1. SOFI Meeting in Rome 2012/01/20

2. © 2012 GigOptix, Inc. All Rights Reserved Confidential 222 Outline  Material: M1 vs M3  Driver Study  GX6255  How to feed the signals to the modulator  Integrated SiGe driver

3. © 2012 GigOptix, Inc. All Rights Reserved Confidential 333 Material: M1 vs M3  Problem: M1 not temperature resistant enough  M3 more temperature resistant  You could get M3 material from Gigoptix  Raluca has the feeling that challenges due to new material are larger than advantages  The change from M1 to M3 will have an effect on the MMI and Bragg Grating performance.  There are some processing challenges related to M3 that may require some additional development work, and or perhaps two masking and etching process steps may be needed to be added to the original process flow that was used for this project. (info from Eric Miller, GM of Bothell Materials & Processes Center)

4. © 2012 GigOptix, Inc. All Rights Reserved Confidential 444 Material: M1 vs M3 Material ID r 33 @1.3 mm (pm/V) Dielectric constant @ 1 MHz Dielectric constant @ 25 GHz Optical Loss @ 1.55 (dB/cm) Tg ( 0 C) RI @ 1.55 mm TETM M1854.13.20.9-1.0138 1.6801 (Unpoled) 1.6612 (Poled) 1.6711 (Unpoled) 1.7433 (Poled) M390-1104.13.21.2-1.3167 1.6913 (Unpoled) 1.6700 (Poled) 1.6748 (Unpoled) 1.7339 (Poled)

5. © 2012 GigOptix, Inc. All Rights Reserved Confidential 555 Driver: SiGe vs CMOS  Study performed  CMOS: 40 nm technology for an f t above 250GHz  In academics: 65 nm possible, but not enough margin for a product  Initial costs of 40 nm CMOS are very high, but fabrication costs lower than SiGe  SiGe cheaper, unless you want to sell millions of chips

6. © 2012 GigOptix, Inc. All Rights Reserved Confidential 666 Driver Study  What could be possible?  Experiments with Gigoptix driver GX6255  Study on how to get the signals first to the chips and then to the modulators  Maybe a SiGe driver with 650mVpp swing at 50Ω

7. © 2012 GigOptix, Inc. All Rights Reserved Confidential 777 Experiments with GX6255  32Gb/s MZ Modulator Driver  Is capable to drive 25 Gb/s  Proven concept  Single ended  V out up to 7.5 V pp  Smaller signals possible with attenuator or maybe with smaller input signal at driver  With small input signal: maybe not fast enough  With this setup it is possible to characterize the modulator, proof of concept and determining the requirements for a driver

8. © 2012 GigOptix, Inc. All Rights Reserved Confidential 888 How to feed the signals to the modulator  We know that we can drive 25Gb/s modulators  But many practical issues are open:  Arrangement of the 4*25G modulators, spacing  How to feed the signals to the driver chip  Connection of chip and modulator  FlipChip bonding, wire bonding, tracks on modulator substrate  Issues: Temperature, crosstalk  Impedance of modulator: how accurate?  Maybe we can rent a 3D field simulator to study the influence of the wiring

9. © 2012 GigOptix, Inc. All Rights Reserved Confidential 999 Integrated SiGe driver  We are currently designing a 25 Gb/s VCSEL driver  Unclear when we will have a tape-out  It may be possible to adjust the output stage so, that we are able to drive the modulator  Limited swing of ~650mV  Increasing the swing means changing the architecture of the output stage  Decision if a chip is made after the first measurement results with the GX6255.

10. © 2012 GigOptix, Inc. All Rights Reserved Confidential 10 Miscellaneous  Dissemination activities:  Giulio Cangini worked on the white paper  Personnel changes:  Jörg Wieland left the company  GigOptix was understaffed in the last half year  Silvan Wehrli now responsible for SOFI

11. 11 © 2012 GigOptix, Inc. All Rights Reserved Confidential Thank you!