Great Connector Inventions: High-Speed Performance without Ground Planes
FCI took the next step in the development of high-speed backplanes when it developed a cost-efficient way to deliver higher speed and density.
While Teradyne and its partner Molex were increasing speed with successive generations of large, expensive, heavily shielded waferized connectors, FCI was listening to its large base of telecom, storage, and low-end computing customers who were asking for solutions that offered greater density and speed, but not at the price premium demanded by the TCS designs. Most of these customers were using 2mm HM and 2mm Metral designs that were under $0.12 per mated differential pair and were not willing to consider the highly engineered TCS solutions that sold for upwards of $0.25 per pair.
Clever FCI engineers, including Danny Morlion, Ab van Zanten, Clifford Winings, Joe Shuey, Stefaan Hendrik, Steffan Sercu, and Steven Smith, combined a lot of creativity, brainstorming, and intellectual horsepower to invent a connector that did not need ground planes and stiffeners. Joe Shuey, engineering leader at the time, said the “inspiration was the twisted-pair wires used in the telecom industry. We realized that by placing conductors in the right position, the electromagnetic fields were tightly contained, minimizing crosstalk without any shielding.”
Achieving 100-ohm impedance on a 2mm pitch required the introduction of lots of air, widening signal conductors, and minimizing plastic. They did experiments with prototype wafers on a micrometer stage to finely tune the offset between adjacent wafers to achieve the minimum possible crosstalk. Prototype wafers to prove the concept cost a grand total of $10K.
To create the offset signals, the team created A and B wafers that located signal and grounds in adjacent columns with the optimum offset, enhancing signal integrity without adding complexity. They called this product AirMAX and promoted the benefits of separating signal pairs, not by ground structures, but by using air separation to achieve similar results. Air has the added benefit of a low dielectric constant (0) to enable connectors with low attenuation.
This was a market-disrupting development, forcing all the backplane connector companies to reassess their thought processes and priorities.
By this time, FCI had acquired Berg. The strategic decision was made to take its competitive cost advantage (half the metal and much simpler assembly) to position this product at about half of the price of GbX connectors. This was a delightful development for FCI’s telecom and storage customers, who were able to get performance at a reasonable price point. FCI further decided to make small modules like they had with Metral. This enabled rapid introduction and scale-up for the new product. The headers, by the way, were dead-simple open pin-field headers, just pins in plastic.
FCI made the interesting design choice of offering backplane vertical receptacles that mated to right-angle headers on daughtercards as the primary AirMAX configuration. Since it had a right-angle receptacle as well, it could offer a very cost-effective coplanar RA-male-to-RA-receptacle connection that quickly became popular for cost-sensitive applications in computing and storage. Other connector companies with shielded wafers found it quite expensive, both in tooling and in piece-part costs to make right-angle males, which gave FCI an even greater cost advantage for coplanar interconnects.
The AirMax was designed into a new storage-controller-board standard architecture called Storage Bridge Bay, a standard specifying controller board-to-backplane interconnects. This standard took advantage of the small AirMAX blocks for signals, power blocks, and guidance.
A similar architecture has been adopted by the Open Compute Platform (OCP) sponsored by Facebook. This standard uses coplanar mating configurations to mate small 1U servers to the coplanar midplane that manages power and communication connections to cables. This configuration significantly increased the compute density in server racks and was ideal for front-to-rear cooling strategies that now dominate the cloud-computing world. These architectures do not use conventional backplanes but do consume a lot of simple connectors.
FCI was able to update the AirMAX platform to handle speeds greater than 12.5Gb/s while maintaining the air dielectric shield-less approach. The Molex Impact system introduced later is also a shield-less system.