49 Semtech

7/29/2019 49 Semtech

1/77

Semtech Patrick Diamond PhD Director Systems Engineering

Network Synchronization

Tel Aviv IsraelJune 14, 2007

Patrick Diamond PhD

Director Systems Engineering

Advanced Communications and Sensing

7/29/2019 49 Semtech

2/77

Semtech Patrick Diamond PhD - Director Systems Engineering

A synchronous network is one that provides a common clock.

- Voice and Video are isosynchronous types of information

- Data is asynchronous

- The circuit that carries these info types is synchronous and

provides the ingress and egress clocks.

Network Service providers implement BITS and SSU machines in their

central and serving offices to deliver the network clock to the network

equipment.

A network is synchronized if all elements operate on the same Epoch.

A network is syntonized if all elements operate at the same frequency.

What makes a Network Synchronous?

7/29/2019 49 Semtech

3/77


Basic System Reference Clock Architectures

Autonomous - No system reference clock or clock distribution system. Line

cards generate local clocks. Depends upon logic buffers and inter-card

memory stores for data stability. No protection switching or frequency/phase

transient control. Not capable of performing SSM or clock transit node

functions.

Centralized - Centralized reference clock generation and distribution.

Protection switching capability typically limited to frequency matching. Not

capable of input to output phase build-out control. Cannot correct for phasedisturbances. Line cards have same limitations.

Distributed - Reference clock generation and distribution with phase build-out

and frequency control on clock card and selected line cards.

7/29/2019 49 Semtech

4/77


Why is Synchronization so Important to Services Delivery?

Transport Networks of the last 20 years have always been built using

the circuit technology E-1, T1, SDH and Sonet.

These transport techniques have a fundamental need for physical

layer clock synchronization.

The BIT rate and the Line utilization are always the same value, ie an E1

has 2048 bit periods as well as 2048 clock cycles.

The Network Clock provides the perpetual heartbeat for the

services using the network. The source of the Network Clock is a caesium beam oscillator located

in a GPS satellite, a BITS or SASE. This is called the G.811 Clock in ITU

standards parlance.

Circuits use a frequency synchronous clock to control the ingressand egress of the information streams to be transported.

7/29/2019 49 Semtech

5/77


Why has Syncronization been so Important to Transport?

The Synchronization standards guarantee the performance of today's

transport networks.

The Currency of Synchronization is TDEV and MTIE.

Time Deviation is measured for 10000 seconds and compares areference clock, G.811 quality to a derived clock. The amount of time

drift allowed from the derived clock is tightly regulated. The

measurement is presented in the form of a graphic mask target and the

actual measured.

Maximum Time Interval Error is measured for 100000 seconds and

accumulates the time error between the above clocks. This

measurement is also presented in graphic form with a target mask and

actual measured plot. Synchronization standards define the limits of reference and nodal

system clock wander, phase gain, phase jump and jitter.

7/29/2019 49 Semtech

6/77


Why has Syncronization been so Important to Transport?

The ITU and ANSI standards organizations have defined the rulesfor Synchronization.

These rules are expressed as a numerical standard. The ITU

standards typically start with a G. and the ANSI standards start with a

GR-.

There are hundreds of standards issued by each organization.

7/29/2019 49 Semtech

7/77


ITU Synchronization Standards

7/29/2019 49 Semtech

8/77


ANSI Synchronization Standards

7/29/2019 49 Semtech

9/77


Synchronization Clock Quality Distribution Rules

7/29/2019 49 Semtech

10/77


Synchronization Distribution Hierarchy

7/29/2019 49 Semtech

11/77


Synchronization Currency in Todays Circuit Transport Networks

10ns

100ns

1us

10us

100us

100ms1s

10s100s

1,000s10,000s

100,000s

Obs. Interval(sec)

PRC (Cs)G.811

SSU (Rb)

G.812

SSU (Rb)

G.812

}SECs (Quartz)

G.813

SECs (Quartz)G.813}

CDR

PLL

Framer

SDHG.823

(Traffic)

G.823(Sync)

BTS/NodeB(50ppb)

BTS/NodeB(50ppb)

2.5us phase tolerance (TDD)Requires GPS to achievePhase alignment

PRCSEC SSU

G.823 Sync

G.823 TrafficE1

E3

E1

7/29/2019 49 Semtech

12/77


SETS - Functionality as Defined by ITU Standards

SelectASelectC

Squelch T4

T0

n x T1p x T21 x T3

OSC

SETGSelectB Squelch

T1 - Clocks recovered from SONET/SDH line

T2 - Clocks supplied from PDH tributary inputT3 - Clock supplied from HSCT0 - Clocks supplied to SONET/SDH equipmentT

