An automated test system and exacting measurement practices have revealed a close correlation between high contact resistance and RF relay VSWR system-level failures.
Coaxial RF relays are used throughout military systems. Systems developers and integrators such as Raytheon Systems Company (St. Petersburg, FL) must rigorously screen defective components from their suppliers in order to avoid system-level failures. Since 1993, the company has tracked failures of coaxial RF relays from two suppliers, identifying the need for design, manufacturing, and test-process changes. To aid in the testing/screening process, an automated low-current contact resistance test set was developed. The test set uses generic test equipment to identify failures (in some cases isolating relay lots with failures exceeding 15 percent). In addition, the system revealed that the contact resistance of failed relays typically exceeded 30,000 Ω, and has learned that three issues continue to impact the reliability of vendor-supplied relays: improper adjustment (lack of overtravel), excessive volatile-material condensation, and contamination.
Raytheon Systems Company (RSC) has developed some simple methodologies for troubleshooting failsafe coaxial relays. The military contractor has used coaxial RF relays in two military applications in particular: a missile Global Positioning System (GPS) antenna and a frequency-agile filter (radio). The missile application uses two relays to switch receive signals only: a null signal for the normally closed-path and antenna-receive signals for the normally open path. The filter/radio application uses nine relays in series that transmit less than 50 W on the normally open side and conduct receive signals on the normally closed side. Both applications operate in dry circuit conditions. The Engineers' Relay Handbook1 states "that the performance of contacts under dry circuit conditions is affected by the following parameters: contact material; contact force; contact wipe; cleanliness; environment; magnitude of current and voltage." Since dry circuit applications at RSC typically pass microwatts of power through the contacts, contact resistance is an important parameter.
For years, the coaxial relay industry has used contact resistance testing to detect RF anomalies. The cost of a simple tester that detects opens (a few thousand dollars) makes that method of testing far more attractive than the investment in an RF test set for $80,000 or more. However, test results indicated that there was a correlation between receive application VSWR failures and contact resistance.
Figure 1 shows typical examples of failsafe coaxial relays for radio (left) and missile (right) applications. System-level ESS testing usually indicates a symptom of high VSWR for a failed relay, with most failures occurring for tests conducted at a cold temperature (−55°C). Failures can occur at high (+70°C) and room (+25°C) temperatures, however, as well as during temperature transitions (from cold to hot or from hot to cold).
When attempting to duplicate receive failures during RF testing, power levels are a prime concern. Because of this, voltage and current levels must also be taken into consideration when measuring contact resistance using conventional digital multimeters (DMMs). Both types of relays were designed to meet the requirements of MIL-S-3928/15.2 Unfortunately, a review of the MIL-S-3928 standard reveals that it does not specify the power levels required during RF testing. Contact resistance is specified, for a maximum of 0.240 Ω. However, the voltage and current levels are not specified for that contact resistance. These shortcomings in MIL-S-3928 required RSC to develop procurement documentation that specifically addressed the needs of a particular application. RSC also took into account that coaxial RF relay suppliers do not always test their products to ensure that they will perform as needed in receive applications
RSC has attempted to simulate the receive application using generic laboratory equipment. The current failsafe coaxial relay test set uses two 6.5-digit DMMs, although it was recognized that test current varies depending on the DMM scale selected. The 300-Ω scale has a 1.7 mA test current compared to 160 µA on the 30,000-Ω scale. Typically, DMM low-range scales (300 to 3000 Ω) exceed 1.0 mA test current. The 30,000-Ω scale was chosen to limit the current and more closely mimic the receive application. When using the 30,000-Ω scale, the 6.5-digit DMMs allow a resolution of 10 mΩ.
The RSC automated test set (Fig. 2) has evolved from testing one part at a time (at cold temperature only) to eight at a time exposed to temperature cycling (+25°C, −55°C, −47°C, +71°C) with each relay actuated 800 times. The missile-application relay incorporates female pins that accept male solder tabs for the contact connections. A male pin-adapter fixture was used to connect to these contacts. Double-shielded cables with SMA (subminiature series A, MIL-C-348) connectors were used to connect between the fixtures and each of the test equipment. The relay test set employs three programmable RF scanners to handle eight relays at a time. The test set measures resistance by the two-wire rather than the (preferred) four-wire method mostly due to lack of reed switches.
The relay test set is capable of measuring each of the four states of the relay: normally closed to common closed (NCCC); normally open to common open (NOCO); normally closed to common open (NCCO); and normally open to common closed (NOCC). To increase the number of measurements during a four-hour test period, modifications were made to the test program to capture only the closed measurements (NCCC and NOCC). After measuring hundreds of relays, it was found that that the normally open data was not very valuable and was not captured.
