Structure of th

    Background Requirements

    Base s.1 Base s.2

    Network time NTP’s.1 NTP v.2 Simple.3

    Precision time Differe.1 PTP O.2 PTP S.3 Compo.4 Best M.5 Messag.6 PTP m.7 Delay .8 Delay .9

    Peer d.10 Metrop.11

    Available Syn Time E.1 Time I.2 Maxim.3 Time D.4 Modifi.5

    Related work IEEE 8.1 IEEE 1.2 Synchr.3 Other .4

    Conte

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................d Abbreviatio

    n .......................................................................................

    y .....................ture Study ......iment Design .Collection andusion ..............his thesis .......

    and related ws for base statistation synchrostation synchroe protocol ....... clock stratumersions ..........

    e Network Time protocol ......ence between

    Operation .......Stack architectonents of a PT

    Master Clock Sges in PTP ....

    message packeComputation request-respo

    delay mechanipolitan Area Nnchronization Error ..............Interval Errormum Time InteDeviation/ minfied Allan Dev......................802.1AS .........1588v2 over Wronization of Grelated work .

    ents

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................ons ..........................................................................................................................................................................d Analysis ..................................................

    work .............ion synchronizonization todaonization in th......................

    m ..........................................

    me Protocol (S......................PTPv1 and P......................ture ................TP network ....Selection ..............................

    et format .............................

    onse mechanissm .................Network .........Metrics ........

    ......................r .....................erval Error ....nTDEV ..........iation (MDEV............................................Wide Area NetGSM and UM......................

    vii

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ..........................................................................................................

    ......................

    ......................zation ...........ay ..................he near future................................................................

    SNTP) ..................................

    PTPv2 ...............................................................................................................................................................

    sm ...........................................................

    ..........................................................................................................

    V) .............................................................tworks ..........

    MTS basestatio.....................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................e ...........................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................

    ons with PTP ........................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ................... ii

    ..................... v

    ...................vi

    .................... ix

    .................... x

    ................. xii

    ..................... 1

    ................... 1

    ................... 1

    ................... 2

    ................... 3

    ................... 3

    ................... 3

    ................... 3

    ................... 3

    ................... 4

    ..................... 5

    ................... 5

    ................... 6

    ................... 7

    ................... 7

    ................... 7

    ................... 8

    ................... 8

    ................... 8

    ................... 8

    ................... 9

    ................... 9

    ................. 12

    ................. 13

    ................. 14

    ................. 14

    ................. 17

    ................. 17

    ................. 18

    ................. 19

    ................. 19

    ................. 19

    ................. 20

    ................. 20

    ................. 21

    ................. 21

    ................. 21

    ................. 21

    ................. 22

    ................. 22

    ................. 23

    i ii v ii x

    xi ii 1

    5

  • viii

    33.1 N

    3.1.3.1.3.1.

    3.2 3.2.3.2.3.2.

    3.3 3.4

    3.4.3.4.3.4.3.4.

    44.1 4.2

    4.2.4.2.4.2.

    4.3 4.3.4.3.4.3.4.3.4.3.4.3.4.3.4.3.4.3.

    4.4 4.4.4.4.

    4.5

    A55.1

    5.1.5.1.

    5.2 5.2.5.2.

    5.3

    66.1 6.2 6.3

    ReferencAppendiAppendi

    Method ........Network Con

    Switch.1 Netwo.2 Traffic.3

    Simulation ... Netwo.1 OMNE.2 OPNE.3

    Emulation ....Experimental

    Goals .1 Enviro.2 Scenar.3 Netwo.4

    Design and ISimulation wTestbed ........

    PTP n.1 Oscillo.2 Traffic.3

    Experimental Scenar.1 Scenar.2 Scenar.3 Scenar.4 Scenar.5 Scenar.6 Scenar.7 Scenar.8 Scenar.9

    Emulating net Delay .1 Packet.2

    Data Collectio

    Analysis .......Metrics .........

    Delay .1 Time D.2

    Evaluation .... Expect.1 Experi.2

    Summary .....

    Conclusions Conclusions .Future work .Required refle

    ces ................ix A ...............ix B ...............

    ......................nfiguration ......h .....................rk Topology ..

    c Load ..................................rk Simulator (

    ET++ .............ET ...................

    ...................... Study .................................

    onment Setup .rios ................rk Characteri

    mplementatiith OPNET ..........................ode Specificaoscope Specific generator .... Scenarios .....rio 0 ...............rio 1 ...............rio 2 ...............rio 3 ...............rio 4 ...............rio 5 ...............rio 6 ...............rio 7 ...............rio 8 ...............twork charact......................t Loss .............on ..................

    ......................

    ......................Variation ......

    Deviation (TD......................ted results......imental Result......................

    and Future w............................................ections ...........

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................(NS) ........................................................................................................................................................................istics Emulatio

    on .............................................................tion ...............

    fication ............................................................................................................................................................................................................................................................teristics using ..................................................................

    ......................

    ......................

    ......................DEV)/ minTDE

    ......................

    ......................ts .........................................

    work ...............................................................................

    ......................

    ......................

    ......................

    ......................

    .....................................................................................

    .....................................................................................

    ......................

    .....................................................................................on Scenarios .

    ......................

    ......................

    .....................................................................................

    ...................................................................................................................................................................................................................virtual bridge..........................................

    ......................

    ......................

    ...........................................

    EV..................................................................................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................e and Traffic C..................................................................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................Control ............................................................................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ......................

    ................... 25

    ................. 25

    ................. 25

    ................. 26

    ................. 26

    ................. 26

    ................. 26

    ................. 27

    ................. 27

    ................. 28

    ................. 29

    ................. 29

    ................. 30

    ................. 31

    ................. 34

    ................... 35

    ................. 35

    ................. 36

    ................. 36

    ................. 37

    ................. 37

    ................. 37

    ................. 38

    ................. 39

    ................. 40

    ................. 42

    ................. 43

    ................. 45

    ................. 46

    ................. 48

    ................. 49

    ................. 51

    ................. 52

    ................. 52

    ................. 53

    ................... 55

    ................. 55

    ................. 55

    ................. 55

    ................. 55

    ................. 55

    ................. 64

    ................. 81

    ................... 83

    ................. 83

    ................. 83

    ................. 84

    ................... 85

    ................... 91

    ................... 95

    5

    5

    5

    3

    5 1 5

  • ix

    List of Figures

    Figure 1-1: PTP over MAN............................................................................................................................ 2 Figure 2-1: General protocol architecture for IEEE1588 ............................................................................... 9 Figure 2-2: Software-based PTP functions .................................................................................................. 10 Figure 2-3: FPGA-based PTP functions ....................................................................................................... 11 Figure 2-4: Microcontroller-based PTP functions ........................................................................................ 11 Figure 2-5: PTP component in network, adapted from [28] ........................................................................ 13 Figure 2-6: PTP header format ..................................................................................................................... 15 Figure 2-7: Master-Slave message exchange in PTP-v1, adapted from [17] ............................................... 17 Figure 2-8: Message exchange with end to end transparent clock, adopted from [16] ................................ 18 Figure 2-9: Message exchange with peer to peer transparent clock, adopted from [14] .............................. 18 Figure 3-1: Traffic Management System (adapted from [34]) ..................................................................... 27 Figure 3-2: BC emulation design ................................................................................................................. 29 Figure 3-3: Overview of network scenario in the experimental study ......................................................... 31 Figure 3-4: GM andslave connected via virtual bridge ................................................................................ 34 Figure 4-1: Suggested new PTP engine process .......................................................................................... 36 Figure 4-2: Scenario 0 topology ................................................................................................................... 38 Figure 4-3: Scenario 1 topology ................................................................................................................... 39 Figure 4-4: Scenario 2 topology ................................................................................................................... 40 Figure 4-5: Scenario 3 topology ................................................................................................................... 42 Figure 4-6: Scenario 4 topology ................................................................................................................... 43 Figure 4-7: Scenario 5 topology ................................................................................................................... 45 Figure 4-8: Scenario 6 topology ................................................................................................................... 46 Figure 4-9: Scenario 7 topology ................................................................................................................... 48 Figure 4-10: Scenario 10 topology ................................................................................................................. 49 Figure 4-11: GM and slave connected with virtual bridge ............................................................................. 51 Figure 5-1: Delay variation, 100usec delay added to network ..................................................................... 56 Figure 5-2: Delay variation in different amount of delay added betwen GM and slave while

    connected through virtual bridge ............................................................................................... 57 Figure 5-3: Absolute value of delay variation between GM and slave while they connected

    together through virtual bridge .................................................................................................. 57 Figure 5-4: Delay variation, GM and Slave facing different amount of packet loss .................................... 58 Figure 5-5: Abstract value of delay variation, GM and Slave facing different amount of packet

    loss ............................................................................................................................................ 59 Figure 5-6: Packet loss in different profiles, five nodes between GM and slave ......................................... 60 Figure 5-7: Packet loss as captured by traffic generator in profiles 3 and 8, with 5 nodes between