4

- BITS clocks supplied back to HSC source

7/29/2019 49 Semtech

13/77


Synchronization Solution - Upstream / Downstream

Network Input SEC References

Line Card 622M

LineOutputFrequencyMultiplier

Line Card DS1

LineOutput

Frequency

Translation

Line Card 2.5G

LineOutput

Frequency

Multiplier

SSM InputReference

SourceSelection

Priorities

Backplane SECs

Line InputFramed

Data Mb/s

Line Interface1

Line InputFramed

Data Mb/s

Line Interface2

Line InputFramed

Data Mb/s

Line InterfaceN

SETS ClockGenerator

Slave

BITS SEC

ClockPDH SECClock

SECMFrSync

SETS ClockGenerator

Master

SECMFrSync

7/29/2019 49 Semtech

14/77


Traditional Home Grown SETS Implementation

High component count means: high component cost board space consumption layout problems

PhaseDetector

LowPassFilter

Digital Filter,TemperatureCompensation,Control

DAC LowPassFilter

MUX

ADC

Temp.Sensor

ADC

TCVCXO

PDH

BITS

Ref Clk

MC68xx16

uP

Restricted choice givesinadequate flexibilityand back-up

Sensitive analog circuitryneeds special care inlayoutDSP uP runs

complex, proprietary

algorithm

Temp. compensationmust be calibrated- ignores differentialaging coefficients

Extremely difficult to get manufacturing

repeatability and achieve standardcompliant performance.

7/29/2019 49 Semtech

15/77


Oscillator Performance Requirements

0.01

0.001

4.6 + 4.6

4

0.012

0.01

0.001

0.001

4.6

4.6

Stratum 3E

GR1244-III

G.812

0.010.01UIJitter (generation) RMS

1-101-10 (3)HzFilter bandwidth

4.6 + 4.64.6 + 4.6ppmPull-in/Hold-in range

24ppmAging (20 years)

2.06

2.0

0.01

0.05

0.37

0.28

0.04

0.05

ppmHoldover stability (total)

Temp

Drift (24 Hrs)

Initial Offset

4.64.6ppmFreq offset (total)

4.64.6ppmFree run accuracy

ETSI

300-462-5

Stratum3

GR1244-II

G.813

UnitParameter

The difference between initial offset and drift is that drift directly relates to the performanceof the external XO. The initial offset is directly tied to the performance of thetiming control when switching from a locked state to holdover.

7/29/2019 49 Semtech

16/77

7/29/2019 49 Semtech

17/77


ACS8520/22/30- Two independent PLL paths for ITU Compliance

Two PLLs, known as the T4 & T0 path are completely independent

They are different for good reasons

The T4 path is to select a clock to output to a local BITS or SSU

It is not filtered, no phase buildout

BITS/SSU will filter it to a much higher degree than our SETS (T0) function

The T0 path is used to generate the clock for the egress ports

PRC

SSU

SECs

Syn

cTrail T0 Path

Syn

cTrail

T0 Path

T0 Path

T4 Path

T0 Path

SSU may wellprovide time from

GPS or othersource. But can

simply clean upthe linerecovered clock

SSU

SECs-

line timed

SEC & co-

located SSU

ITU Sync

Trail

7/29/2019 49 Semtech

18/77


ACS8530 Block Diagram

7/29/2019 49 Semtech

19/77

amondPhD

-DirectorSystems

Engineering

E

xistingSyncCardFamily

1041

010Number of Outputs

1441

414

Number of Inputs

Revertive/Non-revertive mode

OutputConfigurations

PBO on currently locked reference

Stratum 3E compliance

N x 8 kHz (N = 1 to 12499)

6.48, 19.44, 25.52, 38.88, 51.84, 77.76 MHz

N x E1 or N x DS1, N = 1 to 16

2, 4, 8 kHz

155.52 MHz

Programmable pulse width and polarity

2 kHz and 8 kHz Frame Sync Pulses

1.544 MHz or 2.048 MHz (E1 or DS1)

19.44 MHz

44.736 MHz or 34.368 MHz (E3 or DS3)

6.48, 25.52 MHz

38.88 MHz, 51.84 MHz, 77.76 MHz

311.04 MHz

1.544 MHz or 2.048 MHz (E1 or DS1)

Holdover (with appropriate TCXO, OCXO)

Activity Monitoring

Los of Lock Indication

Output Phase Adjustment

Phase Measurement

InputConfigurations

General

Features

Features

155.52 MHz

Frequency Monitoring

8 bit micro-processor modes

Hit-less Source Switching

7/29/2019 49 Semtech

20/77


High Rate LineCard WithHoldover

Recovered Clock

Clock A

Sync A (Optional)

Framer

XO

Clock B

Sync B (Optional)

Data SERDES

ACS85xxLC/P

ACS89xx

ClockDistribution

High-Rate (dddd16161616TDM Line card,

oldoveroldoveroldoveroldover

ACS8525/26/27 for stability and Holdover

Supports SEC frequencies from 2kHz to 155MHz

ACS894x for OC-12 & to STM-16 outputs

7/29/2019 49 Semtech

21/77


To/FromNetwork

High Rate Line Card (

7/29/2019 49 Semtech

22/77


To/FromNetwork

Low Rate Line Card(

7/29/2019 49 Semtech

23/77


Existing Line Card Part Family- Feature Chart

2 & 83155MHz2kHz23ACS8525

622MHz19MHz11ACS8944

FE

C

Rates

Sh

ortT

ermH

old

ov

er

(ms)

LongT

ermH

old

ov

er

(indef)

2

2

2

3

Num

ber

ofin

puts

Features

Ph

aseB

uild

out

Avail

ability

Performance

Level

2 or 8*1622MHz19MHz4ACS8946

Part No.

