Microwave test systems often include one or more RF/microwave switches to allow automated testing as well as measurements on multiple devices under test (DUTs). When configuring a highfrequency switching system for test applications, it is important to understand the type of switch and multiplexer options available as well as the critical switch specifications when choosing a switch.

Switch systems for RF and microwave applications can range from simple to quite elaborate. For example, a single-pole, doublethrow (SPDT) switch can be used to route signals to two different devices under test. It can be expanded further into a multiplexer configuration so that a single instrument can be routed to many different DUTs. In addition, multiple instruments can be routed to multiple DUTs. In this case, the switch system is known either as a multiplexer-demultiplexer or a blocking matrixwith only one signal path active at any given time. Figure 1 shows the application of an SPDT and a 1-to-16 multiplexer.

A switch system with a "blocking configuration" (Fig. 2) can connect multiple instruments to multiple DUTs. Any instrument can be connected to only one output at a time. To switch any signal to any DUT at any time, a "non-blocking" configuration (sometimes called a switch matrix) can be used. Although this switch configuration has the highest flexibility, it is also the most expensive. The minimum number of interconnecting cables in a non-blocking matrix is a product of the number of inputs and outputs. Therefore, a 10 10 nonblocking matrix requires 100 cables and single-pole, 10-throw (SP10T) relays. Assembling such a system is a major challenge. Furthermore, the use of a large number of components opens the way for potentially reduced reliability. Figure 3 shows a 4 4 nonblocking switch configuration, and Fig. 4 shows an expanded 4 6 switch configuration.

Any kind of switch can be thought of as an "interruption" in a signal path through a measurement system and, as such, it is impossible for the switch to have no effect on the system. In fact, the use of a switch will inevitably degrade the performance of a measurement system, so it is important to consider several critical parameters that may affect system performance significantly. Two types of specifications are very important when considering a switch for a measurement application: electrical and mechanical characteristics.

In terms of electrical specifications, such switch characteristics as impedance matching, insertion loss, isolation, and return loss can impact the overall performance of an RF/microwave measurement system. Because a switch will be positioned between the measurement instruments and the DUT, it's critical to match the impedance levels of all three system elements. For optimal signal transfer, the impedance of the source must match that of the switch and the DUT. In RF testing, different impedance levels are used to achieve different purposes. The most common impedance in high-frequency systems is 50 O and all RF/microwave relay manufacturers make relays with characteristic impedance of 50 O. This impedance represents a trade off between the theoretical impedance at which maximum power transfer occurs, 30 O (30- components are not available in the marketplace), and minimum theoretical attenuation at 75 O (75- components are used in cable television). Also, 50- components are easily constructed.

Insertion loss is a measure of unintended signal attenuation that occurs due to dissipative losses in dielectric materials, impedance mismatches in connectors and circuit leads, and a variety of other contributing factors. Any component added to the signal path will cause some degree of loss, but the amount of loss is especially severe at higher or resonant frequencies. When the signal level is low or noise is high, insertion loss is particularly important. The insertion loss is reflected as a decrease in the available power to the DUT as compared to the test instrument source value. Normally, it is specified as the ratio of output power over the input power in dB at a certain frequency or over a frequency range:

Insertion loss (dB) = 10log(Pout/Pin)

Each component in the signal path will have its own insertion loss specification with some, such as couplers and power dividers, contributing an amount of loss above and beyond the decrease in signal level from a power coupling or power division.

Isolation is an important consideration since, at higher frequencies, signals traveling on different paths can interfere with each other or create "crosstalk" due to capacitive coupling between the paths or through electromagnetic (EM) radiation. This is especially severe when signal paths are not properly shielded or decoupled from each other. Crosstalk is problematic when a weak signal is physically adjacent to a strong signal. In such cases, the highest possible isolation is desired to minimize interference.

