comparison of experimental data with theoretical prediction of edc...

Comparison of Experimental Data Comparison of Experimental Data with with

Theoretical Prediction Theoretical Prediction of of

EDC Performance EDC Performance Martin LobelMartin Lobel

Jesper HanbergJesper HanbergBenny ChristensenBenny Christensen

Standards and Advanced TechnologyStandards and Advanced Technology

Optical Networking Components Division (OND)Optical Networking Components Division (OND)Intel Communications Infrastructure GroupIntel Communications Infrastructure Group

Intel CorporationIntel Corporation

May 2004May 2004IEEE 802.3 meeting IEEE 802.3 meeting –– Long BeachLong Beach

[email protected]@intel.com

Objective of presentationObjective of presentation

Establish correlation between measured and Establish correlation between measured and simulated EDC performance of individual fiberssimulated EDC performance of individual fibers

–– Allows the extension from the specific case to the Allows the extension from the specific case to the general casegeneral case

OutlineOutlineExperimental results of EDC performanceExperimental results of EDC performance–– 2 different fibers (Corning 2 different fibers (Corning ® ® and and SiecorSiecor ®®))–– Establishing the characteristics of the fibers Establishing the characteristics of the fibers

Simulation results based on impulse responsesSimulation results based on impulse responses–– Estimation of path penalty for correlationEstimation of path penalty for correlation–– Description of model Description of model

Other brands and marks are the property of their respective owOther brands and marks are the property of their respective owners.ners.

MMF link experimentMMF link experimentEx. Ratio: 5.8 dB

Rise/fall time: 41ps/50ps 400m/200m 62.5um/125 um MMF Optical

Att.EML

IEEE Offset Patch cord

1310 nm

10Gb/s

2^31EDC chip

CDRTIA PIN

1:1 CDR

BER

testerLinear Rx:

Sensitivity < -15 dBm

Rise/fall time: 39ps/47ps

Scope

Trace 2^7

CorningCorning®® 400m MMF fiber400m MMF fiber

Corning® 62.5 MMFCorning® 62.5 MMF–– Backup slide for spec detailsBackup slide for spec details

Overfill launch bandwidth*Overfill launch bandwidth*–– 751 [MHz*Km] @ 1300 nm751 [MHz*Km] @ 1300 nm

Measured Bandwidth (1310nm):Measured Bandwidth (1310nm):–– Center Launch: >5000 MHz*KmCenter Launch: >5000 MHz*Km–– Offset Launch: 1600 MHz*KmOffset Launch: 1600 MHz*Km

Bandwidths and impulse response Bandwidths and impulse response are extracted from tracesare extracted from traces

–– 2^7 PRBS (~128 bit periods)2^7 PRBS (~128 bit periods)–– See backup slides for detailsSee backup slides for details

Impulse response for offset launch with 400 meter fiber

(2 measurements)*According to data provided by fiber manufacture

EyeEye--diagrams after 400 m MMF diagrams after 400 m MMF

Center launch: >5000 MHz*km Offset launch: 1600 MHz km

Sensitivity measurements Sensitivity measurements –– 400 m MMF400 m MMFCorning 62.5 um MMF 400 m

1,E-121,E-111,E-101,E-091,E-081,E-07

1,E-06

1,E-05

1,E-04

1,E-03

-22 -21 -20 -19 -18 -17 -16 -15 -14 -13 -12 -11 -10 -9 -8 -7 -6Received optical avg. power [dBm]

BER

400 m, CL+1:1 CDR

400m OL+1:1 CDR

0 m, 1:1 CDR (BtB)

0 m, EDC (BtB)

400 m, OL+EDC

400 m CL+EDC

Extrapolation

CL= center launch OL= Offset Launch BtB= Back-to-back

Reference

OL+EDC

OLCLCL+EDC

Error floor

2.8 5.1

Relative PenaltyRelative Penalty

+2.8 dB+2.8 dB--1.1 dB1.1 dB--1.0 dB1.0 dBWith EDCWith EDC

+7.9 dB*+7.9 dB*+1.4 dB+1.4 dB0 dB0 dBStateState--ofof--artart

1:1 CDR1:1 CDR

Offset Offset Launch Launch

400 meter400 meter

Center Center LaunchLaunch

400 meter400 meterBackBack--toto--BackBackPenalty at Penalty at

BER=10^BER=10^--99

*Extrapolated value

Correlation between 3dB bandwidth and Penalty: Weiner_1_0504

Offset launch on 400 m of 1600 MHz*km 4 GHz Reciprocal 3 dB = 0.25 ns

Conventional Receiver: ~ 5-8** dB penalty <~6 dB>**

With EDC: ~1.8-3.5** dB penalty <~2.2dB>****Values are taken from the graph shown in Weiner_1_0504