STM-64

2

2

6

Number

of

outputs

LOS2 & 8155MHz8kHzACS8527

2 & 8

2 & 8

S

yncoutputs(kHz

)

155MHz

155MHz

Maxim

umo

utput

Fre

quen

cy

2kHz

2kHz

Min

imumin

putfre

quen

cy

LOS

Aut

omaticR

ef

Switchin

g

STM-4

STM-1

6

ACS8526

3ACS8595

STM-1

Mic

ropro

cessorIn

terfa

ce

Nu

mber

of

Syn

cin

puts

7/29/2019 49 Semtech

24/77


PICMG AMC.0

Advanced Mezzanine Card

AMC defines

A modular add-on or child card

Extends the functionality

Allows multiple line cards off a single

carrier blade

Mix and match line cards e.g. DS1/E1/J1

DS3/E3

OC-3/12/48

GbE WAN Cards

10 GbE Optical WAN Card.

Unique ACS8595 drop-in solution

ATCA AMCCarrier blade-

Vendor 'A'

Clock A

Sync A (Optional)

XO

Clock B

Sync B (Optional) ACS8595LC/P

RecoveredClocks

RecoveredClock

Selection

AMC Slot 1

AMC Slot 4

AMC Slot 3

AMC Slot 2

Vendor 'B' AMC Card

AMC Slot 2

Vendor 'C' AMC CardEach Clock to AMC Module:

8kHz, E1, DS1 or 19.44MHz

Vendor 'A' AMC Card

7/29/2019 49 Semtech

25/77


Synchronization Status Message Processing per G.781

The process of selecting a synchronization source from the set of physical ports is performed inthree steps:

Figure 6/G.781 - Visualization of the synchronization source selection process(es)

7/29/2019 49 Semtech

26/77


Synchronization Status Message Carriage SDH

Do Not use for synchronization111115

Reserved111014

Reserved110113

Reserved110012

Synchn Eqpt Time Src. SETS101111

Reserved101010

Reserved10019

Rec. G.812 Local10008

Reserved01117

Reserved01106

Reserved01015

Rec. G.812 Transit01004

Reserved00113

ITU Rec G.81100102

Reserved00011

Quality Level Unknown00000

SDH DescriptionS1 bits

b5 b8

Qual Lev

SSM Allocation Definitions for SDHFrom G.707 (03/96) Table 5

RSOH

MSOH

Pointer

POH Payload

1 32 4 5 76 8

Synchronization Byte

A1

D2E1

D3F1

A2 J0A2A2A1A1

D1B1

D8

K1B2B2B2

D5

E2

D12

D9

D6

K2

D4

D7

D10

M1

D11

S1

Admin Unit Pointers

STM-1 Section Overhead

7/29/2019 49 Semtech

27/77


Synchronization Status Message Carriage SONET

Do Not Use for Synchronization11117

Reserved for Network Synchronization

Use

1110Use Assignable

Reserved1101-

SONET Minimum Clock Traceable11005

Reserved1011-

Stratum 3 Traceable ST310104

Reserved1001-

Reserved1000-


Reserved0110-

Reserved0101-

Reserved0100-

Reserved0011-

Reserved0010-


Synch Tracability Unknown - STU00002

SONET DescriptionS1 bits

b5 b8

Qual Lev

SSM Allocation Definitions for SONETFrom Bellcore GR253 CORE(12/97), Sect 5.4.2

SectionOH

LOH

Pointer

POH Payload

1 32 4 5 76 8

Synchronization Byte

STS-N Transport Overhead

A11

D2E1

D3F1

J0

D1B1

D8

K1

D5

E2

D12

D9

D6

K2

D4

D7

D10

M1

D11

S1

SPE Pointers

A1N

B21 B2N

A21 A2N

Z1 Z1 Z2 Z2

Z0 Z0

7/29/2019 49 Semtech

28/77


Synchronization Status Message Carriage E1

TS0 TS1 TS16 TS31

Frame Payload

E1 Frame - Time Slots 0 to 31 - 256 bits - 125 us One Time slot = 8 bits

E1 Time Slot 0 (TS0) Multi-Frame (MF)(SMF = Sub Multi-Frame ) - From G.704 10/98 Table 5D

