Status Of Silicon Photonics Reliability Test
Status of silicon photonics reliability testIchiro Ogura, PETRARichard Pitwon, AIO CoreIEEE 802.3cz Multigigabit Optical Automotive Ethernet Interim Task Force29th June 2021
Clarifying misconceptions around silicon photonics1310 nm is the dominantwavelength for current andfuture data centre Ethernetstandards and will drivecommunication laser demandSpeedEthernetNomenclatureLane no xNominal rateLink distance /mediumWavelengthStandardsgroup100 GbE100GBASE-SR1010 x 10G NRZ150m OM4 10 pairMMF850 m802.3ba100GBASE-SR44 x 25G NRZ100m OM4 4 pairMMF850 nm802.3bm100G-PSM4100G-CWDM4500m on 4 pair SMF2km on duplex SMF1310 nmCWDM (1271nm,1291nm, 1311nm,1331nm)MSAMSA100GBASE LR410 km on duplex SMFLAN-WDM (1295 nm,1300 nm, 1305nm,1310 nm)802.3ba100m on 4 pair OMFMMF850 nm802.3cd200 GBASE-DR4500m on 4 pairparallel SMF1310 nm802.3cd200 GBASE-FR42km on duplex SMFCWDM (1271nm,1291nm, 1311nm,1331nm)802.3cd200 GBASE-LR410 km on duplex SMFLAN-WDM (1295 nm,1300 nm, 1305nm,1310 nm)802.3cd200 GbE400 GbE200 GBASE-SR44 x 50G PAM4400 GBASE-SR1616 x 25G100m on 16 16parallel OM4 MMF850 nm802.3bs400 GBASE-FR88 x 50G PAM42km on duplex SMF4.36 nm WDM gridover 1310 nm802.3bs10km on duplex SMF4.36 nm WDM gridover 1310 nm802.3bs500m on 4 pairparallel SMF1310 nm802.3bs400 GBASE-LR8400GBASE-DR4Status of Silicon Photonics Reliability Testing – 29th June 20214 x 100G PAM42
Clarifying misconceptions around silicon photonicsIndustry supportMultiple MSAs around siliconphotonics including:PSM4, CWDM4, 100GLambda, CWWDMMainstream international foundriesmanufacture silicon photonics Status of Silicon Photonics Reliability Testing – 29th June 2021Global FoundriesTSMCST MicroelectronicsIntelTowerJazzVTTSkorpiosiHP CompoundTekSilexAIM PhotonicsAPMCEA LetiIMECAMFSkywater3
Clarifying misconceptions around silicon photonicsCost depends on volume, chip size and assembly Silicon photonics benefits from mature CMOS processThe cost of silicon photonics is determined by chip size. A 1-ch silicon photonicstransceiver would be almost equal to a transceiver and receiver ICCoupling to GI-MMF (OM3) requires very low-cost passive alignment.Multimode interfaceStatus of Silicon Photonics Reliability Testing – 29th June 20214
Clarifying misconceptions around silicon photonics There is a wide spectrum of silicon photonics transceivers targeting different applicationsand cost sensitivities. High-end silicon photonics for telecommunications At the high-margin telecommunications end, advanced silicon photonics basedtransceivers incorporating coherent, DWDM, PQSK and other advanced features areexpensive. These are the devices traditionally associated with silicon photonics. Competitive silicon photonics transceivers for data centres For the past 6 years, front-pluggable transceivers such as QSFPs and AOCs based onsilicon photonics have been commercially available for wide-spread deployment indata centres competing with VCSEL based devices. The data centre is a highly costsensitive environment and purchase decisions are based primarily on cost. Co-packaged optics driving low relative cost silicon photonics For the next 5 years, co-packaged optics will continue to drive silicon photonics costsfurther down and reliability further up.Status of Silicon Photonics Reliability Testing – 29th June 20215
Co-packaged optics (CPO) is driving silicon photonics evolutionOver the next few years the need for co-packaged optics(CPO) solutions will drive costs further down andreliability further up in a higher temperature environmentthan has been typical of data centres.Silicon photonics is the only serious contender to meetthis challenge.VCSEL based CPO were also considered, however whereCPO environments are now expected to exceed 100 Cthey are not considered viable given the high reliabilityrequirements.Optical fibre ribbon to front-endHigh speed electrical PCB trace length range 1 cmStatus of Silicon Photonics Reliability Testing – 29th June 20216
