Signal management

Also, don't underestimate the importance of signal management, or the design and configuration of the signal-switching system. A good signal-management system allows a functional tester to share test resources, to test multiple units under test (UUTs) simultaneously, and to control the environment of one or more UUTs. Properly configured, it helps the system designer minimize the tester footprint and therefore save on the bottom line.

Before you can design a functional test system, you must first fully understand the UUT. Study not only the product's intended use but also how many units will be produced, what faults to anticipate, and so on. You'll need to define these factors:

• type of product,
• configuration (one PCB, panelized PCBs, final product),
• test specifications,
• planned test points,
• anticipated quantity (per line/day/shift), and
• anticipated fault spectrum.

You also need to consider component density. The NEMI (National Electronics Manufacturing Initiative), in its 2002 roadmap, indicated that the amount of I/O per square centimeter will increase from an average of 23 in 2002 up to 55 by 2011. Clock speed will rise to nearly 20 GHz by 2011.

On the surface, component density does not appear to be an issue for functional test. After all, the primary concern is basically "something goes in, something comes out." Defined stimulus is applied to the inputs of the UUT, and a specific data set should come out of the UUT. Access to the I/O connectors should be the only access issue.

But component density is a factor for various reasons. For example, higher density is a good indication of high functional complexity—which in turn can stem from circuit complexity or high I/O count or a combination of both. Also, an increasing functionality per UUT, coupled with shrinking geometry, creates a test challenge. Smaller PCBs tend to be built in groups, or palletized, and they may be tested while still in the pallet, so you must find a way to share test resources among all the UUTs.

Because signal management can greatly affect the functional test system's wiring, chassis configuration, overall footprint, and cost, it should be one of the first factors you address in test-system development. You have to answer these questions about a PCB before specifying a signal-management system:

• What is the I/O count?
• How many of the pinouts are active at any one time?
• What types of signals (digital/analog) are present at the connector(s)?
• What instrumentation is required?
• Will access for calibration be necessary?
• Are diagnostics down to the component critical?
• Will any probing be performed by a human or some robotic means?
• Will automated handlers be used?
• Are the I/O connectors used easily probed or connected to?

The specifications of the analog channels—including voltage, frequency, and current requirements—and the bus width and speed of the digital I/O help to determine what types of instruments you need. Budget limitations coupled with the beat rate of the production line will dictate that limits be placed on the number of simultaneous stimulus and response channels.

Your first step in specifying a signal-management system is to select a signal switch. Switching can be as simple as an on/off connection. SPST (single pole, single throw), DPST (double pole, single throw), and other switch configurations (Figure 1) connect devices including power sources, loads, and mechanical actuators. In most cases, the primary issues are voltage, current, and power—bandwidth is hardly ever a problem for power and mechanical-actuator connections. As the frequency goes up, though, you should use shielded (or "screened") switches.

In most cases, the frequency response of the mux is also important. Defining the maximum number of connections needed is important if cost is a factor in your system configuration, but be careful not to underestimate your needs at the beginning: Expanding a multiplexer later on in the system design involves added cabling and modules, which can limit system bandwidth and add cost.

Figure 3.  A cross-point switch is flexible, although it can exhibit bandwidth limitations.

But matrices do have limitations. The number of possible connections and potentially long unterminated lead lengths can create unwanted "antennas," or "stub lengths," that limit frequency response. Matrices can also be expensive. The ability to make all possible connections between a UUT and an instrument or series of instruments requires numerous relays. For example, if you want to connect a four-wire measurement device to 96 possible connections, you'll need up to 1536 relays.

Once you select a signal-switching method, you can implement signal management. Figure 4 shows a typical system block diagram—the PC controller and stimulus and response instrumentation. The signal-management subsystem sits between the test adapter and the instrumentation. Whether this subsystem sits in the same instrument rack or stands alone will be determined by the type of instruments used—for example, PXI or IEEE 488.

Figure 4.  In a typical test system, the signal-management subsystem resides between the instrumentation and test adapter.

Another issue to bear in mind when you design a test system is how the instrumentation connects the UUT to the signal-management subsystem. Figure 5 shows a "pass-through" configuration. This design connects any instruments in the system directly to the test adapter. Wiring internal to the test fixture connects the instruments to the signal-management subsystem. This configuration is the most flexible as you can reconfigure the test system by changing test fixtures. The downside is that wiring lengths are much longer. In addition, each signal goes through the test adapter three times before being connected to the UUT. This can induce undesirable losses and affect signal integrity.

Figure 5.  In a pass-through setup, wiring connects instruments to the signal-management subsystem.

Figure 6.  Hard-wired configurations keep wire lengths manageable and provide the best signal integrity.

Many engineers make the common mistake of not considering the overall system specifications when configuring a test system. For example, a measurement channel may have to exhibit particular characteristics—such as line loss and capacitance. The engineer configuring the system looks at the specifications of the switching subsystem and forgets about the fact that cables and connectors, as well as the characteristics of the instruments, all add to the characteristics of the measurement channel. The end result is that either the test is compromised or upgrades become necessary, resulting in delays and added costs.

Another problem relates to the currents being switched. Too often, test engineers miss the different specifications of a switching system for "cold switching" (closing and opening a relay before applying power to the UUT) and "hot switching" (closing and opening relays while the UUT is powered up). Hot-switching specifications are typically about half the level of cold-switching specifications, and attempting to hot switch currents that are too high can damage the switching system. Hot switching can also cause arcing, which greatly reduces the expected life of the relays in the switching system.

I've described signal management from a standpoint of sharing resources. There are other reasons to consider signal management in your test strategy, ranging from diagnostic capabilities to coping with varying operator skill levels. For more information on these additional applications, see the "Justification for signal management."