16
Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

Considerations for building accurate

PTP networks for now and the future

Thomas Joergensen

ITSF 2013

November 2013

Page 2: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

Why Timing Is So Important

2

Poor Phase Synchronization has severe impact on TD-LTE and LTE-A

Page 3: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

IEEE1588 Network Timing Is Required For LTE

Indoor Picocells

Picocells or Enterprise Femtocells for LTE indoor

coverage & capacity

Can’t get timing to Femtocells using GPS

3

Outdoor Small Cells

Small Cells to deliver LTE capacity

Often no line of sight to GPS satellites

More vulnerable to attacks at street level

Page 4: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

The Biggest Misconceptions and Problems

• “So it’s ok if my equipment is only 100ns accurate…”

• Base Stations, Routers, Gateways, Microwave Equipment – all say that

Timing Error Budget is ±1μs

• This won’t be true in real networks

• Base Stations, Routers, Gateways can be much higher accuracy

• MW and MMW Backhaul will need more slack

Every Hop has the same TE Budget

• Right now, no operator can figure out if they will meet the timing budget before the network is installed and can be tested!

And the Biggest Problem?

4

Page 5: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

PTP Equipment Performance

The PTP performance of

equipment can be divided

into 2 parts

Constant Time Error (cTE)

– Caused by inaccuracy in

the timestamp points

Dynamic Time Error (dTE)

– Caused by Timestamp

resolution, Local oscillator

noise, Other sources

– Can to some degree be

filtered off by the next

node

5

+30

-30

0

0 8 20 24 16 4

Time (Minutes)

Tim

e E

rro

r (n

s)

12 26

+18

-18

0

0 20 40 50 30 10

Time (Minutes)

Tim

e E

rro

r (n

s)

Page 6: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

ITU-T G.8273.2 T-BC Performance Classification

ITU-T G.8273.2 will specify two T-BC equipment classes

Class A: Maximum Constant Time Error of 50 ns

Class B: Maximum Constant Time Error of 20 ns

More accurate classes can be added in the future if needed

Constant time error accumulates linearly (worst-case)

Unknown fiber asymmetries also cause constant time errors

Happens if two fibers in an RX/TX pair has different lengths (~2.5 ns cTE per

meter skew)

Can be manually measured and compensated for at installation

6

Page 7: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

Other part of the Time Error Budget…

The GPS/GNSS/Grandmaster has been allocated 100 ns

Dynamic time error (caused by timestamp resolution, noise from the local oscillators/SyncE) will also exist but most of it can be filtered out by the end application. ITU-T is working on equipment specification limits for the dynamic time

error as well

Some part of the timing budget might be allocated for short-term holdover, depending on the application

The end application (base station) also needs a piece of the budget <150 ns for LTE-TDD

<50 ns for LTE-A

7

Page 8: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

LTE-A MIMO/CoMP Budget Example

Budget Component Budget allocation

PRTC (incl. T-GM) 100 ns

Equipment cTE 250 ns

Link asymmetry 50 ns

Holdover 0 ns

Error caused by dynamic time

errors

50 ns

End Application 50 ns

Total 500 ns

8

BC/TC performance only has 60% of the 500 ns budget

Page 9: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

• Equipment vendors can continue selling older

equipment but must specify cTE and dTE

performance

• Operators can mix and match older and new

equipment, and can foresee what time accuracy

they will achieve

So what about these Equipment Timing Classes

Most legacy PTP Switches and Routers today only achieve max cTE of 100ns

MW/MMW links with 1588 support can achieve cTE well below 100ns

10-20ns accuracy can easily be achieved with PHY-based timestamping for both Boundary and Transparent Clocks

9

Page 10: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

How That Would Look Visually…

10

PRTC

500ns LTE-A MIMO Time Error Budget

A) 6 hops Class A (50ns) – Link budget of 10 ns (350 ns cTE)

B) 14 hops Class B (20ns) – Link budget 5 ns (345 ns cTE)

C) 3 hops Class A & 7 hops Class B – Link budget 5/10 ns (345 ns cTE)

