A bus called "Bob"
The latest news in stackable architectures is actually what’s “old.” The ISA bus is still the volume leader in custom and standard stackable cards, and for small form factor military and avionics applications, many stackable products have been introduced over the last several years with the PC/104 ISA bus, providing seamless upgrades during a time of reduced Pentagon budgets. In this space, if you can’t lick ‘em, join ‘em.
Stackable SBC technologies evolve more slowly than any other type. This is because CPU design teams are either small companies or small teams within very large companies. Same goes for I/O design teams. Truly this niche market operates like an unusual First-In, Last-Out (FILO) buffer. So, why is there a groundswell movement to force the migration of ISA-based CPUs and I/O cards to PCI and PCI Express (PCIe)?
A number of arguments have been made over the years in the off-the-shelf board community about why ISA needs to go. For one, it is no longer included (“free”) in the latest x86 SoCs and chipsets. But bus bridges can be added. Another point is that such bridges support serialized interrupts but do not support Direct Memory Access (DMA) transfers. This was a big deal back in the sound card days, but few current I/O cards need DMA. The strongest argument seems to be that there is only enough space and proper connector registration for two expansion bus connectors, and those need to be PCIe and parallel PCI because those buses are newer and included on many processing platforms. One hole in that argument is that parallel PCI is going away in favor of PCIe, which is software compatible with PCI. Another is that PCIe-to-PCI bus bridges are more expensive than LPC-to-ISA bridges for the Low Pin Count (LPC) bus; so ISA is, in fact, easier to keep around as PCI goes away.
A new name for an old bus
As an industry, how do we get past the negative stigma that the ISA bus has of being old? In mil/aero and defense circles, we really like to use well-established, proven technologies. We also like the vast ecosystem of proven PC/104 I/O cards that use the ISA bus, including MIL-STD-1553 bus cards. Certainly, our purchasing dollars carry weight with I/O suppliers who continue to fill orders since their boards do not have End Of Life (EOL) component problems. They do not need to redesign what already works. Why re-engineer and re-validate hardware and software for a newer bus that does not change the specs of the actual I/O in the first place?
Rather than churn all the hardware and software, we should just ask the marketing folks for a new name for the ISA bus and leave it in production. How about the name “Bob?” With apologies of course to Ms. Melinda Gates, the former product manager for an application software package bearing the same name that never took off, so the name is available.
A bus bridge for the “Atomic” age
Intel’s new Atom family of processors are attractive for military applications due to low power (below 10 Watts), extended temperature operation (-40 ~ +85C) for certain models, and enough performance to upgrade legacy military systems. In this case, the “Atomic” age means that Atom processors are positioned for legacy I/O with the LPC bus. The LPC bus allows easy attachment of low-cost ISA bus bridges and low-cost UART chips to it. Some SBC manufacturers have even implemented ISA core logic within an FPGA so that the ISA bus can avoid EOL forever. An example of an SBC that supports ISA with extended temperature for mil/aero applications is shown in Figure 1.
Many new SBCs allow the continued use of the ISA bus directly off the SBC for technology refreshes to legacy systems in an era of tight defense budgets. With a new name like “Bob,” the ISA bus could shake the stigma of being an ancient bus that threatens its longevity in the market. Regardless, some of the ISA-friendly SBCs and I/O cards from SFF-SIG members can be found at.
Small Form Factor Special Interest Group