Any component added to the highfrequency signal path will not only cause insertion loss, but will affect the voltage standing wave ratio (VSWR) of the signal path. This standing wave is formed by the interference of the transmitting electromagnetic wave with the reflected wave. This interference is often the result of mismatched impedances in different parts of the system or connecting points in the system, such as connectors. The VSWR is specified as the ratio of the standing wave's highest voltage amplitude to the lowest voltage amplitude in the signal. VSWR is often also expressed as return loss:

Return loss (dB) = 20 log

As the size of a test system is expanded, signals from the same source may travel to the DUT via different paths of different lengths. When the lengths of the signal paths differ the result is phase distortion or propagation delay. For a given conducting medium, the delay is proportional to the length of the signal path. Different signal path lengths will cause the signal phase to shift. This phase shift may cause erroneous measurement results. Therefore, techniques to ensure the same phase or path length can be used to compensate for such effects.

Reliability and repeatability are important considerations for switches in test systems. Typically, a switch relay should provide a lifetime of at least one million closures; many relays offer rated lifetimes of five million closures or greater. The repeatability of the switch performance is an equally important issue. Repeatability is the measure of the changes in the insertion loss or phase change from repeated use of the switch system. In making RF/microwave measurements, the effects of cycle-to-cycle changes in a switch relay closure can impact the accuracy of the system.

In some applications, particularly when low-power signals are being switched, solid-state PIN diodes can be used as the switching elements. With no moving parts, they offer a much longer life than electromechanical switches with their moving parts. Where high volume testing is being performed, and test tolerances are not that tight, solid-state switches can be very cost-effective. Their disadvantages are much higher insertion loss at a given frequency (typically more than 1 dB at 4 GHz for a PIN diode switch compared to 0.3 dB for an electromechanical switch). The isolation is much lower for a PIN diode switch compared to an electromechanical switch. An electromechanical relay can provide 100 dB or more isolation through 6 GHz, where a PIN diode switch may achieve less than 30 dB isolation below 3 GHz.

In terms of a switch's mechanical specifications, the switch's physical form factor may limit the choice for a particular test system application. For a switch system with a medium number of signal pathways, a system such as the Model S46 RF/ Microwave Switch System from Keithley Instruments can offer numerous flexible configurations with up to 32 control ports. For a large-scale switch system, the firm's System 41 RF/Microwave Signal Routing System offers dual 1 36 multiplexers or a single 1 72 multiplexer. The System 41 also can be configured as a 6 6 or a 10 10 nonblocking multiplexer. When both low frequency and microwave switching are required, the company's model 7116-MWS combines a 1 16 microwave multiplexer with 40 channels of lowfrequency DC switching and control capability.

Configuring a measurement system with multithrow switches also involves selection of cables and connectors, and the number of choices is many. The bandwidth of the test system, for example, will determine the type of coaxial connector used on the cable assemblies, with standard formats such as Type N connectors operating to about 12 GHz and SMA connectors to about 26.5 GHz. Smaller-dimensioned connectors, such as 2.4-mm connectors, can support a continuous bandwidth of DC to 40 GHz. The choice of cables must be made based on whether a test installation is fixed, and can use semi-rigid cables with their excellent loss characteristics, or can sacrifice some additional loss for the versatility provided by flexible coaxial cable assemblies. The signal frequency, the system impedance, power rating, and test fixture/handler compatibility, etc. should all be taken into consideration when choosing connectors and cables.

Continue to page 2.

Page Title

Several other factors must be considered carefully because they can easily degrade system performance even when high-quality components are used. At high frequencies, unless all signals are properly terminated, unacceptable amounts of power will be reflected from the terminating point. This is seen as an increase in VSWR. It may even cause damage to the source if the reflected portion of the signal is large enough. Switching paths must maintain the system's characteristic impedance, either 50 or 75 O, and the signal path must be terminated in the characteristic impedance. Unlike at low frequencies, where relay contacts can be connected in parallel or connected in matrix crosspoints, microwave switches cannot be connected in parallel because the characteristic impedance would not match the standard characteristic impedance of the system. This requirement to maintain the same impedance throughout the system is why non-blocking microwave matrices are such large systems.

