Avoiding the Obsolescence Trap

A 20-year industry veteran provides insight into designing a system that will survive the ever-changing PC.

Test systems based on PCs have been the standard for nearly two decades. In fact, it was 1984 when most of us had our first glimpse of graphical programming with the introduction of the Macintosh. At the same time, the market for data acquisition add-ons for PCs was just taking off.

The benefits of PC-based systems are so obvious they almost are not worth mentioning anymore: low cost, a wide variety of hardware solutions, multiple-choices of software environments, inexpensive computing, display, communications, and data-storage options. Prior to PC-based test systems, almost none of these options were available in the relatively closed world of mini computers and dedicated instrument controllers.

There is a less obvious downside to PC-based systems that may not become evident until years after a system is installed and operating effectively. Consider the following example. During the mid 1990s, you developed a test system based on Windows, Visual Basic, and ISA plug-in data acquisition boards for your I/O. There were more than 3 million ISA data acquisition boards used during the ï¿½90s, so if this profile fits you, youï¿½re not alone.

If you wanted to replicate those systems today to add more testing capacity or replace a nonworking system, you might be in for a surprise. First, many ISA data acquisition board suppliers have discontinued making their boards, are no longer in business, or have been acquired and the acquirer has discontinued the ISA boards. It is unlikely that they offer an exact PCI equivalent because most PCI boards provide new features to be competitive with other suppliers in the marketplace.

Secondly, if you wanted to keep your existing ISA data acquisition boards and upgrade the PCs, you would have a hard time finding a PC with ISA slots. Finally, if your existing hardware is fine but you need to upgrade your PCï¿½s operating system, itï¿½s unlikely that your existing ISA-based PC data acquisition hardware has drivers for the latest operating systems.

These are only a few examples of the variety of problems that you can encounter when designing a system based on the ever-changing PC. You also could be faced with the same issues relative to upgrading your software environment. As a result, test-system designers must deal with several considerations related to obsolescence.

What is the anticipated life of the test system?
After you answer this question, add 25% because most systems are used longer than originally anticipated. If your adjusted answer was three to five years, then almost assuredly you can ignore the remainder of this article since your likelihood of encountering obsolescence problems is small. If your answer was greater than eight years, then read on.

Hereï¿½s an example of why 10 years should cause you concern. Letï¿½s consider the case where you designed your system in 1995 based on one of two hardware platforms that were popular at that time.

If you had developed your system based on a PC, an ISA IEEE 488 card, C, and IEEE 488 instruments, then today you would be in a good position to upgrade your system to todayï¿½s standards without substantial redesign. Hereï¿½s why. First, nearly every supplier of ISA IEEE 488 boards offers a PCI equivalent that is code compatible. The reason is simple: There still are thousands of IEEE 488 instruments in existence so the market demand to support them is vibrant and lives on.

Second, nearly every IEEE 488 instrument supplier continues to support its new products with IEEE 488, although they may have added other interfaces along with IEEE 488. Agilent Technologies, for example, includes IEEE 488, USB, and Ethernet on its new products.

If instead, in 1995 you had designed your system using ISA data acquisition boards and wanted to upgrade it today, you could be in for a surprise. Your ISA board supplier may be extinct. Or even if it still is in business, it likely no longer offers your ISA boards nor provides support in the newest version of Windows for your 10-year-old board. In short, you would likely be redesigning your system if you wanted to upgrade it to todayï¿½s standards.

Will the platform you select be viable 10 years from now?
Letï¿½s review todayï¿½s platforms and discuss the likelihood of their viability in 10 years.

IEEE 488
IEEE 488, also known as HP-IB and GPIB, has been around for more than 30 years, and it is very likely that IEEE 488 instruments still will be offered in 10 years due the sheer number of products currently in the marketplace. Unlike plug-in boards that can be designed in a matter of months, it takes instrument companies years to develop a new product. Consequently, it is not cost-justifiable to redesign instruments every time a new interface comes along.

At some point, however, the advantages of new technology so greatly outweigh the incumbent technology that the incumbent begins to disappear. That very thing happened to parallel port, which today is all but impossible to find on a PC or a device that attaches to a PC despite the fact that parallel ports were ubiquitous just a few years ago.

Because of the age of IEEE 488 and its limitations relative to todayï¿½s standards, it is possible that new instruments will not have IEEE 488 in 10 years. It also is likely that many IEEE 488 instruments available today will be unavailable in 10 years. As for what they will have in lieu of IEEE 488, that answer is not as obvious although USB and Ethernet are likely candidates.

PCI
Now more than 10 years old, PCI still is going strong, and there are specifications for faster versions that are compatible with existing slower PCI boards. So like Ethernet and USB, PCI may have an extended life as new incarnations are backward compatible.

There is, however, a different reason for concern regarding PCI. With the proliferation of notebook PCs driven by low-cost LCD technology, the concept of slotless PCs is gaining momentum. As consumer electronics suppliers stop making tube TVs and the market for LCD or other flat-panel display technologies continues to expand, the PC without slots will conceivably replace the classic box PCs that most of us have hiding under our desks.

