ATX AND PCs
Application Is King; Understanding End Use Is Critical to Embedded Platform Options
Deciding on compute power, size and variety of custom and standard I/O–whether to go with a COM module or an embedded motherboard–requires an in-depth understanding of the needs of the application.
CURTIS CHANG AND CHRISTINE VAN DE GRAAF, KONTRON
Page 1 of 1
Embedded designers face more computing demands than ever before–processing power, thermal considerations, time-to-market–and their efforts are both supported and challenged by an incredibly broad and expanding range of options for shaping their designs. Computer-on-Modules (COMs) and small form factors can address specialized I/O, increased computing power and space constraints with design flexibility. They are ideal for a broad range of embedded applications where they fit mechanically, economically and functionally. Standardized solutions like embedded ATX–ATX with an embedded chipset–can deliver performance, cost efficiency and a shorter design cycle, in turn suitable for applications that can leverage what the board has to offer without extended design time or big design budgets (Figure 1). Because each approach has its own advantages and disadvantages, application-specific requirements are the deciding factor and the first step in unraveling a complicated landscape of design options.
Selecting a computing platform can in fact be the most critical step in the design process. Both ATX and COM solutions offer full features and their own versions of design flexibility, either through complete customization or by deft use of standard but robust industrial board components. Almost all implementations, whether they are customized COMs or ATX, offer the high-end communication capabilities that are becoming a requirement for most products. In many cases, it comes down to what designers are familiar with, what level of make vs. buy applies and most importantly, what the end application dictates as its highest priority form and function.
Making Good Choices
Just about any embedded design platform has high-end graphics, Gigabit Ethernet, SerialATA, PCI, and for the most part PCI Express. What makes the difference is how the technologies that are available in each of these products are used in the end application. For example, ATX is a single board solution, and because it is a motherboard it is larger. COMs can be small, and used anywhere from something that fits in the palm of your hand to the size of an ATM machine. But they can still use the same technology at their cores, so they both have their strengths and their unique place in a design. They are very different and used for different reasons, frequently based on a range of external factors that impact design choice.
For ATX, the customer wants or needs to spend less time and money on design, and needs to fully leverage what the board has to offer with its fully developed features and low cost. ATX is a perfect fit in stationary systems that don’t require customization, but need flexibility from one generation to the next, as well as readiness for interface change from one design to the next. Hardware customization adds to cost and doesn’t always make sense to do, especially with smaller runs of product. With ATX solutions, however, designers don’t need to be concerned with whether or not they will be built in quantities of thousands or if they can maintain a three-year lifecycle; they can be smaller quantities because they represent a standard off-the-shelf product.
A designer would choose a COM solution if their requirements included a lot of application-specific customization and they could afford a two-board solution (module plus custom carrier board), a high run of product and the need for some scalability from generation to generation. COMs work well for devices that not only require scalability from generation to generation, but also within a single generation. When an application requires something special that is not typically found in a standard motherboard, those computing issues can be customized into a COM’s accompanying carrier board, which allows for easy transitions to future generations (Figure 2).
Modules contain a standard off-the-shelf product within the module core itself. Customization is designed into the module’s carrier board and can last for generations with various CPU cores. If the design plan includes multiple variations of a product within the same generation, perhaps with different performance capabilities, designers can use the same carrier board for those variations by just changing the module used with it.
Function Defines Solution
Industrial automation–for example, an automotive production floor or even a CPU manufacturing environment–requires a look at end use in order to determine which solution goes behind the design. In many instances, big heavy robots are completing some process, and the automation factor refers to how they are being told what to do and how to do it. These robots need to be “smart”–meaning they must be able to provide information back to the operator, including data on the number of items completed as well as successes and failures in the production line.
Once a product is completely manufactured, it is moved off the production line into a storage area. The tools that are used to move it from one place to another may involve tracking and continued feedback to the operator indicating how many completed products are being moved and are now considered “in stock.” There are elements within these machines that are small, for example, the subsystem, and elements that are much bigger. That tool that goes on the vehicle that moves product off the line and into the storage area might be a module. The actual arm that is moving to build the unit communicates with a bigger box, whether it’s in a server or maybe a special kind of display, and it may be large enough to contain an ATX motherboard. In contrast, displays that a person has to touch are likely candidates for modules or perhaps a smaller single board computer.