Sa84

Sa83

Sa82

Sa81

Sa84

Sa83

Sa82

Sa81

Sa74Sa64Sa54Sa44A1E15

1101100C414

Sa73Sa63Sa53Sa43A1E13

1101100C312

Sa72Sa62Sa52Sa42A1011

1101100C210


1101100C18

II


1101100C46


1101100C34


1101100C22


1101100C10

I

87654321

Bit Number 1 to 8Frame

Num

S

MF

MF

Do Not use for synchronisation111115

Reserved111014

Reserved110113

Reserved110012

SETS101111

Reserved101010

Reserved10019

SSU-B (Was G.812 Local )10008

Reserved01117

Reserved01106

Reserved01015

SSU-A (Was G.812 Transit )01004

Reserved00113

ITU Rec G.81100102

Reserved00011

Quality Level Unknown00000

DescriptionSan1, San2,San3, San4,

- Note 1

Qual Lev

SSM Allocation Definitions for E1From G.704 10/98 Table 5C

Note 1: n = 4,5,6,7 or 8 ( ie one Sa bit only ) depending onoperator selection

7/29/2019 49 Semtech

29/77


Synchronization Status Message Carriage DS1

Bit 3Bit1Bit 2 Bit 192 Bit 193--------------------------------------------------------

F-Bit Frame Payload

DS1 Frame - 193 byte 125 us

Assigned SSM in DataLink acc G.704 10/98 Table 2,ANSI T1.403 Table 4, Bellcore GR-253 CORE (12/97),

Sect 5.4.2.

Note:All messages have the form11111111 0 P1 P2 P3 P4 P5 P6 0

So each is one code of a basic 6 bit message

11111111 00100000

11111111 00110000

11111111 00111110

11111111 00001000

11111111 00010100

11111111 01000100

11111111 0001111011111111 00010000

11111111 00001100

11111111 00000010

G.811

G.812 Type II

G.812 Type III

G.812 Type IV

Stratum 4 (Note)

G.813 Option 3

G.812 Type VSynchronization Traceability

Unknown

Do not use for Synchronization

Provisionable

ESF FrameESF Frame

NumberNumber

11

22

33

44

5566

77

88

99

1010

1111

1212

13131414

1515

1616

1717

1818

1919

2020

2121

22222323

2424

Each F-Bit occurs in Bit 1 of each

frameFPS = Framing Pattern Sequence

(001011)CRC6 = CRC6 word C1 to C6FDL = Facility Data Link, m = FDL

bit

FPSFPS

--

--

--

00

----

--

00

--

--

--

11

----

--

00

--

--

--

11

--

----

11

CRC6CRC6

--

C1C1

--

--

--C2C2

--

--

--

C3C3

--

--

--C4C4

--

--

--

C5C5

--

--

--

C6C6--

--

FDLFDL

mm

--

mm

--

mm--

mm

--

mm

--

mm

--

mm--

mm

--

mm

--

mm

--

mm

--mm

--

FF--BitsBits

7/29/2019 49 Semtech

30/77


SSM-Based Network Resynchronization Flow

NE NE NEPRC - ST1

G.811

SSM carriage. Value = G.811/ Stratum 1

NENE

BITS ST2G.812

NE NE

SSM carriage. Value = G.812 /

Stratum 2

NE NE

PRC FAIL

Timing direction

NEPRC - ST1G.811

BITS ST2

G.812

7/29/2019 49 Semtech

31/77


SSM - Network Synchronization Reconfiguration

NE 3

NE 4 NE 5

PRCG.811

NE 2 NE 1

All info before failure. THEN X happens

ST1

ST1 ST1

ST1

ST1

ST1

DUS

DUS

DUS

DUS

NE 3

NE 4

NE 5

PRCG.811

NE 2 NE 1

LOS

DUS DUS

ST1

ST1

ST1

DUS

DUS

ST1ST1

Secondary

Primary

NE 4 NE 5

LOS

ST1 SMC

ST1 DUS

DUS

NE 4

LOS

DUSSMC

ST1 ST1

DUS

1. NE 1, 2, 3 cfgd so if P&S = same, use Secondary fortiming.

2. NE 4,5 cfgd so if P&S = same, use Primary for timing.

SSM Definitions :DUS: Do not Use for SynchronizationST1: Stratum 1SMC: SONET Minimum Clock

7/29/2019 49 Semtech

32/77


Mobile Wireless Synchronization Requirements

Overview of sync requirements for cellular/mobile access networks: Clock rate synchronisation: Deviation of base station clock rate (= clock

frequency) needs to be within certain limits from reference clock present incentral radio control.

Time or phase synchronization: Offset of base station clock phase (or time

offset) needs to be within certain limits from reference clock phase (or time)present in central radio control.

Sync required to ensure seamless connection hand-over of handsets movingbetween coverage areas of different base stations

TDD (time division multiplex) requires both clock rate and phase/timesynchronization to reference clock. Spectrally efficient.

FDD (frequency division multiplex) requires only clock rate synchronization toreference clock. Spectrally inefficient.

Classification of international mobile radio systems:

CDMA2000 (3GPP2, US ,Asia): TDD system with a 50ppb frequency sync

and time/phase within +/- 1.25us WCDMA (3GPP, Europe ME, Asia) and GSM: FDD system with a 50ppb

frequency and no particular time/phase requirement & TDD system with a50ppb frequency sync and time/phase within +/- 1.25us

Pico RBS (WCDMA and GSM): FDD 100ppb frequency and no particular

time/phase requirement

7/29/2019 49 Semtech

33/77


Break time

Coffee Break, standup and walk around, make your calls and eat lunch

7/29/2019 49 Semtech

34/77


Next Generation Networks are Packet Networks

Does synchronization change in PacketNetworks?