Co-packaged optics (CPO) is driving silicon photonics evolutionCo-Packaged Optics Collaboration *Founded in March 2019 by Microsoft and FacebookFirst specification for 3.2T CPO released in 2021Fibre Array UnitTransceiverASIC* http://www.copackagedoptics.comStatus of Silicon Photonics Reliability Testing – 29th June 20217
Example of competitive silicon photonics transceiver This silicon photonics micro-transceiver design is simplified and accommodates amultimode fibre interface in order to minimise relative cost, targeting short-reach andhigh-temperature applications. Although this specific example would probably not be used for automotive, it serves as atestbed or demonstrator for how silicon photonics can be cost competitive.Optical couplinginterfaceSilicon photonicsintegrated circuitOptical I/O Core(25 mm2)LD (passivelyaligned)Dime(251 mm2)5mmFlip chipbonded ICStatus of Silicon Photonics Reliability Testing – 29th June 20218
Short reach multimode silicon photonic transceiverFunctional DescriptionDriverQD lasersourceQD lasersourceQD lasersource1310 nmQD 10011Glass topOptical pinlaser sourceSilicon waveguideVertical gratingcouplerGlass topTaperedoptical pinSi GePD10011Transimpedance AmplifierStatus of Silicon Photonics Reliability Testing – 29th June 20219
Short reach multimode silicon photonic transceiverComponent overviewMMFQuantum Dot Laser DiodeElectrical inputDriver ICSi photonics substrateModulatorWaveguideGratingcouplerMultimode interfaceMMFStatus of Silicon Photonics Reliability Testing – 29th June 202110
Redundant optical circuit for increased reliabilityBuilt-in dual redundancy in Si photonics TransceiverMZI-modulator configuration allows for dual light sources to be operated as redundancy1) 1st generation QD-LD and MZ configurationLDSilicon Photonic substrateCh4TxOut4Ch3TxOut3Ch2Ch13dB CouplerTxOut2TxOut13dB CouplerPower monitor2) 2nd generation configuration with redundancyLDSilicon Photonic itch3dB CouplerTxOut2TxOut13dB CouplerPowermonitorStatus of Silicon Photonics Reliability Testing – 29th June 202111
Internal optical communication in automotive applicationsSensor clusters /camerasProcessing hubReliability of optical communication links in automotive applications iscritical and not to be compromised.Status of Silicon Photonics Reliability Testing – 29th June 202112
Ongoing reliability tests of Quantum Dot Laser Diodes (QD-LDs)80 ,120mA/2ch（Tj 93.8 ）N 10p100 ,120mA/2ch （Tj 113.8 ）105 ,120mA/2ch （Tj 118.8 ）N 10pN 5p 20%Test chip: Integrated 2ch-LDSample size: N 66 pieces (equivalentto 132 pieces of single-LD)No sudden failures-20%105 ,220mA/2ch （Tj 130.3 ）N 7p100 ,300mA/2ch （Tj 134.5 ）N 17p100 ,390mA/2ch （Tj 144.9 ）N 17pAutomotive mission profiles are wellcovered by test conditions.If 2ch vs Tj Test and Use conditionRange of test conditions80 ,120mA/2ch（Tj 93.8 ）N 10p100 ,120mA/2ch （Tj 113.8 ）N 10p105 ,120mA/2ch （Tj 118.8 ）N 5pRange of use conditions105 ,220mA/2ch （Tj 130.3 ）N 7p100 ,300mA/2ch （Tj 134.5 ）N 17p100 ,390mA/2ch （Tj 144.9 ）Max testconditionMax useconditionN 17pΔPo, Po were measured at 85 ,100mAStatus of Silicon Photonics Reliability Testing – 29th June 202113
Wear-out failure estimates for QD-LDsUnreliability formulaMethod of estimating wear-out lifetime𝐹 𝑡 Φ Estimate linearly extrapolated lifetimes (EOL) for -20% power drop Apply following reliability models: Lognormal distribution Acceleration factors：Junction temperature (Tj), Current (I)μ μ0 𝑒𝑥𝑝Lognormal probability plotMaximum Likelihood Estimate(MLE) analysis was conducted usingthree test conditions, in which testsreached more than 3000 hours andaccurate lifetimes could beestimated.MLE analysis yielded followingparameter values:Ea 0.97 eV, σ 0.66, μ 5.5 104 h @Ta 105 , 140 mAStatus of Silicon Photonics Reliability Testing – 29th June 2021ln 𝑡 ln(μ)σ(normalized to 105 C, 140mA）Sample N 22p𝐸𝑎 𝐼 𝑛𝑘𝐵 𝑇𝑗F(t)：Unreliability function, t：TimeΦ：Standard normal distributionEa, n, σ, μo：ConstantskB: Boltzmann constantModel line105 ,140mA2ch-LDEOL: ΔPo -20%Lognormalσ 0.66μ 5.5 104 h14
