Designing an electronic test system can require significant efforts. Here are a few useful tips geared to both the novice and veteran designer for improving measurement quality, minimizing rewiring hardware, and reducing test software rewriting.

Attracting Unexpected/Unwanted Components

In any electronic test system, some signals must be measured to establish the proper or improper operation of the Device Undergoing Test (DUT). In a perfect world the connections or probes measuring these signals would pick up only the signal of interest and nothing else. In reality, the signal to be measured is usually accompanied by other signals from various sources. Obtaining a high-quality measurement means being aware of these other signals, and designing the test system to eliminate them in one way or another. Understanding the sources of signals and the operation of the test equipment will enable selection of the best technique for a given situation.

Sources of Interference

Signal interference from a DUT can stem from many varied sources. The ability to identify an interference signal source is an important way to improve measurements. As a general rule, it is best to eliminate these additional signals at their source. This is usually much more effective than allowing the test signal to become contaminated and attempting to clean it up later.

Power-line noise is pervasive in developed areas of the world, and represents the lower frequency range of interference. Going up in frequency, computer-generated noise appears, including not only from the computers and video monitors, but also local area networks, printers and other peripherals. These can generate interference, especially if they are connected to the test equipment.

At the high end of the frequency spectrum are cellular phones and on-site radio transmitters for paging, site operations or security. These sources can generate considerable interference due to their RF power and also can be very difficult to identify due to their mobility and intermittent operation. These RF signals may be mixed with other signals in the DUT to produce unwanted lower-frequency signals, or they may be rectified by the DUT’s semiconductor components to produce dc voltage offsets. These intermittent signals can cause transitory test failures that are hard to identify.

Protecting test systems from external interference usually involves careful examination of equipment grounding and the shielding of power supply and measurement cables. However, it is often easier to eliminate the interference at the source.

In one real-life example, an electronics manufacturer experienced a manufacturing line shutdown when, for unexplained reasons, one sensitive product began to fail its final manufacturing tests. The test system rejected every device for a variety of different test failures. After several weeks of attempting to identify the problem, it was discovered that the plant security office had installed a new radio transmitter on the roof of the production facility. This transmitter had sufficient power to interfere with the test system’s operation, resulting in these test failures. Reducing the transmitter’s power solved the problem, but only after production had been halted for a costly length of time.

The most likely place to look for the noise source is in the DUT itself. Switching power supplies and digital circuitry can produce signals that do not necessarily degrade the DUT’s operation, but can affect the measurement system testing the device. It becomes a matter of economics to decide whether to reduce the noise in the product being tested, or to eliminate the noise somewhere in the test system. The test equipment itself often has a number of features which allow the test engineer to reduce or eliminate the interference if he or she is aware of them. Used well, these features can solve a wide variety of common problems.

Compensating for Power-Line Noise

Perhaps the most common interference comes from 50- or 60-Hz power-line signals. If the test signals can be measured at low to moderate speeds, using the right measurement equipment setup can very effectively remove the power-line noise. Setting up the integration time of voltmeter-type ADCs to one power-line cycle can reduce power-line interference by 40 dB or more.

A simple lab experiment will demonstrate this effect (see Fig. 1.) Connect a function generator output to a voltmeter input, and set the function generator to a low frequency (e.g. 3 Hz) sine-wave output. Set the voltmeter to a fast integration time and watch the voltmeter display as the measurements vary within the peak-to-peak values of the sine wave. Next, set the voltmeter to one cycle of integration time, and watch the reading settle down to a constant value, equal to the dc offset of the function generator output. This shows the powerful filtering effect of integrating ADCs.

With some newer instruments, such as the HP 34970 Data Acquisition/Switch Unit (see Fig. 2), integration time is adjustable not only in number of power line cycles (NPLC), but also is adjustable in 10-m s increments up to 4 s using aperture commands. This allows the test system engineer to null out interfering signals at a very wide range of frequencies. Any signal with a period that is an integer multiple of the integration time will be filtered out, so careful selection of aperture time may be able to remove multiple signals.

