Fieldbuses and I/O
Challenges in Providing Optimal I/O Solutions for Small Systems
As systems get smaller yet more powerful and feature-rich, the challenge of bringing all the I/O out of the box increases. Various methods are available to bring that I/O out of the system despite SWaP challenges and reduction in number of slots, all while retaining the use of military-style connectors.
RAM RAJAN, ELMA ELECTRONIC
Page 1 of 1
The most obvious challenge in bringing I/O out of electronic packaging is the lack of real estate. The same powerful processors that are enabling systems with smaller footprints are also supporting greater numbers and types of I/O from each card in the system. Thus, density is being driven both by advanced silicon that can handle more inputs and by the smaller form factors that the more powerful processors are allowing. This is especially challenging in rugged defense environments, where large military-grade connectors are still necessary to meet human factor requirements and to withstand the rough environment. There are multiple ways of addressing these I/O challenges. An overview of considerations for different cable types is presented in the sidebar titled, “Cable Type Considerations.”
Five general approaches are regularly used to bring I/O out of embedded systems based on standard backplane architectures:
• Hard-mounted I/O connectors incorporated into the backplane itself that support all types of discrete custom cables as well as all the standard commercial cable connectors.
• Special I/O slots that bring the necessary signals to a faceplate, much like other plug-in cards.
• Sidewall I/O panels that support multiple connectors and are connected to the main backplane with discrete wire cables and or flat flex cables.
• Rigid-Flex-Rigid (RFR) assemblies that support multiple connectors at the enclosure sidewall on rigid printed circuit boards. These are similar to the previous approach, but they are actually extensions of the backplane and manufactured at the same time, with a flexible multilayer section that allows the I/O panel to be positioned at a different angle than the backplane itself.
• Connectorized (Direct Plug Interconnect) risers and mezzanines that support multiple connectors at the enclosure sidewall, as in the previous two approaches but are directly attached using board-to-board connectors to one of the edges of the backplane, or even on the rear of the backplane as a mezzanine.
Hard-Mounted I/O Connectors
In this method, individual bulkhead connectors are mounted to the wall of the enclosure and multiple discrete wires connect it to one or more cable connectors mounted directly on the backplane. This method includes cable connectors that are designed to plug directly into the backplane slot connectors from the rear. Such cable connectors exist for most standard backplane architectures such as VME, CompactPCI and even the ATCA architecture. The VPX architecture has a particularly dense backplane cable connector system that supports very high speed signals. Hard-mounted I/O connectors also include flexible coaxial cables, semi-rigid and rigid coaxial cable. Fiber optic cables are also included in this approach.
The challenge with RF cables is that coaxial cables require a bend radius for the cable to reach the sidewall. Right angle connectors will help in some cases, but if there are too many connectors arranged in a small area, right angle connectors cannot be used. Insertion loss numbers need to be very low with RF cable, so this determines the type of cable selected. When using rigid or semi-rigid coaxial cables in a chassis, 3D modeling is required to determine the actual layout of all cables needed for a chassis. The cables need to be laid out in a manner that can be converted into a 2D drawing for each cable assembly. Tight tolerances are needed for this application, as are special manufacturing sequences when installing these types of cables into a chassis. This can, however, be an expensive solution, particularly in the prototyping stages.
Optical cabling often needs special trays to support the cables within the enclosure. In the case of fiber ribbons, consideration must be given to providing additional fibers for future repair and replacement. One of the important considerations for fiber optic I/O cabling is the minimum allowable bend radius. Fiber optic signal integrity can be affected if the bend radius is too small. Although there are some new fiber cables on the market that allow much tighter bends, a good rule of thumb is the “soda can” rule, which suggests a bend radius of no less than 35 mm, equivalent to the curve of a typical soda can.
