Power Matters.

ITU Standards for Frequency and Phase

and Network Deployment Guidelines

© 2013 Microsemi Corporation. COMPANY PROPRIETARY

Outline

▪ The Need for Synchronization in Carrier Ethernet networks

▪ Synchronization Technologies: BITS, SyncE, GNSS, PTP
▪ The Challenges of PTP
▪ The ITU Standards and Profiles
▪ Network deployment guidelines
▪ Field Measurements

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 2

The Need for Sync in Carrier Ethernet Networks

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 3

Mobile Technologies – Sync Requirements

Application Frequency: Phase

Transport / Air Interface

GSM / UMTS / W-CDMA 16 ppb / 50 ppb

N/A
UMTS/ W-CDMA Femtocells n/a / 250 ppb

CDMA2000 16 ppb / 50 ppb +/-3 – 10 µs

TD-SCDMA 16 ppb / 50 ppb +/- 1.5 µs

LTE (FDD) 16 ppb / 50 ppb N/A

LTE (TDD) 16 ppb / 50 ppb +/- 1.5 µs to UTC
LTE-A (eMBMS, CoMP, eICIC) 16 ppb / 50 ppb +/- 1.5 µs to UTC
5G - eCPRI 16 ppb / 50 ppb +/- 750ns to UTC
HetNet 16 ppb / 50 ppb +/- 750ns to 1.5 µs to UTC

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 4

Frequency, Phase and Time Synchronization

Frequency Synchronization

TA=1/fA Leading edge of the

A t pulses are at same
pace, but not at the
TB=1/fB identical moment.
B t
fA=fB

Phase Synchronization

TA=1/fA

A t Leading edge of
the pulses are at
TB=1/fB identical moment.
B t
fA=fB
Time Synchronization
TA=1/fA
Leading edge of
A t the pulses are at
TB=1/fB the identical
moment and
B t
fA=fB identical time.

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 5

Mobile Network Synchronization Trends
Macro, cRAN, Hetnet, IoT

▪ LTE-A & 5G NR adding functions that impose tighter sync requirements

▪ Network will have multiple sync requirements at the edge from relatively lax (IoT,
WiFi) to very stringent edge cRAN clusters

Phase/time
+/-40-100nsec CPRI cRAN
ePRTC Phase/time
PRTC
eCPRI cRAN
+/-30ns
+/-750nsec or
Phase/time
better
PRTC +/-100ns Macro Network
16pb WiFi
PRC Frequency
G.811
+/- 1.5usec Hetnet
Small Cells
PRTC
SSU Frequency +/-750nsec –
syncE
Phase/time
+/-100ns
+/-1.5usec
IoT

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 6

Edge PTP Timing Models for eCPRI

1) Co-located REC and RE -- Same as macro towers today

PTP
RE
Slaves

PRTC REC
1PPS/ToD
Backhaul (Few meters)
vRAN server
Data Center

PRTC

Backhaul REC
RE
Fronthaul
PTP
Slaves
2) Separated REC and RE PTP/Ethernet
Same as CPRI cRAN today 10s of km

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 7

CPRI vs eCPRI Main Differences

CPRI eCPRI
Radio Access Point RRH RE
Remote Radio Head Radio Equipment

Servers BBU - Baseband Unit REC: Radio Equipment Control

Network Transport Dedicated fiber Ethernet
Scalability Poor: Max 17 KM Excellent (10s of km)
dedicated BBU Virtual Servers
dedicated Fiber Ethernet SDN Model

Clock Transport from CPRI PTP

cRAN to Radio (embedded TDM sync) G.8275.1 or G.8275.2

eCPRI puts time sensitive functions into the Radio Head (RE)
▪ Timing changed from CPRI to PTP or GNSS at the Radio
Head
▪ eCPRI over Ethernet
▪ Improves scalability, bandwidth efficiency,
▪ Enables SDN and Virtualization

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 8

Synchronization Technologies for Carrier
Ethernet Networks

IEEE1588-2008 SyncE
GNSS NTP
PTP (or old SDH)

Frequency & Phase Frequency Only Frequency & Frequency

Phase

Layer 2 and 3 Physical Layer Physical Layer Layer 3

Legacy & Greenfield Greenfield LAN / WAN

LTE /Carrier Deployments Legacy & IMS, Billing /
Ethernet LTE / Carrier Greenfield Logging / SLA /
IEEE 1588, ITU-T Ethernet IPTV / Femtocell
G.8261, G.8265.1, ITU-T G.811
IETF RFC 1305,
G.8275.1, G.8275.2 ITU-T G.8261/2/4
RFC 5905-8