ObservationsObservations (w/Corning fiber)(w/Corning fiber)

BackBack--toto--back sensitivity with EDC is better than back sensitivity with EDC is better than conventional receiverconventional receiver

–– Compensation for Compensation for TxTx/Rx bandwidth limitations/Rx bandwidth limitations–– In line with theoretical predictions (not shown)In line with theoretical predictions (not shown)

Sensitivity curves can be reproducedSensitivity curves can be reproduced–– Leave setup with minimum disturbance during testLeave setup with minimum disturbance during test–– Not possible in case of fiber stress/long term changesNot possible in case of fiber stress/long term changes

Different offset patch cords can give very Different offset patch cords can give very different resultsdifferent results

InfineonInfineon* measured dataset* measured datasetOne fiber typeOne fiber type

–– SiecorSiecor ®® 62.5 62.5 µµmm–– Approx. 500 MHz km bandwidth (Overfill launch)Approx. 500 MHz km bandwidth (Overfill launch)

Fibers with lengths from 50 m Fibers with lengths from 50 m -- 550 m 550 m –– Taken from same fiber spoolTaken from same fiber spool

Two types of test setup (impulse response)Two types of test setup (impulse response)–– Directly modulated DFB laserDirectly modulated DFB laser–– External modulated laser (EML)External modulated laser (EML)

Pulse pattern 0000000100000000Pulse pattern 0000000100000000Dataset includes calibration pulse measurementsDataset includes calibration pulse measurementsFor details see hanberg_1_0304 (March meeting)For details see hanberg_1_0304 (March meeting)

*[email protected]

EDC simulation pathEDC simulation pathInput data•Trace (≈ 128 bits)•Impulse response

Extract impulse response

Convolute on random data

sequence

Calculate optimal tap coefficients

using mmse

Apply EDC filter to data sequence Do BER estimate save

BER vs. SNR

add noise

Do this for a range of noise levelsFilter is idealFilter is ideal

–– Textbook FFE/DFE configurationTextbook FFE/DFE configuration

–– no bandwidth limitationsno bandwidth limitations–– no noise contribution from filterno noise contribution from filter

Jitter not considered in BER estimatesJitter not considered in BER estimatesOnly receiver noise is considered (5 GHz BW). Noise added after Only receiver noise is considered (5 GHz BW). Noise added after PINPIN--TIATIA

Siecor MMF fiber

-2

-1

0

1

2

3

4

5

6

7

0 50 100 150 200 250

link length (m)

Pena

lty (d

Bo)

(SN

R @

10-9

BER

)

DML without EDCDML with EDCEML without EDCEML with EDC

Fiber length [meters]

Penalty [dBo] @

BER

10^-9 Penalties are calculated based on impulse responses obtained for different fiber lengths (50… 300m)

DML

EML

2.5-5.2 dB penalty is expected at 200 m with EDC

Simulations/Penalty with EDC include finite number of taps equal to EDC chip

200m 200m SiecorSiecor fiber (spool 9+10@100m) fiber (spool 9+10@100m)

0 100 200 300 400 500 600 700-0.2

0

0.2

0.4

0.6

0.8

1

1.2

time (ps)

Extracted Impulse response

0 1000 2000 3000 4000 5000 6000 7000 8000 9000-0.2

0

0.2

0.4

0.6

0.8

1

1.2

time (ps)

Trace: MEASURED

Bandwidth of 600 Bandwidth of 600 MHzMHz··kmkm obtained from impulse obtained from impulse response extracted from bit sequenceresponse extracted from bit sequenceSee backup for eye diagram and impulse res. in freq. See backup for eye diagram and impulse res. in freq.