So while PCI has a migration path that will allow current-generation and next-generation PCI boards to reside in the same PC, the prevalence of slots in which to install those boards could be minimal as slotless PCs take over. The net result may be that PCs with PCI slots will be hard to find and more expensive than their closed-box alternatives.

PXI
The PXI plug-in board standard has gained modest success in test-system development and is based on the same architecture as PCI. Despite the amount of advertising and publicity, the market share of PXI today still is a small fraction of what IEEE 488 and some of the other architectures enjoy. Consequently, you should seriously consider whether PXI has the market share and technological foundation to be a thriving and viable standard in 10 years.

Also ask PXI suppliers what their upgrade path is relative to the new higher performance PCI standards such as PCI-X. While PXI may be a perfectly sound architecture, if the total market size of PXI is not large enough to attract a substantial number of suppliers long term, then the prospects of it being a thriving and well-supported platform in 10 years are low.

Ethernet
The long-time Ethernet standard has the advantage of replacing itself with backward-compatible versions every decade or so. Ten years ago, most Ethernet installations were using 10Base-T and possibly considering upgrading to 100Base-T. (100Base-T is 10 times faster but 100% compatible with 10Base-T with an off-the-shelf 10/100 Ethernet switch.) Today, new systems are moving toward 1000Base-T, which is 10 times faster than 100Base-T but 100% compatible with both 100Base-T and 10Base-T with an off-the-shelf 10/100/1000 Ethernet switch.

Given this upward mobility, it is highly likely that a system based on Ethernet today will be viable and maintainable 10 years from now. New instruments with Ethernet are constantly being introduced. Combine this with the proliferation of slotless PCs, and Ethernet has a bright future in the test-system world.

LXI
An emerging instrumentation stand-ard called LAN eXtensions for Instrumentation (LXI) uses Ethernet as its communications backbone and will take a stab at being the next IEEE 488 for the instrumentation world. LXI is to Ethernet as PXI is to PCI. With backing by market leaders like Agilent Technologies, LXI could very well be the next long-term standard for test-system development.

LXI has the advantage of being based on Ethernet. So even if the LXI standards are not embraced long term, an LXI instrument still would be capable of attaching to any PC with an Ethernet interface.

USB
Like Ethernet, USB has taken the route of enhancing its performance while maintaining compatibility with its predecessor. Combine this with the proliferation of USB and newer, higher speed USB 2.0, and it is very likely that a system based on USB will remain viable 10 years from now and beyond.

USB may be the next RS-232, which has been in existence for more than 40 years and still is very commonly used. USB and USB 2.0 also are experiencing a proliferation of new data acquisition devices and instruments, so the trend is certainly in the positive direction. Unlike Ethernet and LXI or PCI and PXI, there has yet to emerge a movement to adopt and embrace USB as an instrumentation standard on a large scale.

Is the platform you select available with multiple form factors and PC connectivity alternatives?
How likely will your platform of choice today adapt to changing technology over the next 10 years? For example, IEEE 488 interfaces have been available for nearly every PC system starting with RS-232 to IEEE 488 converters 20 years ago and moving to IEEE 488 interfaces for ISA, PCI, parallel, USB, and Ethernet.

Some data acquisition PCI board suppliers offer 100% compatible products in Ethernet and USB platforms as well. This means that if you design a system today around their PCI plug-in board you can easily migrate your system using their Ethernet or USB equivalent with little or no changes to application code.

This virtually ensures you a migration path in the event that your platform faces obsolescence before your requirements for the system are exhausted. This offers the same advantages as in the IEEE 488 example.

Is your software environment likely to be around in 10 years?
It is very costly for software suppliers to upgrade their environment as new operating systems become available. If the environment that you choose does not have a critical mass of users that enables the supplier to invest in continuous enhancements and migration to new operating systems, then you could find yourself with an obsolete or unsupported operating environment in five to 10 years.

Some sure bets to be around in 10 years would be Visual Basic, C++, and LabVIEW, based on the size of their installed base. If youï¿½re considering an environment other than one of these, it would be prudent to learn from the supplier whether, for example, they have support underway for Longhorn, the next 64-b operating system from Microsoft. Some currently available platforms are, in fact, still only available in the older 16-b foundation, which means that they hobble around on todayï¿½s 32-b operating systems and probably will be even worse when Longhorn debuts in the coming years.

Conclusion
If the intended life of your PC-based test system is more than five years, then it is important to pick a hardware and software platform that has a solid foundation if you expect your test system to be upgradeable and maintainable. There are no sure bets. But if you consider these issues, youï¿½ll decrease your chances of having to redesign your system.

About the Author
Tom DeSantis is the founder and president of IOtech. He holds a bachelorï¿½s degree in electrical engineering and has more than 20 years of experience in the test and measurement industry. IOtech, 25971 Cannon Rd., Cleveland, OH 44146, 888-890-1322, e-mail: [email protected]

December 2004