    GM and slave ............................................................................................................................ 61 Figure 5-8: PTP PDV, five nodes between GM and slave, no traffic in the network .................................. 61 Figure 5-9: Histogram, five node between GM and slave, no traffic in the network ................................... 62 Figure 5-10: PTP GM and Slave versus relative time at sniffer, five node between GM and Slave,

    no traffic in the network ............................................................................................................ 62 Figure 5-11: Delta PTP GM and slave versus relative time at sniffer, five nodes between GM and

    Slave, no traffic in the network ................................................................................................. 63 Figure 5-12: Delta GM and slave versus delta relative time, five node between GM and Slave, no

    traffic in the network ................................................................................................................. 63 Figure 5-13: TE between GM and Slave while no traffic running ................................................................. 64 Figure 5-14: Absolute value of delay variation – No traffic running ............................................................. 65 Figure 5-15: Delay varioation when there is no traffic in the networks, no delay_req and

    delay_resp ................................................................................................................................. 66 Figure 5-16: Delay Variation in Different Profile load, No node between GM and Slave ............................ 67 Figure 5-17: Absolute value of delay variation in scenario 0 ........................................................................ 68 Figure 5-18: Delay Variation in Different Profile load, one node between GM and Slave ........................... 69 Figure 5-19: Absolute value of delay variation in different profile load, one node between GM

    and Slave ................................................................................................................................... 69 Figure 5-20: Delay Variation in Different Profile load, two node between GM and Slave ........................... 70 Figure 5-21: Absolute value of delay variation, two nodes between GM and Slave ..................................... 71 Figure 5-22: Delay Variation in Different Profile load, three node between GM and Slave ......................... 72

  • x

    Figure 5-23: Absolute value of delay variation, three node between GM and Slave ..................................... 72 Figure 5-24: Delay Variation in Different Profile load, four node between GM and Slave........................... 74 Figure 5-25: Absolute value of delay variation , four nodes between GM and slave .................................... 74 Figure 5-26: Delay Variation in Different Profile load, five node between GM and slave ............................ 75 Figure 5-27: Absolute value of delay variation, five nodes between GM and slave ...................................... 76 Figure 5-28: Delay Variation in Different Profile load, six node between GM and slave ............................. 77 Figure 5-29: Absolute value of delay variation – six nodes between GM and slave ..................................... 77 Figure 5-30: Delay Variation in Different Profile load, seven node between GM and slave ......................... 78 Figure 5-31: Absolute value of delay variation, seven nodes between GM and slave ................................... 79 Figure 5-32: Delay Variation in Different Profile load, ten nodes between GM and Slave ........................... 80 Figure 5-33: Absolute value of Delay Variation in Different Profile load, ten nodes between GM

    and Slave ................................................................................................................................... 80

  • xi

    List of Tables

    Table 2-1: Synchronization accuracy required of various radio standards (inspired by the table in [3]) ........................................................................................................................................... 5

    Table 2-2: PTP messages type and specification ........................................................................................ 15 Table 2-3: Header fields of the PTP packets [26] ....................................................................................... 16 Table 3-1: PTPd vs Linux PTP project ....................................................................................................... 28 Table 3-2: Hardware Specification ............................................................................................................. 30 Table 3-3: Network Interface Card (NIC) Specification ............................................................................. 30 Table 3-4: Summary of Scenarios and Tests (with × indicating a combination of number of

    nodes and load that was tested) ................................................................................................. 32 Table 3-5: Traffic pattern and traffic load specification ............................................................................. 33 Table 4-1: PTP node specification .............................................................................................................. 37 Table 4-2: Channel configuration of the osciloscope ................................................................................. 37 Table 4-3: Scenario 0 test description ......................................................................................................... 38 Table 4-4: Scenario 1 test description ......................................................................................................... 39 Table 4-5: Link bandwidth in scenario 2 .................................................................................................... 40 Table 4-6: Scenario 2 tests description ....................................................................................................... 41 Table 4-7: Link Bandwidth in scenario 3 ................................................................................................... 42 Table 4-8: Scenario 3 tests description ....................................................................................................... 42 Table 4-9: Link Bandwidth in scenario 4 ................................................................................................... 44 Table 4-10: Scenario 5 tests description ....................................................................................................... 44 Table 4-11: Link Bandwidth Scenario 5 ....................................................................................................... 45 Table 4-12: Scenario 5 tests description ....................................................................................................... 45 Table 4-13: Link Bandwidth scenario 6 ........................................................................................................ 46 Table 4-14: Scenario 6 tests description ....................................................................................................... 47 Table 4-15: Link Bandwidth scenario 7 ........................................................................................................ 48 Table 4-16: Scenario 7 tests description ....................................................................................................... 48 Table 4-17: Link Bandwidth scenario 8 ........................................................................................................ 50 Table 4-18: Scenario 8 tests description ....................................................................................................... 50 Table 4-19: Amount of delay manually added.............................................................................................. 52 Table 4-20: Profile ecplanation - Emulationg paclet loss ............................................................................. 52 Table 5-1: Standard variation, Delay added while GM and slave connected through virtual

    bridge ........................................................................................................................................ 58 Table 5-2: Standard variation in different profile, GM and slave connected through virtual

    bridge and facing packet loss .................................................................................................... 59 Table 5-3: Standard Deviation .................................................................................................................... 65 Table 5-4: Delay Variance in different vendors .......................................................................................... 65 Table 5-5: Test topologies baased on number of nodes .............................................................................. 66 Table 5-6: Standard deviation of delay variation with different load profiles- Scenario 0 ......................... 68 Table 5-7: Standard deviation of delay variation with different load profiles, Scenario 1 ......................... 70 Table 5-8: Standard deviation of delay variation with different load profiles, Scenario 2 ......................... 71 Table 5-9: Standard deviation of delay variation with different load profiles, Scenario 3 ......................... 73 Table 5-10: Standard deviation of delay variation with different load profiles, Scenario 4 ......................... 75 Table 5-11: Standard deviation of delay variation with different load profiles, Scenario 5 ......................... 76 Table 5-12: Standard deviation of delay variation with different load profiles, Scenario 6 ......................... 78 Table 5-13: Standard deviation of delay variation with different load profiles, Scenario 7 ......................... 79 Table 5-14: Standard deviation of delay variation with different load profiles, Scenario 10 ....................... 81 Table 5-15: Synchronizarion result for different number of nodes between GM and Slave in

    different load profile .................................................................................................................. 81 Table 6-1: D-link switch specification ........................................................................................................ 91 Table 6-2: HP switch specification ............................................................................................................. 92 Table 6-3: NETGEAR switch specification ............................................................................................... 93 Table 6-4: Westermo switch specification .................................................................................................. 94 Table 6-5: Mean, maximum and minimum delay variation (in microseconds), GM and slave

    connected through virtual bridge and facing different amount of delay(in microsecond). ............................................................................................................................ 95

  • xii

    Table 6-6: Mean, maximum and minimum delay variation (in microseconds), GM and slave connected through virtual bridge and facing different percentageous of packet loss. ............... 95

    Table 6-7: Mean delay variation (in microseconds) with different numbers of nodes between GM and slave with different load profiles. ................................................................................ 96

