Small Form Factor Standards Co-Exist in Harmony

Don’t expect a shake-out any time soon; stackable small form factor standards will continue to co-exist, each serving certain classes of applications.

By Robert Burckle, WinSystem

The consumer market has taught us to think that only a few platforms or even just a single main platform can exist. Whether it’s a device such as a PC, smartphone or tablet, or operating system such as Windows or Android, the market very quickly consolidates for the masses.

While it’s tempting to view the small form factor (SFF) embedded computing market in the same manner, the truth is quite the opposite. Unique requirements for each application don’t allow excess baggage. There is simply no space for any I/O circuits that are not absolutely needed.

While full custom or a computer-on-module (COM) mounted on an optimized carrier board can yield perfect-fit solutions, most system manufacturers using a SFF single board computer (SBC) cannot afford the design time or non-recurring engineering (NRE) cost of such an approach. COMs are optimized for cost, short height, CPU upgradability and minimal cabling. Stackables are optimized for flexibility and I/O expandability. The two approaches are analogous to capitalizing versus expensing a design, or owning versus renting. Commercial off-the-shelf (COTS) is more practical than COMs for quick-turn, moderate-volume designs since a custom baseboard design is not necessary. A stackable SBC – along with application-oriented I/O cards from the stackable ecosystem – allows systems to be configured and placed in service quickly.

The PCI and ISA legacy parallel bus architectures are yielding to higher-speed serial buses such as USB and PCI Express. During the last several years, some new standards have been launched as replacements for the incumbents. But for stackables, which standard is better? The simple answer is: it depends upon the application.

Stacking architectures need to offer sufficient bandwidth and flexibility to serve a myriad of I/O bandwidth and software interfacing needs. I/O requirements may be as simple as a slow relay or switch closure to a high-speed vision system. Yet as diverse as these are, they both must be able to be mounted on a stackable expansion connector on top of a SBC. PC/104 (ISA) and PC/104-Plus (PCI) are the established stackable standards for a variety of environmental conditions. They have scaled well from low- to moderate-bandwidth I/O devices.

Three different PCI Express solutions for PC/104 stackable modules are available in the market today, each addressing a different application segment. PCIe/104 Types1 and 2 target medium- to high-performance processing applications and are becoming popular for military and computation-intensive applications. SUMIT-ISM targets low- to medium-performance requirements for instrumentation, pipeline, communications, utility, medical and transportation applications. Each has pros and cons for each application.

The stackable SUMIT connector architecture offers PC/104 legacy support as well as support for new high-speed I/O. SUMIT supports ultra-low power chipsets from Intel while providing a lifeline for the vast PC/104 I/O ecosystem that constitutes 70-80 percent of all stackable I/O module production volume. SUMIT can be used on industry-standard 90 x 96mm modules as well as for I/O expansion on EPIC, EBX and Pico-ITXe SBCs. It handles the entire range of low-speed SPI and LPC buses to USB and PCI Express on the same connector since its bandwidth supports up to 5 GT/s.

Intel’s sub-10W processors are easily adapted into a stackable, low-power, 1.66GHz legacy-friendly, 90 x 96 mm ISM board with both SUMIT for high-speed I/O and PC/104 for legacy I/O. A SUMIT-ISM product is the best evolutionary, legacy-friendly step for mainstream applications.

Don’t expect any shake-out of SFF specifications or standards any time soon. More trial balloons are being floated and pin-out types are being created to address the many niches in this market. Innovative ideas, although quick to generate, play out in an incredibly slow fashion. Specification definition, SFF module design, sampling / design-in, qualification / certification and production combine to take ten years or more. Nevertheless, stackable SFF standards will co-exist, each serving certain classes of applications.



Robert A. Burckle is vice president of WinSystems. Burckle’s background includes over thirty-five years of experience as a project engineer in such applications as biomedical, test and computer design, interface and simulation. Burckle earned an MBA at North Texas State University and both master’s and bachelor’s degrees in electrical engineering from the University of Louisville. He can be reached at bburckle@winsystems.com.