Many test-and-measurement users have to deal with dramatically increased production rates these days. Manufacturers are pushing for higher production throughput. At the same time, product life cycles and the cost of manufacturing continue to decline, while the cost of testing hasn't typically kept pace.
Increased product complexity and advances in technology and product integration are escalating requirements on production test equipment. Products are also undergoing evolution during production. That is driving the need to modify test systems during production runs.
In situations where product life cycles are long, there are significant issues with the obsolescence of test assets. Moreover, improved product performance drives the need for higher quality system-level calibration and self-tests that execute faster and at run time.
Good Old Rack-and-Stack
Back to top
Typically, conventional test-ystem providers take a combination of benchtop instruments or instrument-specific modules, and they rack them up with the appropriate interconnect cabling and connectors between the instruments and a product. They then add software that makes calls to the functional capabilities embedded in these instruments. This is known as the rack-and-stack approach to test-system development.
Table 1 lists examples of such instruments.
Table 1 - Conventional Instruments
(most are also available as instrument modules)
Figure 1 illustrates an example of a simple rack-and-stack RF (radio frequency) test system.
Figure 1 - Block-diagram example of a conventional RF test system based on benchtop instruments
Benchtop instruments are usually designed for use in a lab environment. They provide standalone capabilities, but some suppliers also include the option of a rack-mount kit for integration into these test systems.
Some instrument modules include the same core hardware as benchtop instruments, but they don't include the user interface, power supply, or other ancillary hardware elements. Many instrument modules plug into a backplane and rely on a remote computer for user interface and control functions.
Conventional test systems using these elements are common, but they have a great deal of difficulty addressing the challenging problems facing test-and-measurement users today.
Limited by Design
Back to top
Rack-and-stack approaches are usually limited by design. They typically test one component, module, or subsystem, with limited ability to support any testing of a broad range of products. At best they can test a closely related family of products.
As product life cycles and the cost to manufacture products continue to shrink, it becomes less cost-effective to justify dedicated test systems. To keep product test costs down, a test system must now test many products in support of a reasonable Return on Investment (ROI) for each of the individual products.
Many benchtop instrument-based systems also rely on low-speed communications buses, such as the venerable IEEE-488 General-Purpose Instrument Bus. GPIB significantly limits test throughput.
However, this is only part of the problem. Significant problems also stem from the fact that system integrators are generally limited to using the functions and test algorithms built into the individual instruments at the time of manufacture. In many instances the configuration of the product slated for testing will permit optimizing the test algorithms for speed and performance, but with traditional instruments a system integrator doesn't necessarily have access at this level.
Many times a signal waveform or data protocol used in a product will change over time. These changes can result in the need to change one, or even all, of the instruments in the test system. The conventional RF test system (see Figure 1) illustrates this point.
If the original product, or unit under test (UUT), is operated below 3 GHz, a test-system integrator would typically select spectrum analyzers, synthesizers, signal generators, or vector signal analyzers covering this frequency range. This would be done to reduce initial procurement costs—especially in a competitive bid for a test system.
However, if the next generation of the product were to move into the 6 GHz range or higher, for example, it would require reprocuring and re-integrating all of the instruments with the limited 3 GHz frequency coverage, despite the fact that the processing power and instantaneous bandwidths of the 3 GHz elements may still prove adequate for testing the upgraded product.
This process can result in significant nonrecurring costs, and it contributes to the proliferation of test systems specific to individual products.
All of these issues affect military, aerospace, and commercial test system applications. However, the problems associated with instrument obsolescence challenges military applications to the highest extent. That's because of their long operational life cycles.
Military test systems typically remain in service for more than 20 years. Obsolescence issues in these systems arise largely because test software is written around instrument-specific hardware. The software drivers are also correspondingly specific, making evolutionary enhancements more like revolutionary redesigns.
The Obsolescence Conundrum
Back to top
Replacing obsolete devices can also result in a situation where no functionally equivalent device exists. In some circumstances, system integrators must replace one instrument with several to achieve equivalent function, and conventionally architected test software can make this task difficult.
Last, but not least, system calibration and self-test are becoming more important in terms of the total cost-of-ownership of a test system. The instruments used in rack-and-stack approaches are designed primarily for standalone operations and calibration. In many cases, when these instruments are integrated into a system, calibration and functional test requirements become difficult to perform—and in some cases become impossible.
The performance levels of today's products require the execution of calibration routines at the integrated system level. This approach is necessary in order to have any chance of meeting the standard 4:1 or 10:1 performance requirements typically levied on the test system.
Challenges for Suppliers
Back to top
Test-and-measurement suppliers are being challenged to supply flexible, scalable, and efficient test-system approaches that solve these problems. The goal is to cost-effectively meet the demands of today while preserving the investments of the future. As such, a technique dubbed synthetic instrumentation has emerged.
Synthetic instruments synthesize the stimulus or measurement capabilities found in conventional instruments. They do this through a combination of software algorithms and hardware modules that are based on core instrumentation circuit building blocks.
The concept of synthetic instrumentation finds its roots in the well-accepted technologies and techniques behind so-called software-defined radios (SDRs), mobile phones, and other similar communications systems.
To better describe and define the concept of synthetic instrumentation, let's review SDRs. An SDR (see Figure 2) consists of a digital signal processor (DSP) engine, a generic transmitter and receiver front-end, and some form of transmission antenna.