Projected MTTF values based on wear-out failures2ch-LD EOL: ΔPo -20%Ta (deg C)If 2ch (mA)Tj (deg C)ML( μ)2ch (h)MTTF2ch (h)105120118.89.02E 041.13E 0595120108.81.91E 052.38E 058512098.84.22E 055.26E 055512068.85.99E 067.47E 06105140121.15.46E 046.80E 0495140111.11.15E 051.43E 0585140101.12.51E 053.12E 055514071.13.44E 064.29E 06Tj of QD-LDs is lower than that of VCSELs due to low thermal resistanceStatus of Silicon Photonics Reliability Testing – 29th June 2021The temperature difference (ΔT) between Taand Tj for Si-photonic modules is less than 20 ΔT 15 ＠140mAfor actualmeasurement15
QD-LD lifetime estimate for automotive mission profileWear-outRandomEa (eV)0.969Lognormal σ0.664 -ln(μ) Base cond.If (mA)Tj （ ）Total operation time tm (h)10.907 -Random FIT Base cond. Base cond. Ta（ ）Operation Current (mA)052ch-LD @tmIf 140mA 2ch-LD, Random 5FitCumulative failure rate：161 ppmMean FIT：5.0 Fit14032000CumulativeMean FITfailure (ppm)1615.0Wear-out failure rate is negligible small.Failure rate is determined by random FIT.105140Temperature profile121.132000 hCumulative failure rate vsOperating Time320ppm2ch-LDEOL: ΔPo -20%2ch-LD TotalRandomWear-outStatus of Silicon Photonics Reliability Testing – 29th June 202116
Signal integrity analysis at 128 Gbps (32G x 4ch), 105 C operationRxTa 25 Ta 105 TxTa 25 Ta 105 Ch1Ch1Ch2Ch2Ch3Ch3Ch4Ch432 Gbps PRBS31 - TX,RX All lanes activeOperational current region32 Gbps PRBS31 - TX,RX All lanes activeKurata, K. et al. "Short reach, high temperature operation and high reliability silicon photonic micro-transceivers for embedded and co-packagedsystem integration," Proc. SPIE 11692, Optical Interconnects XXI, 1169204 (16 March 2021); https://doi.org/10.1117/12.2576670Status of Silicon Photonics Reliability Testing – 29th June 202117
Advantage of QD-LD: Resistance to DLD induced sudden failureQD-LDVCSELActive layer: InAs/GaAs quantum dot (QD) isolated in a planeDislocation motion is reduced by localized carriersin QDs, which suppresses DLD induced sudden failure.¹）Active layer: (In)GaAs quantum well (QW)spreading in a planeDislocation motion is enhanced by diffusedcarriers in QW layers, which leads to suddenfailure due to DLD formation.2)QWQDThreading dislocationStay in the QDVCSEL structureThreading dislocationExtend along active layerExtend along active layerDBROrigin of dislocationActive layerDBR1) T. Kageyama, 27 Nov.,2019 Optoelectronics ＆TechnologyDevelopment Association Workshop2) Materials and Reliability Handbook for Semiconductor Optical and ElectronDevices(2013) pp150-205The main advantage of QD-LDs are resistance to sudden failure and low thermal resistance.Status of Silicon Photonics Reliability Testing – 29th June 202118
Quantum Dot laser reliabilityFabry Perot Quantum Dot laser Strong and increasing industry support due to superior performance, stability at high temperatures Intel paper: J. C. Norman et al., "A Review of High-Performance Quantum Dot Lasers on Silicon," inIEEE Journal of Quantum Electronics, vol. 55, no. 2, pp. 1-11, April 2019, Art no. 2000511, doi:10.1109/JQE.2019.2901508FIT rate of QD-LDs is estimated to be 1.7 FIT based on total shipment of3.2Mpcs and total 1.4 1011 component hours.Status of Silicon Photonics Reliability Testing – 29th June 202119
