The Industry Standard Architecture (ISA) bus has served the test-and-measurement community well for 15 years—a long time in today’s quickly changing world of electronics. Since its introduction, the ISA bus has become the universal standard, permitting the development of new products and uses that still dominate the PC-based data acquisition market.
But next year marks a turning point. 1998 may be the last year you can readily buy a computer with an ISA bus inside it. This could make your ISA-bus-based data acquisition plug-in board destined to join the rotary phone and the vinyl LP in the scrap pile. As we put up next year’s calendars, this is a good time to look back at the achievements of the ISA bus and forward to its successors.
A Quick History Lesson
IBM developed the ISA bus to serve as the internal bus for the original IBM PC in 1981. Despite the fact that the first PC had about as much power as a wristwatch, PC plug-in boards became a popular tool for data acquisition (Figure 1). The appeal of plug-in boards widened in 1984 when IBM introduced 16-bit Advanced Architecture for the PC. This doubled the typical throughput of the ISA bus from 1 MB/s to 2 MB/s.
The industry responded by making a wide variety of ISA-based plug-in boards for many different types of data acquisition applications. Today’s data acquisition plug-in boards feature sampling speeds up to 1 GS/s, up to 64 channels, and various combinations of digital I/O, analog output, and counter/timer channels.
Over the years, competing bus structures were introduced but none of them reached critical mass. IBM improved on the ISA bus with a new 32-bit bus called Micro-Channel. It solved most of the problems associated with the ISA bus: Micro-Channel was faster and was plug-and-play. But it was not compatible with previous ISA plug-in boards. And IBM did not widely license the bus, leading to a perception that it was an IBM proprietary bus. Consequently, Micro-Channel never became popular.
Extended ISA (EISA) came along in 1989. Unlike Micro-Channel, it was backward-compatible with ISA boards. In other words, you could plug an ISA card into an EISA slot although you could not mix and match ISA and EISA cards on the same PC.
EISA was faster than ISA, but demand for the bus never developed. Because of its cost, it was not widely implemented in PCs, and many plug-in boards did not require the increased performance that it offered. Eventually, EISA was used in file servers and high-speed graphics—applications where the increased performance was needed.
Three years later, when the VL bus became available, manufacturers preferred it over EISA because VL offered EISA-like performance at lower cost. Originally, VL was used primarily in high-end graphics applications. A consortium of companies formed the Video Electronics Standards Association (VESA), and together they developed the VL bus to improve the performance of video cards.
As an expansion bus, it did not replace the PC’s internal ISA bus, but it did fit on top of it, on the 32-bit pathway between the processor and the plug-in card. VL was the first non-ISA bus to gain wide acceptance.
Because vendors were not implementing VL consistently, the industry sought to standardize its implementation. The standardization effort broke into two groups: one led by VESA and one led by Intel.
Five years ago, Intel and its partners offered a new bus that improved on the VL bus design and had the capability to evolve in step with Intel processors—the Peripheral Connect Interface (PCI) bus. The rapid acceptance of the PCI bus has made it the new de facto standard for the PC industry.
As a major manufacturer of PC chip sets and motherboards, Intel has a captive market for its PCI bus, and it is widely available on PCs today. As if Intel’s support for PCI was not enough, Microsoft dealt the coup de gras for the ISA bus in April. The company announced that ISA was too slow to use with Microsoft operating systems such as Windows® 95 and Windows™ NT.
A Better Bus
By combining marketing muscle with a captive market, Intel’s new bus succeeded where other buses failed. But that’s not to say that this bus is being forced on a reluctant data acquisition industry. There is a high demand for a better bus, one that will take advantage of today’s increased computing power, cope with GUI-intensive operating systems, and meet user expectations for ease of use.
Compare the PC of 12 years ago with the PC of today. Back then, a 286 PC had a clock rate of 8 MHz to 12 MHz. Today’s Pentium PCs have a clock rate of 150 MHz to 300 MHz. Compared to where we started, speeds have increased 25-fold, and processing power has increased by a factor of 50 to 100.
Despite this remarkable advancement in computing technology, the ISA bus remains basically the same as it was 12 years ago. The disparity between computation capability and communications is enormous.
At relatively slow processing speeds, ISA-based data acquisition plug-in boards are reasonably efficient. The boards and other peripherals wait their turn for processing time, issuing interrupts, and waiting for replies. Under an ISA-bus I/O cycle, there are only a few wait states at 12 MHz. But the new Pentium processors subject the boards to 100 to 200 wait states at 200 MHz. This slows the performance of the entire system.
The operating systems have changed, too. Before the days of Windows 95, the typical PC did only one thing at a time. It did not matter, then, if the ISA bus kept the PC’s processor from doing other tasks during data acquisition.
But in today’s multitasking environment, it is inappropriate for a single application to consume so much of the PC’s resources. No wonder Microsoft and Intel are steering their customers away from ISA-based peripherals. Compared to the speed of events happening in the modern computer, the ISA bus has become a bottleneck.
Besides hardware/software considerations, you want peripherals that are easy to use. PCI, Universal Serial Bus (USB), and FireWire buses offer plug-and-play capability. The ISA does not. You have come to expect a certain level of automatic configuration from the computer products you buy. You do not have time to mess around with an ISA-based plug-in board if a PCI-based plug-in board will set the configuration for you.