Figure 2 - A software-defined radio, or SDR
The generic transmitter and receiver front-end converts digital data to and from modulated radio waves for wireless communications. A high-speed DSP that provides most of the radio function sits behind these components.
In essence this combination provides a generic radio. The radio designer programs the functions of the radio into the DSP. The designer writes software algorithm modules and control modules that generate or process digitally represented signals at the inputs and outputs of the DSP. If the communication protocols or the processing algorithms need modification, or if the radio must now communicate as a different radio type, the designer need only modify the radio's software.
The SDR approach removes much of the necessity for redesigning and manufacturing new hardware, as with earlier dedicated radio designs. The processing speed and power available in today's small packages makes SDR implementations possible with relative ease.
Let's look at a block diagram of a synthetic instrument (see Figure 3).
Figure 3 - Block diagram of a synthetic test system
It looks very similar to that of an SDR, doesn't it? The main differences are the replacement of the antenna with interfaces to the product that will be tested, as well as the addition of multiple levels of circuitry to support even more flexible signal conditioning.
There are also signal paths that permit reconfiguration or bypassing of circuit elements as desired. These simple modifications enable the real power behind synthetic instrumentation.
To walk through the fundamentals of synthetic instrumentation, let's look at just the stimulus path (see Figure 4). If the test system is modularized around the functional building-block circuits commonly found in many signal-generating instruments, you'll generally find some type of DSP engine to create the signal waveform, followed by a D/A converter.
Figure 4 - Synthetic-instrument signal stimulus path
In the case of RF signal generation, an up-conversion block will be found. Its primary purpose is getting the signal from baseband to RF. In an integrated system, the RF output will then connect to the UUT through some cabling-and-switch matrix.
In a conventional instrument these circuit elements are internalized, locked up inside a box with no possible way to access intermediate circuit capabilities without significantly compromising the integrity of the actual fixed instrument. In a synthetic instrument these functional blocks represent separate modules with access to the inputs and outputs.
Functional Partitioning
Back to top
This so-called functional partitioning permits a user to put signal switching between functional blocks, and to exploit the fundamental capabilities of the individual circuits as well. For example, if you connect all of the blocks serially, you'll have an RF and microwave signal generator. If you tap off the signal between the D/A converter and the up-conversion module, you'll have an analog signal generator or function generator. If you take the output directly from the DSP, you'll have a digital pattern generator.
By combining various signal-generation, capture, and signal-conditioning modules, a synthetic test system may include the synthesis of digital, analog, power, RF, microwave, and many other signal types.
Note that a synthetic instrument may also include redundant or parallel paths (see Figure 5). For example, a system can have a high-resolution and narrow-bandwidth D/A converter module, as well as a low-resolution and high-bandwidth module.
Figure 5 - Synthetic-instrument signal stimulus path
with the addition of some parallel paths
The synthetic-instrument architecture focuses on partitioning the system by the fundamental circuit building blocks necessary to construct the required stimulus signals and measurement analysis. The design can include as many different signal-conditioning modules as required.
The digital processing modules can be duplicated as well for situations where multiple simultaneous signals are required. This permits you, as a system designer, to provide as many parallel stimulus or measurement paths as required. However, since the architecture supports such high levels of re-use, it also minimizes the redundant elements to only those required.
By maximizing the use of individual modules and reducing redundancy, a synthetic-instrument supplier can procure higher performance elements to satisfy tougher requirements. Since the user isn't paying for unnecessary duplicate copies of many functional elements, it's possible to reduce the cost of the overall system while still increasing system performance.
For example, if three different instruments include a DSP, D/A converter, filter, and attenuator circuit blocks, it's possible to apply a portion of the money saved toward the purchase of one higher performance set. These factors are highlighted in Figure 6.
Figure 6 - Synthetic instrumentation reduces the need to procure redundant
hardware and permits the insertion of higher performance circuit elements
These reductions in test assets have the added benefit of reducing the size of the test system, as well as reducing the number of hardware spares required to support it.
Aiding the Calibration Process
Back to top
This access to low-level circuit building blocks greatly aids the calibration process. Figure 7 shows a basic synthetic instrument block diagram, with the addition of calibration and system functional test (SFT) loopback circuitry. This circuitry can include simple loopback switch paths, along with calibrated sensors and related hardware.
Figure 7 - Basic synthetic-instrument block diagram with the addition
of calibration and system functional test loops
It isn't possible to break into the middle of the circuitry chain of a conventional instrument. This makes such instruments difficult to calibrate. More importantly, it makes conventional instruments difficult, if not impossible, to calibrate effectively at run time.
The synthetic-system approach provides the ability to calibrate each functional element, and allows for the tailoring of calibration routines for varying types of measurements. In most cases this results in greatly improved system performance.
The synthetic architecture also enhances the ability to upgrade the test system. It also enhances the systematic handling of obsolescence issues. When an upgrade or obsolescence situation arises, it's only necessary to add or replace the functional blocks directly impacted—not the entire instrument suite. This reduces the cost of handling obsolete instruments, in addition to reducing the technical risks associated with the effort.
In Part 2 of this two-part ChipCenter feature, author Marvin Rozner, Jr., continues the discussion of synthetic instrumentation by looking at how the technique treats hardware in the much the same way that object-oriented programming treats software. Synthetic implementations use software objects that correlate one-to-one with the functional hardware modules, and they also implement stimulus and measurement algorithms as software objects.
