With the advent of digital high-speed serial communications, the development of connectors for such applications has evolved into the design of very carefully crafted transmission lines. Where in the past the electrical performance of a multi-pin connector usually referred primarily to its current-carrying capacity, now such connectors are characterized in terms previously used almost exclusively within the microwave community. Characteristics such as return loss, insertion loss, and S-parameters are now routinely considered within the digital signal integrity community when discussing the performance of a high-speed connector.
A connector is but one passive component within a communications link, but can become a critical factor as speeds increase. The connector system's design and its interface to the PCB is the juncture that is important to assure that the highest level of signal integrity is maintained in high-speed datacom and telecom applications, in that signal loss and cross talk must be kept as low as possible.
To achieve the required performance, most backplane connectors have been constructed as either stripline or microstrip transmission lines. Such constructions consist of two contacts acting as a differential pair with a vertical groundplane or shield on one or both sides of the pair. The shields are required to maintain the differential impedance as well as to minimize noise by providing shielding between different sets of differential pairs in adjacent columns within the connector.
The use of shields in the connector can present some disadvantages, however. The most significant is the increased cost for more material as well as the cost of assembly associated with shielded connectors. With high-speed connectors, the assembly can be quite critical because the location of the shields relative to the signal pairs is a critical dimension and must be held very tightly in order to maintain the correct differential impedance.
There are two types of connectors — those with metal shields and those without. By removing the metal shields, the performance and flexibility of the connector is vastly improved.
In the older, shielded technology, the differential pairs are usually embedded in plastic by an insert molding process. The shields are then positioned and attached to one or both sides of the plastic. In most cases, each column of differential pairs will be constructed separately with a shield on one side. When the columns are juxtaposed, the shield from one column also serves as the second shield to its neighboring column; hence a stripline geometry is achieved for all except one of the end-most columns. With one shield missing, this end-most column usually suffers from an impedance that is far higher than nominal, and perhaps cannot adequately carry high-speed signals. Even the internal columns will typically suffer from an impedance mismatch when the signal passes through the connector-mating region.
This occurs because in order to maintain controlled impedance, it is necessary to enforce uniformity in the cross sections of the conductors and between the shields. In the contact region, however, it is usually necessary to change this geometry for the purpose of making some type of mating beams. The inclusion of more conductor material in this region increases the cross-sectional area of the conductors as well as the surface area facing the shields, and the impedance in this region will dip much lower than nominal.
In response to this challenge, technology has been developed in which a structure similar to a coplanar waveguide can provide excellent impedance control, low loss, and low crosstalk, and has the additional advantage of not requiring vertical shields, thus lending itself quite easily to implementation within a low-cost, high-speed backplane connector.
Those that have not yet converted to the benefits of shieldless technology point to non-issues such as lack of robustness or ineffectiveness for single-ended signals. In reality, shieldless connector technology is a win-win situation — a win due to greater flexibility of design options and a win due to the proven electrical performance for both differential pair and single ended signals. High speed shieldless connectors have been in the marketplace for over two years with a quality record that is outstanding. Beginning with the premise that the shields can be eliminated, a cost advantage is immediately realized because the absence of shields reduces material usage and simplifies assembly.
Advanced Design
Some have described this technology as shielding done via air. While air is used as the dielectric in this design, air is not achieving the shielding. At the core of this technology, is the AirMax VS connector system from FCI, an insert molded leadframe assembly, which supports both differential pair and single-ended signal lines or a mix of both using air as the dielectric both between signal paths within an IMLA column and between adjacent IMLAs. The contact design and the edge coupling technology, concentrating the fields in the air gaps within the IMLA as well as the staggered assignment — ground-signal-signal-ground versus signal-signal-ground on the adjacent IMLA — provide virtual shielding.
The use of air as opposed to plastic provides additional flexibility. There's plastic in every connector, but some need to add metal shields and some do not. Other connectors may now use AirMax's edge-coupled technology and stagger, but they don't utilize air. By continuing to use plastic, those connectors trade extra performance for a perception of robustness. Suppliers claim this design is better in harsh environments, but it's not.
