Shielded vs. Unshielded High-Speed Backplane Connectors
By Bob Hult, Bishop & Associates Inc.

In the beginning, backplane connectors had the relatively simple task of providing electrical interconnect between daughtercards. Parallel bus structures simplified the routing pattern, and relatively large connector centerlines allowed a few hundred contacts to fit in conventional card cages. One-piece edge connectors on .156” centerlines slowly evolved into two-piece pin and socket backplane connectors on 0.100” and 2.0mm centerlines to achieve increased density, as well as improved system reliability.

The market for high-speed backplane connectors has undergone a series of transformations over the past 10 years, driven largely by advances in the semiconductor industry. As system speeds began to evolve into the gigabit/second arena, the performance limitations of existing parallel structured backplanes became apparent, and accelerated the transition to low-voltage differential pair signaling.

In some cases, standard open pin field connectors, such as Metral, 2mm HM®, and FutureBus+, were utilized in applications where the required signal density would allow the use of a greater number of ground pins per signal.

These standard connectors had become commodity products in the market and offered the advantages of low cost per mated line, as well as the availability of multiple sources. PCB designers were familiar with their layout patterns, while assembly shops had installed automated tooling which made them the preferred choice wherever possible.

As system speeds continued to increase, signal density began to suffer as a greater percentage of contact positions were assigned to ground. PCB routing of high-speed signals through a “picket fence” of ground pins became more difficult, and in some cases necessitated additional backplane layers that added to the cost. User demand for improved high-speed performance resulted in greater pin counts that created increased mating and unmating forces. The solution was to replace multiple ground pins with integrated ground structures located between single-ended or differential pairs.

The first high-speed connectors specifically designed for differential signaling, such as Tyco Electronics Z-PACK HS3™, FCI Electronics Metral HS®, and Amphenol TCS VHDM- HSD™, offered improved performance and signal density, but they were priced higher than their open pin field relatives, and were only available from a limited number of sources.

These connector families utilize internal metallic shielding structures to minimize crosstalk between differential pairs, establish a return ground plane for power, as well as assist in impedance control.

The general market assumption was that integrated shields were an essential element of high-speed connectors.  Subsequent product introductions from Amphenol TCS (GbX®), Tyco (Z-PACK HM-Zd™), and Winchester (SIP-1000), featured internal shielding structures.

Many of these connectors were initially introduced with a published bandwidth of up to 3.125 Gb/s, with the intention of supporting the emerging XAUI specification. A body of circuit design experience began to be built-up, which allowed the development of successful circuits operating at multi-gigabit speeds.

Over the past several years, the published bandwidth of these interfaces experienced a significant expansion into the 10+ Gb/s arena, enabled by both accumulated channel design experience, as well as major improvements in SERDES performance. Dynamic equalization and pre-emphasis has allowed the discrimination of reliable data streams in noisy channels and pushed the capabilities of every element of a circuit, including the PCB material and trace design, as well as separable connectors.  Shielded connectors continued to be refined and prices remained relatively high. The introduction of the FCI Electronics AirMax VS ® dramatically changed both the performance and price paradigm that had been established.

AirMax VS uses contact placement to create a “virtual shield” to achieve high-speed performance without the use of internal metallic shield assemblies.

Intuitively, one could assume that shielded connector systems have greater bandwidth capability than one without. Grounded metallic shielding structures located between closely coupled differential pairs should provide greater isolation between pairs, resulting in reduced crosstalk, a key performance criterion. This is the assumption that the industry accepted, as high-speed differential signaling became the norm. The actual performance of unshielded connector systems continues to confound conventional wisdom. System designers are divided on the relative merits of both shielded and unshielded backplane connectors, and strong advocates can be found in each camp.