The integration technique requires time to complete a measurement, inversely proportional to the frequency being rejected. Integrating to reject 50- or 60-Hz frequencies will limit measurements to approximately that rate. If a faster ADC is being used, averaging can be used to approximate integration. Averaging is also effective against random noise.

When averaging, many signal measurements are made (usually at higher rates), with the results then totaled and divided by the number of samples. This technique decreases noise by a factor of the square root of the number of samples taken. For example, if the test system measures a signal 64 times and averages the results, the noise can be reduced by a factor of eight, essentially increasing the resolution of the measurement by three bits. Of course there are limits to this technique, and it’s imperative to note that the resolution -- not the accuracy -- will increase. Some instruments include calculation functions like "CALCulate:AVERage" which allow averaging to be done in the test equipment itself.

Compensating for Dc Offsets

If the test points in the DUT have a small dc voltage present during testing, this voltage can cause measurement error. The voltage may be due to thermoelectric effects of connecting dissimilar metals in connectors, wiring and probes, or it may be due to electrochemical (battery) effects of minor contamination of the DUT or test wiring. This small offset is simple to remove in the measurement system if the test equipment supports "Offset Compensated" measurements.

For example, to test a resistor in circuit, use the offset compensated ohms measurement mode in the ohmmeter. This mode operates by first measuring the voltage across the resistor with the ohmmeter current source applied, then measuring it again with the current source turned off. The two voltages are subtracted and the resistance calculated inside the ohmmeter. The process is fast and is calibrated for the current source used in the ohmmeter without any external calculations required.

If the test equipment did not provide this measurement mode, the user would have to know the exact value of the current source inside the ohmmeter to calculate this externally. If performed within the test equipment, these values are known and correctly used in calculations to obtain valid resistance measurements without external data handling.

Switching Test Signals

In most test systems, many signals are switched between signal source and measurement points, and switching equipment may comprise a large part of the test system cost. To obtain good measurements over the lifetime of the test system, the operation of the switching system should not be ignored. Both transient and lifetime considerations should be investigated before completing a design.

Transient conditions affecting a switch include inductive spikes caused by opening an inductive signal path, and inrush currents stemming from closure of a capacitive signal path. In the first case, high voltages may be generated when switch (relay) contacts are opened, and this may cause arcing of relay contacts, or damage to measurement equipment downstream. In the second case, the relay contacts may be damaged by a sudden current inrush as capacitors downstream represent a low-impedance path while they charge.

The switch subsystem may not experience a catastrophic failure from these effects, but the cumulative effects can degrade the signal measurement over time, resulting in premature failure of the switch subsystem. One strategy to avoid the degradation is to use a pair of relays, wired in parallel, to switch the difficult signals (see Fig. 3.) One "sacrificial" set of contacts is closed first and opened last to handle any transients, and a second set of contacts (closed last, opened first) is used to assure a high-quality signal path through the test equipment. The first set of contacts may degrade somewhat, but can still protect the good, low-impedance contacts on the second relay.

Even if a test system does not have high reactance signal paths, users should be aware of the finite lifetimes of relay contacts in the test system. Lifetimes will be determined not only by the relay design and manufacturing quality, but also by the switch-on of current in the test system. Higher current typically leads to shorter lifetimes -- sometimes by orders of magnitude.

For example, one relay may last 10 million operations in a low-current application, but its lifetime may drop to 100,000 operations under full-rated load. Automated test systems can operate relays so frequently that one full lifetime can be consumed in a few hours of relay operation. If the test system switches that relay 10 times per second, its rated lifetime will be consumed in only three hours of use.

To extend relay life, care should be given to the sequence of operation so relays are not changed unnecessarily. For example, it may be a common programming practice to set all relays to default between major parts of a test, which may rapidly consume relay lifetimes. Instead, design these tests so that similar switching paths are used for sequential tests, minimizing the number of relay actuations required for the whole test. This will also speed testing, since much time can be consumed waiting for hundreds of relay actuations.