Questions about 850 nm and 980 nm VCSEL suitability850 nm VCSELs – strong questions about reliability for target environmentProsCons850 nm has been deployed in the field as a Based on different contributions, 850 nm VCSEL reliability in thedatacom transceiver for over 20 years.target environment seems strongly questionable.It has benefited from two decades ofdesign and manufacturing optimisation.Shows excellent reliability in data centreenvironmentsResults from calculations vary between failure and pass dependingon small changes in choice of activation energy Ea figure and deltabetween junction temperature Tj and Ta.Environmental testing is non-conclusive.940nm / 980 nm VCSELs - more reliable, but no field deployment history for optical datacomProsCons980 nm shows better reliability resultsthan 850 nm in target environmentIt has only been used for sensing. It has no field deployment historyfor datacom. No standards advocate 980 nm, no transceivers use980nm VCSELs, its performance has only been shown in labsPlease avoid false equivalency between 850nm and 940nm or 980 nm VCSELs.There is no composite “850nm/980nm” VCSEL, which combines the long field deployment history of850 nm and the higher reliability of 980 nm. 850 nm and 980 nm are not the same.Status of Silicon Photonics Reliability Testing – 29th June 202120
Exclusion of silicon photonics difficult to justifyQuestionIn 2026, when this standard will most likely be sought in earnest, silicon photonics will bethe dominant optical transceiver by revenue* and is inherently more reliable thandirectly modulated VCSEL solutions. How will we justify having completely excluded itfrom consideration and choosing only VCSEL based PMDs?ProposalNo need to choose all 28 PMDs, but for the first batch of PMDs we should at least choosethe most useful PMDs in terms of reliability or cost to give automakers the choice.Propose that the first batch of PMDs include VCSELs for at least lower data rates butshould also include silicon photonics at least for the higher data rate 25G over OM3 andPOF to give automakers the choice between the best of both worlds.* Recommended source: Online Lightwave article: nics-overhyped-nahStatus of Silicon Photonics Reliability Testing – 29th June 202121
Contributions and technical referencesI. Ogura (PETRA) and K. Kurata (AIO Core) havepublished many scientific papers and madenumerous contributions to IEEE 802.3cz on theirdevices alone (shown below)Generally there is a vast body of technicalpublications on 1310 nm silicon photonicstransceiversContributions to 802.3cz– March 2021: Response to Proposal AssumptionsTechnical publications1.Kurata, K. et al., “"Short reach, high temperature operation and high reliabilitysilicon photonic micro-transceivers for embedded and co-packaged systemintegration", Proc. SPIE 11692, Optical Interconnects XXI, 1169204 (16 March2021); https://doi.org/10.1117/12.25766702.Kurata, K. et al., “Short reach, low-cost silicon photonic micro-transceivers forembedded and co-packaged system integration,” Proc. SPIE 11286, OpticalInterconnects XX, 112860R (28 February 2020); https://doi.org/10.1117/12.25466263.Pitwon, R., O’Faolain, L., Kurata, K., Lee, B., Ninomiya, T., "Hyperscale IntegratedOptical and Photonic Interconnect Platform," 2020 IEEE Photonics Conference(IPC), Vancouver, BC, Canada, 2020, pp. 1-2, doi:10.1109/IPC47351.2020.9252246.4.Nakamura, T. et al., “Fingertip-Size Optical Module, “Optical I/O Core”, and ItsApplication in FPGA” IEICE TRANSACTIONS on Electronics, Vol.E102-C, No.4,pp.333-3395.Mogami, T. et al., "1.2 Tbps/cm2 Enabling Silicon Photonics IC Technology Basedon 40-nm Generation Platform," J. Lightwave Technol. 36, 4701-4712 (2018)6.K. Kurata, I. Ogura, K. Yashiki and Y. Suzuki, "Chip-scale si-photonics opticaltransceiver for a photonics-electronics convergence system (invited paper)," 2016Tenth IEEE/ACM International Symposium on Networks-on-Chip (NOCS), Nara,Japan, 2016, pp. 1-6, doi: 10.1109/NOCS.2016.7579338.– March 2021: A proposal of Si-photonics for automobile– January 2021: Thoughts on PMD baseline proposal for automobilebased on Si-Photonics– January 2020: A study for highly-reliable optical transceiver basedon Si Photonics technology– November 2019: Introduction of SI Photonics transceivertechnology with High temperature operation capability and MMFtransmissionStatus of Silicon Photonics Reliability Testing – 29th June 202122
THANK YOU
