Test's promised land?

The test industry serves two major markets: military-equipment makers, who often need test systems to last for 20 years, and manufacturers of consumer products, whose product lines can change every few months.

Figure 1. A synthetic test system uses signal conditioners, data converters, and a processor to form a generic hardware set. Signal conditioners include amplifiers, level shifters, attenuators, and delay lines. Calibration includes traceable sources and sensors.

"Synthetic" instruments could be the answer for both industries. These instruments could make tester obsolescence obsolete for the military and make hardware reuse a daily occurrence for commercial manufacturers.

The concept of synthetic instruments grew out of the military's need for replacement instruments that are compatible with existing models. Over the years, standard commands for programmable instruments (SCPI), virtual instrument software architecture (VISA), and interchangeable virtual instruments (IVI) partially eased the software compatibility problem. Synthetic instruments, which make use of a collection of components and automated software, could finish the job that these other technologies started.

Right now, synthetic instruments are still in their infancy. Only a handful of companies—including Aeroflex, Honeywell, and Teradyne—build test systems with synthetic instruments. A few others, such as Phase Matrix and Agilent Technologies, make components or subassemblies that are used in such test systems. Despite offering a promising alternative for manufacturers seeking test equipment that can serve for many years, synthetic instruments won't gain widespread acceptance until they become easier to develop or obtain.

Currently, engineers build automated test stations with a set of instruments—oscilloscopes, DMMs, spectrum analyzers, etc.—that come in boxes or on VXI or PXI cards. In contrast, synthetic instruments use a common set of modular hardware components and then implement measurement-specific tasks in software (Ref. 1). Synthetic instruments differ from virtual instruments in their purpose and implementation. (See "Synthetic or virtual?  ".)

Figure 1 shows the concept for an RF/microwave test system based on synthetic instruments. A measurement channel's hardware consists of an RF downconverter and an analog-to-digital converter (ADC). The downconverter filters the incoming signal and mixes its carrier frequency down to an intermediate or baseband level where the ADC can digitize it. If you remove the RF components from the system, it becomes a reconfigurable analog baseband instrument such as an audio analyzer. If you use the digital signals directly from the processor, you can make a logic analyzer or digital pattern generator.

Software modules—running on a host computer, a VXI or PXI slot-0 controller, or a separate processor—contain all the code needed to produce measurement results. In RF applications, a synthetic instrument's measurement channel resembles a software-defined radio, but software produces measurement information rather than an audio signal.

To synthesize a stimulus signal, the processor generates a digital representation of an analog signal and feeds it to a digital-to-analog converter (DAC), thus producing a baseband analog signal. In RF applications, an RF upconverter shifts the baseband signal's frequency to the proper band.

The RF converters, ADCs, and DACs reside on VXI or PXI cards or on other modular components. A set of switches, which also generally reside on cards, route signals from the I/O channels to the unit under test. A system may also include traceable sources and sensors for calibration.

Because they're reconfigurable though software, synthetic test systems should be easy to upgrade. To add a measurement function to a system to accommodate a new product, you just add a software module rather than add an instrument. To increase capacity, you simply add I/O channels. When a better ADC or DAC card comes along, you can just replace your existing channel hardware with the new converter card.

Most synthetic-instrument test systems need a set of ADCs, DACs, upconverters and downconverters. "One size doesn't fit all," reports Mike Granieri, VP of business development for aerospace and defense at Phase Matrix (San Jose, CA; www.phasematrix.com), a company that makes upconverters and downconverters. "A synthetic test system may need different ADCs for different applications based on accuracy, resolution, and bandwidth." For example, a system may need a 24-bit, 100-ksamples/s converter for an audio measurement and an 8-bit, 5 Gsamples/s converter for an RF or digital data stream measurement.

RF and microwave synthetic test systems may need several upconverters or downconverters to satisfy the unique filtering and frequency translation needs of the application. Granieri points out, "If you don't specify and implement your application's upconversion and downconversion requirements properly, the resulting measured signal may be inaccurate and not represent the stimulus or measurement signal of interest."

A synthetic instrument needs system software and measurement-specific software modules. The software may run on a processor such as a common PC, an embedded slot-0 controller, a digital signal processor (DSP), or a field-programmable gate array (FPGA). Each type of processor has advantages and disadvantages. Running software on a PC, for example, may produce the most flexible and lowest-cost system, but it may also produce the slowest one.

Figure 2. Processors force tradeoffs between speed, flexibility, and cost.

Suppose you need to measure a signal's spectral content. In a traditional test system, you might use an oscilloscope and its built-in FFT function or a spectrum analyzer. You can also digitize a signal with an ADC card and use an FFT routine that comes with your test programming language. Early adopters of synthetic instruments, though, will most likely develop FFT software modules themselves.

"Synthetic instruments bring measurement science closer to the test engineer," says Frank Angelo, general manager of Agilent Technologies' Systems Products Operation (Santa Rosa, CA). "Software will be complex, especially at first." Early adopters of synthetic instruments who want to implement their own functions may need to learn DSP or FPGA programming and the tools that make it possible.

Over time, software functions that implement specific measurements will become commercially available, thus eliminating the need for you to learn DSP or FPGA programming. The industry will devise tools that will let you create functions and applications with software-development tools you already know how to use, such as LabView, Mathcad, Matlab, Vee, or Visual Basic.

Marvin Rozner, VP of Business Development at Aeroflex (Bowie, MD, www.aeroflex.com) says, "Test-system integrators will develop software modules that contain a measurement or stimulus function, calibration coefficients, command and data set descriptions, and signal-processing functions. Modules will also contain diagnostics information, default parameters, and any other information necessary for the module to operate."

Rozner continues, "Test executive software uses this information to assemble the modules in the proper sequence to construct the desired stimulus and measurements circuits. This assembly process can take place in preprocessing or at run time depending on the test executive's capabilities."

To specify the test sequence in a synthetic system, you'll probably fill in a table with the measurement parameters you need. The tester's "operating system" will assemble the function by figuring out which hardware and software modules it needs in order to implement measurement or stimulus signals. Initially, proprietary software may perform the assembly, but not for long.

Already, test-system developers and people from the military have formed a group whose goal is to standardize the language used for programming automated test systems. The group is developing extensible markup language (XML) schemas specific to the ATE industry, thus creating automated test markup language (ATML). (You can learn more about the ATML project at www.atml.org.)

Through a common hardware set and measurement functions implemented in software, synthetic instruments may eliminate the need for individual box or card-based dedicated instruments in ATE systems. "Synthetic instrument obsolescence and upgrade issues require just a few modular hardware blocks, not a myriad of instrument types," notes Granieri of Phase Matrix. "In an ideal synthetic system, hardware components won't employ application-specific firmware, thus enabling you to easily upgrade and add hardware."