Candidate experience overview (2013)
1 like686 views
Huawei has expertise in both Si and InP photonic integration. For Si integration at Luxtera, they designed optical chips and interconnect systems on an SOI wafer. For InP integration at Infinera, they developed 10 channel transmitters and receivers using OOK and advanced modulation formats like DQPSK. Potential future work includes PAM modulation for 500m reach and 3D packaging for chip-to-chip optical interconnects.
Candidate experience overview (2013)
- 1. Hybrid Photonic integration @ Huawei: for optical transport & C2C Photonic integration in Si and InP: Overview of experience and expertise Roman Malendevich Confidential and proprietary 02/23/13
- 2. Outline Si integration at Luxtera InP integration at Infinera • OOK • Coherent detection & advanced modulation Summary Potential future aspirations • PAM for 500m • 3D packaging for C2C Confidential and 2
- 3. At @ Luxtera: Si photonics integration Confidential and 3
- 4. Si optical chips and interconnect systems Optical Filters - DWDM Flip-chip bonded lasers 10G Modulators 4 km 10G Rx Fiber cable Ge detectors, TIAs Ceramic Package copper SOI transistor Optical “wire” SOI transistor (waveguide) poly Si WG Field oxide "Active" Si Si Buried Oxide (BOX) "Handle" Si Confidential and 4
- 5. Design experience Grating couplers Designed SOI production wafer structure • Integration challenge • Need best system compromise of performances of individual components “Photonics Spectra”, March 2006 – Grating couplers (IL and bandwidth) – Waveguide loss – Modulators (speed and IL) – AWG (cross-talk and IL) – Manufacturability and yield Confidential and 5
- 6. General rigorous device design methodology Functionality Architectural choices Design inputs and outputs • Input parameters interactions Theoretical limits Key trade-offs Design: • theory • modeling & simulations • experimental DOEs (tape-outs) Design and process sensitivities • corner cases, Monte-Carlo statistics • manufacturability, yields Skewing test results • Repeatability, reproducibility, accuracy, Gage R&R Correlations and building models Confidential and 6
- 7. Wafer level testing: Design and architecture Million of multi-port devices on ea. wafer fibers RF • 6D alignment Multiple input-output fibers • separated by only 127um DC Finds ea. device in seconds • not minutes Initial aligning structures: • one-port reflecting Littrow gratings Capacitor Z-axis spacing sensor • tight 5um air gap across 8” wafer Operator-free 24/7 Still in production (10 years) • 9 patents 8” Si wafer “Photonics Spectra”, March 2006 Confidential and 7
- 8. At @ Infinera: InP photonics integration Confidential and 8
- 9. 10ch. x 10G OOK Tx and Rx PICs Optics and Photonics News (2009) CS MANTECH Conference (2006) Confidential and 9
- 10. 10ch. x 10G OOK Transmitter module 200G spacing 0 Normalized Power (dB) -10 -20 -30 -40 -50 Tx PIC -60 1.525 1.530 1.535 1.540 1.545 Wavelength (µm) Fiber Systems, Lightwave Europe (2006) OPN (2009) Confidential and 10
- 11. Advanced modulation formats Confidential and 11
- 12. Pol-Mux DQPSK Reference Tx box: Performance HW development and integration Amplitude • Ref Tx to Ref Rx: test data shown • OSNR-loaded long-term (24/7) stability <0.1dB Q Phase 90o 0o TE 180o 270o TM Confidential and 12
- 13. Reference Tx for PM-DQPSK and coherent QPSK (typical HW configuration) Confidential and 13
- 14. Typical test set for BER test (simplified) OFC 2011 (OML7) Confidential and 14
- 15. Non-coherent DQPSK Rx Passive PIC: uses 1-bit delay Pros: • No LO frequency mismatch || Pol. demux done by MIMO analog control • No need for coherent DSP ASIC || Saves 1-2 yrs development time || Saves power Cons: • Reduced Rx power sensitivity • TE/TM paths must be bit aligned – need integrated PBS and pol. rotator → High losses in InP • MIMO circuits’ zero offsets complicate pol. tracking || Reduced Q performance MIMO ASIC 1-bit delay OFC 2011 (OML7) Confidential and 15
- 16. Coherent Rx Active Rx PIC: uses DFB LO Pros: • Pol. tracking done by DSP ASIC digitally • DSP compensates for PMD, CD • PBS can be external • Better Rx power sensitivity Cons: • Requires development time of complex, power-hungry DSP • Tx DFB and LO DFB must have limited linewidth, phase noise “10-channel, 28 Gbaud PM-QPSK, monolithic InP Terabit Superchannel Receiver PIC” M. Kato, R. Malendevich, D. Lambert et al., 2011 IEEE Photonics invited, & OFC 2011, invited Confidential and 16
- 17. 1 Tb/s Tx PIC Each PIC: >400 integrated functions Each λ: “Terabit Superchannel Coherent InP PM-QPSK Transmitter PIC” P. Evans, M. Fisher, R. Malendevich et al., Opt. Express 19 2011 & OFC 2011 (post-deadline) Confidential and 17
- 18. InP DQPSK modulator “Terabit Superchannel Coherent InP PM-QPSK Transmitter PIC” P. Evans, M. Fisher, R. Malendevich et al., Opt. Express 19 2011 & OFC 2011 (post-deadline) Confidential and 18
- 19. Tx DFB Frequency Noise Power Spectral Density (PSD) Fundamental problem: phase noise σ = 2π ⋅ LW ⋅ ∆t Linewidth < 1MHz Linewidth < 1MHz Ref. on slide 18 Confidential and 19
- 20. First live network coherent PIC demo (2010) Confidential and 20
- 21. et al. Linear Tx driver From coherent DSP Confidential and 21
- 22. * et al. * O. Gerstel (VZ) et al, “Elastic Optical Networking: A New Dawn for the Optical Layer?”, IEEE Comm. Magazine, 2012 Confidential and 22
- 23. Gridless super-channel transmission Same ref. as on slide 23 Confidential and 23
- 24. “Elastic optical network” (spectrum defragmentation) Same ref. as on slide 23 Confidential and 24
- 25. Potential future aspirations Confidential and 25
- 26. PAM for 500m Ref.: IEEE 100G taskforce Confidential and 26
- 27. 3D Packaging for C2C Ref.: Computer Systems Based on Silicon Photonic Interconnects, Sun, IEEE Proceedings (2009) Confidential and 27
- 28. Summary The field of photonics integration is blossoming • With applications in long-haul, metro, local, FTTH and datacenter networks, as well as C2C and MCM communication – Even in your living room! • A number of start-ups (Aurrion, Skorpio Technologies, Luxtera) and giants (Intel, Huawei, Cisco/Lightwire, Oracle/Sun, IBM, Finisar, Samsung) intensifying the field • The future of photonic integration looks as promising as ever • I have personally worked on OOK and coherent PICs, modules and systems, integrated both in Si and InP – From concept to real world, customer demos and production Confidential and 28
- 29. Addendums Confidential and 29
- 30. Academics CREOL (Center for Research and Education in Optics & Lasers) Univ. of Central Florida Kiev National University • 1st runner-up National Physics Olympics Confidential and 30