Transition boards bring COMs to save I/O stacks
Embedded board stacks are celebrating two decades of use in a variety of applications. Typically, there is a single host CPU card at the top, bottom, or even the center of the stack, driving the expansion buses upward and/or downward. Previously, that host CPU had to follow the same form factor standard as the I/O cards. But with Computer-On-Module (COM) products taking the Small Form Factor (SFF) market by storm, it was just a matter of time before COM CPUs for custom systems penetrated stackable systems as well. In this first of two columns on transition boards, the focus is on how to extend the life of legacy stacks.
In the late 1980s, the ISA bus appeared in a horizontal stacking format. Originally just for CGA graphics expansion – but in a more compact and rugged format than desktop CGA cards – the stacking bus gained traction for other types of I/O and became the PC/104 bus standard five years later. The 8-bit and 16-bit real-world I/O cards found homes in applications including real-time motor control, data acquisition, medical devices, and transportation. Two decades later, these I/O cards are still in production, even as generations of CPUs have come and gone.
Well over a dozen manufacturers used to build SFF CPUs for these stacks. These days, however, the overall market for stackables has narrowed a bit, and tighter budgets for military, Unmanned Aerial Vehicle (UAV), and transportation applications is threatening to turn stackable CPU vendors into endangered species.
An unlikely lifeline
In spite of the best efforts of standards trade groups to promote stackable architectures, the very members of these groups began to move their higher volume stackable customers over to a new architecture – namely, COM. The benefits and ROI of this move were clear: Invest in a custom carrier board up front, and shrink the system size while saving build costs for years to come. When inevitable processor End-Of-Life (EOL) issues arise, simply unplug the CPU module and replace it with the latest one. Because COM processor modules have a much broader market appeal, they reached critical mass for low-cost Asian manufacturing quickly, and the cost savings have been staggering.
One by one, CPU and I/O stacks in medical and other applications were lured into COM architectures such as the ETX form factor with custom-designed carrier boards. The total available market for stackables stopped growing for years until stackable CPU manufacturers finally realized that COMs were a direct threat. Finally, it was the stackable I/O vendors who answered the “S.O.S.” – that is, “Save Our Stacks” – with an unlikely lifeline: embracing COMs with transition boards. One of the first such transition boards is shown in Figure 1.
The ETX-NANO-104 from ACCES I/O looks like a tiny embedded motherboard with VGA, RS-232/422/485, USB, audio, PS/2 mouse and keyboard, and Ethernet ports. It features ISA and PCI bus expansion upward from the top surface, as shown. Noticeably absent is a large processor chip on the top. That’s because it isn’t on the top, and isn’t on the bottom either. On the bottom surface (not shown) is a connector array for a pluggable CPU that gets cooled by a flat heat spreader plate. The processor, chipset, LAN controller, RAM, and power supplies for the core voltages reside on the COM, which uses the high-volume, low-cost ETX form factor. Many legacy-friendly ETX modules are still in the market, with single- and dual-core Gigahertz x86 processors to provide a wide range of performance, price, and power-consumption solutions.
Quite a diverse set of applications can be covered with this approach. The transition board, often called a baseboard or carrier board, also provides PC-style connectors for the I/O that comes off the ETX module, which can simplify system cabling substantially. The ETX expansion buses pass through to the upward stacking bus interface. System OEMs can now use an off-the-shelf transition board to extend the life of their legacy I/O stacks and legacy software while tapping into the wide offering of ETX CPUs. The amount of engineering effort to move these stacks to new bus technology is sizable, and fortunately there is a viable alternative to expensive new stackable CPUs in the form of a transition board.
Next quarter, we examine how I/O vendors are turning the tables on full-custom designs, thereby breathing new life into mezzanine and stackable I/O architectures.
Small Form Factor Special Interest Group 408-480-7900