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 9

The Challenges of Packet Synchronization
IEEE1588-2008 PTP

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 10

Two Kinds of Network Impairments

▪ Packet Delay Variation

• A challenge for Frequency and Time/Phase synchronization

▪ Network Asymmetries
• A challenge for Time/Phase synchronization

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 11

Packet Delay Variation

Voice
FIFO Buffers Video
Voice Data
Video
Data Packet
Environment

 Main Delay Variation Causes

▪ Waiting time jitter in network elements
▪ Routers/switches congestion
▪ Extended packet loss, Network outages/re-routing: may cause holdover from
lack of information
 Note: absolute delay, even high, is not a problem for sync technologies

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 12

Typical PDV Profile

PDV Tail
Distribution

Minimum Delay
Packets

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 13

Packet Network Characterization

10 switches, 20% load 10 switches, 60% load

10 switches, 40% load 10 switches, 80% load

Key characteristics:
Packets experiencing minimum delay • variance of minimum delay
• frequency of packets with minimum delay
© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 14
 Asymmetry – 1588 assumes symmetrical delays in
the forward and reverse paths
Master Clock Time Slave Clock Time

t1 Sync message Data at • Each “event message” flow (sync,

Slave Clock
delay_req) is a packet timing signal
t2 t2
• Master frequency determined by comparison
Follow_Up message of timestamps in the event message flows
containing true value of t1 • e.g. comparison of t1 to t2 over multiple sync
messages, or t3 to t4 in delay_req messages
t1, t2 • Time offset calculation requires all four
timestamps:
t3 t1, t2, t3 • Client time offset = (t1 – t2) + (t4 – t3)
Delay_Req message
2
• assumes symmetrical delays
t4 (i.e. the forward path delay is equal to the
Delay_Resp message reverse path delay)
containing value of t4 • Time offset error = fwd. delay – rev. delay
2
t1, t2, t3, t4
time

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 15

Static Asymmetries

Path Asymmetry Fiber Asymmetry Lambda or microwave

asymmetry

•Asymmetries cannot be resolved by PTP, and can easily reach 100’s of ns or

even be >> 1µs
•They can be avoided by placing the master as close as possible to the slaves
© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 16
ITU Standards for Frequency and Time/Phase
synchronization

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 17

 Structure of ITU-T Sync Recommendations
G.8260: Definitions and Terminology for
Definitions /Terminology Synchronization in Packet Networks

Frequency Time/Phase
Basic Aspects G.8261: Timing and Synchronization Aspects in Packet G.8271: Time and Phase Synchronization Aspects in Packet
Networks (Frequency) Networks

G.8261.1: PDV Network Limits Applicable to Packet- G.8271.1: Network Requirements for Time/Phase Full
Network Based Methods (Frequency) on Path Support
Requirements G.8271.2: Network Requirements for Time/Phase
G.8261.2: Reserved for future use
Partial On Path Support

G.8262: Timing Characteristics of a Synchronous G.8272: PRTC (Primary Reference Time Clock)
Ethernet Equipment Slave Clock (EEC) G.8272.1: enhanced PRTC

G.8262.1: enhanced EEC G.8273: Packet-Based Equipment Clocks for

Time/Phase: Framework
Clocks G.8263: Timing Characteristics of Packet-Based G.8273.1: Grandmaster (T-GM)
Equipment Clocks (PEC)
G.8273.2: Boundary/Slave Clock (T-BC/T-TSC)
G.8266: Timing characteristics of packet master G.8273.3: Transparent Clock (T-TC)
clock for frequency synchronization G.8273.4: Assisted PTS Telecom Time Slave
Clock

G.8264: Distribution of Timing Information through

G.8274: Reserved for future use
Packet Networks
Methods
G.8265: Architecture and Requirements for Packet- G.8275: Architecture and Requirements for Packet-
Based Frequency Delivery Based Time and Phase Delivery

G.8265.1: Precision Time Protocol Telecom G.8275.1: PTP Telecom Profile for Time/Phase
Profile for Frequency Synchronization Synchronization, Full OPS
Profiles
G.8265.2 PTP Telecom Profile for Frequency #2
G.8275.2: PTP Telecom Profile for Time/Phase
Synchronization, Partial OPS & APTS

agreed ongoing options

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 18
G.8265.1 Telecom Profile for Frequency

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 19

G.8265.1 – Prime Objectives

▪ To permit the distribution of frequency using PTP over existing

managed, wide-area, packet-based telecoms networks
▪ To allow interoperability with existing synchronization networks
(such as SyncE and SDH)
▪ To define message rates and parameter values consistent with
frequency distribution to the required performance for telecom
applications
▪ To allow the synchronization network to be designed and
configured in a fixed arrangement
▪ To enable protection schemes to be constructed in accordance
with standard telecom network practices