Firstly the definition of and performance requirements for

synchronization will not change. In Packet networks synchronization will become an access or last mile

service aide to delivery of traditional services. In this model the service

being delivered will determine the quality of synchronization required.

The financial drive to switch to a sharedtransport versus dedicated transport is

overwhelming.

The ratio in cost per megabit for native ethernet backhaul

service versus traditional private line service is 1 to 6 or greater.

7/29/2019 49 Semtech

35/77


A Financial Example

Traditional Leased Line Access pricing model;Assumed 100Mbps of an STM1 is used at RNC and 3xE1 at each Node B (giving 17 Node Bs in RNS).

Assumed distance between RNC and provider network POP to be 5 miles, and average distancebetween provider network POP and Node Bs to be 5 miles.

Then, yearly rental costs:RNC: 100,295 + 18,000 = 118,295

Node Bs: 100M/E1 x (4098 + 1000 + 1192) = 314,500

Total annual cost: = 432,795

Simple Ethernet Backhaul pricing model;Assumed 100Mbps Ethernet at RNC: = 14,000

Assumed 6Mbps Ethernet at each of 17 Node Bs: 17 x 2600 = 44,200

Total annual costs: = 58,200

Potential ANNUAL SAVING : 432,795 - 58,200 = 374,595. ($700k, E560k)

NOTE - While this is an example and cannot be promised by Semtech, it provides a glimpse into thenew paradigm of cost structure for mobile wireless network backhaul services.

7/29/2019 49 Semtech

36/77


Synchronization Currency in Packet Networks

10ns

100ns

1us

10us

100us

100ms1s

10s100s

1,000s10,000s

100,000s

Obs. Interval

(sec)

PRC (Cs)G.811

SSU (Rb)

G.812

SSU (Rb)

G.812

}SECs (Quartz)

G.813

SECs (Quartz)G.813}

CDR

PLL

Framer

SDHG.823

(Traffic)

G.823(Sync)

BTS/NodeB(50ppb)

BTS/NodeB(50ppb)

50ppb freq alignmentMinimum Node B FDDperformance

PRCSEC SSU

G.823 Sync

G.823 Traffic

Minimum node BTDD

+/-1.25uS adjacentnode phasealignment

E1

E3

E1

Packet Network

2.5us phase tolerance (TDD)

Requires GPS to achieve

Phase alignment

7/29/2019 49 Semtech

37/77


Can circuit emulation services solve the

Packet Backhaul synchronization problem?

What are circuit emulation services anyway?

Alternatives for Synchronization Delivery over Packet Networks

7/29/2019 49 Semtech

38/77


Circuit emulation is the process of packetizing bit streamservices like a T-1, E-1 for transmission over a packet network.

The network can be native Ethernet or Ethernet trunksbetween MPLS nodes. There are several standards for CES.

The IETF has one called PWE3. The Metro ethernet forum has onecalled CESoE. The ITU has the SAToIP or USAToIP.

They all take in a bit stream and create a packet stream and theyare point to point.

What is Circuit Emulation in Generally understood Terms

7/29/2019 49 Semtech

39/77


Timing in Circuit Emulation Services

IWF Processor

~SDiwf

To & From Metro

Ethernet Network

CESIWF

TDMline EXT

FREthline

Tx

Rx

Data

Clock

Data

Clock

External Timing

Reference

MEF 3 Figure 13 : CES IWF Sync. Ref. Model

TDM Line: line-timing, from CE

EXT: From BITS/SSU

1588

7/29/2019 49 Semtech

40/77


Timing in Circuit Emulation Services Adaptive Clocking

~

TDMservice

~

Pktr DPktr

Mpll

Fr Clk.

Rec. ~

FsFs

Data DataTDM

service

Inter-workingFunction

Inter-workingFunction

CustomerPremises

Equipment

CustomerPremises

Equipment

Clk.Rec.

Adaptive-clocking can provide line-timing.

Goal is to match the ingress service clock Fs, to the egress Fs while

maintaining G.823 MTIE

Adaptive-clocking has problems when :- packet-loss rate significant

- packet delay variation significant

- path delay changes due to network reconfiguration or restoration

Clk.Rec. ~

Fr

7/29/2019 49 Semtech

41/77


Timing in Circuit Emulation Services Network Clocking

Network clocking uses a common network clock at each end of circuit:

Network-clocking has been providing line-timing (or external timing!) to

NEs since the dawn of digital networks. Can still be used at end-points of

Packet Networks.Noise is kept to G.823 limits by existing Network Clock distribution

But network operators want to replace the TDM links with packet links:

~

~

TDMservice

~

Pktr DPktr

Mpll

Fr Clk.Rec. ~

FsFs

Network Clock

DataData

TDMservice

Inter-working

Function

Inter-working

Function

Customer

PremisesEquipment

Customer

PremisesEquipment

Will j t Ci it E l ti f G 823 T ffi d th j b?

7/29/2019 49 Semtech

42/77


Will just Circuit Emulation of G.823 Traffic do the job?

