IBM team reveals 'world's fastest' optical chipset
Scientists from IBM have developed a prototype optical transceiver chipset capable of reaching speeds of 160 Gbit/s.
The transceiver, which is making its debut at the Optical Fiber Communication (OFC) conference in Anaheim, CA, is largely based on a CMOS chipset that performs all driver and receiver functions.
But the transmitter function of the chipset still relies on III-V technology, and a 4x4 VCSEL array is integrated within the transceiver.
In theory, it would be fast enough to reduce the internet download time for a typical high-definition feature-length film to just one second, compared with the usual half an hour or more for the best connectivity available today. IBM says that the development will transform how data is accessed, shared and used across the web.
According to the research team, this is the first time that such high-speed communications management has been achieved in a single, integrated transceiver chip. The chip is intended to be coupled with optical printed circuit board (OPCB) employing waveguides for computing-intensive applications, and measures just 3.25 x 5.25 mm.
"This compact design provides both a high number of communications channels as well as very high speeds per channel," said Marc Taubenblatt, senior manager of Optical Communications at IBM s TJ Watson Research Center in Yorktown Heights, New York. "This results in an amount of information transmitted per unit area of card space taken up by the chipset that is the highest ever achieved."
The transceiver IC was fabricated in the standard IBM CMOS8RF 130 nm foundry process, and consists of 16 independent laser-diode driver circuits and 16 receiver-amplifier circuits arrayed in two separate 4 x 4 blocks with a 250 x 350 micron pitch.
VCSELs emitting at 985 nm were then flip-chip bonded to the transceiver IC. The wavelength allows the optical signal to pass through the substrates of the optoelectronic devices and reach integrated lenses on the back side of the chips.
"This level of integration is unprecedented," Taubenblatt added. "Getting the channels in the device as closely packed as possible, but also doing that in a way that complements regular CMOS manufacturing techniques: that is what has enabled us to achieve 160 Gbit/s bi-directional communications."
"By comparison, the best commercial transceivers already available, such as the POP4 modules that will handle four channels of 5 Gbit/s each, give a total of around 20 Gbit/s, which means that the IBM chipset is effectively eight times faster."
Other commercial alternatives are available from several companies, including Emcore and Avago. Taubenblatt says that the key advantage of the IBM prototype lies in its single-chip-like assembly, whereas the POP4 MSA design is based on a combination of separate III-V chips and other optoelectronic functions.
POP4 is a multi-source agreement providing a common specification for four-channel, pluggable, parallel fiber-optic transceivers. These transceivers support applications such as next-generation scalable switch-router backplane interconnections in conjunction with 10 Gbit/s very short reach (VSR) physical layer interfaces.
While there is a clear demand for technologies to speed download times for data-intensive applications such as video-on-demand, IBM s transceiver is still very much at the prototype stage.
Although the relatively economical CMOS technology at its heart will ultimately make these high-speed chips widely affordable, the company is only in the early stages of exploring plans for use in commercial products.
"There is certainly room for growth in the core optical network, and the new chipset could support that," Taubenblatt said. "Ultimately, it s a serious possibility to provide 100 Gbit/s services to the home. We are seeing a tremendous increase in growth due to the usual quad-play developments, especially mushrooming content on the internet.
"One could say that this 160 Gbit/s chip anticipates an all-fiber optical communications network. The other side of the equation is that if customers [increasingly] wish to do video-on-demand functions then the bandwidth needs to be that much greater."
IBM s work was partially funded by the US government s Defense Advanced Research Project Agency through the Chip to Chip Optical Interconnects (C2OI) program.
•The report on this work, "160 Gbit/s, 16-channel full-duplex, single-chip CMOS optical transceiver," by C L Schow, F E Doany, O Liboiron-Ladouceur, C Baks, D M Kuchta, L Schares, R John, and J A Kash of IBM s TJ Watson Research Center will be presented on Thursday, 29 March, at OFC in Anaheim, California.
About the author
Matthew Peach is a contributing editor to optics.org and Optics & Laser Europe.