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 20

Key design decisions

▪ No on-path support, (e.g. boundary and transparent clocks), because these

are not generally available in existing networks
▪ IPv4 was adopted as the network layer due to its ubiquity, rather than
operation over Ethernet or other lower-layer protocols
▪ The PTP Announce message was adapted to carry the Quality Level (QL)
indications defined in G.781, for continuity with SONET/SDH and SyncE
synchronization status messaging.
▪ Unicast transmission was adopted over multicast, since it could be
guaranteed to work over wide-area telecoms networks
▪ BMCA (Best Master Clock Algorithm) was replaced by static provisioning,
allowing the synchronization flow to be planned, rather than dynamically
adjusting itself
G.8265.1 Telecom Profile has enabled the deployment of PTP
synchronization by many operators
Microsemi has ~250 such deployments in the world
© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 21
1. Frequency Synchronization:
G.8265.1 Basic Architecture
Managed Ethernet backhaul
consistent, known performance
CORE AGGREGATION ACCESS

Rb
PTP GM

Macro eNodeB

Small Cell
Aggregation
PTP GM
Metro Small Cells

• Central deployment of high capacity PTP grandmaster

• Common deployment model today: 100s of networks worldwide

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 22

2. Frequency Synchronization:
G.8265.1 Edge Architecture
In case of uncertain performance, noisy backhaul
Multiple technologies, many hops/paths, high packet delay variation, 3rd
party access vendors, etc.

CORE AGGREGATION ACCESS

High PDV / 3rd Party

Edge Master
Clock
PTP GM
Microwave Macro eNodeB

PON Edge Master

OLT ONU Clock

Small Cell
DSL modem
Aggregation
DSLAM
PTP GM
Metro Small Cells

• Deploy PTP Grandmaster at or near the edge to eliminate backhaul

performance issues. May be supported with existing, centralized GM.
© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 23
Rate of Timing Messages

▪ The rate of timing messages required is dependent on several factors

• Amount of noise in the network
• Local oscillator stability
• Efficiency of clock servo algorithm
▪ The Telecom Profile defines the range of message rates Masters and
Slaves should support
Message rates Minimum Maximum Default
Announce 1 msg. every 16s 8 messages/s 1 msg. every 2s
Sync 1 msg. every 16s 128 messages/s Not defined
Delay_Request 1 msg. every 16s 128 messages/s Not defined

▪ It is not expected that a slave will achieve the required performance at

all message rates
• Slave must request the message rates needed to maintain performance

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 24

Packet Timing Signal Fail

▪ Profile defines three types of signal failure:

• PTSF-lossAnnounce, where the PTP Slave is no longer receiving Announce
messages from the GM
– This means there is no traceability information for that master
– Slave should switch to an alternative GM after a suitable timeout period
• PTSF-lossSync, where the PTP Slave is no longer receiving timing messages
from the GM (i.e. Sync or Delay_Response messages)
– This means there is no timing information for that master
– Slave should switch to an alternative GM after a suitable timeout period
• PTSF-unusable, where the PTP Slave is receiving timing messages from the
GM, but is unable to recover the clock frequency
– This means there is no recoverable timing information for that master
– Action is undefined

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 25

Master Selection and Protection

▪ Telecom slave clock consists of several logical protocol instances,

each communicating with a different grandmaster
▪ Selection process follows G.781 selection rules:
• Availability, Traceability, Priority

Slave Telecom
PTP GM
1
Protocol
Instance 1
Slave Clock

G.781-
Slave based
PTP GM Packet Protocol Master
2 Network Instance 2 Selection
Process

Slave
PTP GM
Protocol List of N
N Instance N
Grandmasters

Separate© PTP
2013 domains
Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 26
Additional Protection Functions

▪ Non-reversion function
• By default, a slave should switch back to the original master once the
failure condition has been rectified
• Optionally, this automatic reversion function can be disabled
▪ Wait-to-Restore Time
• Follows an initial protection switch, e.g. due to loss of traceability or signal
failure
• Time waited before switching back to the original highest priority master,
once the failure condition has been rectified
• Implies slaves must continually monitor the original master following a
protection switch

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 27

Additional Traceability Functions

▪ Forced traceability
• If the PTP GM is connected to a reference by a signal with no SSM QL value,
the input can be manually “forced” to a suitable value
▪ Output QL Hold-Off
• Applies to slave output timing signals that carry an SSM QL value (e.g. SyncE)
• Change of QL in the incoming PTP clockClass input should be delayed before
being applied to the output
• Allows time for synchronization to a new reference
• Avoids any unecessary switching in downstream equipment
▪ Output Squelch
• Output clock signal of a PTP slave should be “squelched” in case of holdover
• Prevents end equipment attempting to synchronize to a clock in holdover
• Only applies to signals that do not carry a QL value (e.g. a 2.048MHz
unframed timing signal)
© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 28
G.8272
Primary Reference Time Clock (PRTC)

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 29

G.8272 Primary Time Reference Clock
(PRTC)