Data

RNC

PSN

CES

IWF

CES

IWF

CES

IWF

G.823 Traffic ports:

CES is point to point so an ethernet path for each link is necessary

G.823 service is frequency only so can only be used in FDD service

To keep frequency transient below 50ppb needs a 20,000-second time

constant for each millisecond of delay change (needs expensive, ultra-stable OCXO)

G.8261 says CES impractical as a timing-transport mechanism using G.823 Traffic ports

2.5us allowance

+/-50 ppb freq.

Eth t T ti S h N t k S i ?

7/29/2019 49 Semtech

43/77


Ethernet Transporting Synchronous Network Services?

RNCNode B

Real Time Service such

as node B backhaultraffic

Real Time service suchas node B traffic

Metro Ethernet Packet Network

Voice & Data

Video

Voice & Data

VideoNative Ethernet is an Asynchronous Service!

U GPS i t d t S h i th N t k?

7/29/2019 49 Semtech

44/77


GPS ReceiversGPS Receivers

Use a GPS receiver at every node to Synchronize the Network?

GPSReceiver

GPSReceiver

Node B RNC


Voice & Data

Video

Voice & Data

Video

This is a very expensive to install and operate. Typical GPS receiver withHigh precision holdover oscillator is $500.00 USD +. Antenna installation is

typically $$thousands$$ of USD and requires annual calibration.

Use a free run oscillator at every node to Synchronize the

7/29/2019 49 Semtech

45/77


y yNetwork?

Free runoscillator

Free runoscillator

Node B RNC


Voice & Data

Video

Voice & Data

Video

This will not work unless the oscillators are caesium beam type whichTypically cost $50,000.00 each.

Precision free run

oscillators

Impairments Packet Networks Present to Synchronization?

7/29/2019 49 Semtech

46/77


Impairments Packet Networks Present to Synchronization?

Most importantly traditional clock path continuity, the circuit, isbroken due to Ethernets asynchronous timing. Line recovered timing with

PRC traceability does not exist!

In layer 3 IP networks the bi-path symmetry at any point in time is

unknown and can change via route flap or RIP, BGP, or IS-IS mandatedchanges? The ITU in G.8261 has stated layer 3 networks are un-fit for

circuit emulation services or cell site backhaul.

Every packet is delayed by a differing amount of time as it passes

through packet network switch queues. The result is a self-similar orfractional Brownian distribution model of PDV or Packet Delay Variation.

There are 2 factors at play, switch queue depth and serialization delay.

Techniques for Synchronization over Packet Networks

7/29/2019 49 Semtech

47/77


Techniques for Synchronization over Packet Networks

The IETF has an RFC named Pseudo-wire emulation end to end, orPWE3. This RFC deals with encapsulation and transport of any circuit

based service over an MPLS network. Everything from the basic T/E1

circuit emulation to ATM vc/vp emulation and VLAN emulation is

included in this, except sync! The ITU-T standard G.8261 deals with delivering pseudo-wires over

packet networks targeting the MTIE traffic mask for G.823. This standard

does not cover delivering the MTIE sync mask for G.823 required for

node B and BTS operation. A standard for Synchronous Ethernet does NOT exist as yet. This

will eventually lead to pre-standard products in 2009 and the earliest.

Standards Bodies Activities

7/29/2019 49 Semtech

48/77


Standards Bodies Activities

ITU meeting this week in Geneva is modifying G.8261 for the third

time to include IEEE 1588.

The draft standard for Synchronous Ethernet is to be completed at

this meeting. Will take years for deployment.

There are 2 elements to this modification.

First is the replacement of the 100ppm crystal on the PHY.

Second is the slow protocol to carry the SSM like information for clock quality definition.

Semtech is contributing the concept of network profiles as a means of

determining traditionally synchronous services delivery over packet

networks.

Measuring each path is useless as networks changes continuously.

IEEE-1588 A Standard for Delivering Sync

7/29/2019 49 Semtech

49/77


g y

What is IEEE 1588?

Originally developed by Agilent Technologies for the Test

and Measurement community.

V1 Approved by the IEEE-SA Review Committee on

September 12, 2002

Published as IEEE 1588-2002 on November 8, 2002

Approved as IEC standard IEC 61588 on May 21, 2004

1588 V2 is Telecom friendly and WG draft standard spec is

out to ballot on June 7th.


7/29/2019 49 Semtech

50/77


g y

What does IEEE 1588 do?

Official Title is Precision

Time Protocol

IEEE 1588 is a protocoloriginally designed to

synchronize real-time

clocks in the nodes of a

distributed system that

communicate using a

network.

IEEE 1588 calls for the use

of the physical layer for

timestamp inclusion.

NETWORK

Implementing IEEE 1588 Synchronization for UTRAN over PacketN t k ?

7/29/2019 49 Semtech

51/77


Networks?