    Table 6-8: Minimum Delay variation(in microseconds) with different numbers of nodes between GM and slave with different load profiles. .................................................................. 97

    Table 6-9: Maximum Delay variation(in microseconds with different numbers of nodes between GM and slave with different load profiles. .................................................................. 98

  • xiii

    List of Acronyms and Abbreviations

    Advanced LTE Advanced Long Term Evolution AOCM Adaptive Oscillator Correction Method BC Boundary Clock BMC Best Master Clock CDMA Code Division Multiple Access CPU Central Processing Unit FDD Frequency Division Duplexing FIFO First In First Out FPGA Field-Programmable Gate Array GB Ggigabyte GHz Ggigahertz GM Grandmaster GPS Global Positioning System GSM Global System for Mobile LAN Local Area Network LTE Long Term Evolution MAC Media Access Control MAN Metropolitan Area Network Mbps Megabits per second MIB Management Information Base minTDEV Minimum Time Deviation ms Millisecond NA Not Applicable NS Nanosecond NTP Network Time Protocol OC Ordinary Clock OCXO Oven-Controlled Crystal Oscillator OS Operating System PCI Peripheral Component Interconnect. PLL Phased Locked Loop PPB Pulse per Billion PPM Pulse per Million PPS Pulse per Second PQ Priority Queuing PRC Primary Reference Clock PIM-SM Protocol Independent Multicast-Sparse Mode PTP Precision Time Protocol PTPD Precision Time Protocol Daemon QoS Quality of Service RBS Radio Base Station RTC Real-Time Clock RP Rendezvous Point SNMP Simple Network Management Protocol Sync-E Synchronous Ethernet TC Traffic Control TC Transparent Clock

  • xiv

    TDD Time Division Duplexing TDev Time Deviation TDM Time Division Multiplexing TLV Type-Length-Value UDP User Datagram Protocol UMTS Universal Mobile Telecommunications System UTC Coordinated Universal Time WAN Wide Area Network WCDMA Wideband Code Division Multiple Access WiMAX 2G

    Worldwide Interoperability for Microwave Access second generation

  • In1This

    will be of the th

    1.1 OWh

    synchroattachincommotime ofsynchroentire n

    In pnetworkused to or interused togway del1 ms wsecond achievehas beetiming Time (U

    1.2 GFigu

    timing i(GM) cGM closynchroits netwto discip

    ntrodus chapter prgiven, follo

    hesis structu

    Overviewen building

    onize all of ng a highlyonly done uf day and onization is network are

    packet switk time protosynchroniz

    rnal time sogether withlay is below

    when used insource is a

    e an accuracn successfuaccuracy of

    UTC).

    Goals ure 1-1 shoinformation

    clock. The aock acts as aonized acts work interfapline its loc

    ction rovides a geowed by theure.

    w g various tyf the base sty accurate sing a glob

    to discipldone with synchroniz

    tched netwocol (NTP)ze computerources (suchh local, metrw 128 millisn a metropavailable (scy of ~50 m

    ully used to f better tha

    ows an overn (1 pulse peactual granda master fora slave cloc

    ace (for examcal clock.

    eneral introe problem d

    ypes of witations withclock to eal positioniline an osan atomic ced.

    works two p) and the prrs to externh as those uropolitan, resecond (ms)olitan area

    such as frommicrosecondsynchronizn 100 nano

    rview of ther second (Pdmaster clocr all of the cck at the edmple an Eth

    1

    oduction to description,

    de area celhin a givenach radio bing system cillator. Inclock, to w

    protocols hrecision tim

    nal time souusing quartzegional, and)). Howevenetwork (M

    m a GPS rd (μs). PTPe clocks wi

    oseconds (n

    he planned uPPS) and timck might beclocks withidge of the Mhernet inter

    this thesis. , initial goa

    llular radion system. Tobase station(GPS) radi

    n traditionawhich all of

    have been ume protocol urces (such az oscillatorsd even natio

    er, NTP is liMAN); unlereceiver), inP, describedthin a local

    ns) with resp

    use of PTPme-of-day) e a GPS recin this netwMAN and rerface). The

    First, an ovls, limitatio

    o networks oday this isn (RBS). Io receiver t

    al circuit sthe elemen

    used for sy(PTP). NT

    as one or ms). NTP hasonal networimited to aness an accun which cad in the IEE

    area netwopect to Coo

    P over a MAare provide

    ceiver or an ork. A RBSeceives timiRBS will u

    verview of ons, and a su

    there is a s typically In the fieldto provide b

    switched tents of the op

    ynchronizatTP has beenmore atomics been succrks (where n accuracy urate one pase it is posEE 1588 staork with an ordinated U

    AN. In thised by a gran atomic clo

    S which neeing informa

    use this info

    the area ummary

    need to done by

    d this is both the

    elephony perator’s

    tion: the n widely c clocks) cessfully the one-of about ulse per ssible to andards, absolute

    Universal

    s system ndmaster ock. This eds to be ation via ormation

  • 2

    Figure 1-1: PTP over MAN

    Different types of RBS (e.g., GSM, Wideband Code Division Multiple Access (WCDMA), and Long Term Evolution (LTE)) have different timing requirements (these requirements will be described in detail in section 2.1.1). While PTP can normally meet these timing requirements in a local area network (LAN), in this thesis project we will examine whether PTP can meet these requirements in a MAN. To do this we will explicitly calculate a delay budget and examine the expected jitter in a MAN to which the various RBSs are attached. The main scientific question of this thesis project is to understand how well PTP can keep the slave within a bounded time of the GM and for what period of time the slave is within this bound.

    1.3 Limitations This section states the limitations of this thesis with respect to the study and

    implementation that have been carried out within this project.

    Initially, the planned thesis was to include a comparison between a simulation of the PTP protocol and experimental results using real hardware. However, after several months of working with a PTP implementation in OPNET, this effort was terminated because some parts of the code were missing and unavailable from the original author of the code.

    Following this the thesis aim shifted to an experiment with both full time aware networks and partial time aware networks. Unfortunately, budget limitations prevented the acquisition of a real hardware boundary clock (BC) or the emulation of a BC.

    As a result the experiments were done with older borrowed PTP devices that do not support PTPv2 (which is currently used in market). Moreover, those devices only support a special kind of GPS receiver. As a result the experiments were limited to measuring the time difference between a GM and slave where the GM stratum is 255 instead of 1. This means that the GM did not have a GPS disciplined oscillator and this GM was not synchronized with UTC; the results is that it was only possible to measure the relative deviation of the slave from this GM, rather than the absolute deviation from UTC.

    A further limitation of this thesis was that the industrial team, I was working with, did not have any competence in the area of synchronization; hence I have done this thesis completely independently. Having a larger budget and support from team mates, who had competence in synchronization would have been helpful and reduced the time necessary to carry out this thesis project.

  • 1.4 MThis

    were dedue to tmethodtiming r2.1, theRBS to

    1.4.1

    Thetechnolprovidenetworkrequiremthe mos

    1.4.2

    In ththe chacharactenetwork

    Thisdescript3.

