A virtual instrument reality check: Where's it really at?

Edited by Senior Tech Editor Alex Mendelsohn

eeProductCenter (11/12/2004 12:58 PM ET)

Leaps in processing power, I/O speeds, and local memory characterize today's system-ready instruments. Here's why they're a major evolutionary phase, packing every bit as much impact as IEEE-488/GPIB did when it was introduced back in 1972. This in spite of so-called virtual instruments.

By Marsh Faber, Program Manager for Agilent's System Components Division, Agilent Technologies Inc., Palo Alto, Calif.

What image comes to mind when you hear the word instrument? Is it an image of a box with knobs and a display, or a plug-in card for a PC?

Great effort has been spent of late to evolve the PC plug-in card into a respectable measuring device. Indeed, as the card migrated from the hostile environment inside a PC mainframe to the more compatible surroundings of an external card-cage, the result has been a series of products capable of reasonable accuracy and excellent digital speed.

These have been dubbed virtual instruments. They're virtual because they lack a front panel, and rely heavily on the PC for brain power and a user interface.

However, the term virtual is inexact. Certainly there is nothing virtual about the skin you scraped off your knuckles when you install the card into the PC. The hardware is real because there must be a physical interconnection between a unit under test and the measuring device.

The Software Isn't The Instrument

The term virtual stems from the software's ability to re-define what the card does, within certain physical and electrical boundaries. The software isn't the instrument.

In the fervor over this matriculation of PC cards into virtual instruments, some folks assume that system performance of real standalone rack-and-stack instruments has gone unchanged. But while PC cards have been evolving, rack-and-stack instruments have been doing some morphing of their own. Let's examine the evolution of both.

PC Plug-in Evolution

For the engineer whose main analysis tool is the PC, a natural step was to add some measurement capability. The first cards were unsophisticated by today's standards, lacking important elements such as isolation.

The hostile EMI environment of a PC also limits a card's capabilities, and to this day restricts the noise-floor of PC plug-ins. This issue is mitigated by moving the card outside the PC and into a more suitable card-cage environment, where EMI (electromagnetic interference) and shielding are considered. The final step is to place a specially designed PC inside the card-cage for the best backplane speed.

Virtual Instruments evolved as PC plug-ins (left), with brain power existing primarily in the PC. The move to an external card-cage (center) still requires the PC to do most of the work. A third architecture (right) puts the PC inside the card-cage.

The tradeoffs surrounding these different configurations shown above are well known. On the downside, the PC plug-in exposes the PC to damage from the measuring device, and is subject to hostile EMI signals. On the flip side of the coin, plug-ins are reasonably priced and offer good backplane performance.

There are several card-cage versions. Two of primary interest to you as a test engineer are VXI and PXI. VXI accommodates large cards, for maximum function and well-specified shielding. PXI has less real-estate, hence lower functionality per card, but PXI is also less expensive than VXI.

Both have high-speed backplanes and the ability to add a PC-on-a-card. Both are also supported by hundreds of manufacturers.

Where's The Brain?

It could be argued that the central difference between virtual instruments and real instruments is simply where you locate the test instrument's brain. In the case of a purely virtual instrument, the intelligence resides in the PC.

That means the PC can potentially re-define what the instrument does, but it also means the PC is burdened by making all the realtime decisions for the instrument, as well as doing all of the GUI (graphical user interface) and analysis.

If the test system has multiple instruments, the PC has to juggle all the decisions for the system, and that can be a programming nightmare. As shown in the previous figure, the majority of the brain power resides in the PC itself.

When the PC is installed directly into the card cage, it brings the intelligence closer to the point of measurement for improved triggering, but it still forces the PC to make all the realtime decisions. Newer plug-in cards can support FPGAs (field programmable gate arrays) on the board for more distributed gray matter, but this approach puts the burden of instrument development on the user.

Standalone instruments initially had little processing power onboard, but today's new system-ready instruments are smaller and smarter, freeing the PC for analysis. Some can download instrument personalities, and store local lists and data to improve system speed.

Before the days of the PC, the standalone instrument had to work without a computer and carry all its own brain power (as shown in the figure above). As PCs and microprocessors became available, the instrument not only connected to a PC, but also added to its own internal processing power. This onboard intelligence permitted the instrument to act autonomously, and not slow down a system during a measurement.

