Most PCB designers—except those that work on optical transceivers—are probably not aware of the coming revolution in silicon photonic integrated circuits (PICs), electronic-photonic integrated circuits (EPICs), and greater proliferation of embedded optical systems outside of telecom. Applications outside of telecom (for example, military and aerospace systems) that require massive data transfer rates already make use of fiber optics for embedded computing. Modern data center architecture, advanced embedded systems, advanced sensors, and future automobiles will leverage huge amounts of data moving between different subsystems. This means you can expect optical interconnects to play a much greater role in these systems.
How deep can we expect optical interconnects to penetrate into PCBAs? Currently, optical fibers are accessed through an SFP connector that interfaces with a fiber optic transceiver module. However, research over recent decades has looked at bringing optical interconnects directly into the PCB. Photonic integrated circuits have progressed slowly, but when these product do reach the market, a new physical interconnect architecture will be needed to route optical signals in an assembly without using bulky fiber optics.
As electrical signals switch at faster rates, signal integrity problems such as crosstalk and radiated EMI become more severe, and losses on standard substrates increase at higher frequencies. Replacing the electrical infrastructure in PCBs for networking equipment with optical interconnects relieves many signal integrity problems. With multi-mode fiber, the number of channels can be increased in a single interconnect without increasing routing density. This allows data rates to scale without significantly scaling board sizes or component sizes.
If you think that all this sounds like it came from an episode of Star Trek, rest assured that this technology is closer to becoming commercialized than you may think. Organizations like AIM Photonics are supporting development photonic microprocessors, electronic-photonic integrated circuits are being developed by the research community and private companies, and many in the community have already created proof-of concept boards that include optical interconnects for interfacing between optical and electronic components.
At some point, the greater use of optical signalling alongside standard electronic equipment will take up so much space that placing cables inside a chassis is simply impractical. Think about the space required to form 50 or more optical connections inside a chassis with fiber optic cables, and all without bending cables past the minimum bend radius...it’s simply not possible to meet form factor and performance requirements. This means electronics manufacturers will need to accommodate direct printing of optical waveguides directly on PCBs.
The two best options for optical interconnects in PCBs are to embed glass fibers in the interior layers of a multilayer PCB. The other option is to deposit polymer waveguides on the interior layers or surface layers. Glass fibers could also be placed on the surface layer, but using polymers allows greater control over geometry.
This becomes important for interfacing with optical components as the geometry and coupling optics must be precisely defined on the surface layer. Regardless of which method is used, the design process will not change considerably as optical interconnects do not suffer from the same signal integrity problems as interconnects.
Glass optical interconnects will likely be easiest to integrate into standard multilayer PCB manufacturing processes as they can be embedded in the core or prepreg layers. The right material between FR4 laminates can act as a cladding layer for glass waveguides. There is no reason that glass or polymers could not be used simultaneously; standard glasses for optical fibers can be used in the inner layer, while polymers would be easiest to deposit on the outer layer.
Circuit board with embedded glass optical interconnects. Image credit.
Glass or polymer waveguides placed in the interior layers of a PCB require transmission back to the surface layer and coupling optics for EPICs/PICs, or using a photodetector with fast response time (normally a PIN photodiode). Especially in optical BGAs for EPICS and PICs, an optical interconnect requires some type of coupling optics in the form of 45-degree mirrors. This requires extremely precise micromanufacturing. Otherwise, the laser diode and receiver in an optical interconnect will need to be embedded in the substrate.
Transparent polymers for use in optical waveguides at infrared or visible wavelengths were been demonstrated in the literature as early as 2001. Prior work focused on more general materials development and the use of optical glass in innovative ways. Polymer waveguides are a major advantage over glass waveguides for several reasons:
I think the natural approach with polymer waveguides is to use them on the surface layer so that they have direct access to electro-optical modules and PICs/EPICs. However, they have also been studied for use in the internal layers of a PCB, similar to having an optical interconnect layer made from embedded optical fiber. The image below shows some coupling methods for polymer waveguide optical interconnects in a PCB. These are most likely to be used on a surface layer through an interface with a PIC/EPIC package.
The remaining challenge is mass manufacturing and greater integration of optical components and waveguides into PCBs. This requires scalable printing techniques for manufacturing dielectric waveguides directly on PCBs for interfacing between various optical components, electronic ICs, EPICs, and PICs. Using polymers for optical interconnects directly on an FR4-grade PCB or PTFE layer is preferable as they can be patterned using standard lithography techniques.
As data rates scale higher, these optical interconnects will need to scale smaller in order to accommodate shorter wavelengths, although modal dispersion will become a problem with continued scaling and as more modes are packed into a given fiber. This doesn’t mean that copper will go the way of the dinosaur; copper will still be necessary to isolate optical interconnects, particularly with radio over fiber applications. The research community and some companies were producing proof-of-concept boards with multimode waveguides that operate at 12.5 Gbps and higher data rates. Since 2019, there has been renewed focus on polymer interconnects given the most recent publication I cited above.
These fibers might be found in the core of a multilayer PCB in the future
The huge data rates required in some of the most advanced systems may cause implementation with SFP modules to run up against serious form factor constraints in many different application areas. Other technologies that may leverage integrated optics in the structure of a PCB include:
As the photonics ecosystem develops, it should be clear this list of technologies will surely grow to support implementation on modules and PCBs. It is my opinion that optical interconnect implementation in PCBs will most likely involve formation of polymer interconnects, a technology that has been provide over and over again over recent decades. Now the semiconductor ecosystem needs to catch up and deliver PIC/EPIC solutions.
Anyone that works with optical components and who is looking to commercialize new products with optical interconnects can benefit from the full suite of PCB design tools in Altium Designer. The industry-standard layout and simulation tools are ideal for applications in electro-optical embedded computing and high speed networking. The data management and documentation tools can help you quickly prepare your new designs for manufacturing.
Contact us or download a free trial if you’re interested in learning more about Altium Designer. You’ll have access to the industry’s best layout, simulation, and data management tools in a single program. Talk to an Altium expert today to learn more.