Most switch system users would like to have a switch with as wide and as flat a bandwidth as possible. However, wide-bandwidth switches are costly. If it is not absolutely needed, a narrowband switch can achieve the same objectives at a significantly lower cost. Another factor to consider is that when higher frequency is used, the bandwidth will depend on the type of connectors and cables used. More expensive types of these products are typically needed to ensure adequate system performance.

Another important consideration is the system's ability to transfer the RF/microwave power from a test instrument to the DUT. Due to insertion loss, the signal may require amplification in order to reach a level required at the DUT (for example, for measuring the saturated output power of an active device). In some applications, it may be necessary to reduce signal power to the DUT. An amplifier or attenuator may be needed to ensure proper control over the level of power transmitted through the switch.

Anyone testing, characterizing, or performing environmental studies on multiple digital devices driven with pulses at gigahertz rates or higher will need a high-frequency, microwave switch system to transfer undistorted test pulses. Because the devices are being driven by digital pulses rather than sinusoidal-based signals, the switch system must be capable of switching the DC component of the pulses as well as the higher frequency components of the excitation pulses. That means the frequency response of the switch must be DC coupled, reaching from DC to the maximum frequency for the required bandwidth of the test system. Coaxial microwave electromechanical switches are specified to operate to DC. Users may assume that the specified life of the relay is the same at all frequencies, but not all switches perform as repeatably at DC as they do at RF frequencies.

Figure 5 shows the plot of a microwave coaxial relay's performance at the mid-point of its warranted 1-million- cycle operating life. The relay's DC contact resistance is plotted versus switch closure activations. Note how the contact resistance varies. Particularly note that the DC contact resistance does not fail, exceed its specification of 100 mO, and then remain in a failed state. It actually recovers and operates within specification, and then it fails again. Variations like this can significantly affect test data. For applications that require switching pulsed high-frequency signals, a switch system manufacturer should be able to supply test data to ensure that the relay will provide repeatable performance over its specified life for DC as well as for microwave signals.

Although microwave relay manufacturers provide specifications over the frequency range of the relay, they usually provide insertion loss, isolation, and VSWR data at only a few frequency points. A microwave switch system may have paths that include multiple relays, cables, and other components that complicate determining the insertion loss or return loss at specific frequencies important to a particular test. If signal magnitudes are a critical aspect of the test results, request data from the switch system manufacturer or test the system for this critical performance data.

Figure 6 is a vector network analyzer (VNA) plot of insertion loss as a function of frequency for seven paths of a 4-GHz multipath switch system. Note that the insertion loss changes by about a factor of eight over the frequency range of the switch system. Also note that the insertion losses can vary significantly depending upon the path. In this case, the insertion loss variation between paths can be as much as a factor of three. If accurate determination of power levels is required, this characterization data is essential. If equal insertion loss paths are necessary, either fixed or variable attenuators can be added to the paths to achieve this. In cases where small differences exist between paths, cable lengths can be adjusted to equalize the insertion losses.

The VSWR plot in Fig. 7 shows how the VSWR varies over frequency, which has an impact on the amount of power delivered through the system. Complex switch systems with multiple components in a pathway may require the addition of isolators to protect input devices such as instruments or DUTs from excessive amounts of reflected power. (Refer to the sidebar for information on estimating VSWR and return loss.)

To minimize VSWR, use these recommended construction techniques:
Minimize kinks in cables.
Ensure that the cable shields are free of holes or discontinuities.
Confirm that all connectors are properly torqued. SMA connectors should be torqued to 7 to 10 in.-lbs. Proper torquing of connectors minimizes the insertion loss through the interface of the two connections and minimizes VSWR degradation. Proper torquing ensures that the connections are stable and helps to maintain repeatable results through each switch pathway. Keep connectors clean. Oils and salts from the skin can contaminate insulation and reduce component isolation.