The Evolution of Standalone Instruments

The latest generation of standalone instruments typically has one, two, or several microprocessors. An oscilloscope, for example, might carry one processor for the instrument's I/O, another for its GUI, and a third for controlling measurements---in addition to DSPs (digital signal processors) that take care of data reduction.

That's an amazing array of brain power in the modern real instrument. It significantly reduces the amount of programming, too, because the user can take advantage of algorithms inside the instrument, developed by measurement experts.

Measurement throughput also requires more than just sheer I/O speed and processing power. For leading-edge measurements, the processor must be located close to the source of data. Although today's PCs operate with impressive speed, the real issue becomes how to move megabytes of data into a non-deterministic environment, and then turn around an answer in time to make the next set of measurements.

Evolution of instrument I/O from standalone (left) to GPIB (center) to PC standards (right). New system-ready instruments take advantage of modern low-cost and fast PC I/O, giving test engineers many new possibilities in system architecture.

By locating the intelligence inside the instrument, and equipping it with sophisticated data-handling hardware, the system-ready instrument can apply time-tested algorithms and impressive processing techniques so the PC can concentrate on managing the system and analyzing results from multiple instruments at the same time.

The System-Ready Instrument

I/O speed is frequently mentioned when comparing card-cages against standalone instruments. Things are about to change dramatically.

The previous figure shows the evolution of I/O for standalone instruments. Initially, there was no I/O, or the I/O that did exist was proprietary and slow. Then Hewlett-Packard (now Agilent Technologies) engineers invented GPIB (the General Purpose Interface Bus, defined by IEEE-488).

GPIB changed the world of test equipment by enabling multiple instruments of varying speeds to exist on the same communications bus. While GPIB is still the choice of most engineers, the PC industry doesn't recognize it as a popular standard.

The PC industry has recently spent vast sums of money coming up with new communications vehicles that also happen to answer the test-and-measurement industry's needs for low latency and high transfer speed.

The Open Door

That leaves the door open for the test industry to adopt true PC-standard I/O, along with its inherent advantages of fast data transfer, small lightweight cables, and very low cost.

A LAN (local area network) controller card is built into most PCs, and the cable might cost five dollars. Compare that with a $500 GPIB card and $100 cable, or a $1500 MXI card. In addition, the use of standard I/O means the standalone instrument can now become a NET object, opening up all sorts of architectural opportunities. As we combine multiple local processors inside the standalone instrument with PC-standard I/O, the power of the test system expands dramatically.

And that's where we're heading with new instrument design. The system-ready instrument not only has the advantages of fast and cheap I/O and local brain power, but it maintains its standalone capability. That leaves it available to troubleshoot the system and provide an extra locally viewable display, or one that can be used for setting up the test system by manually verifying DUT (device under test) tests and the fixturing, before software is completely written.

Radical Change For Standalone Instruments

While lots of media attention is focused on virtual instruments, the standalone instrument has simultaneously been undergoing similarly radical changes, making it more system-friendly, fast, and smart.

The system-ready instrument is real hardware with a local human interface and local intelligence, designed in a smaller package for system use, optimized for measurement performance, and ready to converse with a PC over standard high-speed PC I/O protocol.

Because of its high manufacturing volume, the standalone system-ready instrument can typically be bought for a lower price than open-standard card cage instruments, too.

System-ready instruments have made significant leaps in processing power, I/O speeds and local memory as well. As such, system-ready instruments are a major evolutionary phase in the test-and-measurement industry, and they will have every bit as much impact as GPIB did back in 1972.

About The Author

As Program Manager for Agilent's System Components Division, Marsh Faber has many years of experience and a varied background, first with HP and currently with Agilent Technologies. His roles have included production engineering, test engineering, research and development, applications, product planning, and marketing. Faber helped initiate the VXI business with HP, and later established the HP marketing center in Hong Kong, where he and his family lived for several years. He was also responsible for starting the Education Marketing program at Agilent.

Related Stories:

» Why Ethernet reigns supreme for test-and-measurement instrumentation