Where the Market Is Headed
What will happen to the $200+ million market for ISA-based data acquisition plug-in boards? It’s too simple to assume that all of the market will convert to the PCI-bus structure.
More likely, it may convert to a combination of PCI-based boards and external solutions that connect to the PC via one of the new high-speed communication ports, such as USB (Figure 2). In any case, it is not clear that PCI plug-in boards are the obvious successors to the ISA plug-in market. There are pros and cons in each direction.
For one thing, there are plenty of applications using three-year-old computers containing ISA buses that will still need ISA bus plug-in boards. Applications developed under Windows 3.1 or DOS will not run under the operating systems that ship with the new computers. Compared to rewriting the data acquisition application from scratch, using an existing ISA plug-in board is an attractive alternative.
Familiarity is another advantage of the ISA bus. You can create your own plug-in board when one is not commercially available. It is far easier to create a driver in DOS or Windows 3.1 than it is to program for a more sophisticated operating system such as Windows 95 and NT.
Windows 95 and NT will not permit easy and direct access to the hardware. While it may have been common in the DOS days, you almost never write custom board-level drivers anymore. Instead, you expect the plug-in board vendors to provide the drivers for you.
Unfortunately, plug-in board vendors will not always continue to make DOS or Windows 3.x drivers for their products. Consequently, to use the latest plug-in board, you must install the new operating system, and the new operating system will not run well on an old PC.
Also, system expansion is easier with ISA. ISA-bus-based PCs can have up to 20 slots for plug-in boards, but PCI-bus-based PCs can have only four slots (up to six with Compact PCI) and require a bridge for further expansion. Nevertheless, for most PC-based data acquisition applications, four slots are plenty.
It seems ISA has as few advantages as PCI has many. Most of you will look for PCI plug-in boards to replace existing ISA boards. But this is only part of the picture of data acquisition in the future.
PC Cards and External Solutions
Formerly known as PCMCIA cards, PC Cards add functionality to a notebook or laptop PC which cannot accept ISA or PCI plug-in boards. If you look hard enough, you can find a few PC Cards and laptop PCs using a new bus called CardBus. CardBus is essentially PCI in a PC Card format.
Although better than ISA for the same reasons, this new bus does not solve the limitations inherent in the PC Card format: a small space on which to build functionality and a generally fragile cable connection. For this reason, the emerging external-port-based data acquisition systems probably will be preferred over PC Card systems, new bus or not.
Of course, data acquisition does not have to occur inside the PC. Many data acquisition systems reside outside it, interfacing via a standard communications port.
External solutions also avoid the noisy internal PC environment and are well-suited for low-level signal applications. They offer more physical space for electronics, better isolation, and convenient connectivity so the data acquisition system can be located close to the signal source.
The USB already is shipping on most new computers. It is a neat, plug-and-play way to attach peripherals such as joysticks, mice, and keyboards. The USB standard calls for 500 mA of power at 5 V to be conveyed over the same cable that provides communications with the PC. That means that a USB-based data-collection device does not need a separate power supply.
The USB specification also allows the cable to be up to 5 m long, making it more convenient to place the device closer to the signal source. Typically, the USB offers 500 kB/s throughput (theoretical limit is 1.5 MB/s). Using a 16-bit A/D converter, this translates into about 250 kS/s, more than sufficient for most data acquisition applications that typically require 100 to 200 kS/s. As time goes on, the industry will embrace USB as an excellent platform for data acquisition.
FireWire is a faster, external, high-speed serial bus that, like USB, also supplies power. Its widespread implementation is still to come. By the end of 1998, it will begin to show up in chip sets. Its first major application will be for the home digital TV market.
FireWire will communicate large amounts of data at high speed, promising a 200-MB/s throughput. With options like FireWire and USB, a significant portion of the market for plug-in boards could be replaced by external solutions.
Remember that data acquisition is a small part of the total PC peripherals market. The real driving force is the convergence of computing and home entertainment. The manufacturers of data acquisition plug-boards will have to adopt whatever technology the PC embraces.
Watch for the trend toward making the PC more like an appliance that has no slots and which you cannot open. This trend will make external solutions the primary method for attaching data acquisition to a PC.
About the Author
Tom DeSantis is the founder and president of IOtech. He holds a bachelor’s degree in electrical engineering and has 20 years of experience in the test and measurement industry.
Allen Tracht is the principal engineer at IOtech. He earned B.S. degrees in electrical engineering and physics from Massachusetts Institute of Technology and a master’s degree in electrical engineering from California Institute of Technology. Before joining IOtech, Mr. Tracht completed a medical degree and passed the Ph.D. qualifying exam in biomedical engineering.
IOtech, 25971 Cannon Rd., Cleveland, OH 44146, (216) 439-4091, www.iotech.com.
Throughput (in MB/s)
2 to 3
2 to 3
On the way out, many types available
Plug-and-play, fast, new standard
1 to 6
Fragile connection, limited space
0.1 to 3
0.1 to 3
Noise immunity, bus independent
Noise immunity, plug-and-play
High speed, automatic configuration
Copyright 1997 Nelson Publishing Inc.