FREQUENCY TIME/PHASE

• Frequency distribution ▪ Time/Phase distribution

requires a Primary Reference requires a Primary Reference
Clock (PRC) Time Clock (PRTC)
• a.k.a. - PRS ▪ This reference must come from
• The PRC is usually a Cesium GNSS
clock ▪ The PRTC output is a timestamp
▪ The PRC output is a telecom / time code
frequency such as E1/T1, ▪ This timestamp must be linked
10MHz, 5MHz to UTC.
▪ Delay from input of GNSS signal
to time output must be
<= 100nsec
* ref:
© 2013 Microsemi Corporation. COMPANY Figure 7b/G.8275
PROPRIETARY Power Matters. 30
G.8272 Primary Time Reference Clock
(PRTC)
▪ PRTC with physical layer frequency input/output and embedded
T-GM output function * * ref: Figure 7b/G.8275

<100 nsec

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 31

G.8272.1 Enhanced PRTC

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 32

Enhanced PRS/PRC/PRTC
• A standard PRC’s provide 10-11 frequency stability, i.e. 864 ns phase drift per
day, and a G.8272 PRTC provides +/- 100ns accuracy to UTC

• The ePRC/PRTC combines GNSS and 1 or 2 Cesium clocks for high stability

• GNSS + 2 Cs ensembling gives up to 5.7 10-14 stability, and 30ns accuracy to

UTC. With this frequency stability the phase will only drift 6 ns/day,
therefore the 200ns holdover budget of the G.8275.1 or G.8275.2 standard is
consumed in 25 days instead of 1 day with a standard PRTC.

• 1 or 2 ePRTC’s in a network provides very high phase resilience

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 33

G.8272.1 - ePRTC

▪ Provide 30ns alignment to UTC

▪ 2MHz with 5.7 10-14 stability

SDH network
- Ideal for Frequency
- No use for transporting PTP

Core
Sites

Will Hold 250nS with the ePRTC traceable

input for up to 50days Distribution/
Edge Sites
© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 34
TS 3550 enhanced PRTC

Stability 250nS Drift

Two Way Timing Antenna. 1 x 5071 Standard 5.70E-14 50 Days
300m Coax max

2MHz with stability 5.7 10-14

1PPS & TOD

UTC to 30 nS

10 MHz from Caesium(s)

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 35

 EPRTC Single Standard Tube Cesium B One Week Steady State

14 nS

Frequency output MTIE

Mask in use: G82721r1
Absolute Minimum Mask Clearance: 3.9952 ns, tau=1 s
Relative Minimum Mask Clearance: 6.5198 ns, tau=131072 s

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 36

EPRTC Dual High Performance Tube Cesium Ensemble
One Week Steady State

4 nS

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 37

ePRTC - Leading performance

ePRTC Cesium Configuration (results in nanoseconds)

Standard Standard
High Performance High Performance
ITU ePRTC Metric Performance Performance
Dual Single
Dual Single

TDEV (10K seconds)

0.04 0.15 0.3 0.5
1ns specification

MTIE (1000K seconds)

3.5 5 12 17
30ns specification

1) High Performance Dual = Operation with Dual High Performance 5071 Microsemi Cesium Clocks
2) High Performance Single = Operation with a Single High Performance 5071 Microsemi Cesium Clock
3) Standard Performance Dual = Operation with Dual Standard Performance 5071 Microsemi Cesium Clocks
4) Standard Performance Single = Operation with a Single Standard Performance 5071 Microsemi Cesium Clock

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 38

G.8275.1 – Telecom Profile for Time/Phase #1

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 39

G.8275.1
PTP Elements for Full On Path Support
▪ PTP options
• Node types : Ordinary Clocks and Boundary Clocks. Note than Transparent
Clocks are not permitted.
• Both one-step and two-step clocks are permitted.
• Transport mechanism is IEEE 802.3/Ethernet.
– Both non-forwardable multicast and forwardable multicast (well known
addresses) must be supported
• PTP messages are sent multicast, unicast mode is not permitted
• The Alternate BMCA implementation is required
• The delay request/delay response mechanism is used

▪ No other PTP features are currently permitted

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 40

G.8275.1 Time Profile: Full On Path Support

Packet Network with T-BC

Time Slave
GNSS Clock

T-BC T-BC T-BC T-BC

PRTC GM T-SC

PRC
End Equipment

Attributes
• PRTC
• PTP Multicast over Layer 2 network
• “Time Boundary Clock” (T-BC) on every network element
• Physical layer frequency support from syncE (needs a PRC)
• Hops: 10 to 20 hops depending on precision of BC
• Attempts to provide “hop” count engineering guideline for time/phase
• Issue: does not take care of network asymmetries; needs calibration

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 41

 G.8275.1 Boundary Clock Time Error Budget
dynamic time error

±100 ns ±200 ns
PRTC
10 “Class A” T-BC
network
±550ns ±250 ns asymmetry
time error
budget
±380 ns depends on
±420ns BC type
20 “Class B” T-BC

