Packaged, small form factor computing

October 30, 2005

Duncan Young

GE Intelligent Platforms, Inc.

With the growth in Commercial-Off-The-Shelf (COTS) adoption over the past 10 years, more and more rugged, mission-critical, high-reliability applications have migrated to off-the-shelf solutions typified primarily by VME. While VME may be the market le...

With the growth in Commercial-Off-The-Shelf (COTS) adoption over the past 10 years, more and more rugged, mission-critical, high-reliability applications have migrated to off-the-shelf solutions typified primarily by VME. While VME may be the market leader for high-end military applications, there are many more applications that VME doesn't fit well because of its price points, high power dissipation, and relatively large size. To satisfy the demand for smaller and lower cost, a number of real alternate solutions have emerged to fill many of the niches that don't require the design sophistication, the overall set of project-oriented capabilities, the cooling and mechanical standards, and the logistics trail that have been developed for VMEbus products.

Excluding 6U CompactPCI, which has similar characteristics as VME, the options when space and budget are limited are many and varied. Typical of these are 3U CompactPCI, stacked PMC, ETX and ETXexpress, PC/104, PC/104-Plus, PCI-104, EBX, and Computer-On-Module (COM). With the exception of 3U CompactPCI, these are all quite similar in their size and complexity and, as will be seen, 3U CompactPCI and PMC modules are supported by ANSI/VITA standards for ruggedization. The list is by no means exhaustive and could include, for example, DIMM and similar concepts; however, these should be considered too fragile for use in military environments.

Choosing a form factor
But how does one make the selection? First, forget about price, which is irrelevant at this stage. Once the decision has been made to use off-the-shelf modules, price doesn’t drive the solution. This is because of commoditization of the basic sets of components that drive performance: processors, memory, flash, and so on. The price to be paid will be directly related to the performance required by the application – MIPS, MFLOPS, memory size, memory bandwidth, I/O bandwidth, I/O signaling frequencies, and the deployed environmental requirements – and the space available for it all to fit within. This performance requirement has a direct bearing on the overall power dissipation at the equipment’s maximum operating temperature. Performance requirements, if not negotiable with the end customer, will flow down into myriad technical requirements that will point to a few alternative solutions available from a number of vendors. This is where price variability exists.

Physical size is the other factor to consider. It has two components that equate to performance density or Watts/in.3:

• Available platform space
• Required system performance

A universal law seems to be that designers of off-the-shelf computing modules will fill however much space is available in their product with as much functionality as possible, using leading-edge technology and gate geometries available at the time of design. Functionality and performance are equivalent to power dissipation. The effect of this law is that, at any point in time, direct comparisons of performance can be made between various types of computing modules just by looking at the volume they occupy. Table 1 uses a conduction-cooled VME card, dissipating today's practical limit of 100 Watts as the baseline; this is the limit of performance density (100/46.4 = approx. 2.2 Watts/in.3) that is equally applicable to all the different sizes of board, unless the designer resorts to exotic cooling methods such as liquid flow-through or evaporative spray cooling. The maximum power dissipation of a number of small form factor modules is derived from this baseline by comparison of their physical volume with the baseline. This may seem blindingly obvious, but the table also helps define the potential application areas for each of the various sizes of computing modules:

* 75% of optimal dissipation

Table 1.  Power dissipation derivations

A Synthetic Aperture Radar (SAR) requiring the performance of four or more Freescale Semiconductor MPC7447A PowerPCs is not a candidate for PC/104 or ETX, but is obviously in VME’s space. Similarly, a battery-powered, handheld, single-channel software defined radio doesn’t need VME. Sometimes the volume available in a particular application is enough to support high performance, but the space is just the wrong shape to fit VME cards. In these cases, it will be necessary to choose a smaller form factor and accept the less-than-optimal packaging.

Another conclusion from the table is that many military applications require operation in enclosed environments for protection from external contaminants. This usually precludes the use of fans blowing cooling air directly over the components. VME, CompactPCI, and PMC all have standards for conduction cooling for use in these harsh environments, hence their realistic power dissipation is considerably higher than those without standards-based conduction-cooling options.

Beneficial platforms and applications
There are many types of platforms and applications where the use of off-the-shelf, rugged, small form factor computing will be beneficial: missiles, torpedoes, intelligent munitions, handheld and portable equipment, combat aircraft wingtips and tail, pods, pylons, UAVs, particularly mini and micro-sized UAVs and unmanned combat aerial vehicles. Typical of these embedded computing applications will be guidance, targeting, IFF, mission computing, missile approach warning, navigation, control functions, communications, and sensor processing. However, whatever the choice of form factor, the system’s performance will be ultimately limited by performance density.

