As test engineers continue to maximize production efficiencies, they need to become more aware of ways to decrease test time and the cost of test. Evaluating the speed vs. accuracy trade-offs involved in making multichannel measurements is essential when selecting the right type of DMM and switching hardware.
DMMs and relay-based switching are the key building blocks for many test applications and the core elements of many ATE systems. There is a wide variety of implementation options, three of which are the most prevalent: stand-alone instrument systems, chassis-based systems, and integrated solutions.
Stand-alone solutions, often referred to as rack-and-stack, consist of two or more instruments connected together with cabling and networking. The instruments have separate communications ports that are not necessarily compatible, meaning that often times you must integrate the pieces.
The main advantage of this implementation is the variety of vendors from which to choose. There are more than a dozen vendors that make stand-alone DMMs or relay-based switching instruments. With this variety come many choices, allowing you to mix and match the best combination of instruments for the application. This is especially advantageous with switching, because the signal level and topology, such as multiplexer or matrix, can become very individualized.
The chief disadvantages include the need for hardware instrument coordination and the demands of programming individual instruments.
Chassis-based solutions, such as VXI- or PXI-based systems, consist of a chassis with a high-speed communications bus and a number of slots that accept plug-in cards, such as relay or DMM cards. The chassis concept, with many individual slots, provides flexibility to interchange and adapt configurations. In addition, a PC card can be plugged into the chassis or connected externally.
But chassis-based solutions have a few drawbacks, including a limited number of vendors. In addition, like the rack-and-stack solution, there is the hardware disadvantage of maintaining the signal routing between the DMM and switching. And in most cases, there is no analog or signal backplane within the chassis so cabling between switching cards must be considered.
Lastly, integrated solutions include both a DMM and switching within a single instrument. A single communications port controls both instrument functions while embedded firmware automates connections and measurements to eliminate operator error. Much of the system operation is automated and optimized to allow maximum throughput and accuracy.
The integrated solution also has an analog backplane or signal bus enabling an automated connection from switching to DMM and from switch card to switch card, leaving only the DUT cabling for you to complete. The chief disadvantage is the limited number of vendors making this type of all-in-one instrument.
A good predictor of system performance is the DMM specifications including key parameters such as accuracy, throughput, and noise. Simply put, the amount of noise will have an impact on system accuracy and, ultimately, on throughput. Signal noise can come from a variety of sources, including RF signals, AC power lines, lighting, motors, and transformers.
DMM accuracy is defined as how close the measurement result is to a known standard. Resolution represents the smallest portion of a signal that can be observed.
Figure 1 depicts the accuracy term when plotted over the full range of an input signal. The full range is signified by 0% full scale (FS) on the left and 100% FS on the right. The offset error and gain error specifications determine the intercept and slope of the red lines.
Figure 1. Offset and Gain Contributions to Accuracy
For specifications with small offset error and larger gain error, the slope is steeper and the gain error is the more significant factor. If the offset error is relatively larger and the gain error relatively smaller, the line is flatter and the shaded area more rectangular. In some cases, the accuracy specification is stated as a percentage of full scale, which implies the shaded area of uncertainty would be a rectangle.
When considering the uncertainty as a percentage of the reading value, the graph shows that the offset error is the more significant factor for lower signal levels, and the sum of offset and gain errors defines the uncertainty for larger signal values. Because of offset errors, relative uncertainties are larger for lower signal inputs and lower for higher signal inputs.
The other important item to note is the specific number of power line cycles (nPLC) in which the accuracy is given. In all cases, you will find the nPLC to be at least 1.
Specifications are given in integer multiples of power line cycles for many reasons. Manufacturers typically use a technique called power line-cycle (PLC) integration to minimize the effects of 60- or 50-Hz line pickup, which is the most common noise element for DMMs. Integrating over a complete power line cycle causes cancellation of the positive and negative terms, resulting in a more accurate reading.
If an application calls for readings faster than 50 or 60 Hz, then the line-cycle integration technique won’t work, and the measurement may be subjected to more noise. But if reading accuracy can be sacrificed, a number of items will need to be verified, including the speed capability of the instrument and the impact on measurement accuracy.
Typical reading-rate specifications include writing the measurements into the buffer, meaning how fast the instrument can take readings and store them in internal memory, and writing measurements to a PC, which indicates the overall throughput including measurement time and bus transfer. Because it includes bus transfer, this specification will vary for different types of communications buses such as Ethernet, GPIB, or USB.
