Test AOCs during design and production
As communication data rates increase, maximum propagation distances over copper cable decrease. This trend is driving the use of fiber-optic links to shorter distances. Already well established in the telecom and datacom markets, optical fiber is now poised to find applications in the consumer and industrial markets.
Consumer protocols such as USB and Thunderbolt are achieving data rates of 10 Gbits/s today. Thus, the reach of traditional copper interconnects has become limited to a few meters. Optics can remove the distance limitations and enable longer-reach applications with a thinner and lighter cable. These data rates are more than adequate for most applications in the consumer and industrial space because their reach ranges from a few meters to tens of meters. Active optical cables, which maintain electrical connections, are ideal to address the higher-speed longer-reach portion of the consumer and industrial market.
To be successful in the consumer or industrial market, optical cables need to be robust—both optically and mechanically—and low cost. The various elements of the optical link—including the fiber, the cable, the coupling optics, and the optoelectronics—can be engineered to work together at a system level to meet those requirements. Testing and analysis done at critical junctures during the design process can further reduce costs by optimizing the design for high yield given specified manufacturing tolerances. Additionally, testing during manufacturing can catch any defective parts early to reduce cost of materials and assembly.
At Corning, we developed the ClearCurve VSDN fiber and we perform optical testing in-house. Optical systems (the optoelectronics, coupling optics, and fiber) at minimum have to be sufficient to meet the electrical standards imposed by consumer specifications. We test for optical power, modes, and noise and correlated those results to tests of the electrical signal. Therefore, we can impose a minimum standard for power and noise on the optical path despite the overall system (e.g. USB and Thunderbolt) recognizing only electrical signals. The optical parameters are also tested throughout manufacturing and production. Key tests catch any issues on the manufacturing line while also minimizing overall costs.
While the data transmission performance of optical links is still important in consumer and industrial applications, the prerequisites for the market—size, robustness and low cost—require certain tradeoffs during design. Testing during the manufacturing process and during system design further ensures each element of the design performs to spec despite manufacturing variability, which reduces waste and cost.
The fiber
To be widely accepted in the consumer market, optical solutions have to be small, rugged, and low-cost while still maintaining a minimum optical power throughput or maximum optical loss (i.e. the “optical link budget”). Mechanical robustness of an optical fiber can be improved by reducing the overall fiber diameter as it reduces bend-induced stress. The fiber diameter is made up of an inner core and an outer cladding. Reducing the fiber diameter effectively improves the mechanical robustness without having a strong impact on the optical throughput.
Similarly, the core diameter can be reduced to slightly improve the optical bend loss by increasing confinement of the optical signal, but at the severe penalty of reduced ease of optical coupling. As an extreme example, SMF (single-mode fiber) with typically an 8 µm core would need 10× or better alignment accuracies to maintain the same optical coupling compared to MMF (multimode fiber) with typically a 50 µm core. SMF can go very long distances with nearly zero optical loss while MMF is limited to a few hundred meters for the same data rates. Consequently, there is a clear tradeoff between optical propagation distance and ease of optical coupling for a given data rate in the fiber’s design. Tailoring the fiber design helps meet the application needs for cost, bend sensitivity, bandwidth, and distance.
Because consumer and industrial applications cover distances of less than a few hundred meters, MMF—with better optical coupling and shorter propagation distance—is the better choice. To improve the bend performance and optical coupling for these applications, Corning developed the ClearCurve VSDN fiber. This fiber has a reduced diameter from the typical 125 µm to 100 µm to mechanically provide a bend radius as low as 1.5 mm.
To achieve low optical bend loss with this new radius, we developed a fiber with an unusual refractive index profile. The ClearCurve series, including VSDN, has a low-index "trench" around the core, which reduces the optical bend sensitivity. These two attributes together allow for the 1.5 mm bend radius.
To also improve the optical coupling performance, the core diameter was increased from the typical 50 µm of a standard MMF to the 80 µm for VSDN. While that increase leads to a slight reduction in the maximum distance of propagation, it significantly improves the misalignment tolerances allowed and thus improves the manufacturing costs. (Figure 1 compares the attributes of VSDN fiber to those of standard MMF).
