What are the Parallelization Technologies for Cost and Energy Efficient 1.6Tb Links?

Tuesday, 08 March 14:00 – 16:00


Tomoyuki Akiyama, Fujitsu Laboratories Ltd., Japan
Long Chen, Acacia Communications, Inc., USA
Gloria Hoefler, Infinera Corp., USA
Zuowei Shen, Google LLC, USA


The deployment of 400G optics is ramping up quickly. On the short reach side, the direct-detection 100G/lambda PAM4 in parallel (DR4) or CWDM (FR4) are implemented at low cost and high energy efficiency with either EMLs or silicon photonic transmitters. On the longer reach side, 400G ZR / ZR+ / Open ROADM coherent pluggables bring fast, streamlined network architectures to hyperscale data centers and carriers. The next generation 800G PAM4 optics with DR8 is expected to deploy next year, followed by 800G coherent solutions.

What would the generation of 1.6T optics look like? One frequently cited viewpoint is that as the data rate goes up, coherent becomes more advantageous because of its higher link margin. Would simplified coherent, or coherent light, become energy efficient to compete against direct-detection in short reach? For coherent, 1x 1.6Tb, 2x800G, 4x400G, which one is more cost and energy efficient? For direct-detect, would the 200G/lambda still be PAM4, or higher format such as PAM6 at lower baud rate? Would it be 8 parallel fibers, or a mixture of parallel fibers and CWDM (2xCWDM4 or 1xCWDM8)? Also, as the chip-to-chip interconnects gain more visibility, what would the optics look like for ultimate efficiency?

The panelists would share their viewpoints on these topics and open discussion with the audience.  


Arash Farhoodfar, Marvell, USA

Xiang Zhou, Google, USA

Osa Mok, Innolight, China
1.6T optics for Next Generation DC

Karl Muth, Broadcom, USA

Sergey Shumarayev, Intel, USA