The relays are actuated individually using a programmable power supply that is fed through a programmable RF matrix switch, which applies the actuation voltage. A third meter monitors the case temperature of one of the relays and stores this information with the contact resistance data.
One cycle of the temperature profile used during temperature cycling consists of:
The total temperature cycle exposure lasts about 2 h. The test equipment operates under computer control and measurement data are stored in a serial log file. Test data are loaded into a spreadsheet, parsed, then a macrofunction automatically distributes the data into columns by relay, contact measurement type, and case temperature. Charts are automatically generated as soon as the macro is run.
The relay test set required modification to accommodate the radio relays due to the SMA connector and the built-in logic circuitry. Modifications were performed as follows: A power supply was used to constantly apply 28 V to the 28-V input terminal of each relay. The logic input voltage (0 or 5 V) was supplied from a programmable power supply that employs an RF scanner to individually actuate each relay. The test system's adapter fixtures were removed and relays were attached directly to double-shielded coaxial cables by means of male SMA connectors. Three shielded cables were needed for each relay (NC, NO, and COM terminals); each was tightened with a torque wrench set to 14 inch-pounds. The test set was programmed to null cable resistances and incorporated ohms compensation for thermal effects over temperature. Shorting fixtures were fabricated using the RF decks of disassembled relays. A short length of buss wire was soldered across the probe contacts (with one wire bridging the normally open side to the common and then to the normally closed side). The shorting fixtures were connected to the test cables and used for measurements conducted over each temperature cycle to determine the stability of the test system.
Figure 3 offers a graphical representation of the test data. The bold line shows the NCCC contact resistance; the medium density line is the NOCC contact resistance and the very thin line shows the relay case temperature. The left resistance scale is 0.250 Ω full scale, so the test set shows only 0.020 Ω variability with temperature.
By replacing the buss wire in the shorting fixture with precision resistors, the accuracy of the test system could be determined (see table). This use of the shorting fixtures also makes it possible to measure the contact resistance of each RF scanner reed relay as well as the device under test (DUT). The slight offset between the normally closed and normally open readings (0.010 to 0.030 Ω) could have been equalized using the nulling feature of the DMM. However, an intentional offset was left to allow visual separation when displaying the data on charts.
The relay test set was developed using generic lab equipment. It was limited to two-wire resistance measurements, although four-wire measurements are known to be more accurate. For this application, however, the two-wire measurements provide adequate accuracy, as demonstrated by test-set accuracy and repeatability. Measurement delay times were identified as an issue because of an apparent resistance decay problem. Using gross contact resistance failures as test specimens, the measurement delay times were varied to determine the optimal setting. The test data showed contact resistances repeatedly measured less than 10-Ω spikes with 0.5 to 1.0 s delay times. We noted that if the delay was reduced from a 0.5 to 0.25 s the contact resistance measured over-range (more than 30,000 Ω). This probably can be attributed to the 1.5 V and 160 µA generic digital multimeter test current and voltage punching through a thin film.
If a dry circuit tester (0.020 V and 100 µA) was used, the delay time might not be a factor. The model 2010 multimeter from Keithley Instruments (Cleveland, OH) was identified as having the dry circuit test features listed above. However, it appears settling on a 0.050-s delay (the switches were specified to be 0.015 s maximum) alleviates the resistance decay concerns.
The test set used standard reed switches that were guaranteed to have less than 0.5 Ω contact resistance after two million actuations. Typical mercury wetted relays are guaranteed to have less than 0.25 Ω contact resistance after two million actuations. Even after two million actuations, there has been no noticeable degradation to the contact resistance of the scanner reeds, probably due to using such a low voltage and current (1.5 V and 160 µA). Test-set repeatability is periodically evaluated by installing the shorting fixtures, performing a contact resistance run, then reviewing the resulting data.
It was learned that since resistance measurements are source-type measurements (a voltage is output to the DUT), the adjacent meter's measurement could be affected. Excessive noise (about 0.075 Ω) was generated in the measurements because these parts have a common port that is shared between the normally open and normally closed meter inputs. This problem was overcome by programming one meter to read (but not measure) amperes while the other meter measured ohms. Since the ampere measurement requires a different terminal and is a sense-type measurement, the meter error voltage is cleverly removed. The data in this report includes this added noise (before it was removed via programming changes).