Accuracy only incorporates internally generated instrument noise because there is no external noise rejection when operating sub-line cycle. For that reason, working in a noisy electrical environment without good wiring procedures could significantly increase measurement uncertainty.
Many times in resistance-measurement applications you must consider the four-wire or Kelvin connection method of measurement. This approach is preferred for low-resistance measurements because it compensates for lead and relay resistance. Such four-wire measurements have a considerable impact on throughput.
One of the newest techniques that maximizes throughput and accuracy for four-wire measurements is called line synchronization. This technique uses the power line zero crossing as an event to kick off two A/D measurements when operating sub-line cycle. This technique can almost produce the same type of noise cancellation found in 1-PLC measurements while improving the reading rate.
The other half of the equation in DMM and relay switching test setups is the switching aspect. The type of relay selected generally will have the largest impact on throughput. However, it also is important to consider controller functionality to verify that there will not be any unexpected delays caused by the design.
The most basic consideration for speed in switching systems is the type of relay technology. Figure 2 outlines the basic types of relays and their key specifications.
Figure 2. Relay Types
Basic relay designs include electromechanical, reed, and solid-state. The most widely applicable is the electromechanical relay because it has the largest signal range capability of the three types. However, it also has the slowest actuation time and the shortest useful life.
Electromechanical relays are ideal for high-voltage, high-current, and RF applications. In addition, a latching feature makes this relay very good in low-voltage applications where contact potential can interfere with the measurement.
Reed relays have actuation times of 0.5 ms to 2 ms and long life. By design, reed relays can only handle a portion of the signal range that electromechanical relays can offer. But they do provide a great trade-off between increasing speed and maintaining signal integrity.
Lastly, solid-state relays can switch the fastest and have essentially infinite life. However, they can only handle small signal ranges and suffer from high on-resistance and high offset currents in the nanoamp range compared to picoamps of offset current for the other two types.
After deciding on the relay type, the next consideration is the capability of the switching controller. The first area to examine is the open and close times required to execute commands. If a fast relay is selected but the controller takes milliseconds to carry out a command, throughput will be greatly impacted.
Another common feature in switching control is stored switching patterns that increase throughput in switching hardware. These memory patterns are preprogrammed groups of channel closings that reside at the instrument level.
A scanning option is another typical feature. Scanning often is the fastest method to open and close a list of channels because the list is predefined and can become optimized within the switching hardware. Generally, throughput improvements of up to 2x are possible when operating in this mode. A scan list using memory patterns allows closure of multiple channels in every scan step.
The successful integration of the DMM and switch involves tight triggering between instruments and paying close attention to system effects such as relay settling time. Settling time is an important consideration especially in electromechanical and reed relays. It determines when the DMM can be triggered to measure. Having an accurate understanding of the settling time is critical to obtaining good measurements.
Figure 3 shows the relationship among the stimulus, the response, and the measure command. The top row indicates the stimulus while the second row shows the response curve at the DUT. The lagging response is due to cabling capacitance or the DUT itself.
Figure 3. Stimulus/Response/Measure Graphic
The bottom line shows when the DMM is commanded to take a measurement. So if measurements were taken when the stimulus was activated, the system would be measuring a wrong value. This is one of the most common errors in test systems.
Large systems with much cabling and cable capacitance or that measure high impedance and have large RC time constants may require delays of 5 ms to 10 ms or special signal handling techniques such as guarding. With an integrated DMM/switching solution, the consideration of settling times is built in.
A final system element is the use of triggering. This is in the form of a dedicated hardware trigger located on each of the instruments to accommodate very fast execution and deterministic operation.
A flow chart would begin with the switch closing a channel, then activating a trigger to the DMM that begins the measurement. When the measurement is complete, the DMM issues a trigger to the switch indicating it has completed measuring. The switch then opens the previous channel and closes the next channel and continues in this fashion.
Also, with an integrated switching system, there is no additional need to program two components to interact properly.
An integrated DMM/switch solution offers many advantages in both hardware and software. When optimizing the DMM, understanding the instrument specifications to get a handle on the speed vs. accuracy trade-offs is important. For switching, consider the type of relay and controller functionality. And, lastly, take into account the system aspects of the solution such as settling times to ensure that the results are accurate and throughput is sufficient.
Jerry Janesch is a marketing manager at Keithley Instruments, where he has been employed for seven years. Mr. Janesch earned a bachelor’s degree in electrical engineering from Fenn College of Engineering at Cleveland State University and a master’s of business administration from John Carroll University. Keithley Instruments, 28775 Aurora Rd., Cleveland, OH 44139, 440-248-0400, e-mail: firstname.lastname@example.org