There has been a lot of speculation of late that VME has reached its natural limits as a fresh enough architecture to continue to capture the large number of new design wins necessary to sustain its market momentum. Certainly CompactPCI is an alternative, but it does not have the broad market acceptance, including defense and aerospace, that VME enjoys. VME is still ideally suited in medium to high-end applications that require multiple processors, high-speed intelligent I/O, shared memory, and the maximum leverage of open standards, including software products. Simpler applications have migrated away from VME to lower cost alternatives, such as industrial PCs, PC-104, or Computers on Modules (COMs), which all make good economic sense if they meet the majority of the other requirements for a customer’s program.
VME has now become pervasive as the embedded computing choice in defense and aerospace applications. But, while the basic functional characteristics of the products remain the same, there are many differences due to customers’ environmental and maintenance requirements that affect how solutions for ruggedization are devised and implemented. Defense and aerospace requirements vary from the benign, air-conditioned environment found in the Naval Operations Room to a missile or pod slung beneath the wing of a single-seat combat aircraft. However, these environmental requirements must also be linked to the characteristics of the end user and the system integrator’s needs for longevity of supply, support, and technology insertion throughout the life of a program. The needs of customers are characterized in Table 1.
|
|
Just a quick glance at the table shows how difficult it would be to offer just one type of VMEbus product to meet all of these needs. If it meets the needs of a combat aircraft then it will inevitably be overengineered and overpriced to meet the benign Naval requirements. VME manufacturers overcome this by offering varying grades of components and choice of cooling methodologies for their products if they want to cover as many of the market opportunities as possible. As an example, Figure 1 shows a conduction cooled (left) and an air-cooled (right) SVME/DMV-179 board from Dy 4 Systems.
|
|
In addition to the environmental requirements, there are very often requirements for maintainability to be considered. Connection of external I/O to VME cards can be made in two different ways, either with a cable to a mating connector on the VME card’s front panel, or through the backplane to J1, J2, or J0. Connection to the front panel requires extra space to accommodate the cables and connectors. Plus, to remove a card for maintenance will entail not just the disconnection of its own cables but also the cables to many of the other cards in the rack to gain enough space for removal. This is often acceptable where space is not a constraint, but is completely untenable where space is limited in, for example, an avionics system packaged in an ATR chassis. The disconnection and replacement of cables is also a reliability hazard as well as the very real risk of replacing cables in the wrong place.
As a result, most defense and aerospace applications for VME use backplane connection of I/O with all I/O signals routed via the P1, P2, and P0 connectors. This poses its own set of issues. Take a look at a state-of-the-art, VME single-board computer (SBC). Most are fitted with two PMC sites. With the 64 pins of I/O from each PMC routed to P0 and P2 respectively, most of the user-defined pins on the backplane are already used up. Then add to that the typical complement of two Ethernets, six serial ports, SCSI, two USB ports, 14-bits of digital I/O and the available pin count is easily exceeded. Manufacturers are offering optional pinouts of I/O functionality to fit customers’ needs and to maintain backward compatibility with pinouts of previous generations of less densely-packed SBC products.
The backplane connectivity issue has developed to the point where changes must be made to enable VME to continue to evolve further. In addition to the pin-out limitation, there is also a pressing need to accommodate high-speed signals such as digital video through the backplane, in addition to the upcoming high-speed switched fabrics. The defense and aerospace VME manufacturers are addressing these issues through VITA and the expectation is that future VME direction in this area will be announced very soon.
Connectors are a good lead in to other aspects of ruggedization as they form one of the key mechanical interfaces between a VME board and its chassis or enclosure. Today, the P1, P2, and P0 connectors are defined by the VME specifications. But what the specifications don’t address is the application environment of the VME board and its mating backplane. This is a deliberate policy, as it allows different manufacturers to differentiate their products by market space. So, even though the connector set is identical between a commercial/benign application of a VME card or a high-vibration military application, the devil is in the details. Under high levels of vibration, if either the VME board or the backplane are not stiff enough there will be movement between the mated halves of the connectors. This will eventually wear away the contact faces leading to intermittent connections and random system failures. Adding stiffness to both the board and the backplane alleviates the problem, but this has to be done up front in the design process, not added as an afterthought. This also impacts the mounting of PMC modules onto VME cards, as fretting of the connector contacts is equally undesirable at this interface as well.