Figure 1. Comparison between Corning’s ClearCurve VSDN fiber and standard 50 µm multi-mode fiber.
With the increased core diameter and the index contrast for the ClearCurve VSDN fiber, the loss is reduced to 1 dB at a bend radius of 1.5 mm. The ClearCurve VSDN fiber was able to achieve error-free transmission of up to 10 Gbits/s over 100 m. Figure 2 shows different fiber profiles and their corresponding loss as a function of bend diameter.
Figure 2. Index profile (a) engineered for low bend loss (b).
Figure 3 compares optical eye diagrams of 10 Gbits/s after transmission through 3 m and 100 m of VSDN fiber showing that it is possible to achieve minimum dispersion-induced signal degradation.
Figure 3. Eye diagram of 10 Gbit/s optical signal after propagation through 3 m (left) and 100 m (right) shows virtually no degradation.
The cable
The cable provides another opportunity to improve on robustness. With a fiber designed for robust optical transmission and mechanical flexibility, the protective cable was then designed to allow for bends as small as 1.5 mm radius. The cable cross-section (Figure 4) shows a slotted interior design that protects the fiber while the cable remains thin and flexible to allow for inadvertent pinching.
The slot lets the fibers move freely, which minimizes their degree of bending. As a result, the cable can be bent back on itself at 180 degrees with a nominal zero bend radius while keeping the fiber radius of curvature greater than 1.5 mm, preventing breakage or significant optical loss. Thus, the cable can be pinched without breaking or a loss of performance. During manufacturing, occasional checks of the fiber and cabling are performed to ensure it remains within specification of the design.
Figure 4. A slotted design (cross-section, top left) of a VSDN cable lets fibers move freely, allowing the cable to be tied into knots (right) and even pinched (bottom left).
Coupling optics
One challenging aspect to optical systems in consumer applications is the small packaging requirements. The fiber, fiber connection mechanism, any coupling optics, optoelectronics, and electronics all have to fit in a small space defined by the USB or Thunderbolt specification. Compared to electrical cables, active optical cables have a cost and size disadvantage because they require additional components such as the fiber connection mechanism, coupling optics, transceiver, and optoelectronics (VCSELs and photodiodes). For cost considerations, the VCSEL (vertical-cavity surface-emitting laser) is a better choice than an edge-emitting laser in part due to the optical profile, but mainly due to the cost of the components. Consequently, the design of the optical system was defined by using a VCSEL and the VSDN fiber.
There are two methods to connect the fiber to the VCSEL or PD (photodiode), one that requires an electrical turn and the other an optical turn. For USB 3.1 Gen1, with speeds up to 5 Gbit/s, an electrical turn enables the lowest cost design. At 10 Gbits/s Thunderbolt and USB 3.1 Gen2 requires a slightly costlier design that utilizes an optical turn. With either implementation, the biggest constraint is fitting within the low headroom of the cable end. An electrical turn, as implemented in USB, provides for a lower cost option due to the lack of coupling optics needed with direct coupling between the fiber and the VCSELs/PDs (Figure 5).
Figure 5. Optics add to the cost of a cable as opposed to electrical connections.
Direct coupling typically requires mounting the VCSEL and PDs on a plane perpendicular to that of the PCB. Due to the low headroom within the connector, this design imposes a certain electrical path distance between the photodiode and the TIA (transimpedance amplifier), increasing the amount of noise on the signal. One solution is to mount the TIA on the same vertical substrate as the photodiodes, though space constraints typically prevent such a solution. In addition to the noise from the electrical turn, as the data rate increases, the PD aperture becomes smaller, which degrades the SNR (signal-to-noise ratio) to the PD itself (reduced sensitivity). Less light detected through the smaller aperture using direct coupling reduces optical power. For these reasons, the direct coupling approach is typically limited to lower bandwidth or data rate communication links with speeds under 10 Gbits/s.
