Ideally a test engineer would connect a separate instrument to each test point of a Device Undergoing Test (DUT), thereby providing the highest performance and most accurate measurements. But this is costly and often unnecessary. That is why switch systems are so appealing for they can be very effective at reducing the cost of creating a test system. Too often a switch subsystem is not designed until the last minute and then "thrown" together. What's more, a switch subsystem that is not carefully designed or used as intended can produce invalid test results, delaying product development or allowing a product to be shipped that does not meet its design requirements.

A generic test-system architecture is shown in Figure 1. Note that the switching subsystem is central to the entire system, connecting the many test points to the measuring instruments, routing signals, simulating contact closures and connecting power to the DUT. The switch enables a level of automation that greatly diminishes the time to test. What's more, it reduces the monotony and resulting human error so often encountered when conducting a complete test.

Figure 1:  A generic test-system architecture. With the exception of the bus interconnections nearly all analog and digital signals and power pass through the switch subsystem.

The control function orchestrates the application of stimuli required to perform a test. For example, switches used to control the automatic seat adjustment in automobiles can be simulated during the test of the microcontroller that controls that function. The control function includes simulating a contact closure with a general-purpose relay switch, or simulating a logic level—such as a TTL signal from a sensor or other microcontroller—with a digital I/O device.

As long as the amplitude and frequency of the test signals are of moderate value and the measurement accuracy requirements are not stringent, it is reasonably easy to construct a switching system from off-the-shelf parts. But potential savings in the cost of hardware obtained by building a custom switching device is likely to be consumed, perhaps many times over, by design time, documentation and support costs. These costs tend to make the purchase of well-specified commercial hardware quite attractive.

The most common simple relay configurations are the familiar form A, B, and C contact configurations. These can be linked together to form binary switching networks that guarantee that only one point can be connected to any other point at a time—an important safety consideration. Similarly, these switches can be configured to form measurement buses for the connection of many points to one point at a time.

A binary switching ladder is depicted in Figure 2. It can be used to ensure that only one instrument at a time is connected to a test point and that no two test points are connected together.

Figure 2:  A binary switching ladder using form C relays

Two-wire multiplexers, as depicted in Figure 3, are useful for floating measurements. Inductive coupling can cause the generation of ground loops in the low lead. To break these loops it is necessary to switch the "low" as well as the "high". This is particularly useful for capacitance and inductance measurements at frequencies below 1 MHz. Devices such as thermocouples and other DC transducers can also be connected to voltmeters while preserving good common-mode noise rejection.

Figure 3:  A two-wire multiplexing topology. This 10-channel relay multiplexer can be used for floating measurements because both conductors, the signal and the ground return, are connected to the common bus at the same time.

Single-wire configurations are useful for high-frequency applications because most measuring instruments for higher-frequency signals are single ended (common ground or low with all the inputs). Also, at high frequencies it is difficult to design a switch where both "high" and "low" are switched without affecting the characteristic impedance through the switch.

Three-wire multiplexers are designed for guarded voltmeter applications. The third connection can shunt noise current away from the input measurement terminals and give the user up to 120 dB of common-mode noise rejection—almost 1000 times more than the two-wire measurement. Four-wire multiplexers can be used for four-wire resistance measurements that employ resistive bridges. Five- and six-wire switching can be useful for simultaneous application of driven grounds for in-circuit component isolation and measurements. It is often advisable to select a flexible switch architecture that allows many types of switch topologies. For example, the Agilent N2260A 40-channel, multiplexer module can be configured as an 80-channel, single-wire; 40-channel, two-wire; or a 20-channel, four-wire multiplexer.

T-switching, depicted in Figure 4, is a multiplexer enhancement that reduces unwanted signal coupling into the measurement channels. With T-switching an additional relay is connected to ground via a low-impedance path, so that unwanted signals, which would otherwise be capacitively coupled to the measurement, are shunted to ground. This establishes excellent channel-to-channel signal isolation at high frequencies on the same multiplexer.

Figure 4:  A simplified circuit diagram illustrating T-switching. Conductor 2 is open, so its "C" switch is closed, providing a low-impedance path to ground.

A matrix topology may be required when more than one instrument must be connected to the same test point at the same time. In theory, a matrix switch topology enables anything to be connected to anything else in the system, simultaneously. For instance, a full crossbar matrix such as the Agilent N2262A, 4 x 8 matrix module, enables two-wire connection from any point to any other point. This configuration can be expanded to much larger systems. However, the switch capacitance increases dramatically as the system increases in size.

What's more, multiple matrix cards can be configured so as to increase the size of a switching system. For instance, with five 4 x 4 matrix modules, four instruments can be simultaneously connected to any of 20 test points. But because the inter-channel capacitance has increased due to longer conductor paths, the overall capacitance increases more than just the sum of the additional switch capacitances. This makes high-frequency signal integrity difficult to maintain for large matrix configurations.

Shown in Figure 5 is how multiplexers can be used to effectively expand the number of inputs or outputs of a matrix. This arrangement is not a true complete crossbar matrix because any combination of inputs cannot be connected to any combination of outputs. For example, channel A cannot be connected to channel 2 while channel C is connected to channel 1. But this "4 x 40" topology is an economical way to provide flexible multiple channel closures for four different signals.

Figure 5:  Multiplexers can be used to effectively and economically expand the number of inputs or outputs of a matrix, although this switching topology does not have the unlimited flexibility of a full crossbar matrix.

Selecting the appropriate relay is also important for maximizing the useful life of the relays. This is covered in detail in Test System Signal Switching, Agilent Technologies Application Note 1441-1.

The topology of the switch subsystem is important for ensuring that all connections that need to be made can be made in an appropriate manner. It may be necessary to optimize the switch subsystem for high-speed measurements, low noise, or high-frequency signals.

Selection of the relays used to implement the switch topology is very important, because it affects the type of circuits and systems that can be tested, as well as test-system reliability and accuracy. But whatever topology is selected, the switch must be able to achieve the level of automation required.

Chuck Reynolds has 35 years experience in the field of test and measurement applications. He is currently the VXI Program Manager for Agilent Technologies located in Loveland CO. He has a BSEE and a MBA from Colorado State University. His email address is chuck_reynolds@agilent.com.