Some may claim this solution's virtual ground and noise cancellation doesn't work for single-ended signals, but the technology is being chosen by datacom providers precisely because it does. The gap design does concentrate even the single-ended fields.
Flexibility
One of the greatest benefits of shieldless connectors is the signal density and increased flexibility available to system designers when unencumbered by the additional weight and space requirements of shields. The IMLA design offers more versatility in structuring backplane signal routing, as it has been adapted into coplanar, right-angle, mezzanine and BGA versions; offers 3-pair, 4-pair or 5-pair options; and provides 2 mm, 2.5 mm and 3 mm spacing. Because system designers can mix differential pair and single-ended signals within one connector, systems can grow from 2.5 Gb/s to 25 Gb/s without necessitating the redesign of the basic platform. Because of this, the datacom market has incorporated the technology into server and storage applications.
This makes the technology an ideal solution in designing multi-service switches, particularly in regard to headroom. It eliminates the need to redesign the system when migrating to much higher speeds and provides backward compatibility.
Cost Savings
While most advanced technology requires customers to pay more to get more, removing shields from connectors saves money in a number of ways. The improved electrical performance and scalability allows the connectors to be designed in now without concerns about expensive re-designs and upgrades later when faster speeds are required.
Additionally, shielded connectors have built-in additional costs due to the additional material required to provide the shields. And the shields are terminated to the ground plane, which means additional vias to the board, adding to overall cost.
As stated, the flexibility of shieldless connectors allows designers to choose column spacing greater than 2 mm, thereby allowing routing of two sets of signals through a channel and lowering the PCB layer count.
Reliability
Shielding does not improve long-term reliability. Most high speed backplane connectors have two points of contact on the same side of the header blade. The separable interface within the AirMax VS connector contains a mating interface with two opposing receptacle beams, offering two points of contact, one on each side of the blade. A contact system with opposing forces optimizes reliability during exposure to shock, vibration, and elevated temperatures.
A high-speed signal traveling through a contact interface with same-side beams may be interrupted by vibrations, such as those present inside most types of machines, due to both beams lifting from the header blade simultaneously. Such disruptions could wreak havoc depending on the type of signal being transmitted. A contact interface with opposing forces is less likely to lose both points of contact at the same time during exposure to mechanical vibrations.
Backplane connectors that utilize a mating interface with both points of contact on the same side of a mating header beam rely heavily on their plastic components for structural rigidity to maintain proper normal force throughout the product's life. Because plastic will creep under load it is likely that the long-term normal force could be compromised leading to reliability concerns. In some applications connectors may exposed to high temperatures that could accelerate a normal force reduction because the mechanical rigidity of plastic is reduced at elevated temperatures.
However, connector systems that employ a mating interface with opposing beams rely on stamped metal to maintain normal forces. Plastic components are not exposed to constant stresses in this design configuration. Because mechanical properties of metals are far superior to those of plastic materials, the long-term reliability of this type of mating interface is preferred due to stability of normal forces exerted by receptacle beams.
Durability
Shieldless connectors are not subject to pin-bending and dust contamination. Shields do not protect the pins, but exist to affect signal integrity. Shieldless connectors are robust and durable because they exhibit very low mating forces, especially in high pin-count applications. This can be attributed to the absence of shields, which greatly contribute to the mating forces. If the card sliding into a backpanel has a lot of pins, higher force is required to insert the card into the system, and there is a greater risk of damaging the pins (and the connector).
A by-product of the mating force issue also affects the board design because of the need for levers on the card. Applying the force is most commonly done by two levers installed on card; the more force necessary, the bigger the levers, and the bigger the levers, the more space they take up on the card and the less space available for other components. The lower the mating forces, the better. If you compare shieldless connectors to those that are shielded, mating forces are significantly less in connectors without shields.
John Burkett is Senior Product Manager for FCI; john.burkett@fciconnect.com.