    1.4.3

    Thisquantitafor eacoscillosEach teincreaserequiremcollectinpresente

    1.4.4

    Theused in and anythe GMatomic mention

    Methodols section deescribed in the desire fods were rulerequiremen

    en industry be connect

    Literature

    e first step inogies. Next

    ed the essenk. Two majments descrst relevant r

    Experime

    his thesis paracteristicseristics conk.

    s experimetion of how

    Data Colle

    s thesis prative researh of the sc

    scope was uest was repe our confidments of thng data is ed in Chapt

    Conclusio

    e result of ma small MA

    y slave. NotM is not co

    clock. The ning we exp

    logy escribes thesection 1.2

    or quantitatied out as a

    nts in a MANcould adopted to a GPS

    e Study

    n this thesist the standantial backgrjor synchroribed in secresearch reg

    nt Design

    roject severs of a largnsidered we

    ental studyw the experim

    ection and

    roject is mrch method.cenarios (dused to comeated two tdence that the most dem

    described ter 5.

    on

    measuremenAN, specifite that this t

    onnected to evaluation

    pect increas

    thesis fram2). A quantiive values fa goal of thN. If PTP it PTP for R

    S receiver.

    s project waards for synround knowonization mction 2.1, thgarding PTP

    ral scenarioge networkere: the num

    y is based ment was d

    d Analysis

    measuremen Data was

    details of thmpare the retimes to dethe time demanding RBin section

    nts carried ically a MAthesis projea primary with a GMing the accu

    mework useitative rese

    for the maxihe project wis able to mRBS synchr

    as to study tnchronizatio

    wledge regarmethods werhe main focP (as defined

    os were conk on the Pmber of no

    on the ddesigned and

    nt oriented,collected fr

    hese scenarelative timeecrease the eviation betwBS technol3.4.2. The

    out in this AN with a mect has only

    reference cM connected

    uracy by co

    d to achievarch approaimum delay was to dete

    meet the requronization –

    the timing ron in packerding synchre considercus was on d in IEEE 1

    nsidered in oPTP protoc

    odes in the

    deductive rd carried ou

    which rerom an isolrios are give difference

    confidenceween the sllogy. The ee analysis o

    thesis projmaximum oy shown thatclock (PRC

    d to a PRC ionnecting GM

    e the thesisach was chovariation. Q

    ermine if PTuirements d

    – eliminating

    requirementt networks

    hronization ed: NTP anPTP. Secti588).

    order to evacol. The twnetwork an

    research aput will be de

    sulted in tated networ

    ven in sectibetween a interval oflave and theexperimentaof the coll

    ject suggestof four nodet this sugge

    C), such as is left as fuM to a PRC

    s project’s gosen for thQualitative rTP can me

    described ing the need

    ts for differewere studiin packet s

    nd PTP. Giion 2.3 sum

    aluate the etwo main nd the load

    pproach[1]. escribed in

    the selectiork that wasion 4.3). Aslave and t

    f the resulte GM is wial setup uselected data

    ts that PTPes between estion is val

    a GPS recuture work. C.

    3

    goals (as is thesis research

    eet these n section for each

    ent RBS ed. This switched iven the

    mmarizes

    ffects of network

    d on the

    A full Chapter

    on of a s created A digital the GM. t, i.e., to ithin the ed when

    will be

    P can be the GM

    lid when ceiver or It worth

  • 4

    1.5 Structure of this thesis The first chapter of this thesis has presented the problem that is to be solved. Chapter 2

    provides the reader with the necessary background to read the rest of the thesis. Chapter 2 also summarizes related work that is directly relevant to this thesis. Chapter 3 describes the method and methodology used. Chapter 4 describes the design and implementation of a prototype solution. Chapter 5 evaluates this prototype with the goal of determining the bounds of operation for PTP with respect to the requirements for synchronization of different types of RBSs. Chapter 6 summarizes the conclusions that can be drawn from this thesis project, proposes future work, and offers reflections on some of the economic, social, and ethical considerations relevant to this thesis project.

  • Ba2This

    the rest RBS, an(startingthe diffresearch

    2.1 RA fr

    uplink technoltechniquby GlobWidebaand Wo

    Unlslots fofrequen

    The

    Table 2-

    RadCDMA

    GSM [4UMTS FDD[5]UMTS [6] UMTS FemtocTD SCDLTE FDLTE TD

    WiMAX

    * NA s

    ackgros chapter prof this thes

    nd then desg with NTPferent elemh.

    Requiremrequency diand downlogies, the ue, both thebal System and Code Dorldwide Int

    like FDD, tior uplink anncy, the mob

    e requiremen

    -1: Synctabl

    dio StandarA2000[3]

    4] WCDMA ] WCDMA T

    WCDMA c[3] DMA[3] DD[3] DD[3]

    X TDD[3]

    tands for Not

    ound arovides the sis. It beginscribes somP and PTP).

    ments of the

    ments forivision dupllink. In orproper fre

    e mobile hafor Mobile

    Division Muteroperabilit

    ime divisionnd downlinkbile device a

    nts for sync

    chronizatione in [3])

    rd

    TDD

    2

    st

    applicable.

    and relbackgrounds with the re of the tec The chapte

    e network.

    r base stlexing (FDDder to sup

    equency muandset and t(GSM), Un

    ultiple Accety for Micro

    n duplexingk. Thereforeand the RBS

    chronization

    n accuracy

    Frequenc50 ppb

    50 ppb 50 ppb

    50 ppb

    50 ppb

    50 ppb 50 ppb 50 ppb

    3 ppm a50 ppb bations

    5

    ated wd informati

    requirementchnologies ter also descThe chapte

    tation syD) network

    pport the hust be usethe RBS muniversal Moess (WCMDowave acce

    g (TDD) use in these nS time must

    n for various

    required of

    cy Accurac

    absolutebetween

    work on necessarts for synchrthat will be cribes how er conclude

    ynchronizk uses two shandoff prod. In all rust use the

    obile TelecoDA), Long ess (WiMAX

    ses the samenetworks, int also know

    s radio stan

    f various ra

    cy

    e base (de

    ry for the reronization oused in ourPTP needs

    es with a su

    zation eparate sets

    ocess in difradio standsame freqummunicatioTerm Evol

    X) FDD [2]

    e frequencyn addition t

    w when the t

    dards are sh

    adio standar

    Time±3µs (sh±10 µs NA* NA

    ±2.5µs

    NA

    ±3µs NA ±1.5µs s±0.5µs l±1-8 µs epending on

    eader to undof different r proposed to be addreummary of

    s of frequenfferent FD

    dards that uuency. FDDon System (lution (LTE.

    y, but differto using thetimeslot occ

    hown in Tab

    rds (inspired

    e Accuracy hould) (shall)

    small celarge c

    n implemen

    derstand types of solution

    essed by f related

    ncies for D radio use this

    D is used (UMTS) E) FDD,

    ent time e correct curs [2].

    ble 2-1.

    d by the

    ells, cells

    tation)

  • 6

    2.1.1

    In cspecificthese nRBSs. Fbecausefrom inthat useclocks a

    Tradused a systemsin the pCircuit-strict inthe RBSsynchroHowevetransitio

    2.1.1.1

    In ssynchrothe neeshown i

    2.1.1.2

    In Tsynchro

    2.1.1.3

    WCneeded shown i

    2.1.1.4

    In athat alsynchro

    2.1.1.5

    LTETDM

    Base stati

    cellular radically the netnetwork nodFrequency e the base snterfering we single freqare aligned

    ditional mocircuit swits all elemenprevious se-switched nn their needS will (in tonization soer, due to thoning to pac

    1 Synch

    second genonization. Reded frequein Table 2-1

    2 Synch

    TD-SCDMAonization re

    3 Synch

    CDMA does[3]. The s

    in Table 2-1

    4 Synch

    addition to fll base staonization re

    5 Synch

    E requires 1links may

    ion synchr

    io networkstwork itselfdes. Moreosynchronizastations nee

    with other cequency techin phase [7]

    obile backhtched netwonts of an opection differnetworks ned for time sythe case of ource via thehe increasincket based t

    hronization

    neration (2GReference cl

    ncy synchr1.

    hronization

    A, the RBquirements

    hronization

    s not rely synchroniza1.

    hronization

    frequency sations swiquirements

    hronization

    100 megabitnot be s

    ronization

    , synchronif, the nodesver, there aation is neced to stay wells). Phase

    hniques, suc].

    aul (used toork, typicallperator’s netrent types ed time synynchronizata circuit swe backhaul

    ng demand ftransport ne

    n in 2G/GS

    G) mobile nlocks are disronization [

    n in TD-SC

    BS needs bfor TD-SC

    n in WCDM

    on time-of-ation requir

    n in WIMA

    synchronizaitch betwewere show

    n in LTE/A

    ts per seconsuitable fo

    n today

    ization is neof the netware differencessary to swithin their

    e/time synchch as TDD,

    o connect tly time divtwork were of RBS nenchronizatiotion [3]. Hewitched baclink (in the for greater letworks base

    SM

    networks, sstributed ar[8]. The fre

    CDMA

    both frequenDMA were

    MA

    f-day informrements for

    AX

    ation, WiMAeen uplink/

    wn in Table 2

    Advanced

    nd (Mbps) aor this bac

    ecessary in work, and thnt kinds of support the r allocated hronization in order to

    the RBS toision multipsynchroniz

    eed differenon, but packence based ckhaul link)

    case of a plink bandwied upon IP/

    such as GSround the neequency acc

    ncy and tie shown in T

    mation, but r the differ

    AX needs t/downlink 2-1.