As an example of the failure testing and analysis performed during this study, two radio relay failures were discovered during the course of ESS testing at room temperature—high VSWRs were found for both parts. The two relays were evaluated with the automated test system and problems were found with the normally open contact in both cases. Peaks of 100 Ω were found at cold temperature (Fig. 4, top) while a complete open-circuit condition was found at cold temperature (Fig. 4, bottom) during the first temperature cycle.
Both parts were returned to the supplier who was asked to retrieve the original cold and hot contact resistance data for these two relays. No anomalies were found upon reviewing that data, although it was noted that the original drawings called for a contact resistance test current of 1 mA. The supplier found no failures when testing at this test current. To simulate the RSC testing, the supplier changed the contact resistance test current to 100 µA and again no failures were found during testing.
The supplier temperature-cycled the relays overnight (10 cycles from +85 to −55°C with half-hour dwell times). The supplier then retested the relays and again noticed no failures. The parts were then returned to RSC for evaluation on the automated test system, where measurements indicate both were in good shape in spite of having failed earlier on the RSC equipment (Figs. 5, top, and 5, bottom). It could only be surmised that the supplier's testing may have inadvertently cleaned the contacts.
By analyzing the differences in the two test methodologies and equipment, three findings surfaced:
Differences in the two measurement methods appear to affect test yields since the contact-resistance screening at RSC consistently detects defective relays that have nonetheless passed normal screening procedures.
Within a two-week period of isolating these first two relay "failures," a third radio relay failure was found. The failure symptom was reported to be high VSWR after ESS at room temperature. The relay was tested on the automated contact-resistance test system and the normally open contact side failed at ambient temperature, hot temperature, and cold temperature. After testing a second time, similar results were found.
Because of the appearance of a system-level relay failure, detailed analysis was performed, including x-ray inspection of the three failed relays as well as 24 relays taken from received stock. The relays show similar reed deflection indicating adjustment is balanced (i.e., equal overtravel on both sides). Overtravel is defined as the amount of extra travel after electrical contact has been made and has a direct correlation to contact force.
The left-hand side of Fig. 6 shows a x-ray view of a radio level failure and identifies some of the critical parts within, while the right-hand side of Fig. 6 shows the same relay in an energized state. Inspection of the x-ray revealed similar reed deflection on both the normally open and the normally closed sides indicating adequate over-travel adjustment. Analysis by the supplier of the two system-level failures confirmed proper over-travel adjustment. Internal visual inspections revealed no evidence of contamination or unusual wear, and microsections of the reed contacts showed adequate nickel- and gold-plating thicknesses as well as complete coverage free of voids and pinholes.
After performing a series of additional tests on relays from this and another supplier, it was concluded that there is a high correlation between RF relay VSWR failures and contact-resistance issues. A high first-pass correlation was found to system-level failures. The RSC contact-resistance test (and measurement system) has proven to be very effective in identifying coaxial RF relay lots that are susceptible to the multiple failure mechanisms present in these designs. A recommendation to coaxial RF relay manufacturers is for close monitoring of their products particularly for low-temperature, dry-circuit applications (especially those approaching −55°C). It is recommended that suppliers test and qualify their relays rather than guarantee compliance without performing the necessary tests. Manufacturers must place additional emphasis in their training, process documentation, and testing efforts. In addition, it is recommended that the MIL-S-3928/15 document be modified to incorporate dry circuit testing. At RSC, supplemental coaxial RF relay procurement requirements now include the critical test methods and procedures that help ensure that procured relays meet stringent system requirements.
ACKNOWLEDGMENTS
The author would like to thank John Beach and Skip Dolan for procuring the relays analyzed in this article. The program managers (Ken Shaw and Dick Sutton) that supported the seemingly unending analyses required to complete these builds. I also thank the following people who gave invaluable technical input for this paper: Bruce Arnold and Pete Natiello (for x-ray analysis support), Don Browne, Mike Towry, Lavene Vorel, and Steve Wakefield. Special thanks to Charlie Hill, my relay mentor who is now happily retired. Through all of the issues listed in this paper (and the numerous others not listed) the suppliers have been up-front and honest in their relationship with our company. It was a pleasure working with both under mostly difficult circumstances.
EDITOR'S NOTE
This article represents a condensation of work originally published in 1999 with the Electronic Industries Association under the title "A Military Contractor's Experience with RF Coaxial Relays." That original publication featured numerous case studies of coaxial RF relay failures and failure analyses performed at RSC with follow-up work performed at the relay suppliers' facilities and at RSC. For a copy of that original article, please contact the author or the EIA.