Applications Drive Design
Medical imaging is perhaps one of the most interesting and growing application areas for embedded design. Devices here require high performance. More and more often they require mobility, and they require small size so as not to be intrusive with patient care. For example, an imaging application that in the past has run on a cart, now needs to be developed on a smaller scale. Its carrier board has been customized to talk to the components that run the actual image capture, and now designers want to shrink down that entire process into a smaller device.
Designers know that the CPU, chipset and other existing components have optimal function; they can keep using the COM they are familiar with and modify the carrier board to accommodate a new, smaller size. Core elements remain the same, designers avoid re-spinning drivers for change of hardware, while computing function is simply ported over to the smaller design.
COMs allow a very strong path from medium-sized devices into smaller and smaller ones. But size is not always an issue, as in the case of room-sized imaging equipment. These machines are compute-intensive, but have no requirement for size constraints or considerations. So, for designers deciding on motherboard vs. module for a medical device, a big question is how big the unit is being designed. In room-sized MRI machines, size is not a concern, but in portable or cart-based devices, or ones that need to fit into the pocket of medical personnel, COMs are generally more suitable.
The important trend here is the reduction in size of devices, as well as the reduction in size of components that can go into a design. Size is typically not a limitation with a motherboard; it can easily
have all the components it needs and then some. Ports may go unused, but that will not impact performance or function in any way. Modules are more limited by physical constraints; all its different components must fit within the overall footprint of the module itself, and the connecting pins defined on the bottom side must connect appropriately to the processor and controller signals.
Multicore Is Meaningful
Multicore processors have been a design element in COMs for quite some time due to the size and space demands inherently addressed by their design. These types of processors will soon dominate the processor sockets on many types of embedded boards as well. The first multicore processors arrived in the embedded market two years ago, yet in 2007 just 20 percent of embedded desktop form factor motherboards (ATX, microATX, FlexATX, Mini-ITX and derivatives thereof) had multicore x86 processors on them. That percentage is expected to increase by 2010, reaching close to 50 percent of all embedded desktop form factor motherboards.
Since motherboards have more space, instead of relying on a multicore processor they could simply incorporate more than one processor. With a greater transition to multicore technologies, designers can actually include more than one multicore processor on an embedded motherboard, mirroring trends typically seen in CompactPCI and ATCA blades. Today’s ATX designers now have the option as well and are likely to be looking beyond Celeron and Pentium, toward the Intel Core2 Duo and its follow-on dual- and quad-core processors.
Build, Buy and Call an Expert
Build vs. buy considerations also impact the designer’s choice of platform. This requires evaluating the level of customization required along with the anticipated production volume of the design itself. Designs that require something unique, such as a special connector to communicate with the information-gathering portion of the application or perhaps a certain kind of display, are typically better bought than built. Smaller volumes and highly specialized design expertise are critical elements here, and it may be a much better design option to make a tweak to an off-the-shelf motherboard. As a standard product, the motherboard’s life is well planned out, but simple changes can still be made.
When customization becomes a bit more complex than adapting a connector or an interface, designing with a module carrier board can allow more extensive changes than designing with a motherboard. Modules themselves aren’t hardware-customized as often. The custom elements are commonly built into the carrier board.
Designing an adaptation into an off-the-shelf product solves the problem of small volume, but only if the right resources and design expertise are readily available in-house. If the design will live from generation to generation, or have a lot of customization, it is important for designers to partner with a supplier that has know-how as well as product. In every instance, resources make all the difference, as does the ability to take a design and turn it into something real.
Design expertise and critical evaluation may be a designer’s strongest assets, but they must be paired with an in-depth knowledge and understanding of the end-user application and its key functions. Working with a knowledgeable supplier as partner can help a designer evaluate requirements such as design size, input/output and general processing requirements as well as thermal considerations as a starting point. Understanding technology options requires digging further into regulatory demands, customization needs and lifecycle management, helping designers define the best design platform and, ultimately, the best design.