    LTE

    and faster bckhaul. In

    different pae radio inte

    f clock synccall handovspectrum (is necessar ensure that

    o the operatplexing (TDzed to a singnt degrees oket switcheon the back or will notacket switcdths, today MPLS and

    SM, the RBetwork and curacy requ

    me synchroTable 2-1.

    frequency rent forms

    ime synchrfor TDD

    ackhaul tranLTE tech

    arts of the nerface(s) attachronizationver mechan(this prevenry for base

    at all of the

    tor’s core nDM) links. gle clock. Aof synchroned networkskhaul netwo

    ot have an ached backha

    mobile baccarrier Ethe

    BS need frare used to

    uired by GS

    ronization [

    synchronizof WCDM

    ronization toaccess [3

    nsport netwhnology fr

    network, ached to n in the

    nism and nts them

    stations relevant

    network) In these

    As noted nization. s are not ork used available aul link). ckhaul is ernet.

    equency provide

    SM was

    [8]. The

    zation is MA were

    o ensure 3]. The

    works, so equency

  • synchrophase sy

    2.1.1.6

    To such asor NTPis the cElham Kfor intecould bsynchro

    2.1.2

    Theand femin indoosignal p

    MicThese bphase lfrequen

    Femmakes tThey usstable. Athem toreferenc

    2.2 NNTP

    years aservers betweenclients, more sefor clienof time.

    2.2.1

    In Nbut stra

    onization is ynchronizat

    6 RBS sy

    support syn Synchrono

    P) may be usurrent prefeKhorami [9

    ernal crystalbe used for onization of

    Base stati

    e most recenmto base staor areas, sucpenetration i

    cro and picobase stationlock loop (Pntly connect

    mto cells arethem suitabse a temperAs both pic

    o have a frece should be

    Network P [2] is a w

    ago (in 198and clients

    n clients anservers, an

    ervers as itsnts. Peers c.

    NTP’s cloc

    NTP, the difatum 16 is coStratum 0 Stratum 1 Stratum 2

    needed to tion, and to

    ynchroniz

    nchronizatioous Ethernesed. In WiMerred metho

    9] suggests tl oscillatorsfrequency c

    f multiple ch

    ion synchr

    nt base statations. The ch as a smais weak.

    o base statins use an oPLL) and sted via a Me

    e low poweble for deploature compco and femequency acce accurate a

    time prowidely used5) [3]. NT

    s. This protond servers [nd peers. A reference (

    compare the

    ck stratum

    fferent hieraonsidered tois a referenis a server tis a server t

    keep each provide the

    ation sum

    on among dt (Sync-E)

    MAX-TDD,od of synchthat chip scas. Such a ccontrol, buthip scale ato

    ronization

    tions that arintent of in

    all campus b

    ions are suioven-controsupport pacetro Etherne

    er base statioyment in aensated cry

    mto base statcuracy of upat approxim

    otocol d time syncP provides ocol uses th[11]. Each h

    client is a (for true timeir own time

    m

    archical layo be unsync

    nce clock whthat directlythat connec

    base statione time-of-da

    mmary

    distributed tor packet b, TD-CDMAhronization ale atomic c

    chip scale at there is stomic clocks

    n in the ne

    re introducentroducing sbuilding or

    itable for aolled crystalcket access et Network

    ions that pra weak signaystal oscillattions coverp to 250 pp

    mately ±0.1 p

    chronization synchroniz

    he user datahierarchicalclock that

    me). A servee with other

    yers define dchronized. Thich is consy connectedcts to stratum

    n within itsay [7].

    time referenbased synchA, and TD-[4]. Howevclocks may atomic clocktill a need fs.

    ear future

    ed to markesuch base stresidential b

    a small caml oscillator

    to a high which may

    rovide shortal environmtor (TCXO)r indoor arepb at the RFppm [13].

    n protocol. zation in a agram protol time syncqueries for

    er is a clockr peers in or

    different strThree of thesidered the md [14]. m 1 [14].

    s assigned s

    nces, physichronization (SCDMA RB

    ver, a recentbe cost effek would prfor setting th

    et by vendotations is tobuilding or

    mpus and in(OCXO) wdata rate bbe suitable

    ter range wment, such a) and PLL weas, this shoF output [1

    NTP was hierarchica

    ocol (UDP) chronizationr the currenk that providrder to sync

    ratum. NTPe most impomost accura

    spectrum, to

    cal synchro(such as IE

    RBS, a GPS t master’s tective replarovide a sigthe time of

    ors are micro increase chouse in w

    ndividual buwith an ultrbackhaul. Te for PTP [1

    wireless linkas small houwhich is moorter range 2], hence th

    developed al fashion bfor commu

    n structure nt time fromdes a time rechronize the

    P defines 25ortant strata ate clock [14

    7

    o ensure

    onization EEE1588

    receiver hesis by

    acements gnal that day and

    ro, pico, coverage

    which the

    uildings. ra stable They are 0].

    ks which use [11]. oderately

    requires he input

    over 20 between

    unication contains

    m one or eference eir sense

    56 strata, are: 4].

  • 8

    2.2.2

    NTPmillisecmillisecsub mic

    2.2.3

    As Nis a simsynchroused inmore coNTP, Sprimary

    2.3 PPTP

    among by utilizhave a standardimprovedifferentwo ver

    2.3.1

    PTPaccuraca LAN clock aadminiscompon

    PTPefficienbetter tcommuand mehierarchand map

    PTPa large extensioconfigu“profileto the d[23].

    NTP versi

    P exists inconds. The conds over crosecond a

    Simple Ne

    NTP has a mplified veonization con NTP and omplicated

    SNTP uses ay server is u

    PrecisionP is describdifferent dizing hierarclesser clockdized in twement of vnces betweersions as PT

    Difference

    Pv1 was descy of at leas that suppo

    accuracies, stration [21nents [22].

    Pv2 was puncy in compthan 1 nano

    unication linessage rate,hical synchrps to UDP/I

    Pv2 increasenumber of

    ons by uurations to es”. The usedevelopmen

    ons

    n different vcurrent ve

    a wide areaaccuracy wh

    etwork Tim

    complicatedersion of Nompared wiSNTP for calgorithm

    a single Ethunavailable i

    n time prbed in IEEistributed nechical mastek and will

    wo versionversion one)en these twTPv1 and PT

    e between

    signed to acst sub-microorts multicaand applic]. This prot

    ublished sixparison to thosecond, sunk, added tr, added a mronization, IPV4/IPV6/

    es bandwidf propagatiosing type-be setup

    e of profilesnt of the firs

    versions anersion is Na network (

    hen using a c

    me Protoco

    d implemenNTP. Howeth a full NTchecking timand uses s

    hernet connit can utiliz

    rotocol EE 1588 [14etwork noder and slavebecome a s: version ) was publio versions

    TPv2.

    n PTPv1 an

    chieve the foseconds of ast, applicacable in a tocol was a

    x years laterhe previous upport unicransparent cmechanism separated s/Ethernet [2

    dth efficiencon errors by-length-valuin specific

    s helps PTPvst IEEE 158

    nd is capabNTP [11] ve

    (WAN), subcesium osci

    ol (SNTP)

    ntation, hencever, this sTP implememing errorsseveral servnected time e a backup

    4]. PTP wades [15]. Syne clocks. Easlave to a none was p

    ished in 20in [20]. In

    nd PTPv2

    following obf real time cable in hetenetwork w

    also designe

    r and added version. It

    cast rather clocks, prov

    m for messasynchroniza23].

