Acceptance will require overhauls to PWB and assembly operations.

Optoelectronic interconnect is an alternative to copper that can provide increased bandwidth and other advantages for special applications. For example, OE interconnect does not have the problem of “noise” that copper interconnect can (a significant issue for high-speed communications). OE technology is also appealing for use in aircraft because it eliminates the weight of thousands of copper cables.

Several companies and laboratories are currently working on new waveguide technology; therefore, OE substrate technology is continually changing. However, despite all the good research that has been accomplished, this technology is commercially stagnant. In North America, fiber cable is at the curb, but it has not entered the home, office or technical institutions. In Japan, fiber to the home is growing at a rapid pace.

One reason for the lack of movement is that continued improvements in (less expensive) copper technology have kept pace with circuit bandwidth needs. Good design practice has also helped; however, as the thirst for faster signal processing continues, and home electronics or distribution systems become the workhorse of every household, there may be applications willing to pay for the initial additional cost.

Optoelectronic substrates with embedded waveguides are still years from production, but continue to be discussed and compared with traditional substrates. Many manufacturing dilemmas associated with optoelectronic substrates have to do with the cost and reliability of the optical fiber polymer interface. OE substrates will not become prevalent until the cost-performance benefits are proven (Figure 1).


This article discusses the current state of optoelectronic substrate technology and highlights some key issues surrounding its implementation, based on information from the 2007 iNEMI Roadmap.

Higher data rates to support growing bandwidth requirements are certain to continue, and electrical transmission of signals will, presumably, soon run up against its limits. Telecommunications systems appear to be the primary driver for OE interconnect technology. There are currently optical wide area and local area networks using fiber-based OE technology for infrastructure and hybrid fiber/organic substrates in the supporting backplanes. Today’s systems are generally operating at 10 Gbps, with 40 Gbps coming soon and 100 Gbps being discussed.

Signal conditioning technology has achieved bit rates on copper of 10 Gbps over high-performance and FR-4 boards. While this permits electrical 10G line-speed on the backplane, it also adds cost and complexity. Furthermore, power dissipation is increasing and edge density is limited. So optical interconnect solutions are still promising, but cost and technology reliability will determine the breakpoint.

For future generations of data and telecommunication, there is growing demand for higher data rates and increased performance. For telecom PCBs, there is a growing need for better base materials and circuit board technologies for transmitting high-speed signals. It is clear that further advances in speed and bandwidth can be achieved only through new optical technologies for board-to-board and chip-to-chip interconnection on board.

The situation today can be described as follows:

• Increased data rates to support growing bandwidth requirements are certain to continue.
  
  
• The timing by which applications will convert to optical technology remains unclear, but not expected to happen within iNEMI Roadmap’s 10-year horizon. The optoelectronic substrate technologies that will support the applications are also unclear.
  
  
• Some experts believe that the waveguide needs to be embedded or laminated between conventional base materials; other experts are developing waveguide technology external to the PCB.
  
  
• Optical interconnect is expected to compete with copper interconnect technology for backplane and daughtercard applications where data rates are 10-15 Gbps and higher.
  
  
• Both electrical and optical technologies suffer from signal attenuation and degradation problems that can be improved through circuit design and materials.

Optical Substrates Interconnects

Optical interconnects are in use today. Optoelectronics are currently used as the backplane interconnect if there is some architectural reason the signal needs to remain optical as it goes board to board, or if the system is distributed and a distance (typically >1 to 10 meters at 5>5 Gbps) exists between connections.

Optical interconnections used in backplanes are currently fiber-based and exist as separate physical layers from the electrical backplane. The mechanical connection of the optical interconnect off the line card is done through a cut-out in the electrical back panel with an adapter placed in the cut-out. Optical jumpers or circuits are then plugged into the adapters to create the fiber connections in the backplane (the connections between cards). Issues with this type of interconnect include difficulties with cleaning and inspection, difficult fiber routing/handling, and high cost.

Several different types of optical interfaces between OE components and circuit boards are being developed:

1. The optical path on the PCB or backplane. In current backplane technology, the optical path is generally provided through use of optical fiber loops linking components to connectors or other optical or optoelectronics packages. The main issues with this approach are that it is impossible to perform any signal manipulation and is difficult to achieve high interconnect densities because of the limited bend radius of the optical fiber. Also, difficulties in manufacturing and handling make this a costly and often low-yield approach. Because of limitations of radius bending, this technology is limited to large boards (backplanes).

2. Optoelectronic module (component) connection to optical board (PCB) with integrated waveguides. Two coupling methods are being considered. The first is “free space” (without waveguide) interconnection using micro lenses and special connectors; and the second is direct butt coupling. Direct butt coupling takes advantage of in- and out-coupling without additional micro optical elements, such as lenses and mirrors. The VCSEL-arrays/PIN-diode-arrays have to be positioned directly in front of the waveguide end. On the other hand, there are thermal and alignment problems and the modules cannot be assembled using surface mount technology processes (Figure 2).


3. Guided wave, 90° beam deflection. Out-of-plane light deflection (Z-direction) can be accomplished using gratings (incoupling) or mirrors. A number of publications have demonstrated mirror fabrication by cutting the end of the waveguide with a dicing saw, wet chemical etch, or laser to create a 45° facet. The facet can be metallized to improve the mirror’s reflection properties. Mirrors have the advantage of being wavelength independent, but the mirror surface’s roughness can cause high losses. Aligning the mirror to the waveguide and active device poses significant problems and will have to meet similar tolerances as required for the transmitters and receivers (Figure 3).


