This week I had the privilege of representing one of my clients at the AIM Photonics road map meetings held at MIT in Cambridge, MA. While this was a closed meeting for AIM members, it’s no secret that AIM Photonics, which stands forAmerican Institute for Manufacturing Integrated Photonics, is working diligently to bring up a viable ecosystem for integrated photonic design. While it was heartening to see all of the progress being made, it also became very clear to me that the term “integrated photonics” means a lot of things to a lot of different people. When asked by Dan Nenni what we should call this subject area on SemiWiki I wrongly followed my 30+ years of electronic IC design thinking and replied, “Silicon Photonics”.
While Silicon Photonics is the holy grail for fully integrating photonics and electronics I quickly realized that most photonics wasn’t being done in silicon yet. Upon realizing my mistake, I started referring to it as “Integrated Photonics”.
This week my eyes were opened yet again to the fact that I am a product of my upbringing. I had foolishly been thinking integrated would mean ‘on a chip’, whether it be made of silicon, indium phosphide, gallium arsenide or some other materials. The first half day of the sessions at MIT proved this to be woefully short sighted. Photonics has been around for the last 20+ years and it has slowly but surely been working its way from bulky discrete devices and glass-core fibers towards integration onto the chip. What I failed to remember is that you don’t get to the chip unless you go through a printed circuit board (PCBs) and some form of package.
Today active optical cables (AOCs), which are optical cables with built-in transceivers on each end, convert optical signals into electrical signals which then run through connectors and PCB traces to the electronic ICs (see figure from an OSA paper by IBM). The move now is to put the “active” optical parts of the transceivers closer to the electronics (i.e. on the board), leaving the “passive” optical parts in the cables. To do this however, you must have a way to move light around the board.
Make no mistake, there are already production photonic ICs (PICs) being put on boards but till now this has mostly been accomplished by using somewhat cumbersome connections called optical pigtails that carry light directly from the photonic I/O of the PIC to a connector at the edge of the board. Pigtails usually use fiber ribbons that ‘fly over’ the board. While this works, it is costly for several reasons. First it usually implies some kind of manual connection being made from the pigtail to the connector on the board. Additionally, routing of the fiber/ribbon is problematic and in many cases the ribbon is not mechanically held down to the board which makes for a not-so-robust environment where snagging of the ribbon is an ever present problem and in some cases the ribbon has been known to restrict airflow over the PCB causing heating problems.
Forgoing a lot of details, suffice it to say that significant progress is being made in the area of embedding polymer-based waveguides into chip carriers and the PCBs themselves to guide light around the board, much like what is being done to guide light through waveguides on chip. Additionally, there is focus on best methods to cheaply and accurately transmit the light from fiber into these on-board waveguides and from those waveguides onto their counterparts on the PICs. This includes novel structures for passively flip-chipping the PICS onto the carriers or substrates to make a photonic connection (see second figure from the IBM OSA paper).
More recently a new innovation has come out of Petra (Photonics and Electronics convergence Technology Research Association of Japan) called photonic pins that can transmit light vertically through windows in the package allowing transmission of the light from the PCB to the PIC and back (see figure from Petra’s Chip-scale packaging paper). It’s not as easy as you may initially think as there is a considerable difference in the size of the fiber to the PCB waveguide and then from the PCB waveguide to the PIC waveguides. Add to this is the push to move from multi-mode photonics which is inherently easier to manufacture to single mode photonics to enable wavelength division multiplexing. Moving to single mode photonics is challenging on the PCB as the waveguides must be made much smaller and to tighter specifications to confine the light to a single mode. None-the-less, significant progress was reported with one vendor showing a PCB with 16 electrical layers and 4 optical layers (albeit still in multi-mode) and another vendor showing novel connectors that would allow a PIC to be inserted into a chip carrier with ready-made light channels to make the connections to the board waveguides.
Embedding polymer waveguides into the interposers and PCBs promises to remove the need for pigtails, manual connections and flyover ribbons making for much more cost effective integration and leads me back to where I started which is that we are seeing an unprecedented jump in capabilities across the entire eco-system to truly enable “integrated photonics” from the system all the way down to the PICs including connectors, printed circuit boards, chip carriers and packaging.
Intel and the IFS Dilemma: Stuck Between a Rock and a Hard Place – Should They attempt to Sell Intel Foundry Services