    cy by using y adding tr

    ue (TLV) devices. I

    v2 to be mo88 Telecom

    ble of proversion 4 whb milliseconillator as its

    ce it is not simplicity rentation. Ms and correcvers to estim

    server as aserver [14].

    as designednchronizatioach node canode whichpublished in008. Macé a

    the followi

    bjectives: toclocks in neterogeneous

    with minimued to be sim

    d some impwas designthat multic

    vide redundage extensioation messa

    shorter synransparent c

    parametersIn PTP theore flexibleprofile to a

    viding synchich providnd accuracy precise tim

    suitable forresults in loreover, difcting the timmate the cua reference .

    d for precion is achiev

    an be a masth has a bettn 2002 andand van Keing text we

    o enable syntwork nodenetworks t

    um resourcmple and inv

    provements,ned to achiecast, work dancy, provions, providges from si

    nchronizatioclocks, ands. PTPv2 ese configu[15]. Ericss

    address mob

    chronizationdes an accuy over a LA

    me resource

    r all systemlower qualifferent methme. NTP uurrent timefor time, b

    ise synchroved in this pter for nodeter clock. Pd version empen desce will refer

    nchronizaties, to be suitthat have d

    ces or withvolve minim

    , such as ineve synchroover asymide better s

    de configurignalling m

    on frames, pd provides m

    enables durations areson AB conbile backhau

    n within uracy of AN, and [13].

    s. SNTP ity time hods are

    utilizes a . Unlike ut if the

    onization protocol es which PTP was two (an

    cribe the to these

    on to an table for different

    hout any mal host

    ncreased onization

    mmetrical ampling ation of

    messages,

    prevents message different e called ntributed ul needs

  • 2.3.2

    PTPbased osystem,informa

    Theatomic in GPS differenin PTP interfacused to

    2.3.3

    PTPmessaginternalshows tunit incsoftwarsynchroas possihardwarcombinof the sexist.

    Figure 2

    PTP Opera

    P provides ton the IEEE, it determination betwee

    ere can be dclock whichsatellites),

    nt types of ccan be a m

    ce. In a mastsynchroniz

    PTP Stack

    P protocol ces that prol clock basethe relationscludes a prere unit implonization wible to the hre timestam

    nations of hasystem. The

    2-1: Gen

    ation

    time synchrE 1588 standnes the curren the netw

    different typh is very acor the cloc

    clocks and caster, a slavter-slave re

    ze the slave

    k architect

    can be expreovide time ed on the rship betweeecise real timements the hen using Phardware in

    mping is necardware ande following

    neral protoco

    ronization indard. This srent state of

    work nodes i

    pes of clocccurate, a Gck could be creates a mave, or even alation, the mnode [21].

    ture

    essed with tinformation

    received timen hardwareme clock anPTP protoc

    PTP. So thin order to becessary [24]d software; g subsection

    ol architectu

    n a networkstandard desf the clock,in order to p

    cks in a netGPS receiver

    some type aster and slaact as a mamaster is the

    two main fun between

    me informate and softwnd timestamcol. Timestas timestampe precise. T]. Existing Pthe details

    ns describe

    ure for IEEE

    k by exchanscribes the and provid

    perform syn

    twork. For r (synchronof crystal

    ave relationster on one e more accu

    unctionalitiemaster and

    tion in ordeware when ump unit to gamping is oping operat

    To achieve nPTP protocoof the implesome of th

    E1588

    nging messacharacterist

    des messagenchronizatio

    example, a nized to muloscillator. Pship betweeinterface an

    urate timing

    es: (1) exchd slave ander to stay asing IEEE1

    generate timone the mosion should

    nanosecond ol implemenementation he different

    ages betweetics of clockes to transmon [20].

    a clock coulltiple atomiPTP determen them. Eand slave ong device and

    hanging specd (2) updaat sync. Fig1588. The hmestamps, wst critical acbe located accuracy in

    ntation, useaffect the a

    t combinati

    9

    en nodes ks in the

    mit clock

    ld be an c clocks

    mines the ach node n another d will be

    cial PTP ating the gure 2-1 hardware while the ctions in as close

    n PTPv2 e various accuracy ons that

  • 10

    2.3.3.1 In a

    are realsoftwarin the m

    Figure 2

    2.3.3.2

    As commerprotocoand netfollowin

    An vary, bubased dFPGA d

    Softw

    a software ilized in softre-based PTmarket, of th

    2-2: Soft

    2 Hardw

    implementircial PTP im

    ol in hardwtwork physng paragrap

    FPGA-baseut in all desdesign offedirectly con

    are based

    implementatware. Figu

    TP system. These two op

    tware-based

    ware-base

    ing PTP inmplementatare, such asical layer phs.

    ed realizatiosigns the FP

    ers greater nnects the IE

    d PTP ation of PTPure 2-2 showThere are seen source im

    d PTP functi

    d IEEE158

    n hardware tions in hardas Field-Pro

    (PHY) [25

    on of IEEEPGA is plaaccuracy th

    EEE 1588 c

    P, both the ws the archieveral softwmplementat

    ions

    88

    enables gdware. Theogrammable5]. Each of

    E1588 is shoaced outsidehan a softw

    clock and th

    timestamp itecture of t

    ware-based Ptions will ex

    greater accure several we Gate Arraf these wil

    own in Figue of commuware-based

    he timestamp

    unit and ththe various PTP implemxamined in

    uracy, vendways of impay (FPGA)-l be briefly

    ure 2-3. Theunication in

    PTP implping subsys

    he IEEE158PTP functi

    mentations asection 3.3

    dors have dplementing -based, chipy described

    e exact desnterface. Anlementationstem.

    88 clock ions in a available [25].

    designed the PTP p-based, d in the

    ign may n FPGA-, as the

  • Figure 2

    In amicrocotimestammicroco

    Figure 2

    2-3: FPG

    a chip-baseontroller anmping wheontroller ba

    2-4: Mic

    GA-based PT

    ed design, nd Etherneen packetssed PTP sys

    rocontroller

    TP function

    accuracy iset Media A

    enter thestem [25].

    r-based PTP

    s

    s improvedAccess Cone MAC. F

    P functions

    d because tntrol (MACigure 2-4

    he chip incC) which p

    shows the

    cludes a dperforms he general v

    11

    edicated hardware view of

  • 12

    2.3.4

    Thesynchroboundarhierarchupon [2

    Grandm

    Ordina

    Bounda

    Transp

    Figuexploit between

    Componen

    ere are difonization. Try clock (Bhy in a netw26, 27]):

    master cloc

    ary clocks

    ary clocks

    parent clock

    ure 2-5 shothe concep

    n these com

    nts of a PT

    fferent typeThese compBC), and trawork. These

    cks The granor an

    A noOC.accoexpl

    A nocallealsoIn thbe sdevideem

    ks TCstransintroprotsysteend-introqueuclocthusfaste

    ows an exampt of PTP. Tmponents.

    TP networ

    es of PTPponents are ansparent ce different t

    most precindmaster clon atomic clo

    ode with a sThis cloc

    ording to thelained in sec

    ode with mued a BC. Th provides p

    his clock, onynchronizedice will act med slaves.

    were addesmission doduce an iocol with rem, these c-to-end tranoduces an inues from fk introduce it also preer converge

    mple of theThis figure

    rk

    P componethe grandm

    clock (TC).types of clo

    ise and permock. Generaock.

    single PTP pck can act e Best Mastction 2.3.5.