Integrated vs remote lasersFront panelIntegrated lasersLasers integrated into SiPh chiplet- Serviceability Lowest total power Better testability Lowest CostRemote lasersSeparate on-board modules or front pluggablemodules- Highest laser loss to PIC ( 4dB)- PM fibre required between lasers and PIC Better serviceability Lowest thermal densitySwitchLaser SiPhCarrierPCBFront panelSiPhSwitchLaser OBOCarrierPCBFront panelSiPhSwitchLaser FPPCarrierPCBStatus of Silicon Photonics Reliability Testing – 29th June 202124
Co-packaged optics is driving silicon photonics evolutionFirst deployments at50TStandard 1RU enclosure32 QSFP slots fill up the front fascia51.2 Tb/s(Broadcom:Humboldt)Bandwidth12.8 Tb/sI/O: 512 x 100 Gb/sFPP: 64 x 800GForm factor: 2 RUI/O: 256 x 100 Gb/sFPP: 32 x 800GForm factor: 1 RU(Broadcom:Tomohawk 4)6.4 Tb/sI/O: 512 x 200 Gb/sFPP: 64 x 1.6TForm factor: ?(Broadcom:Bailly)25.6 Tb/s102.4 Tb/sBroaddeploymentsat 100TI/O: 256 x 50 Gb/sFPP: 32 x 400GForm factor: 1 RU(BroadcomTomohawk 2)I/O: 256 x 25 Gb/sFPP: 32 x 200GForm factor: 1 RU20182019202020212022Status of Silicon Photonics Reliability Testing – 29th June 20212023202420252026202725
Mainstream international foundries manufacture silicon photonics . For the past 6 years, front-pluggable transceivers such as QSFPs and AOCs based on silicon photonics have been commercially available for wide-spread deployment in data centres competing with VCSEL based devices. The data centre is a highly cost-
Keywords: Silicon Photonics , high-density photonics integration, Silicon on Insulator SOI, integrated optics, waveguide cross-talk, directional couplers , optical bio-sensors. 1. INTRODUCTION Industry and governments have recently invested and promoted Silicon Photonics, contributing remarkably to its sudden and quick development 1. In .
Photonics technologies for system-level integration System-level: Scalable chip-to-fiber connectivity Chip-level: CMOS silicon photonics Active photonics devices Si photonics provides all required buliding blocks (except lasers) on chip-level: - Modulators - Drivers - Detectors - Amplifiers - WDM filters CMOS electronics 2 1
Silicon Photonics Design and Verification Flow Ref: W. Bogaerts and L. Chrostowski, "Silicon Photonics Circuit Design: Methods, Tools and Challenges," Laser Photonics Rev. 2018, 1700237.
Intel's Silicon Photonics Research Innovating with lowInnovating with low-cost silicon to create new optical devicescost silicon to create new optical devices 1st Continuous Wave Silicon Raman Laser (Feb '05) Hybrid Silicon Laser (Sept. '06) Silicon Modulators 1GHz ( Feb '04) 10 Gb/s (Apr '05) 40 Gb/s (July '07) 8-channel integrated
Dr. Charles Baudot STMicroelectronics Silicon Photonics R&D Datacom & More Than Datacom 5 New opportunities in silicon photonics Silicon photonics cannot rely on data communication professional market only. For the industry to develop: - More applications - Huge volumes - More actors - Bigger competition - Lower prices - More innovation
name of Photonic Integrated Circuits (PICs). Silicon Photonics is a form of PIC that uses silicon substrates and silicon waveguides for the platform. . A key enabler in the silicon photonic revolution has been the development of technologies compatible . [11] L. Chrostowski and M. Hochberg, "Silicon Photonics Design: From Devices to Systems .
semiconductors. Stimulated by a series of recent breakthroughs and propelled by increasing investments by governments and the private sector, silicon photonics is now the most active discipline within the field of integrated optics. This paper provides an overview of the state of the art in silicon photonics and
Tom Sawyer’s observations of his environment and the people he encounters. In addition, students will make their own observations about key aspects of the novel, and use the novel and the journal writing activity to make observations about their own world and the people they are surrounded by. This unit plan will allow students to examine areas of Missouri, both in Hannibal, and in their own .