EDC - BER SensitivitySiecor fiber

1E-121E-111E-10

1E-09

1E-08

1E-07

1E-06

1E-05

1E-04

1E-03

-19 -18 -17 -16 -15 -14 -13 -12 -11 -10 -9 -8 -7 -6

Received optical avg. power [dBm]

BER

BtB without EDCBtB with EDC200m OL without EDC200m OL with EDCSimul. BtB Without EDCSimul. BtB With EDCSimul. 200m OL Without EDCSimul. 200m OL With EDC

•Simulated curves are shifted as a series for overlap at back-to-back point

•Simulations/Penalty with EDC include finite number of taps equal to EDC chip

3.7 dB

Curves w/ red dots are measurements

Reference

Observations (w/200 m Observations (w/200 m SiecorSiecor))Measured penalty of 3.7 dB for Offset launch matches Measured penalty of 3.7 dB for Offset launch matches theory in 2 ways:theory in 2 ways:

–– Within expected range of 2.5Within expected range of 2.5--5.2 dB determined through 5.2 dB determined through direct measurements of impulse responsesdirect measurements of impulse responses

–– Equals 3.7 dB determined through extracted impulse Equals 3.7 dB determined through extracted impulse response from bit sequences response from bit sequences

Bit sequences must be recorded simultaneously with BER Bit sequences must be recorded simultaneously with BER curves to ensure high correlationcurves to ensure high correlation

–– NextNext--day measurements doesn’t workday measurements doesn’t work

Discrepancy between theory and measurements for eye Discrepancy between theory and measurements for eye diagrams having high ISIdiagrams having high ISI

–– Is the BER estimation good enough?Is the BER estimation good enough?

Implementation penalty seems small compared to total Implementation penalty seems small compared to total penalty budget (~5 dB)penalty budget (~5 dB)

ConclusionConclusionGood correlation between measured and Good correlation between measured and simulated data demonstratedsimulated data demonstrated

–– Reproducibility is a challenge due to the nature of MMFReproducibility is a challenge due to the nature of MMF

Simple model with only Rx noise term seems Simple model with only Rx noise term seems sufficient for prediction of EDC performance.sufficient for prediction of EDC performance.

–– Model can be used to investigate expected coverage Model can be used to investigate expected coverage based on family of impulse responsesbased on family of impulse responses

Model needs to be ratified by the groupModel needs to be ratified by the group–– BER estimation may be a weak point.BER estimation may be a weak point.–– Explore limitations of modelExplore limitations of model

BackupBackup

Fiber Fiber data data -- CorningCorning

ProductProduct:: CorningCorning® 62.5/125um ® 62.5/125um

Coating Type:Coating Type: CPC6 (245 CPC6 (245 umum diameter) diameter)

FiberFiber ID ID 100010518479100010518479

LengthLength [meters][meters] 1759 1759

AttenuationAttenuation [dB/km][dB/km] 2.73/0.57 (850nm/1300nm)2.73/0.57 (850nm/1300nm)

BandwidthBandwidth [[MHz*KmMHz*Km]] 182/751 (850nm/1300nm)182/751 (850nm/1300nm)

CoreCore dia.dia. [[umum]] 61.761.7

CladClad dia.dia. [[umum]] 125.2125.2

NumNum. . AperatureAperature 0.2710.271

Back to Back Back to Back -- Center Launch (Corning)Center Launch (Corning)Ideal pulse (used as reference) Ideal pulse (used as reference) Impulse response file: h_0mCLx.txtImpulse response file: h_0mCLx.txt

Frequency response Impulse response

Data trace Eyediagram insnr (dBe): 30

Fiber: 400m Center Launch (Corning)Fiber: 400m Center Launch (Corning)DeDe--convoluted pulse (0m CL used as reference) convoluted pulse (0m CL used as reference) Impulse response file: h_400mCLx.txtImpulse response file: h_400mCLx.txtExtracted bandwidth of impulse response CL : ~ 5000 MHz*km



Fiber: 400m Offset LaunchFiber: 400m Offset LaunchDeDe--convoluted pulse (0m CL used as reference) convoluted pulse (0m CL used as reference) Impulse response file: h_400mOLx.txtImpulse response file: h_400mOLx.txtExtracted bandwidth of impulse response OL : 1600 MHx*km



200m 200m SiecorSiecor fiber (spool 9+10) fiber (spool 9+10)

0 1 2 3 4 5 6 7 8 9 10

x 109

-40

-35

-30

-25

-20

-15

-10

-5

0

Frequency (GHz)

Pow

er (d

B)

normalized responsemeasured responsecalibration curve

2 4 6 8 10 12 14 16-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

Unit [6.25ps ]

Eye diagram

Frequency response

Bandwidth of 600 Bandwidth of 600 MHzMHz··kmkm obtained from obtained from impulse response extracted from bit sequenceimpulse response extracted from bit sequence

comparison of experimental data with theoretical prediction of edc...

Documents