    ultiple PTP his clock is

    precise timinne port is cd with a moas masters t

    ed to PTP indevice. Theinterval delrespect to dclocks can nsparent clnterval delaforming in es propagatevents a qu

    ence in the c

    e network ealso helps

    ents that imaster clock

    Figure 2-5ocks are de

    manent clocally this clo

    port that actas a mas

    ter Clock (B

    ports that ausually a s

    ng for the sechosen by Bore accurateto synchron

    n version twmain task

    lay into thdelay compact as peerlocks. An ay for end-to

    the systemtion delay ueue from fcase of a cha

    elements anto visualize

    nteract in k (GM), ord5 shows thescribed bel

    ck in the sysock will util

    ts as an endster or slavBMC) algor

    acts as a trantandard swiegments tha

    BMC to act e clock. Thnize other cl

    wo. These dk of this tyhe system; ensation an

    r-to-peer traend-to-end

    o-end systemm. A peer-trather than forming anange in the

    nd PTP come the master

    order to rdinary cloce PTP mastlow [26, 27

    ystem is knolize a GPS

    d device is cve in the rithm which

    ansmission ditch or a rouat it is conneas a slave t

    he other porlocks which

    devices can ype of nodthis impro

    nd accuracyansparent c

    d transparenms, thus preto-peer tran

    n an intervand helps to

    network [2

    mponents ner-slave rela

    achieve ck (OC), ter-slave ] (based

    own as a receiver

    called an network

    h will be

    device is uter that ected to. that will ts of the h will be

    act as a de is to oves the y. In the locks or nt clock eventing nsparent al delay,

    provide 1, 27].

    eeded to ationship

  • Figure 2

    2.3.5

    Besnot invofor eachclock w

    In PThese a

    Straprimaryeither a

    Strareferenc

    Stra(greater

    Stra

    Straclock [1

    2-5: PTP

    Best Mast

    t Master Clolve TCs. Th communi

    with higher a

    PTP, clocksaccuracies a

    atum 1 hasy reference an atomic cl

    atum 2 hasce in a PTP

    atums 3 anr than 100 n

    atums 5-25

    atum 255 i18].

    P component

    ter Clock S

    lock (BMCThe main tacation link.accuracy an

    s are categoare specified

    s an accuracclock. Thisock or a GP

    an accuracnetwork [1

    nd 4 can bnanoseconds

    54 are reserv

    is a clock w

    t in network

    Selection

    ) is a distriask of the a. For each c

    nd the slave

    orized accod as:

    cy of withins clock is cPS receiver

    cy of within8].

    be consideres) [18].

    ved [18].

    with a defau

    k, adapted fr

    ibuted algoralgorithm iscommunicais a node th

    ording to th

    n 25 nanoseconsidered t[29].

    n 100 nanos

    ed as other

    ult setting a

    from [28]

    rithm that ins to define mation link thhat should b

    heir accurac

    econds. A sto be a gran

    seconds. Th

    r references

    and can nev

    nvolves BCmaster and

    he master wbe synchron

    cy by differ

    tratum 1 clondmaster cl

    is clock wil

    at a lower

    ver be cons

    Cs and OCsslave relat

    will be the pnized [29].

    rent stratum

    ock is conslock and is

    ll be a seco

    r level of a

    sidered as a

    13

    ; it does ionships port of a

    m levels.

    sidered a s usually

    ond level

    accuracy

    a master

  • 14

    2.3.6

    PTPand slavinclude

    In Pmessagorigin ctimestamthe mesclock up

    FollFollow_origin tand incPTP v1

    In PmessaginformaselectioPTP in messaginfrequesynchro

    PdemessagPdelay_

    2.3.7

    Tabdetail inhave thspecific

    Messages

    P is a synchve clocks. Tan origin ti

    PTPv1, the ee is sent byclock timesmp of this sssage deparp to date.

    low_up, D_up messagtime stamp.cludes the d will be exp

    PTP versiones. The Anation was ion leads to a

    the netwoe. This reflently and tonization.

    elay_resp, Pes in PTPv2_resp and Pd

    PTP mess

    ble 2-2 shown section 2.he same hcation of eac

    s in PTP

    hronization These messimestamp) a

    event messay a master. Tstamp. A Dslave clock.rts the node

    Delay_resp, ge is sent b. A Delay_rdelay receipplained in se

    n 2, Pdelaynnounce meincluded ina smaller Syork. The Sylects the asthat the pr

    Pdelay_follo2. The Signdelay_follo

    sage packe

    ws summary.3.5. Each m

    header. Heach field in t

    protocol wages can beand general

    ages are: (1This messagDelay_req m. An origin e. A Sync m

    and Manby a masterresp messag

    pt timestampection 2.3.8

    y_req and Aessage is ren Sync meync messageync messagsumption th

    rimary purp

    ow_up, and alling messw_up mess

    et format

    y of messagmessage coader formathe header fi

    which relies e categorizel messages.

    ) Sync mesge includesmessage istimestamp

    message wil

    nagement mr right afterge is sent bp. The com

    8.

    Announce mesponsible fssage. Usine, this can de will be shat the netwpose of the

    Signalling sage is used ages are use

    ge types in Pontains headt is shown

    field of the P

    on messaged into two

    ssage and (2 best maste

    s sent by ais the local ll be sent p

    messages r a Sync mby the mast

    mmunication

    messages wfor best mang a differdecrease thesend more work and se protocol

    messages wto send add

    ed in the ne

    PTP and theder, suffix an in FigurePTP packets

    ges exchangtypes: even

    2) Delay_reer selection a slave clocnode’s cloceriodically

    are generaessage and ter in responn between m

    were added aster selectirent message amount of frequently et of clockis to provi

    were added ditional dataew peer dela

    eir purpose and body. Ae 2-6. Tabs [30].

    ged betweennt messages

    eq message.informationck with thck value at to keep the

    al messaged includes a

    nse to a Demaster and

    to the set oion. In PTPge for bestf bandwidth than an An

    ks change revide accurat

    d to the set a using TLVay mechanis

    which discAll of the mble 2-3 sho

    n master s (which

    A Sync n and an e origin the time

    e slave’s

    es. The a precise elay_req slave in

    of event Pv1, this t master used by

    nnounce elatively te clock

    of event V tuples. sm [18].

    ussed in messages ows the

  • Table 2-

    Figure 2

    Messag

    Messagpurpos

    UDP po

    -2: PTP

    2-6: PTP

    e

    ge name SD

    ges se

    gttns

    ort p

    P messages t

    P header for

    event mess

    Sync Delay_Req

    generate antransport timestamps needed for tsynchronisa

    port 319

    type and spe

    rmat

    ages genmeFolDe

    d

    the ation

    usethebetclo

    po

    ecification

    neral essages llow_Up lay_Resp

    ed to measue link delay tween two ocks

    ort 320

    annomess

    ure buildsynchhierar

    unce age

    up the hronization rchy

    signalinmessage

    all otherpurpose

    15

    ng es

    r s

  • 16

    Table 2-3: Header fields of the PTP packets [26]

    Field value Explanation Octets Offset

    Transport Secific 0 IEEE1588 PTP message 0.5 0

    1 802.1AS PTP message 0.5 0 Message type 0 Sync message 0.5 0

    1 Delay_req message 0.5 0

    2 Pdelay_req message 0.5 0

    3 Pdelay_resp message 0.5 0

    4-7 Reserved 0.5 0

    8 Follow_up message 0.5 0

    9 Delay_Resp message 0.5 0

    A Pdelay_Resp_Follow_up ,essage 0.5 0

    B Announce message 0.5 0

    C Signalling message 0.5 0

    D Management message 0.5 0

    E-F Reserved 0.5 0

    Version PTP Version of IEEE1588 0.5 1

    Message length Length of PTP message 2 2

    Domain Number It shows the number of domain the PTP message sender belong to

    1 4

    Correction Field This is a correction value which is ns multiplied by 216

    1 5

    Source Port Identity Indicates clock ID and port which the message sent

    10 16

    Sequence Id Indicates the sequence number of PTP message which relate the messages together

    2 30

    Control Field It uses for compatibility to IEEEv1. Different by messages.

