Comment on this article

Designing CompactPCI into rugged applications

1David discusses the process and challenges involved in designing highly ruggedized CompactPCI systems.

has emerged as the architecture of choice for military and aerospace applications. Its high bandwidth, ruggedness, and familiar computing paradigm have enabled to be deployed in everything from shelters to aircraft to ground vehicles.

I have taken special care to highlight the differences between designing ruggedized and nonruggedized .

Throughout the article, we will use the design of a hypothetical, but very typical, COTS system in a rugged Air Transport Rack (ATR) as a vehicle for discussion.

Designing a navigation computer

When first designing a new system, engineers must determine the desired capabilities of the system (I/O, power, environmental conditions, and the like), figure out which types of software application(s) will be run, and then choose a computing architecture that best matches those needs.

For the purposes of this article, we will assume we have been given the task of design ing a navigation computer for a ground vehicle (Figure 1). The form factor for the computer is specified to be a 1/2 short ATR. It will be cooled by hard mounting to a cold plate, so that all system-generated heat must dissipate through the walls of the chassis and into the cold plate. Note that ATRs, despite their name, are not confined to avionic systems. They are also widely used in vetronic (ground vehicles) and navtronic (naval) applications.

21
Figure 1
(Click graphic to zoom by 2.3x)

Because it is a ground vehicle, the environmental specifications can be derived from existing MIL specs, such as MIL-STD-810F. Although in the "real world" it seems there is always some deviation from these profiles. We'll assume that this system needs to meet the MIL-STD-810 environmental profiles for tracked ground vehicles in terms of shock and vibration. Also, it will need to be operational throughout the -40 ºC to +85 ºC temperature range. It is also safe to assume that 28 VDC vehicle power runs this application, which means it will require a power supply able to handle the peculiar requirements of this type of power input.

Finally, we should note that there would typically be many more environmental requirements, covering everything from weight to salt fog exposure to fungal resistance. These are beyond the scope of this article, but they would factor into your choice of ATR box. Because the electronics will be sealed from the environment, these do not typically affect the selection of your .

Architecture choice

Given the target installation of a rugged ATR inside a ground combat vehicle, CompactPCI and are the most logical architectural candidates. Years ago, VME was the only choice available for truly rugged applications, and VME continues to have a very strong presence in technology refresh and new military applications today. But as the demands continue to change and evolve in this market, CompactPCI has become a reliable counterpart to VME.

In many cases, the limitations of the VMEbus preclude it from being con-sidered. This is especially true for mobile platforms and other small footprint ap- plications requiring the 3U form factor. Limitations of 3U VME in terms of bus width, bandwidth, and rear I/O pins make it impossible to use for many applications. Instead, designers choose for these applications as it supports a much higher bandwidth with plenty of rear I/O.

However, this argument doesn't apply to our system because the chosen ATR will allow five 6U slots. This suggests the CompactPCI's derivative 2.16 or VME derivatives VITA 31 and VITA 41 would all be candidate 6U architectures. All natively support a switched or mesh fabric on the . Table 1 compares these architectures. PICMG 2.16 is the obvious choice for applications requiring a significant number of rear I/O pins, but our system has modest rear I/O requirements, so any of these architectures would meet the application's rear I/O needs.

21
Table 1
(Click graphic to zoom by 2.5x)

Instead, it is the improved reliability and reduction in Total Ownership Cost (TOC) that causes us to select the /PICMG 2.16 architecture for our system. PICMG 2.16 improves reliability in two ways over VITA 31 or VITA 41. First, the PICMG 2.16 architecture provides Intelligent Platform Management Interface (IPMI) on the boards and backplane via an I2C bus. This allows the system to be monitored so that failing boards can be identified and swapped with a replace-ment board. Second, PICMG 2.16 does not have a data bus shared among the cards. Therefore, no single card can go haywire and completely prevent all communication in the system. On the other hand, VITA 31 and VITA 41 include legacy VME64x support, which means that a single bus is shared among the boards, giving the opportunity for a single board to fail the system. In the most pathological case, an electronic failure on a single board could damage all cards in the system.

Combined, these PICMG 2.16 features provide the capabilities necessary for this application, including the ability to create a rugged multiblade supercomputer with reliability benefits that reduce the total lifetime cost of the deployed systems.

Choosing system components

In this case, the choice of the ATR enclosure should be fairly straightforward, given that the environmental and backplane requirements are well-defined. We need a 1/2 short ATR that supports hard mounting to a cold plate for conduction cooling. Running off 28 VDC, the ATR's power supply must generate enough power for the CompactPCI boards. In this case, we'll assume the power supply provides 115 W, which is a typical capacity for an ATR of this size.