The initial reaction to the AirMax VS was a high degree of skepticism among connector users, fanned by competitive claims that only a fully shielded connector system would be capable of providing multi-gigabit performance. Subsequent evaluation and application experience has largely dispelled this impression. After several years of intense evaluation and use in production applications, none of the initial performance concerns have been proven, and unshielded connector systems exhibit many of the same performance capabilities, and limitations, as its shielded competitors. The ability to maintain a consistent low impedance ground path, as well as impedance control, is influenced more by the PCB footprint and termination method rather than the separable interface itself. Those who have had direct experience with the product largely feel that it has met the performance requirements of their applications to date. Unshielded connectors have been successfully implemented in systems operating in the 3 to 4+ Gb/s range, and are looking at higher speeds for next generation products that will be in production within the next two years. The AirMax VS family has profoundly changed the assumption that internal shielding was an absolute requirement to support multi-gigabit signaling.

Today’s 4 to 6 Gb/s applications are expected to require 8-10 Gb/s within the next 12 to 18 months, as silicon to support these speeds becomes more available. A major objective of designers today, is the ability to develop a system with both scalability and performance headroom. The ability to size a current system to the variable needs of many customers is very desirable. Designing a system backplane today for 6 Gb/s that will allow future upgrades by replacing current daughtercards with 10Gb/s daughtercards in the future can greatly increase the design lifetime of a product family. This approach offers tremendous economic advantages to both the system manufacturer and the consumer.

The success of AirMax VS has resulted in a split market between shielded and unshielded camps, both claiming high-speed performance to 10+ Gb/s. The trademark connector continues to enjoy significant market share and has been specified as the backplane interface for the Advanced TCA platform. ERNI continues to be a viable second source for the Zd connector with the ERmet ZD™ family, as well as their unique ERmet zero XT™ line. Amphenol acquired the entire high-speed product line of Teradyne Connection Systems division (TCS), including VHDM, VHDM-HSD, and the flagship GbX, in 2005.  The infusion of Amphenol TCS technology into the extensive manufacturing resources of Amphenol has elevated Amphenol from being a non-participant, to a major player.  Molex has been an aggressive second source for the entire Amphenol TCS high-speed product line. Each of these connector families utilizes integrated ground planes within the connector.

A significantly reduced price per line has been one of the most dramatic changes brought on by the introduction of the AirMax VS system. Users appear to be satisfied that the value and performance characteristics of this connector family, which has caused the entire high-speed backplane connector segment to reduce the market price per mated pair. Users now expect lower prices, and have also been successful in driving the cost of fully shielded connectors down. The perceived value of AirMax VS is what drove potential customers to evaluate the product, and subsequent proven performance has validated that claim. 

After initial resistance from competitive suppliers, several companies have chosen to introduce their own version of high-speed shieldless backplane connectors.

The Tyco Z-PACK Max™ connector utilizes a shieldless wafer construction, while the recently introduced I-Trac™ family from Molex employs a broadside-coupled differential-pair design without internal shields.

Market prices for both of these interfaces are competitive to the AirMax VS. Marketing battles in this segment are being fought on such issues as connector robustness, ease of PCB routing, compatibility with other complimentary product lines, as well as level of customer support. 

Amphenol TCS is in a unique position, having previously acquired a license from FCI Electronics to second source the AirMax VS connector. They now offer an extensive variety of established shielded and unshielded high-speed backplane connectors.

Many designers report that they have not identified any specific bandwidth limitation of unshielded backplane connectors now on the market. The general feeling is that the products currently on the market, both shielded and unshielded, will be capable of supporting performance requirements for at least the next 2-4 years, and that volume sales will allow these products to remain in the sweet spot at least for this period. Several competitive suppliers are reflecting this belief as they adjust their product mixes. Tyco Electronics has tooled various configurations of their Z-PACK Max connector family, while the first internally developed entry into this market by Molex is the unshielded I-Trac connector.

New backplane systems, both shielded and unshielded, continue to enter the market. The recently introduced XCede™ connector from Amphenol TCS is an internally shielded connector which utilizes several technologies to deliver enhanced bandwidth and claims performance to the 20 Gb/s range.