Instruments contain functions that can monitor and report the cumulative operation of individual relays. This allows the test engineer to keep track of the switching portion of the test system and avoid unexpected downtime. Once the recommended relay lifetime has been reached, switching equipment should be replaced to ensure reliable operation of the test equipment.

Even in test systems without relays, transients can be a problem. Semiconductor switches and multiplexers are used in fast measurement systems, and are susceptible to damage from inductive voltage spikes from the opening of signal paths. Another problem may be caused by leaving some semiconductor multiplexer nodes isolated for a time, with neither input nor output loads connected. These-high impedance nodes can be charged by tiny leakage currents in the semiconductor switch circuit. When the switch closes, this charge gets "dumped" into the ADC front-end, sometimes causing a temporary offset in the ADC.

Transient behavior of the DUT does not always damage test equipment, but can cause measurement errors. For example, to measure a resistor which has a parallel capacitor, the ohmmeter current source must be applied long enough to charge the capacitor before the measurement is made. Similar considerations apply to measurements of any reactive circuits, which may even include the capacitance of test system cables.

To solve this problem, instruments permit the engineer to insert a delay between the relay closure and the actual measurement performed on a channel. This delay is performed entirely in-instrument and allows the measured circuit to settle to before it is measured. This means the computer controlling the system does not have to perform this timing, which would add to the test program complexity by requiring the software to be synchronized to the measurement and switching equipment.

Signal Compatibility

When designing signal-measurement sequences, keep in mind that signals of widely differing amplitude should not be adjacent to each other -- either in the switching hardware or in time sequence. Having a high-amplitude signal routed adjacent to very low-amplitude signals can cause crosstalk in multiplexers, contaminating the low-amplitude signals and making measurement difficult. Instead, group similar signals together to avoid coupling effects. In extreme instances, even consider skipping some channels in a multiplexer to gain extra isolation for certain signals, perhaps grounding the unused channels to increase the shielding effect.

In time sequence, it is not wise to try to measure very low-amplitude signals immediately after measuring high-amplitude signals (especially dc.) The measurement equipment has some finite front-end capacitance that must be discharged to change signal levels. This capacitance may be not only actual capacitors, but also the distributed capacitance of the wiring, printed circuit boards and other components. These components also will exhibit some measure of dielectric absorption (DA) which causes the capacitor to act somewhat like a battery, which is much slower to discharge than pure capacitance. The amount of DA in a capacitor depends upon the manufacturing materials, some showing only little DA. In fast measurement systems, the DA left from high-voltage signals can degrade the quality of a low-voltage signal following in the test sequence. While true capacitance will discharge with a v+1/exp(t) characteristic, DA will discharge at only a v+1/t rate -- much more slowly (see Fig. 3.) It should be noted that DA also charges at a 1/t rate, so to avoid a possible problem, high-voltage signals shouldn’t be left on the inputs for long periods, thus avoiding accumulation of DA charge.

To avoid the DA problem, measure low-voltage signals first, and work up to higher-voltage signals. If it’s necessary to do a high-to-low sequence, consider measuring the low-voltage signal twice (keeping the second reading), or inserting a grounded, unused, channel in the measurement sequence. These techniques use time to gain an additional settling of the DA charge.

Conclusion

The process of switching and measuring signals in real test systems is seldom as simple as a schematic diagram would indicate. Care in the design, layout and programming of automated test systems can avoid potential problems and yield the best measurements possible from test equipment.

Biographical Information

Christopher Kelly has been a design engineer with HP’s Electronic Measurements Division in Loveland, Colorado, since 1983. He has worked on the design of a number of test and measurement instruments, and holds two patents in the area of high-speed measurement. He is presently involved in designing a new electronic bench instrument.

		Click here to get your listing up.