IEEE 1588Sync

Master

IEEE 1588Sync

Slave #1

IEEE 1588Sync

Slave #2

Node B RNC

Packet Network

IEEE 1588 delivers a precise copy of the Network Clock toeach end point to enable Real time service & VoATMpackets via some Circuit Emulation Services Technique

Voice & Data

Video

Voice & Data

Video

Interworkingfunction

Interworkingfunction

Network Timing Distribution using IEEE1588

7/29/2019 49 Semtech

52/77


g g

IEEE 1588 V2 (IEC61588) Precision Time Protocol

Distributes time and frequency via WAN ethernet with expected endpoint time alignment precision measured in nanoseconds.

Gets UTC to the network elements and node b without GPS

Distributes the network clock to ingress and egress points ofinterworking function with frequency precision of

7/29/2019 49 Semtech

53/77


How does IEEE 1588 work?

IEEE 1588 distributes a time base around a network using a two-

way time transfer technique. Precise Frequency, Time and

Phase can be generated.

V2standard has Master, Boundary, Transparent and Slave

clocks

IEEE 1588 uses a Master Slave Hierarchy to distribute thetime base.

Slave Clocks are simple if Boundary Clocks and transparent are

used, but there are no Boundary Clocks in a Telecom Network,

This means the Slave Clocks become complex.

IEEE-1588-V2 A Standard for Delivering Sync

7/29/2019 49 Semtech

54/77


IEEE-1588 Code

Network protocol

stack & OS

Physical layer

Sync detector& timestamp

generator

Master clock sends:

1. Announce message

IEEE 1588 V2 A Standard for Delivering Sync

Timestamp

Point

Time at which a Sync message

passed the Timestamp Point


7/29/2019 49 Semtech

55/77


IEEE-1588 Code

Network protocol

stack & OS

Physical layer


generator

Slave clock receives:

1. Announce message

g y

Timestamp Point

Time at which a Sync message

passed the Timestamp Point


7/29/2019 49 Semtech

56/77


IEEE-1588 Code

Network protocol

stack & OS

Physical layer


generator

Slave clock sends:

Delay_Req message

g y

Timestamp Point

Time at which a Delay_Req

message passed the Timestamp

Point


7/29/2019 49 Semtech

57/77


IEEE-1588 Code

Network protocol

stack & OS

Physical layer


generator

Master clock receives:

Delay_Req message

Master clock sends:

Delay_Resp message

g y

Timestamp Point

Time at which a Delay_Req

message passed the Timestamp

Point

IEEE-1588 Standard for Delivering Sync

7/29/2019 49 Semtech

58/77


Synchronization computation in the Slave clock

Offset = receipt time precise sending time one way delay (for

a Sync message)

One way delay = {master to slave delay + slave to masterdelay}/2 (assumes symmetric delay)

Master to slave delay = receipt time precise sending time (for a

Sync message)

Slave to master delay = Delay Request receipt time -precise

sending time (of a Delay Request message)

From this offset the slave corrects its local clock!

BUT: The standard says nothing about how to do this!

Real-World Experience:

7/29/2019 49 Semtech

59/77


Since April 2005, Semtech and Agilent and recently Symmetricom have been

running a field trial on a live Public Metro Ethernet Network owned and operated

by a major carrier.

Agilent and Symmetricom provided the reference and measurement equipment.

TIE data is gathered for every 24 hour period and analysed to get MTIE, TDEV

and Frequency Offset.

Semtech provided two ToPSync Evaluation boards to act as IEEE 1588

Master and Slave

The Master and Slave are test-beds for 1588v2 concepts.

The Master sends 1 Announce message every 2 seconds.

The Slave sends 24 Delay-Request messages every second.

The cost of sync in this test is 20kbps.

Arrangement of Field Trial

7/29/2019 49 Semtech

60/77


Public MPLS Packet Network

VLAN #2

VLAN #1

Atomic Clock

Agilent OmniBer 718

10MHz 10MHz

DS1

To SemtechTo Agilent

Master

ToPSync Slave #145 km from master

ToPSync Slave #220 km from master

Packet delays seen on the Metro Ethernet Network

7/29/2019 49 Semtech

61/77


Single trip delays

Round trip delay

MTIE compared to ITU G.823 PDH Sync Interface allowance

7/29/2019 49 Semtech

62/77


TDEV compared to ITU G.823 PDH Sync Interface allowance

7/29/2019 49 Semtech

63/77


MTIE compared to ITU G.811 PRC allowance

7/29/2019 49 Semtech

64/77


TDEV compared to ITU G.811 PRC allowance

7/29/2019 49 Semtech

65/77


TCXO Slave MTIE compared to ITU G.823 PDH Sync allowance

7/29/2019 49 Semtech

66/77


TCXO OCXO

Sprint Network Slave to Slave 1PPS Alignment

7/29/2019 49 Semtech

67/77


Nortel lab tests of Disciplined IEEE1588PBT with high priority CES traffic load

7/29/2019 49 Semtech

68/77


23%

57%

68%

80%

91%

97%

80% 63%

23%

0%

0%

Medium step load

1h40mins 90ns/1000s (205ns ref. T1 g.823 sync) back-to-back

11ppb > back-to-back

Nortel lab tests of Disciplined IEEE1588PBT with high priority CES traffic load

7/29/2019 49 Semtech

69/77


600ns/1000s (1s ref. T1 g.823 sync) back-to-back

11ppb > back-to-back

57%

68%

80%

91%

97%

0%

Large step loadHeld 9 hours Still under G.8261 mask

Semtech ToPSync Slave Output Frequency Accuracy over ADSL

ADSL1 Alcatel UD with ADLT N Linecard

7/29/2019 49 Semtech

70/77


Fractional Frequency Offset - Norway Test 4, TCXO

-2.00E-07

-1.50E-07

-1.00E-07

-5.00E-08

0.00E+00

5.00E-08

1.00E-07

1.50E-07

2.00E-07

0.00E+00 2.00E+03 4.00E+03 6.00E+03 8.00E+03 1.00E+04 1.20E+04 1.40E+04

Time(s)

