PC/104 is a veteran of embedded computing. Yet, with all the changes in the embedded space since its inception, does it still have a future?
PC/104 has been around almost as long as I have, conceptualized within Ampro almost three decades ago back in 1987. In 1992, the now world-famous form factor was standardized and the PC/104 Consortium was born. This revolutionary thinking created the first multi-stackable computing platform.
The backplane and slot card method, long characteristic of desktop computing, has limitations in its compactness and the ruggedness (thus reliability) of peripheral interconnects via slots – both critical aspects of embedded systems.
The format initially provided a single common bus (ISA), though PCI and PCIe capability was introduced in 1997 and 2008 respectively, following logical demand for increased bandwidth. All standard PC I/O connections were, and are still, satisfied by flying leads from onboard pin headers – a flexibility strength that has served PC/104 well over the years.
New applications for the compactness of PC/104 then emerged by the day and our industry is indebted to the success of the format by the exponential growth that it brought with it. It inspired a generation of engineers and indeed the subsequent generations of Single Board Computers (SBCs) that we see today; yet seemingly it refuses to succumb to old age and we still see new PC/104 boards released each week, albeit satisfying a smaller segment of applications than in its heyday.
So, what future do we see for PC/104, why are companies still spending vast sums to extend its lifetime, and when will its demise be inevitable?
I see a number of reasons PC/104 remains popular today – the first is familiarity. Now, this is not (just) the sepia-filtered nostalgia of a sentimental engineer. Familiarity reduces risk and minimizes development time, which naturally reduces time to market, thus seeking "familiarity" is a commercially sound approach. This aspect of course has a shelf life, as the latest generation of electronic engineer is less likely to pick a PC/104 card for their first project, ad infinitum.
The second aspect I see as a failure of the System-On-Module (SOM) format, ironically a consequence of one of its strengths – multiplexed pins. By SOM form factor specifications permitting individual pins to have multiple functions, selectable at top level, you enable an individual SOM to satisfy a wider breadth of requirements – sound reasoning at product level of course.
Shortsightedly, the ramifications for integrators of SOM products meant that the promise of drop-in replacement scalability simply wasn't reality. As SOM manufacturers were (and are) free to elect individually which multiplexed configuration they would use, despite being mechanically identical, electrically they weren't. An obvious example is the ETX format, with OEM A using the same pins for floppy disk drive support as OEM B directing parallel port data across them, so your baseboard in fact often only supported that specific model!
My point here is that despite SOMs claims, PC/104 remained the de facto choice where scalability and the security of multi-sourcing were needed.
PC/104's popularity is self-perpetuating with legacy re-designs; if a mechanically identical board still exists, this cuts swathes of time and cost and PC/104 remains naturally the successor elect.
The continued availability of data acquisition peripheral cards in PC/104 format has also had its bearing; many OEMs continue to provide Analogue and Digital I/O alongside industry-specific proprietary communications interfaces as such, with the sole alternative often being our old friend the slot card.
I encountered an intriguing development recently: the redesign of a PC/104 stack containing multiple I/O modules. The logical migration seemed to be SOM, laying out all the specialized I/O onto a baseboard and eliminating the costly peripheral modules. In fact, given the multiple I/O configurations of their end product, the stack remained the most cost-effective way of satisfying this huge variation.
The final aspect must be low production quantities. Low quantity, specialist product doesn't have the room to encompass hardware Non-Recoverable Engineering (NRE) – often I see these projects not even using the peripheral bus, delivered with cropped pins. PC/104's success in these projects is owed to its truly "off-the-shelf" nature, but the format's success specifically must lie in the lack of a real competitor. 3.5", 5.25", EBX, EPIC, etc. are less compact and contain few benefits for many, so PC/104, for now, stays!