Hard-mounting connectors and cables is a very flexible approach that works well in low volume applications. There are probably at least one or two such cables in any enclosure system, even if another approach is the primary solution. This sort of approach, however, presents several serious issues. Having a large number of separate cable assemblies is labor-intensive to assemble, test and troubleshoot. Each assembly requires its own test fixture and documentation, and if tested with standard bench equipment, highly trained technicians are essential. In addition to the labor cost associated with assembly, testing and installation, discrete cable assemblies are a known point of failure because over time inevitable movement of the cable bundles can result in failure at the termination point. Enclosures filled with a maze of such cable assemblies can be a nightmare to debug because technicians can inadvertently damage cables while trying to reach other cables. There is a growing need for EMI filtering, which creates additional signal integrity challenges, and using cable mounted filters to route high-count I/O out of the chassis requires a considerable amount of space.
Special I/O Slots for Cards with Faceplates
This is one of the most straightforward solutions. It involves dedicating one or more backplane slots for plug-in cards that are dedicated to bringing I/O to or from all the other cards in the system through the backplane. In typical enclosure systems designed for field or vehicle deployment there is almost never space behind a backplane for rear transition modules (RTMs). Special I/O slots for front removable I/O cards can serve the same purpose as RTMs and work in much the same way. As easy as such cards are to design, they have some serious limitations. First, in most such systems the front panels of cards are behind protective covers, which eliminates any possibility of bringing I/O off the faceplates. In addition, backplane slots are typically too valuable a commodity for any function that does not have to be removable. The primary drawback is related to the issue of front covers. The types of rugged circular connectors used for I/O in these systems are too large for typical front panels, and are designed for panel or bulkhead mounting and not for right angle faceplate applications. The lack of appropriate connectors for faceplate applications is almost certainly because the faceplate of plug-in cards is almost never the desired surface for I/O in deployed rugged systems.
Sidewall I/O Panels Cabled
These consist of three solutions that all support an integrated I/O panel for sidewall mounting. The only difference between the three approaches is how they are connected to the rest of the system. We will first address an I/O panel that is connected by removable cables to the primary backplane. There are a number of advantages to mounting all the I/O connectors on a rigid printed circuit board. This is why this feature is common to all three approaches. An example of a rigid I/O panel is shown in Figure 1. The first obvious advantage is that PCB-mounted I/O connector assemblies require much less labor to assemble than their direct cabled cousins.
Example of a rigid I/O panel.
The I/O panel is easy to test and is quite reliable because the termination is rigid. For cabling to the main system, the I/O signals can be organized as needed and brought to one or more cable connector. This can be a flat flexible cable such as a ribbon cable or a shielded flex-circuit. Alternatively, a discrete wire connector can be used, or finally, some combination of connectors.
The advantage to this approach is that there are cable connectors available that can be mass terminated, which is less costly than individual cables to each sidewall connector. Also, cable assemblies with a large number of wires are more efficient to test and install than many separate cables. Finally, the cable assemblies can be tested separately from the printed circuit panel. The ease of assembling the I/O panel and the use of mass-terminated flat flex cables reduce labor, and the compactness of flat cables uses less interior space than separate cables to each circular connector.
As the name implies, there are three subelements to a rigid-flex-rigid (RFR) interconnect structure: the main board, the flex PCB and the I/O board. The main board is where all the signals are generated, such as in a backplane. The flex element is an impedance-controlled flexible circuit board that is integral to the two rigid boards. The I/O panel is a second rigid section mounted with I/O connectors. All three sections are fabricated at the same time. Figure 2 shows a typical RFR interconnect.
Example of a Rigid-Flex-Rigid interconnect assembly.
There are several advantages to using the RFR approach. The RFR interconnect structure provides a constant impedance from point A to point B on a given signal path. Because it is assembled as a single structure, with the lack of intervening connectors, reflection is minimized. All the signal planes are surrounded by ground or power planes, which minimizes crosstalk, and the structure behaves well under high vibration and shock environments. Finally, this approach is usually the most compact solution possible. The labor cost required for separate cable assemblies and their testing is eliminated. Rear access to the connector pins at each end of the circuit allows accessible test point locations for high speed signal measurement.