Class A L

B

L Class A

L B L B B B L L L B L B B B L L L B L B B B L L L B

L Class A Class A B L B B B L L L B L B B L

Class A L Class A L Class A L Class A L Class A

PRTC

PRTC L

LTE

LTE

LTE

Page 11: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

Network Diagram with cTE

11

Cell Tower B

LTE-A

Accuracy Class 6

Cell Tower A

LTE-A

Accuracy Class 6

Cell Tower C

LTE-A

Accuracy Class 6

Microwave A

Distributed TC/BC

Class A

250ns

Microwave B

Distributed TC/BC

Class A

300ns

T-BC5

Class B

325ns

T-BC5

Class A

355ns

Cell Tower D

LTE-TDD

Accuracy Class 4

Cell Tower E

LTE-TDD

Accuracy Class 4

Cell Tower F

LTE-TDD

Accuracy Class 4

CPE C

Class A

305nsCPE B

Class A

275ns

CPE A

Class 0 (100ns)

355ns

PRTC

100ns

Ethernet with IEEE1588v2

(5 ns per link)

T-BC4

Class 0 (100ns)

300nsT-BC2

Class B

170ns

T-BC1

Class B

145ns

T-GM

Class B

120nsT-BC3

Class A

195ns

T-BC6

Class A

250nsCell Tower I

LTE-TDD

Accuracy Class 4

Cell Tower G

LTE-TDD

Accuracy Class 4

Cell Tower H

LTE-TDD

Accuracy Class 4

T-TSC A

Class B

220ns

T-TSC B

Class B

220ns

T-TSC C

Class B

220s

T-TSC D

Class A

410ns

T-TSC E

Class B

380nsT-TSC F

Class B

380ns

T-TSC G

Class A

380ns

T-TSC H

Class A

380ns

T-TSC I

Class A

380ns

Page 12: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

< 20 ns – Is that really feasible? – YES IT IS!

Vitesse CE Switch reference system

cTE for a single T-BC: < 5 ns

Unfiltered dTE: < 10 ns

Filtered dTE: < 3 ns (0.1Hz low pass filter)

System Level Test results

All links are 10GE (No cable skew)

GM-> 9 x T-BC -> T-TSC (slave)

– Maximum Time Error at slave: -21 ns

– Average Time Error at slave: 2 ns

GM-> 9 x TC -> T-TSC (slave)

– Maximum Time Error at slave: -20 ns

– Average Time Error at slave: -1 ns

12

Page 13: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

BC and TC Systems Testing Using Vitesse VeriTime™ 1588 Technology

13

VeriTime™ Switch/PHY-based Carrier Ethernet Reference Design

Configuration Details PTP Network Equipment:

VSC5621EV Carrier Ethernet

Switch Reference Design

All Interfaces are 10GE SFP+

1-step PTP using Ethernet

Encapsulation

TCs running E2E without

Syntonization

Sync/Delay_req frame rate: 16

FPS

Page 14: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

Test Results – Time Error at slave

14

0 10 20 30 40 50 60 70 80

Minutes

20 ns

0

-20 ns

0 10 20 30 40 50 60 70 80

Minutes

20 ns

0

-20 ns

GM-> 9xBC->SLV:

GM-> 9xTC->SLV:

Page 15: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

Summary and Conclusions

15

TD-LTE & LTE-A require tight control of timing errors per link to

meet timing requirements

New base stations and fiber or copper connected switches and

routers can easily meet 20ns time error per hop

Older 1588-aware equipment and MW link time errors can meet

100ns time error per hop

Operators need an easy way to know if their (heterogeneous)

networks will meet timing (before finishing the build-out!)

With 1588 equipment classes, network timing will be possible even

with diverse equipment and transport technologies

Planning for tight timing requirements are future proofing the

network for advanced mobile technologies

Page 16: Considerations for building accurate PTP networks for now ... · Considerations for building accurate PTP networks for now and the future Thomas Joergensen ITSF 2013 November 2013

Thank You