4. Materials and processes. Cost of the materials and the associated processes need to be considered when developing materials to meet OE performance specifications. A number of inorganic materials can be used for embedded optical interconnects including Si/SiO2, silica, and glass sheets. Of these, only glass sheets have been applied to embedded waveguide technology for PWBs (Figure 4).


SiO2 and silica materials are used in device applications and are processed directly on silicon wafers. Waveguides are formed through a combination of lithographic printing, etching (wet chemical or laser) or ion implantation.

The material of choice, whether inorganic or organic in composition, needs to be compatible with the board manufacturing environment and equipment. The material will be exposed to various chemicals and temperatures, handled most of the time outside a cleanroom environment and cannot be sealed hermetically. It should also be compatible with standard processing (lamination, etching, drilling, soldering, etc.) techniques.

Any material introduced as a waveguide solution for PWB interconnects needs extensive reliability testing, including failure analysis. Similar requirements extend to the electro-optic hybrid laminates and assembled boards (Figure 5).


Techniques used for polymer waveguide structuring include hot embossing, photolithography and laser writing. In particular, for medium-sized boards, and if large amounts of waveguide foils for mass-production are desired, hot embossing seems the most promising technology. However, for waveguides up to 1 m as needed in backplanes, lithographic or laser-writing techniques appear most appropriate. Recently, a working EOCB with hot embossed planar polymer waveguides was demonstrated.

Optical PWB Manufacturing

The integration of optical components will significantly impact the PWB operating environment. Most PCB manufacturing facilities have a limited or “selective” focus on operating environment. Most shops’ imaging areas contain a Class 10,000 cleanroom for exposure operations. Some have improved areas for optical testers and laser drilling.

The PCB fabricator will need to change significantly to integrate OE. This same degree of change will also apply to PCB assembly operations. Waveguide processing will require extremely clean manufacturing environments, and optical component assembly will require precision micron-placement techniques. Both will require improved temperature, humidity and cleanliness requirements. Once boards are assembled, product cleanliness needs to be ensured by active means such as air filters, or passive protections such as connector doors and dust seals.

Emerging Technologies

Several emerging technologies may impact optical interconnects and OE components. These include photonic crystals or optical band gap materials, complex gratings and new material systems. The list is growing continually as more R&D effort is put into OE. It is impossible to predict which technologies will be pervasive in the future. Table 1 summarizes some of the promising technologies being developed that might “disrupt” our current vision and solve the challenges previously discussed.


Conclusions, Gaps and Needs

OE technologies are used in mainstream, high-data-rate applications when they offer lower cost than the alternative of using copper (or sometimes wireless) methods. Since optical technologies are newer than copper methods, their cost is falling faster. As the cost of optical transmission drops, the distance at which optical methods are more economic becomes shorter, and today, optical methods are sometimes more economic for high data rates over distances of 10-100 m. In the future, optical methods are likely to be used at even shorter distances, as demand for high data rate transmission continues to grow. There is increasing interest in optical chip-to-chip connection on the system board, to overcome bandwidth bottlenecks between the CPU (clock speeds up to 10 GHz), and the main memory or I/O bus (running at hundreds of MHz).

The following enablers are needed to make OE a mainstream technology:

Laminated and embedded waveguide interconnect development for high-speed optical backplane and chip-to-chip applications. In addition to the need for optical interconnect, transmitter (VCSEL) to waveguide to receiver coupling and small radius 90° bends will be required.

Improved VCSELs for 1310 and 1550 nm transmission are needed, with sufficient reliability for thousands to be used in systems that have >15-year lifetimes.

Outsourcing to EMS companies will lead to wider dissemination of closely held package, assembly process and test knowledge.

• Network service providers require rigorous reliability and testing to telecom standards, such as Telcordia GR1221, a major impediment to acceptance of lower cost “datacom” components. There is a real need for “lite” standards, suitable for products with 10 to 15 years’ field life. Examples include use of epoxy adhesives inside OE packages and non-hermetic (but acceptably impermeable) materials, such as LCPs, which have been used in electronics packages to MIL spec for years.
  
  
• Measures of the current and future cost of OE interconnect vs. copper, as a function of application, data rate and distance, are badly needed. The cost to provide data rate drives OE. We were unable to find clear measures of the cost of OE versus copper, except at higher levels, such as long-haul telecom systems. Cost details at level 2 would be helpful to determine economic gaps that OE technology might address and would provide a cost-performance basis for the technology roadmap.
  
  
• Emerging technologies such as holey fibers, photonic band gap materials and solutions hold promise of alternative methods of transmitting and controlling optical signals in ways that may someday be commercially important. OE technologies are emerging continually; we need to watch for important developments and keep an open mind.

Ed.: This article was originally published in Printed Circuit Design & Fabrication in September 2007.

Jack Fisher is a consultant with Interconnect Technology Analysis (.i-t-analysis.com) and chair of the 2007 iNEMI Roadmap organic interconnect chapter; fish5er@mindspring.com. For information about the iNEMI Roadmap, visit inemi.org/cms/roadmapping/2007_inemi_roadmap.html.