PRTC/GM Network Equipments eNB (400ns)

Total Time Error Budget +/- 1.5usec

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 42

T-BC Classes

Absolute Time Error is comprised of Constant and Dynamic TE

Constant
TE

Dynamic
TE

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 43

T-TSC Accuracy Requirements

Absolute Time Error is comprised of Constant and Dynamic TE

Constant
TE

Dynamic
TE

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 44

G.8275.2 – Telecom Profile for Time/Phase #2

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 45

G.8275.2 Time Profile Partial On Path Support
G.8265.1 Unicast Segment 1 G.8265.1 Unicast Segment 2

Time Slave
GNSS Clock

T-BC
PRTC GM T-SC

PRC
End Equipment

Attributes
• PRTC
• Optional “Time Boundary Clock” (T-BC) on network elements
• PTP over Layer 3 Unicast mode “Segments” are added together to make an
end-to-end sync chain
• Optional Physical layer frequency support
• Hop count engineering: still Work in Progress in ITU
• Asymmetry issues remain
© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 46
Advantages- Disadvantages
With G.8275.2
▪ Advantage
• Operators with existing L3/L2.5 networks can deploy a standards based PTP
time profile without having to change out every network element
• MPLS is used in 75% of mobile operators
• It uses G.8265.1 which is already deployed in many networks

▪ Disadvantage
• Engineering time transfer over NE without embedded T-BC will almost
certainly go less hops than engineering time transfer over a T-BC enabled
network

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 47

Synthesizing GNSS & PTP Using Assisted Partial Timing
Support (APTS)

Core/Metro Segment Access segment

G.8265.1 Unicast Segment G.8265.1 L3/Unicast or G.8275.1 L2/Multicast
Asymmetry compensation Asymmetry limited area, 3 hops typ
GNSS PDV limited area, 10 hops max
GNSS
APTS Edge Node

T-SC
TP5000 Core
Packet Network
GM
Time Slave
TP2700 with APTS and single Clock
router connection End Equipment

• APTS PRTC maybe placed at any point on the sync chain

• It is expected that most deployment will be at points near to the eNB
where the time error budget can be most easily controlled

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 48

 APTS Operation –
Multiple path support in the Core-Aggregation segment

ASYMMETRY CALIBRATION Access segment

Various/changing paths between the Core (Controlled Asymmetry)

and the Edge PRTC’s

Core PRTC/GM APTS PRTC

TP5000 or TP2700 TP2700
GM T-SC
EM

PTP FLOWS
T-SC

▪ APTS was invented by Microsemi in 2012 and has been

implemented in TP2700 since 2013 End Equipment
(Base stations)
▪ It has been adopted by ITU and is now part of the recently
consented G.8275.2 Profile
© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 49
 Automatic Path Asymmetry Compensation –
(a TimeProvider®2700 unique feature)

• Automatic Path Asymmetry Customer network test environment

Compensation algorithm
supplies external compensation
factor as permitted in IEEE 1588
standard Path Re-arrangement (Ring Topology)

• Algorithm learns path

asymmetries to the north-bound
master … even while system
may be using GNSS as the
primary clock source
• In the event of a GNSS failure,
the system will operate revert to
using asymmetry corrected PTP
BLUE: PPS RED: PPS
• Available with the TimeProvider performance performance
2700 PTP Input SW License without asymmetry with asymmetry
compensation. compensation.

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 50

 The Time Error Budget with APTS and Asymmetry
Compensation
Transport Elements Base Station 26% of budget
dynamic time error

±100 ns ±200 ns
holdover
±150 ns
PRTC ±250 ns

End
Application
±800 ns Constant time error

network asymmetry is
calibrated out of the
system using GPS

Network Equipment End systems

Total Time Error Budget +/- 1.5 µsec

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 51

PTP with Assisted Partial Timing Support

▪ Advantages ▪ Possible Challenges

• Uses “co-operation” between • Use of existing network can
GPS and PTP mean that PTP traffic is non-
• GPS is used to calibrate out PTP deterministic to the point
asymmetry, and to measure its where it is too unstable to
performance use for time transfer
• PTP monitors GPS for antenna • Will have more constraints
failure, spoofing, jamming on the dimension of the
• Operates over existing networks packet network than
G.8271.1 (BC everywhere)
• Operates over third party access
because of the probable
networks
increase in PDV from the
• To be consented in 2015 transport network
© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 52
G.8275.1 Compared To G.8275.2

G.8275.1 G.8275.2
1. Needs PRTC 1. Needs PRTC
2. No APTS 2. Can use APTS
3. Centralized architecture 3. Distributed architecture
4. Boundary Clocks on every 4. Optional Boundary Clocks
network element (may be stand alone)
5. SyncE end to end 5. Optional Frequency support
6. Cannot be deployed on 6. Can be deployed on existing
existing MPLS networks MPLS networks
7. PTP multicast 7. PTP Unicast and multicast
8. No migration from G.8265.1 8. Easy migration from G.8265.1