The type and number of CompactPCI boards is highly dependent on the application, as it is with any other system design. However, we now have the additional constraints of 115 W total board power, conduction-cooling, and the need to meet the environmental specifications.

Recently the VITA standards organization made it significantly simpler to evaluate a board's environmental specifications. VITA 47 defines a handful of environmental classes for boards to meet so that board vendors and customers can work in accordance with a single small set of environmental requirements. In our case, the boards would need to meet the VITA 47 ECC4 environmental class, which is the conduction-cooled class with the harshest temperature range, -40 ºC to +85 ºC.

This simplifies our search to include boards that meet the ECC4 requirements and have a total combined power consumption of less than 115 W. We'll also assume that the software architecture will require as much compute power as possible, which suggests the five slots should be used for four CompactPCI dual-core single board computers and one switch. The blades will use PICMG 2.16 Gigabit Ethernet communication across the backplane.

Figure 2 shows two boards in which CompactPCI's versatility lends itself well to diverse military applications. Kontron's ITC-320 (left) is a 3U CompactPCI solution, while the Kontron CP6001 (right) is a 6U version with PICMG 2.16 connectivity. Both conduction-cooled boards provide high computational power by way of dual-core CPUs.

22
Figure 2
(Click graphic to zoom by 2.4x)

These requirements can be met by using four conduction-cooled Intel Core Duo processor blades plus one conduction-cooled PICMG 2.16 Gigabit Ethernet switch. In this configuration, the blades consume 17 W and the switch consumes 28 W, for a total of 96 W, well under the 115 W constraint.

Rugged application design vs. nonrugged design

Once the boards are chosen, application development can begin. However, application development for rugged and non-rugged systems have one significant difference. Rugged systems do not typi-cally use standard I/O connectors, whether from the front panel or the Rear Transition Module (RTM). In fact, in most cases, there is not enough room for an RTM in the ATR.

As shown in Figure 3, rugged military systems differ from "standard" CompactPCI systems because a custom back- plane is used both for interboard communication and to route the I/O for the system. The I/O from the backplane connects to MIL-style connectors on the ATR box through custom cabling or the use of flex circuit interconnects.

23
Figure 3
(Click graphic to zoom by 2.2x)

This means that a cus-tom backplane and/or flex circuits will be needed for each rug-ged application. Also, the number, type, and placement of MIL-style connectors require the ATR box to be tailored to each application. Combined, this means that it may be weeks or months before the target ATR box is actually available. As a consequence, most of the application development must be done on a nonrugged system until the ATR box is available. In many cases, all development work is done on the nonrugged systems, simply because it can be difficult to properly cool a conduction-cooled ATR box. Also, standard CompactPCI boards tend to be much less expensive than conduction-cooled boards, so using nonrugged boards for software development has the side benefit of reducing the cost of develop-ment systems.

Some rugged board vendors, such as Kontron, fully support this model of application development by manufacturing the exact same board in both conduction-cooled and convection-cooled variants. That is, the Circuit Card Assembly (CCA) is identical for both boards, with the exception of the heat sinking the convection-cooled board will use a finned heat sink, whereas the conduction-cooled board will use a conduction plate and wedge locks to dissipate heat to the chassis.

Using the same board for both the non-rugged development systems and the rugged deployment systems removes the risk associated with software porting. The software can be developed and tested on less costly nonrugged systems, and then ported to the ATR box with confidence once it is available and ready for deployment.

Conclusion

Note that while this was a hypothetical example, it is an accurate, albeit simplified, representation of a typical design scenario and the design choices made throughout. CompactPCI systems have been deployed in a wide variety of military and aerospace applications, including submarines, naval ships, airborne applications, and ground vehicles. CompactPCI's ruggedness and reliability, along with its ability to meet connectivity and rear I/O requirements, give CompactPCI the versatility that makes it so attractive to system designers.

David Pursley is an Applications Engineer with Kontron. He is responsible for business development of Kontron's , , CompactPCI, and ThinkIO product lines in North America and is based in Pittsburgh, PA. Previously, he held various positions as a Field Applications Engineer, Technical Marketing Engineer, and Marketing Manager. David holds a Bachelor of Science in Computer Science and Engineering from Bucknell University and a Master's degree in Electrical and Computer Engineering from Carnegie Mellon University.

Kontron
www.kontron.com
david.pursley@us.kontron.com