Cooling methodology is another key factor to be considered. There are still only two basic methods of construction of VME boards to accommodate the various cooling options available. Convection cooling operates by blowing cool air over the components to remove the heat. This is the conventional form of cooling used almost universally. Providing enough air to cool a high-power system requires a large space for the air to flow both inside and outside of an enclosure and power to drive fans. An alternative method is evaporative spray cooling which sprays an inert fluid with a boiling point of 60°C or so, over the boards. The phase change as the fluid boils provides the cooling required to maintain a constant temperature of operation. Both air and evaporative spray cooling consume quite a lot of space and generally work best in a relatively benign environment.
The simplest way to ruggedize a convection-cooled system is to mount it inside another enclosure, which provides it with a controlled environment in which to operate. An example of this would be a 19-inch rack that is shock and vibration isolated using loops, coils, or an active suspension system, fitted with its own air-conditioning and environmental filtration equipment. In this way, the environment inside the rack is controlled and benign enough to use commercial-grade VME cards in a commercial chassis. This technique is often employed for the benign Naval or land-based environments, but it is not at all efficient in its use of space or weight.
Where space and weight are really key, there is still only one leading contender for cooling a VME card – conduction cooling. Heat from the components passes into the VME card assembly by conduction and is removed from the two short edges of the board into a chassis or heat exchanger. Conduction cooling has evolved significantly over the last few years. Early VME boards used through-hole components with their legs passing through an aluminum heat plate to remove the heat. With the advent of Surface Mount Technology (SMT) the heat path was generally down into the board through thermal vias into one or two copper layers buried within the Printed Wiring Board (PWB) itself. This form of construction was generally good for no more than 30W of heat dissipation. Newer ideas are combining the best of both types with full-size aluminum heat plates, supplemented by copper layers and sophisticated thermal shunts to remove hot spots from high powered PMC modules (see Figure 2).
|
|
Placement of components is also critical for optimal thermal performance of a VME board. Hot components need to be at the edges of the board for the shortest path to the thermal interface. Techniques are even now being developed for really hot devices, such as the newer generation of GHz+ processors, to bypass the conventional thermal paths of the heat plates and copper layers, to take the heat directly to the external thermal interfaces themselves. This means that a typical up-to-the-minute SBC with dual processors and two populated PMC sites can comfortably dissipate 60W or even more. Most VME vendors now use sophisticated thermal and mechanical modeling techniques to predict the characteristics of a product during detailed design. This allows the component placement to be optimized for electrical, thermal, and mechanical performance long before the product is ever realized in fiberglass.
Conduction cooling also confers other desirable benefits for use in military applications with their high levels of shock and vibration. The cooling of the board is contained to within the VME profile. Conduction cooled boards must be fitted into a chassis which will dissipate the heat. This can be done by blowing cooling air through finned channels in the chassis or by conduction from the chassis through a cold plate to the vehicle itself. Blowing air through finned channels in such a chassis is much more space efficient than blowing air across boards mounted in an air-blown chassis. In addition, the construction of a conduction cooled VME board with its heatplate over the surface of the board provides additional stiffness to the completed assembly, which is good for reliability. In order to further stiffen these heat plates vendors are now forming them into complex 3D frames (see Figure 3) the size of the VME card, with areas machined for overall additional stiffness for interfacing to PMC modules and providing localized stiffness around the connectors. The photographs of boards and frames illustrate these points.
|
|
Speculation should be put aside, because VME still has a secure future within its targeted application spaces. VME is available in a complete spectrum of environmental capabilities from benign to the most rugged, but even so, caution should be exercised when evaluating a product for use in a ruggedized application. It is important to understand your end use environment and ensure the product has been engineered to meet it with a reliability margin to spare. In the defense and aerospace arena for ruggedized products, work is ongoing within the industry to resolve the connectivity and switched fabric issues. Every time we think the power dissipation limit has been reached another new technique or use of material technology is invented to keep it moving forward.
Duncan Young is the Director of Marketing for Dy 4 Systems, where he developed the company’s most advanced concepts of COTS and lifecycle management for deployed, harsh environment applications and spearheaded its incorporation into a range of advanced, open-architecture VMEbus products. After early work on airborne radar navigation systems he worked with the European Space Agency on high-speed data links from ground stations to the Control Center in Darmstadt, Germany. He has worked in the defense industry for more than 30 years and was educated in the United Kingdom.