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 53

Network deployment guidelines

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 54

Engineering Frequency in your network

▪ G.8265.1 recommends maximum 10 hops between GM and Slaves

• 10 hops are usually fine on Fiber networks with 1G, 10G,… capacity
• Networks should not be fully loaded, 80% average load is a maximum
• Use highest CoS available

▪ Microwave links are usually a more difficult medium

• Very different performance with TDM, Hybrid, full IP microwaves
– Hybrid are the worse, as Ethernet is often best effort
• Performance also depends on vendor
• Adaptive modulation & coding has adverse effects
• Use highest CoS
• Strict Priority is good if selectively used
• 5 to 7 hops more realistic, sometimes 3 to 5
• Highly recommended to measure PDV and assess compatibility with the
requirements of the slave in the basestations

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 55

Engineering Time in your network
▪ G.8275.2 with APTS is the most reliable and practical architecture
• Use G.8265.1 in the Core/Metro (L3 unicast)
• Use G.8265.1 or G.8275.1 in the access (L2 multicast)

▪ Measure asymmetries and PDV in the access leg

• Measure from the slave back into the network
• Place edge masters when asymmetry reaches 700-800ns maximum

▪ Combine technologies for resiliency

▪ Monitor and trend sync performance

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 56

Ensuring Resiliency: combine technologies
▪ PRTC, Edge GrandMaster locked to GNSS
• PRTC can also be referred to a PRC phase stability

▪ Use PTP as a backup in case of GNSS unavailability

• Outages, jamming GNSS

▪ Combine PTP and SyncE whenever possible

Time
• SyncE provides stable frequency and helps phase PTP
Rb

Phase Osc
accuracy
• Extends phase holdover in case of GNSS and PTP
unavailability
SyncE

▪ Use Rubidium for phase holdover

• Rb oscillator maintains 1.5us for 24 hours
• A good OCXO gives… 1 hour!

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 57

Monitor the Network

▪ Use single EMS to supervise the sync network from end to end
• Core GM, PRTC
• Edge PRTC, Edge Master
• Embedded or Standalone Boundary Clocks
• Embedded Sync clients

▪ Collect performance data/metrics from embedded clients in EMS

• PTP flow availability, stability
• Network jitter
• Frequency accuracy, phase accuracy estimators

▪ Analyze and Trend performance indicators

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 58

Summary Guidelines for Time/Phase

Use an independent source for sync

• Efficient operation in a multi-vendor network
• Use single EMS to monitor Sync End to End

Prepare for Time/Phase even if frequency is the initial concern

• Phase requirements will come with LTE-A

Combine Technologies whenever possible

• GNSS, PTP & SyncE or BITS Stable frequency source
• High performance holdover

Engineer the network from the Edge

• Noisy, high PDV network: requires deployment “at or near the edge”
• Backhaul supports accuracy: allows deployment closer to the core

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 59

Packet Timing Measurements & Metrics

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 60

“TIE” vs. “PDV”

▪ “TIE” vs “PDV”
• Traditional TDM synchronization measurements: signal edges are timestamped
producing a sequence of samples
• Packet timing measurements: packet departure/arrival times are sampled and packet
delay sequences are formed

▪ Phase measurements (TIE) can be made using:

• Frequency/time interval counters
• Time interval analyzers
• Dedicated testsets
• BITS/SSU clocks with built-in measurement capability
• GPS receivers with built-in measurement capability

▪ Packet phase measurements (PDV) can be made using:

• IEEE 1588 grandmaster/probes
• NTP servers/probes
• Specialized network probes

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 61

PDV/Asymmetry measurements
▪ PTP Tests
• To measure the network characteristics (not the client)
• PDV and Asymmetry Tests
• TP5000 in probe mode or SyncWatch collects Timestamps from Sync and
Delay_Req/Delay_Resp messages, and stores them in a PC
• TimeMonitor analyses the timestamps and derives PTP metrics for PDV
(FPP, MinTDEV, MAFE, etc) and histograms for Asymmetry measurements
• TIE Tests
• End to end test including the embedded client if accessible, or a reference client
(like the TP2700 client)
• If SyncWatch measures TIE on E1/T1 signals from Base station or client like
TP2700, and on 1PPS for phase
• TimeMonitor computes metrics from TIE data (MTIE, TDEV)

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 62

PDV/Asymmetry and TIE measurements – Set-up
TimeMonitor

PTP Timestamps TIE data

GPS

PDV GPS TIE

& SyncWatch Probe

Asym
E1 1PPS
TP500
GPS TP5K in Probe mode
or TP2700
or SyncWatch

TimeProvider 5000
or TP2700 PTP
E1 1PPS
GE GE GE
Switch/ Switch/ Switch/
Router N hops Router Router PTP
Base
Station
Core or Edge
PRTC/GM Network under Test