In some cases, such as testing some communication systems and testing radar subsystems, all switch paths must have near-identical time delays. There are limitations to how small the differential can be between multiple switch paths. Even relays have some slight variation in phase. Figure 8 shows the phase variation through a number of SP6T relays. The phase angle differential is a maximum of about 1 deg. The red circles highlight phase angle delays for two of the relays.

Using cables with identical electrical lengths is the key to achieving a phase-matched switch system. Rigid or semi-rigid cables are recommended because they can be specified with delays of 1 ps or better. However, remember that the tighter the tolerance, the higher the cost of the cables.

Figure 9 shows the variation in delay times through an actual switch system. In this case, the delay times through the individual paths were held to within 4 ps of each other since the test system was required to have phase-matched cables. For an even tighter phase tolerance, adjustable phase delay elements can be added to the switch system. However, it's important to weigh the benefit of achieving tighter tolerance in phase matching against the added cost involved.

To achieve an optimal, yet cost-effective system, system designers must weigh a variety of electrical and mechanical parameters when specifying a switch system. Understanding these parameters makes it possible to make informed trade offs between switch flexibility and system cost. Furthermore, the subtleties associated with achieving a high level of system performance require a significant level of detail including system characterization and well-defined specifications for items like relays and cables.

REFERENCE

1. For a discussion of the installation torque requirements for creating and maintaining good radio frequency connections, see "Connector Torque Requirements," by Ted Prema, Times Microwave Systems, Wallingford, CT. Available online at: www.mwjournal.com/ Journal/article.asp?HH_ID=AR_586&tite=Connector%20 Torque%20Requirements.

Continue to page 3.

Page Title

Quantifying Switch System Performance
Estimating switch or overall measurement system VSWR requires identifying sources of reflected power and giving consideration to the effects of reflected power on components in the signal path. The following equations can be helpful in determining the VSWR for a given test setup:

Uncertainty = 20log(1 ΓSΓL) (in dB) (1) where
ΓS= the reflection coefficient of the source and
ΓL = the reflection coefficient of the load.

Return loss = 20log(1/ ) (in dB) (2)

VSWR = (1 + Γ )/(1 Γ ) (3)

The reflection coefficient can be determined from the return loss by Eq. 4:

Γ = 1/antilog10 (4)

Similarly, the reflection coefficient can be calculated from the VSWR by Eq. 5:

Γ = (VSWR 1)/(VSWR + 1) (5)

As an example, Fig. 10 shows a simple RF signal path consisting of a power combiner, an isolator, an attenuator, and cables, with the specifications for each component listed in Table 1. The reflection coefficients for Table 1 were determined using Eq. 5.

To estimate the VSWR at the input of the system, start at the output port. The reflection coefficients of the components in the path are added in a root-mean-square (RMS) sum to yield the VSWR value at the input port. In this case, the VSWR is calculated to be 1.389:1. The steps taken to arrive at this conclusion are detailed in Table 2.

Power losses primarily are a function of the resistive and impedance mismatch losses through a circuit path. Mismatch usually is the largest contributor to power measurement uncertainty and can be calculated from the magnitudes of the reflection coefficients of the source and load by the steps detailed in Table 2.

The best-case input-to-output insertion loss would be no worse than the summation of the components' insertion loss specifications. In this case, that would be four cables at 0.4 dB each, the combiner at 0.4 dB, the isolator at 0.4 dB, and the highest value of attenuation for the variable attenuator, which would be 3.3 dB; these loss contributions add to a maximum of 5.7 dB insertion loss through the system.

The worst-case output-to-input isolation can be estimated by using values of 20-dB isolation for the isolator, 3 dB for the division of power through the combiner, and the lowest value of attenuation for the variable attenuator, which is 2.7 dB; these values add to at least 25.7-dB output-to-input isolation for the system. The isolation between input port 1 and input port 2 should be no worse than the isolation specification of the power combiner, which is 22 dB.