PMC module use is usually associated with VME or CompactPCI base cards, normally single board computers. But a truly innovative approach to small form factor computing, with the advantage of standards-based conduction cooling, is to use PMC modules in a stacked configuration. Stacked PMC modules use their PCI bus to connect to each other. A Processor PMC (PrPMC, VITA 32 standard), with either Intel Pentium or Freescale Semiconductor PowerPC processors, may be used to control the bus and configure it at power-up. The ability of a PrPMC module to be a PCI monarch or non-monarch means that more than one PrPMC can be used in a stack for enhanced performance. The table shows that a PMC module is smaller than 3U CompactPCI, so further savings in space can be made where performance permits. An example of such a stacked PMC product is the Rugged Operational Computer (ROC) product line from SBS Technologies illustrated in Figure 1.

Figure 1: Rugged Operational Computer (ROC) product line from SBS Technologies

ROC uses a carrier card concept to mount the PMCs, allowing conduction-cooled PMCs conforming to the VITA 20 standard to be incorporated, without modification, from many different vendors. The carrier provides a means to interconnect the PCI bus between PMC modules and provides paths for I/O signals from the PMC’s Pn4 connector to military specification, circular connectors on the outside of the enclosure. Carriers can be stacked onto each other, creating a passive PCI “backplane” without a backplane assembly. Carriers are mounted in a modular chassis assembly comprised of slices, one for each pair of carriers. The base of the chassis contains the power supply and the monarch PrPMC module. The complete assembly with five PMCs and power supplies occupies a space of only 6.75" x 4.5" x 3.25".

PC/104 and PC/104-Plus are, of course, designed to be stackable. To facilitate this, there are mounting holes in each corner so that a stack of modules can be fixed together using stand-offs. Mechanically, this is far from ideal in a high-shock and vibration environment as the stack can sway if not properly supported. Care should also be taken to ensure that heavy components are not mounted centrally on the boards as they will put undue stress on the mounting points under vibration. However, a number of solutions exist using cans, containers, or aluminum slices that support the modules compliantly at the corner mounts and provide for the provision of cooling and external connectors. Although there is no standard for PC/104 module conduction cooling, a number of vendors offer heat spreader plates and even thermal gap pads to bond the modules to interior surfaces, coldwall cooling, of an enclosure; however, these operate most effectively with single modules only. A stack of five or six PC/104 modules plus power supply in a container is a practical limit for use in harsh environments and would be equivalent in volume though lower in performance density when compared to a stacked PMC system such as SBS Technologies’ ROC.

Reliability, maintainability, and obsolescence management
Despite many contrary assertions, an embedded military computing application does have special needs when it comes to reliability, maintainability, and obsolescence management. In the context of COTS-based, small form factor computing, maintainability is barely an issue. Because of their relatively low cost, complete assemblies can be inventoried as spares wherever the equipment is deployed and returned to their manufacturers for repair as necessary.

The much bigger issue is software and hardware compatibility and build control. It has to be accepted that semiconductor manufacturers will regularly upgrade their devices with geometry changes, speed bumps, and enhanced functionality. What is much harder to accept is that an embedded computing assembly may have variable performance and functionality depending upon when it was built and/or which replacement parts have been used during repair. For many applications this doesn’t matter, as the user has time to adjust to the new characteristics of a desktop PC or digital camera. But when the equipment is mission-critical and split seconds separate the right from the wrong decision, this is obviously quite unacceptable.

The equipment integrator, if benefiting from the use of COTS by competitively winning new business, must assume the responsibility for mitigating the impact of this variability of performance as technology evolves. Abstraction of the application software from hardware performance and functionality plus an ongoing program of testing are well-accepted ways to reduce the effects of component changes. Equally the end-user must responsibly specify the amount of variability that would be acceptable over the life of the program due to the criticality of the end-use deployment. This is really the point at which the decision on processor family can be made – whether “Wintel” or PowerPC?  PowerPC certainly has cost and ease-of-use penalties, but in return it generally offers much longer component life cycles that translate into less variability of performance for the end user and lower life cycle costs due to extended replacement cycles, fewer spares, less documentation, and less training.

Growth of small form factors in military applications
Small form factor computing is undoubtedly on a volume growth curve for military applications. As usual, there is no ideal solution that fits all, and there are multiple choices of vendor, technology, performance, and size readily available. The recent introduction of stacked PMC into the mix further extends the envelope of COTS-based, embedded computing that offers optimal performance density using standards-based cooling.

. . . . .

Duncan Young has worked in the defense industry for almost 40 years. Duncan was part of the management buyout team that formed Radstone Technology, and he initiated product development of the world’s first conduction-cooled VMEbus modules. He has also served on a number of standardization committees. Duncan is now an independent consultant and writes this column on behalf of SBS Technologies.

For more information, go to the SBS Technologies website at www.sbs.com.

Featured Companies

SBS Technologies

12090 Memorial Pkwy SW
Huntsville, AL 35803

Website