Frac

tionalFrequencyO

ffset

ADSL1 Alcatel UD with ADLT-N Linecard.

Simulated Line length = 2500 metres.

Bandwidth = 0.224 Mb/s upstream, 3.552 Mb/s downstream.

Interleaved path

ACS9590T locked the frequency accuracy achieved never exceeds 60ppb.

Semtech ToPSync Slave Output Frequency Accuracy over ADSL

ADSL2+ Alcatel UD with ABLT F Linecard

7/29/2019 49 Semtech

71/77


Fractional Frequency Offset - Norway Test 4, TCXO

-2.00E-07

-1.50E-07

-1.00E-07

-5.00E-08

0.00E+00

5.00E-08

1.00E-07

1.50E-07

2.00E-07

0.00E+00 2.00E+03 4.00E+03 6.00E+03 8.00E+03 1.00E+04 1.20E+04 1.40E+04

Time(s)

Frac

tionalFrequencyO

ffset

ADSL2+ Alcatel UD with ABLT-F Linecard.

Simulated line length = 2500 metres.

Bandwidth = 0.182 Mb/s upstream, 4.578 Mb/s downstream

Interleaving enabled, which gives a maximum interleaving delay 16 msec.

ACS9590T locked frequency accuracy achieved never exceeds 50ppb

Packet Timing Application Areas

7/29/2019 49 Semtech

72/77


Mobile Wireless

Backhaul

Wireline Access

Aggregation

Femto Cell

WiMAX

Backhaul/Access

Residential

Ethernet

(802.1AVB)

Consumer

TopSync Value

Proposition

Network

Typical Node B System Design Architecture for IEEE 1588 Timing

7/29/2019 49 Semtech

73/77


Network Processor

NetworkProcessor

GbE/FEInterfaces

GbE/FEInterfaces

Processorrunning

ProprietaryAlgorithms

DDSProcess

Stratum 3G.812, G.813,

GR-1244Line Clocks

~TCXO (S3)OCXO (S3E)

MII ports

ToPSyncTM

SPI

Reference

ClocksTimebase

GeneratingEngine

SRAM

1588 PTP Packets

What is the Future of Network Synchronization?

7/29/2019 49 Semtech

74/77


Networks of tomorrow will use the most cost effective, manageable

and reliable technique fitting the service model.

Trunk networks will use traditional BITS based systems.

Metro networks will use a combination of traditional systems and

synchronous Gigabit ethernet.

Access networks endpoints and wireless service delivery points will

use IEEE 1558 V2 or some other packet based synchronization.

How can Synch Influence OPEX?

7/29/2019 49 Semtech

75/77


Enable UTRAN nodes to use lower cost backhaul technologies such

as ADSL and ethernet versus PDH or SDH/SONET.

Eliminate the need for expensive GPS holdover oscillators and

antenna installations for low cost WiMAX in-building and outdoorbasestations as well as for picoCDMA basestations.

Intelligent layer 2 sync end point systems can have SNMP capability

to allow the viewing of sync performance in real time remotely, no

wasted truck roles for sync problems!

Layer 2 synchronization techniques such as IEEE 1588 V2 use the

existing infrastructure and are only required at the ends.

The Future for Sync?

7/29/2019 49 Semtech

76/77


Several possibilities exist for changing the way todays synchronizationclocks and time bases are transmitted from source to sink via packetnetworks without damage to the integrity of the synchronizationinformation itself.

IEEE 1588 permits easy timing distribution of the network clock over a packetinfrastructure with a very low cost per sync stream.

Standard NTP is not accurate enough for real-time applications and was notdeveloped for this purpose.

IEEE 1588 enables native ethernet backhaul for synchronous and time-basedservices.

IEEE 1588 and circuit emulation services are NOT opposing or competingtechnologies.

Summary

7/29/2019 49 Semtech

77/77


We are migrating from a TDM network to packet network, but still needto support TDM services.

Interworking Functions have been defined and offer many clocking

schemes. The only clocking scheme which allows the Public

Switched Network to extend to the Customer Premise and meets thetiming requirements, is IEEE 1588.

Experience with a long-term IEEE 1588 field trial shows that excellent

long term viability, accuracy and stability is consistently available on

an evolving Metro ethernet.

Future services could be based on time rather than frequency. IEEE

1588 delivers time as well as frequency.

49 Semtech

Documents

Transcript of 49 Semtech