Direct coupling has the potential to allow for large misalignment tolerances if a relatively significant amount of optical loss is acceptable. Figure 6 shows the fraction of light coupled into the fiber as a function of the misalignment between a VCSEL and a ClearCurve VSDN fiber core center. The longitudinal distance between them was approximately 150µm. The full width at half maximum of the coupling curve is around ±40 µm—essentially the radius of the VSDN fiber. While direct coupling is forgiving to large misalignments if larger optical loss is acceptable, the technique works best at relatively lower data rates (less than 10 Gbits/s) where photodiodes have large apertures and higher sensitivity.
The addition of focusing optics can achieve better nominal coupling efficiency, especially at higher data rates (Thunderbolt, USB 3.1 Gen2) where PDs have a smaller aperture and reduced sensitivity (i.e. more light is needed to detect the signal over the higher noise floor). A low-cost option is a precision molded lens inserted between the fiber end-face and the VCSEL/PD. The lens reduces optical coupling loss by focusing the light from the VCSEL into the fiber or from the fiber onto the small aperture PD. To enable the TIA to be in proximity to the PD, a turning mirror, such as a TIR (total internal reflection) surface, can also be included in the molded lens module to effectively change the light direction from the fiber axis onto the electronic board. These modules then let VCSELs and photodiodes be located very close to the transceiver to further improve the SNR. While the module increases cost, using the module to both couple light to multiple fibers and to provide optomechanical fiber support makes the overall solution more cost effective. The module includes lenses to focus the optical light, a TIR mirror to steer the light to/from the VCSELs and PDs on the PCB, and alignment V-grooves to hold the fibers.
Optical design plays an important role in lowering assembly costs. For Thunderbolt, the lens module was designed to maximize alignment tolerances (Figure 7).
Figure 7. Measured misalignment tolerances with VSDN fiber using the optical module: VCSEL side (a) and photodiode side (b).
We performed a Monte Carlo analysis to optimize the module design for manufacturing variation using a 70 µm diameter PD, typical VCSEL, and VSDN fiber. The Thunderbolt design shows that large misalignment tolerances of ±25 µm are achievable due to the large core and numerical aperture of VSDN fiber as well as the design of the lenses and optical path in the optical module.
Designing to meet electronic specifications
Following the verification of design for the fiber, coupling optics, and optoelectronics, the optical link is put into the overall AOC (active optical cable) assembly, which includes VCSELs, PDs, transceivers, and electronic chips to meet the protocol and high-speed signal integrity of the USB and Thunderbolt standards. After integrating into the overall system, we tested and optimized the optical link and we correlated the result to the electronic signal. For example, reducing the power output of the VCSEL or the amplification of the PD signal can decrease overall electrical power consumption of the cable, but can also lead to a degraded SNR in the overall optical link. Such noise can translate into increased jitter in the electrical signal. Temperature can also have a strong impact on the optical performance, mainly because the VCSEL’s optical power output inverse correlation with temperature. Increased power consumption of the cable leads to higher temperatures, thus it is important to optimize the transceiver settings (Figure 8).
Figure 8. Optimizing the transceiver laser drive settings yields the largest modulation depth over the temperature range of operation.
The transceiver provides a number of optimization parameters and it can measure the performance of the optical system. Consequently, we can adjust the optical signal while simultaneously measuring the impact on the electrical signal of the end-to-end cable. Some optimization parameters for the VCSEL and PD include emphasis, bias, electrical swing, and amplification. Adjusting these parameters impacts the optical and electrical signal quality as well as the electrical power consumption of the AOC. Thus, optimizing these parameters lets us meet the USB and Thunderbolt specifications for temperature, electrical power consumption, and jitter requirements. Given variability in manufacturing, the settings for the system must be higher than the minimum requirements in order to maintain high yield.
