comparison of experimental data with theoretical prediction of edc...
TRANSCRIPT
Page 1
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
Page 2
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.
Page 3
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
Page 4
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
Page 5
EyeEye--diagrams after 400 m MMF diagrams after 400 m MMF
Center launch: >5000 MHz*km Offset launch: 1600 MHz km
Page 6
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
Page 7
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
Page 8
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
Page 9
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)
Page 10
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
Page 11
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
Page 12
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.
Page 13
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
Page 14
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)
Page 15
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
Page 16
BackupBackup
Page 17
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
Page 18
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
Page 19
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
Frequency response Impulse response
Data trace Eyediagram insnr (dBe): 30
Page 20
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
Frequency response Impulse response
Data trace Eyediagram insnr (dBe): 30
Page 21
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