This is an important target area for next generation equipment, as 10 Gb Ethernet becomes mainstream and designers begin to consider the next step to 100 Gb Ethernet. Designers expect 4 lanes of 25 Gb/s, or 5 lanes of 20 Gb/s, will be the most practical method to deliver 100 Gb Ethernet. Many engineers today feel that bandwidth limitations of unshielded connectors are simply not known at this point, and that more extensive testing and implementation experience will provide the answer sometime in the future.

Another variable will be the need for greater signal density required by future systems. It is possible that more closely spaced differential pair contacts will mandate internal shields to keep crosstalk to acceptable limits, tilting the market away from unshielded interconnect systems.  At some point limitations on molded thin wall sections may preclude the use of separate internal shield structures. This is an area where plated plastic shields or conductive polymers may play a role in future connector design. The need to increase column spacing of unshielded systems is counterproductive to the trend toward greater system density.

As 8-10+ Gb/s systems become more common, it is possible that systems may demand tighter tolerances on crosstalk and impedance control than is common today. These requirements may make unshielded systems less desirable, but it is also unknown if the current crop of shielded connectors will be capable of satisfying these requirements.

Another factor in the connector selection process is that many engineers now feel that connectors have become highly refined, and that the majority of signal distortion can be attributed to the connector launch rather than the connector itself. The connector represents a very small percentage of the total channel path, making the PCB material, trace design, board stack-up, and the connector launch via field, the major culprits.

The ability to directly compare high-speed performance among the leading connectors is currently limited, at best. The market is in serious need of a standardized signal integrity performance test, and verification methodology, which would allow a more accurate and reliable method of predicting component behavior in real world circuits. The lack of such a platform has been recognized as a major pothole in the high-speed connector roadmap. The issue of any bandwidth limitations inherent in any connector system will be determined when neutral test fixtures and methodologies are adopted by the industry, as the published bandwidth numbers currently used to define connector performance are often subject to supplier preferences and individual interpretation of test results.

Bishop & Associates Comments:

• The market for high-speed backplane, mezzanine, and midplane connectors has evolved into shielded and unshielded categories, both claiming performance to support 10+ Gb/s signaling.

• The general perception is that shielded connectors can provide greater bandwidth than their unshielded relatives, but practical application at speeds typical of today’s systems, have proven unshielded connectors capable of acceptable performance.

• Connector users, as well as competitive connector manufacturers, are reluctant to identify a specific level of performance that shieldless connectors will be unable to support.

• The introduction of the AirMax VS connector family has reduced the market price of both shielded and unshielded high-speed connectors.

• The introduction of AirMax VS and its subsequent success in the market has spurred competitors to develop and tool shieldless connector systems, which are promoted to address “cost sensitive” applications.

• Both the Tyco Z-PACK Zd, and Molex I-Trac connectors promote increased robustness as a competitive advantage. The I-Trac design is unique, with a broadside-coupled differential pair configuration.

• The shielded Amphenol TCS XCede backplane connector is the first backplane interface to be promoted as capable of 20+ Gb/s performance.

• Engineers recognize that connectors have become highly refined and contribute a relatively small amount of distortion to the overall channel. The connector launch has been identified as the greatest contributor to signal distortion, and has become the focus of further interface improvement.  Room for improvement still exists in minimizing the stub, intra- and inter-pair skew, signal density, as well as insuring a routable footprint while minimizing PCB layer count.

• At this point, Amphenol TCS, FCI Electronics, Molex, and Tyco Electronics appear to have made a long-term commitment to the high-speed backplane, midplane and mezzanine connector market, with additional suppliers such as 3M, ERNI Electronics, and Samtec continuing to support products in this arena.

Click here to provide your feedback on this article.                            

Robert Hult
Director of Product Technology, Bishop & Associates, Inc.
Robert Hult has been in the connector industry for over 36 years. Hult began his career as a sales engineer for Amphenol. He joined AMP in 1972 and served in several management positions through 1996. In 1997, Hult joined Foxconn as group marketing manager for Intel, Chandler, Arizona, U.S.A. Prior to joining Bishop & Associates, Hult was the regional application engineering manager for Tyco Electronics.

Hult graduated in 1968 from Bradley University with a Bachelor of Science degree in electronics technology and a minor in business.