Increasingly, systems are turning to rigid-flex cabling. The combination of a flex circuit for tight corners paired with a rigid PCB for the I/O connectors is ideal for space-constrained chassis or boxes. Figure 3 shows such a RFR assembly installed.
Rigid-Flex-Rigid assemblies inside an enclosure.
There are however a few limitations to this approach. RFR design is typically very costly, and the assemblies are more expensive because of the specialized materials and processes used in their fabrication. Because design and tooling requirements make this solution more costly, it is best for applications with higher volumes, where NRE costs can be distributed across many units. When volumes are expected to be significant, this approach should always be considered.
Connectorized (DPI) Risers and Mezzanines
The direct plug interconnect (DPI) is implemented with a direct connection between the main board and the I/O panel by means of a board-to-board connector. When attached at a right angle to one of the edges of the backplane, any family of typical backplane connectors can be utilized. The I/O panel can be located at right angles to any one of the four edges of the backplane or even as a rear mezzanine by using stacking connectors. This flexibility together with the varied mounting options for the subrack within the enclosure allow the I/O panel to be placed on almost any desired face of the system enclosure. The compactness of DPI preserves one of the important advantages of the RFR approach. And like RFR, the costs associated with assembling and testing interconnecting cables are eliminated. The important advantage of DPI over RFR interconnect is cost. Figure 4 shows a typical DPI interconnect.
Typical Direct-Plug Interconnect assembly.
This approach is typically more reliable than either of the cabled approaches and equivalent to the reliability of any of the card-to-backplane slot connections. The connectorized DPI approach is more expensive to implement than the cabled approach first mentioned but significantly less expensive to implement than the RFR approach. The DPI offers an excellent solution to intermediate volume programs. It offers a compact reliable solution that is easy to test and efficient to assemble.
Silicon will always move ahead of the rest of the system components, and systems will struggle to keep up with new silicon capability. At the present time there are a number of efforts within VITA to define both higher-speed MIL 38999 connector implementations as well as to define interface strategies for a number of new small form factor standards. In time, new connectors will populate the backplane slots, I/O connectors will get new faster inserts, and the fiber content of I/O cables will increase. Despite such inevitable advances, the five basic approaches to backplane I/O will likely remain for the foreseeable future.
Cable Type Considerations
There are several cabling methods to consider when faced with a challenging package design. Three different types of cables are most common in embedded systems: discrete, RF and fiber optic. See the summary in the accompanying Table 1.
Connector I/O Solutions.
Cable types: Teflon Insulation, Non-PVC, Multi-Conduction Shielded/Jacketed
Connector Types: Mil-Spec Style, D38999, Micro D-Sub, Standard D-Sub, General “Off-the-Shelf” style. Depending on the design of the system, there are several ways to approach discrete system I/O cabling. When space is not an issue, using a more traditional wiring approach is typical and most economical for prototype development. Where space is somewhat limited, a flex circuit might need to be considered, which then eliminates the need for discrete wires. Board-to-board design is the next alternative way to make the custom I/O connections. This is by far the most costly method and is not suitable for prototypes.
Testing: Standard point-to-point testing is done for custom I/O cables using an Ohm Meter. The quantity of systems being built will determine if special test adaptors are required. Custom extender boards have been developed to test custom I/O cables that interface with backplanes, which will test the entire chassis from rear panel connectors through the cables and through the backplane.
Cable types: RG59, RG316, RG402, RG405, Rigid
Connector Types: SMA, SMB, SSMC, TNC, BNC
Testing: Special testing equipment is required for measuring insertion loss for RF cables. In the prototyping stage or for low volumes, this is not practical.
Cable Types: 50/125um, 62.5/125um, 85/125um, 100/140um, Armored
Connector Types: FC, LC, SC, ST. Single Mode, Duplex Mode, Simplex Mode, and Multi-Mode
Testing: Special testing equipment is required for Fiber Optic Cable assemblies.