We performed a number of tests to measure the optical and electrical signals and find the desired operation points. For the optical signal, small signal analysis using a parameter analyzer with two optical ports and two electrical ports let us assess how these tradeoffs affected the maximum bandwidth. The setup would generate an optical signal that we can send into the receiving side of the AOC (PD and TIA) and then measure the electrical output. Additionally, an electrical input could drive the transmitting side (laser driver and VCSEL) and the optical output could be measured (Figure 9).
Figure 9. The test setup for small signal analysis uses a parameter analyzer with two optical and two electrical ports.
For the end-to-end electronic signal, large signal analysis using a bit-error rate tester and a real-time oscilloscope was used to assess how these tradeoffs affected the quality of the electrical eye diagram and error-free transmission. These measurements were performed with a variation of input jitter to understand the sensitivity of the system as well. In this manner, the transceiver parameters were selected to meet the electrical specification while also determining the optical throughput requirements for testing during manufacturing (Figure 10).
Figure 10. The test setup for large signal analysis uses a bit-error-rate tester with a real-time oscilloscope.
Through the manufacturing process
Once the design and parameters were specified, the AOC was ready for manufacturing. It is, however, important to test throughout the manufacturing process, which lets us scrap defective parts early. With this in mind, after assembly of the board, which includes the PDs, VCSELs, and coupling optics (module with the TIR, lenses and v-grooves), the board is tested to meet optical performance metrics using a golden cable assembly. Once two known good boards are verified using the golden assembly, they are assembled together to form the AOC. Prior to placing the mechanical strain relief and overmold that form the end connectors, we test the end-to-end cable again to ensure it was built within manufacturing tolerances. To do this test efficiently, we designed a test box and developed test code to run the cable through a series of electrical and optical tests. Through these tests, we can verify that all lanes operate with correct optical throughput, an open electrical eye, zero bit errors, as well as many other tests specific to the protocol. If the cable passes this series of tests, the overmolds are placed on the end connectors and the final cable assembly steps are completed. The same tests are performed again prior to shipping the cable to the end customer.
As data rates continue to rise, electrical signal transmission over shorter and shorter distances becomes increasingly impractical. As a result, optical fiber is now poised to find applications in the consumer and industrial markets. Protocols such as USB and Thunderbolt, which can achieve data rates of 10 Gbits/s extend the reach of traditional copper interconnects.
Optics can reduce these distance limitations. Although solutions in the consumer and industrial market need to be both small, low-cost, and robust, advancements in fiber optic systems are bringing them closer to cost parity with electrical systems. The entire optical link can be engineered with cost in mind. Testing and analysis done at crucial junctures in the design and manufacturing process can further reduce costs by creating a design that accommodates larger manufacturing variances, making active optical cables a compelling solution for the rapidly-expanding consumer and industrial markets.
About the author
Rebecca K. Schaevitz received her B.S. in Electrical Engineering, Summa Cum Laude, from Tufts University in 2005 and her M.S. and Ph.D in Electrical Engineering from Stanford University in 2008 and 2011, respectively. At Stanford, she pursued her interest in the field of silicon photonics under the mentorship of Prof. David A. B. Miller and funding from the National Science Foundation’s Graduate Research Fellowship as well as an Intel Fellowship. Her thesis topic was on the material properties and device design of silicon-germanium optoelectronic modulators. During her Ph.D, she also interned with Intel to provide understanding of germanium material properties for their silicon photonics program.
Since joining Corning in 2011, Rebecca has contributed to the Thunderbolt program and aided in releasing the first consumer product for Corning in decades. Her initial focus was on the prototype build development, working with a number of contractors to ensure a seamless process flow. After prototyping, Rebecca focused on understanding and testing toward Thunderbolt certification standards by working with Intel and Apple. She helped ensure certification of all cable lengths now available to consumers on Apple.com and other online retailers.
Rebecca is a member of Tau Beta Pi Honor Society for Engineers, Eta Kappa Nu Honor Society for Electrical Engineers and the National Society of Collegiate Scholars. Rebecca was also the recipient of the Tufts University Electrical Engineering Departmental Morris and Sid Heyman Award at matriculation for being the top graduating student in her department.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