    1 32

    Log Message Interval Indicates the interval of PTP messages 1 33

  • 2.3.8

    In Pdelay m2-7 sho

    Figure 2

    s1 =

    m1

    s2 =

    m2=

    t(M

    t(sla

    2.3.9

    PTPsender-rbe usedmessagtranspar

    Delay Com

    PTPv1, the mechanism i

    ws how me

    2-7: Mas

    = slave rece

    = master se

    = slave send

    = master rec

    Master-Slave

    ave-Master)

    Delay req

    P was desigreceiver de

    d to computes are excrent clock.

    mputation

    delay requeis another dessages are e

    ster-Slave m

    ives Master

    ent the Mast

    ds a request

    ceives the re

    ) = s1-m1

    ) = m2-s2

    Delay =t(M

    uest-resp

    gned for symlay is equale end-to-en

    changed be

    m2

    m1

    Ma

    est-responsedelay compuexchanged b

    message exch

    r Timestamp

    ter Timestam

    at this time

    equest at thi

    Slav-Master

    onse mec

    mmetric appl to receive

    nd delay bettween a m

    aster

    e mechanisutation methbetween the

    hange in PTP

    p at this tim

    mp at this ti

    e

    his time

    2t(slave ave) +

    hanism

    plications; ser-sender detween transpmaster and

    Sync

    m1

    Follow_up

    Delay_req

    Delay_resp

    m2

    m is used fhod that wae master and

    P-v1, adapte

    me (via a Syn

    ime (includ

    Master)-e

    so the delayelay. In PTPparent clockslave with

    Sl

    for computias introduced slave to co

    ed from [17]

    nc message

    ded in Follow

    y computatioP v2, this mks [18]. Figh an interm

    s2

    s1

    ave

    ing delay. Ted in PTPv2ompute dela

    ]

    e)

    w_up messa

    Equ

    on assumesmechanism gure 2-8 shomediate end

    17

    The peer 2. Figure ay.

    age)

    uation 2-1

    that the can also

    ows how d-to-end

  • 18

    Figure 2

    2.3.10

    Thea peer-tthis memessagdelay be

    Figure 2

    2-8: Mes

    Peer delay

    e peer delay to-peer TC. echanism. es do not hetween two

    2-9: Mes

    m2

    Master

    m1

    Tra

    ssage exchan

    y mechan

    mechanismPdelay_req

    In contrasthave to be fo nodes [18]

    ssage exchan

    mDela

    r

    Syn

    Follow

    Delay

    m1

    tcA1

    tcA2

    ansparent ClDelay req

    nge with end

    ism

    m is used whq, Pdelay_ret to the deforwarded a. Figure 2-9

    nge with pee

    m2 ay_resp

    nc

    w_up

    y_req

    T

    Pdelay_( tcB

    Pdelay_

    ock A (TCA)uester

    Pdela

    d to end tran

    hen computesp, and Pdelay requesand are sim9 shows this

    er to peer tr

    Transparent

    _follow_up B1, tcB2)

    _req

    Tra

    ay_resp

    nsparent clo

    ting the deladelay_followst-response

    mply used ins mechanism

    ansparent c

    m2

    Delay

    Clock

    Syn

    Follm

    Delay_req

    tcB

    tcB

    ansparent CloDelay resp

    ock, adopted

    ay between w_up messa

    mechanismn measuringm in more d

    clock, adopte

    y_resp

    nc

    low_up m1

    2

    1

    ock B (TCB) ponder

    d from [16]

    two OCs, Bages are invm messageg the point-detail.

    ed from [14

    Slave

    s1

    BCs, and volved in s, these -to-point

    ]

  • tcA

    tcB

    tcB2

    tcA2

    Tranclock Athe enddelay ca

    t(req

    t(res

    2.3.11

    A mKilomeacross afrequen

    Groaggregaaggrega

    2.4 AIn t

    the twoThen, tr(TDEV

    2.4.1

    TE time. TE

    1 = the time

    1 = the time

    2 = the time

    2 = the time

    nsparent cloA (TCA)’s rd of this mean be calcul

    quester-resp

    sponder-req

    Delay

    Metropoli

    metropolitanetres [24]. Aa city. A Mntly used to

    oups of nodeate, and corate part is th

    Availablethis section o traditionalraditional c

    V), will be de

    Time Erro

    is a functioE is shown

    e TCA send

    e TCB recei

    e TCB respo

    e TCA rece

    ock B (TCBresponse. Tessage exchlated as:

    ponder) = tc

    quester) = tc

    y = t(reques

    tan Area N

    n area netA MAN can

    AN is usuaconnect bro

    es in a MANre. An accehe middle pa

    e Synchrothe availab

    l metrics timclock metricescribed.

    or

    on that calcis in equati

    ds a Pdelay_

    ives the requ

    onds to Pde

    ives the resp

    B) sends theThis Pdelay_hange, TCA

    cB1-tcA1

    cA2-tcB1

    respond-rste

    Network

    work (MAbe used for

    ally made upoadband net

    N can be claess network art that conn

    onizationble metricsme error (Tcs, Maximu

    culates the don 2-3[32].

    _req

    quest from T

    elay_req

    ponse

    e Pdelay_fo_follow_up

    A is aware o

    2t(resp der) +

    AN) [31] spr communicp of severaltwork subsc

    assified acck is the partnects the co

    n Metricsin network

    TE) and timum Time In

    difference i

    TCA

    ollow_up afp message cof tcA1, tcA

    requ-erpond

    pans a largcation betwl LANs concribers to a l

    cording to tht of a MANore network

    s k synchronizme interval nterval Error

    in time valu

    fter respondcontains tcBA2, tcB1, a

    uester)

    ge area of een multipl

    nnected via blarger netwo

    he followingN closest to

    to the acce

    zation will error (TIE)r (MTIE) a

    ues of two

    ding to TranB1 and tcB2and tcB2. T

    Equ

    f between 5le buildingsbridges. A

    work such as

    g hierarchy the custom

    ess network(

    be explaine) will be diand Time D

    clocks at a

    Equ

    19

    nsparent 2. So, at Thus the

    uation 2-2

    5 to 50 s or even MAN is WAN.

    : access, mer. The (s).

    ed. First scussed.

    Deviation

    a certain

    uation 2-3

  • 20

    2.4.2

    TIEtwo diff

    Where:

    local cloto calcu[33].

    2.4.3

    MTtime. Ea

    Where:

    Time Inte

    E (shown infferent times

    tisaceπcertaiTIE and TEock to a refulate the fre

    Maximum

    TIE is a metach period’

    , num

    erro

    inde

    total time

    erval Error

    n equation 2s (as shown

    ,ertaintime; nperiodofiE are traditference clocequency erro

    m Time Inte

    tric that mes time inter

    mber of samp

    or sampling in

    ex of the spec

    l number of

    e error sampl

    r

    2-5) computin equation

    T(t+π

    interval; tional metrick (in order or using TI

    erval Erro

    easures the val error ca

    le within a c

    nterval;

    cial period o

    samples;

    le;

    tes the timen 2-4 [32]).

    π)-T(t)

    ics that can to perform

    IE metrics b

    or

    maximum an be associ

    certain period

    over the data;

    e difference

    be used tophase synch

    by calculatin

    interval erated with a

    d;

    ;

    e between t

    o measure thhronizationng the slope

    ror during sliding win

    ,

    two clock v

    Equ

    Equ

    he phase ern). It is also e of a curve

    a certain pndow.

    , , . . ,

    values at

    uation 2-4

    uation 2-5

    rror of a possible e of TIE

    period of

    Equuation 2-6

  • 2.4.4

    Minwhich iwhether

    ,

    2.4.5

    MDfrequeninterval

    2.5 RAfte

    Timing Networstandardactuallyexplain in more

    Somthese ar

    2.5.1

    IEEapplicatbased omechan

    Managmechan

    Path de

    Time Devi

    nimum Timis independer the networ

    , …Modified A

    DEV is a trncy by reducl into n cy

    Related wer PTPv2, t

    and Syncrks” standard has the stry for time

    how PTP we detail.

    me experimere explained

    IEEE 802.

    EE 802.1AStions such on IEEE 1nisms of this

    gement nism

    elay mecha

    iation/ mi

    me Deviationent of the nrk meet a sy

    Allan Devi

    aditional alcing the tota

    ycle clock pe

    work the IEEE 80chronizationrd based onring “for LAsensitive awill operate

    ents have bd in more de

    .1AS

    is a timingas audio/vi

    1588 v2 ans PTP profi

    TMm

    anism Apin

    Aan

    inTDEV

    n (minTDEumber of hoynchronizat

    iation (MD

    lgorithm deal observatieriod [34].02.1AS “Stan for Timn the seconAN and metpp