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 63

Lab Network Configurations

Example Lab Network Configurations

GPS

GPS 1588 Packet

Timestamper

Various Network
1588 Grandmaster Configurations
1588 Slave

Network Configurations:
Crossover
Hub
Enterprise Switch A
Enterprise Switch B Traffic Generator
Enterprise Switch C
Enterprise Switch D
Multilayer Switch E
Wirespeed router M
Router X
Router Y
Multi-Hop: SwA/RtX/SwB
Multi-Hop: SwA/RtM/SwB
Multi-Hop: SwA/RtX/RtX/SwB
Multi-Hop: SwA/RtX/RtY/SwB
© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 64
Interpretation of Measurement Results (“TIE”)

▪ For traditional synchronization measurements, the measurement

analysis used primarily is:
• Phase (TIE)
• Frequency (fractional frequency offset)
• Frequency accuracy
•
•
MTIE
TDEV } All are derived
from phase

▪ MTIE and TDEV analysis shows comparison to ANSI,

Telcordia/Bellcore, ETSI, & ITU-T requirements

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 65

Interpretation of Measurement Results (“PDV”)

▪ For packet synchronization measurements, some of the

measurement analysis used is:
• Phase (PDV)
• Histogram/PDF* & Statistics
•
•
•
•
Running Statistics
MATIE/MAFE
TDEV/minTDEV/bandTDEV
Floor delay Packet Percentage (FPP)
} Derived from PDV
phase

* PDF = probability density function Power Matters. 66

© 2013 Microsemi Corporation. COMPANY PROPRIETARY
Viewing Phase (Packet Delay Sequence)

When graphing packet delay phase it is often best not to connect the dots

Measurement
points
connected

Measurement
points as
discrete dots

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 67

Performance Metrics

•Phase (Packet Delay vs. Time) MTIE

– Basis for all calculations
•MTIE (Maximum Time Interval Error)
– Typically one dimensional for packet
delay data
•TDEV (Time Deviation)
– Useful indicator of network traffic load

Phase
TDEV

Crossover Hub Switch Router

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 68

MATIE/MAFE Packet Metrics

n  k 1
1
MATIE MATIE n 0   max
1 k  N  2 n 1 n
 x
ik
in  xi  , n = 1, 2, ..., integer part (N/2)

MATIE n 0 
MAFE MAFE n 0  
n 0

MATIE: denotes the maximum rate of change of the PDV

MAFE: denotes the maximum frequency error

Reference: Maximum Average Time Interval Error, WD 60, Nokia-Siemens Networks,

ITU-T Q13/15, Rome, Sep. 2008.

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 69

Performance Metrics

Statistics
•Standard Deviation (PDV)
•Mean (Latency) Crossover cable: Switch no traffic:
•Maximum Peak Deviation (PDV) Mean: 287.2818 nsec Mean: 16.75112 msec
Peak to Peak: 10.01 nsec Peak to Peak: 310.0 nsec
Standard Deviation: 4.450 nsec Standard Deviation: 70.10 nsec

Hub: 10% BW Utilization:

Mean: 659.7955 nsec Mean: 17.93500 msec
Peak to Peak: 60.01 nsec Peak to Peak: 121.4 msec
Standard Deviation: 12.13 nsec Standard Deviation: 11.53 msec

Switch: 25% BW Utilization:

Mean: 16.75112 msec Mean: 19.62525 msec
Peak to Peak: 310.0 nsec Peak to Peak: 122.6 msec
Standard Deviation: 70.10 nsec Standard Deviation: 17.61 msec

Router: 50% BW Utilization:

Mean: 277.6874 μsec Mean: 47.99551 msec
Peak to Peak: 212.5 μsec Peak to Peak: 122.8 msec
Standard Deviation: 20.64 μsec Standard Deviation: 50.90 msec

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 70

Performance As a Function of Load

Statistics
Switch (25%) Switch (50%) No traffic:
Mean: 16.75112 msec
Peak to Peak: 310.0 nsec
Standard Deviation: 70.10 nsec

10% BW Utilization:
Mean: 17.93500 msec
Peak to Peak: 121.4 msec
Standard Deviation: 11.53 msec

25% BW Utilization:
Mean: 19.62525 msec
Peak to Peak: 122.6 msec
Standard Deviation: 17.61 msec

50% BW Utilization:
Mean: 47.99551 msec
Peak to Peak: 122.8 msec
Standard Deviation: 50.90 msec

Crossover Hub Switch (0%) Switch (10%)