REFERENCES
As an example of the failure testing and analysis performed during this study, two radio relay failures were discovered during the course of ESS testing at room temperature—high VSWRs were found for both parts. The two relays were evaluated with the automated test system and problems were found with the normally open contact in both cases. Peaks of 100 Ω were found at cold temperature (Fig. 4, top) while a complete open-circuit condition was found at cold temperature (Fig. 4, bottom) during the first temperature cycle.
Both parts were returned to the supplier who was asked to retrieve the original cold and hot contact resistance data for these two relays. No anomalies were found upon reviewing that data, although it was noted that the original drawings called for a contact resistance test current of 1 mA. The supplier found no failures when testing at this test current. To simulate the RSC testing, the supplier changed the contact resistance test current to 100 µA and again no failures were found during testing.
The supplier temperature-cycled the relays overnight (10 cycles from +85 to −55°C with half-hour dwell times). The supplier then retested the relays and again noticed no failures. The parts were then returned to RSC for evaluation on the automated test system, where measurements indicate both were in good shape in spite of having failed earlier on the RSC equipment (Figs. 5, top, and 5, bottom). It could only be surmised that the supplier's testing may have inadvertently cleaned the contacts.
By analyzing the differences in the two test methodologies and equipment, three findings surfaced:
Differences in the two measurement methods appear to affect test yields since the contact-resistance screening at RSC consistently detects defective relays that have nonetheless passed normal screening procedures.
Within a two-week period of isolating these first two relay "failures," a third radio relay failure was found. The failure symptom was reported to be high VSWR after ESS at room temperature. The relay was tested on the automated contact-resistance test system and the normally open contact side failed at ambient temperature, hot temperature, and cold temperature. After testing a second time, similar results were found.
Because of the appearance of a system-level relay failure, detailed analysis was performed, including x-ray inspection of the three failed relays as well as 24 relays taken from received stock. The relays show similar reed deflection indicating adjustment is balanced (i.e., equal overtravel on both sides). Overtravel is defined as the amount of extra travel after electrical contact has been made and has a direct correlation to contact force.
The left-hand side of Fig. 6 shows a x-ray view of a radio level failure and identifies some of the critical parts within, while the right-hand side of Fig. 6 shows the same relay in an energized state. Inspection of the x-ray revealed similar reed deflection on both the normally open and the normally closed sides indicating adequate over-travel adjustment. Analysis by the supplier of the two system-level failures confirmed proper over-travel adjustment. Internal visual inspections revealed no evidence of contamination or unusual wear, and microsections of the reed contacts showed adequate nickel- and gold-plating thicknesses as well as complete coverage free of voids and pinholes.
After performing a series of additional tests on relays from this and another supplier, it was concluded that there is a high correlation between RF relay VSWR failures and contact-resistance issues. A high first-pass correlation was found to system-level failures. The RSC contact-resistance test (and measurement system) has proven to be very effective in identifying coaxial RF relay lots that are susceptible to the multiple failure mechanisms present in these designs. A recommendation to coaxial RF relay manufacturers is for close monitoring of their products particularly for low-temperature, dry-circuit applications (especially those approaching −55°C). It is recommended that suppliers test and qualify their relays rather than guarantee compliance without performing the necessary tests. Manufacturers must place additional emphasis in their training, process documentation, and testing efforts. In addition, it is recommended that the MIL-S-3928/15 document be modified to incorporate dry circuit testing. At RSC, supplemental coaxial RF relay procurement requirements now include the critical test methods and procedures that help ensure that procured relays meet stringent system requirements.
ACKNOWLEDGMENTS
The author would like to thank John Beach and Skip Dolan for procuring the relays analyzed in this article. The program managers (Ken Shaw and Dick Sutton) that supported the seemingly unending analyses required to complete these builds. I also thank the following people who gave invaluable technical input for this paper: Bruce Arnold and Pete Natiello (for x-ray analysis support), Don Browne, Mike Towry, Lavene Vorel, and Steve Wakefield. Special thanks to Charlie Hill, my relay mentor who is now happily retired. Through all of the issues listed in this paper (and the numerous others not listed) the suppliers have been up-front and honest in their relationship with our company. It was a pleasure working with both under mostly difficult circumstances.
EDITOR'S NOTE
This article represents a condensation of work originally published in 1999 with the Electronic Industries Association under the title "A Military Contractor's Experience with RF Coaxial Relays." That original publication featured numerous case studies of coaxial RF relay failures and failure analyses performed at RSC with follow-up work performed at the relay suppliers' facilities and at RSC. For a copy of that original article, please contact the author or the EIA.
REFERENCES