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 71

Not All Devices are Equal
Switch vs. Multilayer switch vs. Router vs. Two Routers
Switch:
Mean: 24.41855 msec
Peak to Peak: 334.8 nsec
Standard Deviation: 52.85 nsec

Multilayer Switch:
Mean: 27.02728 msec
Peak to Peak: 576.0 nsec
Standard Deviation: 76.19 nsec

Router:
Mean: 277.6874 msec
Peak to Peak: 212.5 msec
Standard Deviation: 20.64 msec

Two Routers:
Mean: 477.6874 µsec
Peak to Peak 369.7 µsec
Standard Deviation: 43.18 msec

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 72

Multilayer Switch with Traffic
No traffic:
Mean: 26.9586 μsec
Peak to Peak: 620.4 nsec
5% Standard Deviation: 73.20 nsec

5% BW Utilization:
Mean: 26.9462 μsec
Peak to Peak: 1.209 μsec
Standard Deviation: 79.12 nsec

10% BW Utilization:
Mean: 28.9450 μsec
Peak to Peak: 34.77 μsec
10% Standard Deviation: 7.008 μsec

20% BW Utilization:
Mean: 31.2810 μsec
Peak to Peak: 40.41 μsec
Standard Deviation: 9.426 μsec

30% BW Utilization:
Mean: 33.6201 μsec
Peak to Peak: 41.70 μsec
20% Standard Deviation: 10.88 μsec

50% BW Utilization (2 Traffic

Sources):
Mean: 80.8216 μsec
Peak to Peak: 206.6 μsec
Standard Deviation: 47.06 μsec

Zooms
50%

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 73

Multilayer Switch with Traffic

20% 30% 50%

TDEV

No load 5% 10%

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 74

TDEV with Selection Algorithm: minTDEV

Lower levels of noise with the application of a MINIMUM selection algorithm

TDEV at various traffic levels on a switch (0% to 50%) converge

50%

35%

10%

No load 5%
© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 75
Loaded Multilayer Switch:
TDEV and minTDEV
Mean: 48.3 µsec / Peak to Peak: 50.9 µsec / Standard Deviation: 9.43 µsec

TDEV

minTDEV

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 76

Effects of Different Approaches to Traffic Generation on
PDV

Many aspects of traffic generation, from the choice of equipment to the way the equipment
is configured, can have a great impact on packet delay variation and by extension on the
performance of devices timing from the packet flows, such as IEEE 1588 slaves.

Network Setup
GPS

Analysis IEEE 1588

SW Master/Probe

Traffic Generator

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 77

Effects of Different Approaches to Traffic Generation on
PDV
Total BW BW Port 1/2 Frame Size Burst Length Interburst Gap
1 40% 20% / 20% Different 200000 Same
2 40% 20% / 20% Same 200000 Same
3 40% 20% / 20% Same 50000 Same
4 40% 20% / 20% Same 50000 Different
5 40% 21% / 19% Same 50000 Different

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 78

Floor Packet Percentage

▪ Pertains to Hypothetical Reference Model 1 (HRM-1)

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 79

Floor Packet Percentage (FPP) Metric
▪ The Packet Delay Variation network limit at the point C of figure
3/G.8261.1 for the HRM-1 shown in figure 1/G.8261.1 is defined
as follows:
• With window interval W = 200s and fixed cluster range d = 150µs starting at
the floor delay, the network transfer characteristic quantifying the proportion
of delivered packets that meet the delay criterion should satisfy
FPP Criterion: FPP (n, W, d) ≥ 1%

FPP window

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 80

Asymmetry measurements

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 81

Asymmetry: Layer 2: 1 switch hop (no TC, no BC)

Asymmetry: 200ns.
PDV jitter – peak to peak 24 usec
Time offset at the client.- around 100ns.

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 82

Asymmetry : Layer 3 – 1 router and 2 switch hops

▪ Router introduces 8µs asymmetry

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 83

Asymmetry : 1 router and 3 switch hops

▪ About 10 µs asymmetry

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 84

Asymmetry : Worst case L3 result found: 1 router and 3
switch hops

▪ 1 router and 3 microwave hops

▪ The router introduces a huge asymmetry: 200 µs
© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 85
Conclusions on Asymmetries

▪ Layer 2 much preferable over Layer 3 as routers tend to introduce

high dynamic asymmetries
• In Layer 3 deployments, use VLAN switching rather than IP routing;
• 200-300ns dynamic asymmetry per hop with VLAN switching
▪ Layer 2 asymmetries measured mainly on microwave hops show
typically 200ns asymmetry per hop
▪ Asymmetries depend on equipment vendors
• Some vendors (microwave mainly) include transparent clocks in their
equipments, which help reduce dynamic asymmetries

© 2013 Microsemi Corporation. COMPANY PROPRIETARY Power Matters. 86

 Power Matters.

Thank You!

© 2013 Microsemi Corporation. COMPANY PROPRIETARY