Two simple tests can go a long way toward ensuring the reliability of today�s reed relays.
The reliability of today�s dry-reed relays can approach or exceed that of electronic components in many applications. As a result, the tests that predict whether these parts will fail prematurely must be very accurate.
Two parametric tests have shown considerable accuracy in predicting early life failures: the magnetostrictive twist test (Twist) and the contact resistance (CR) stability test (RDEL).1 These tests examine the contact quality to make life expectancy predictions. Contacts that are smooth, clean, and well aligned will provide superior long-term reliability.2
What is Magnetostrictive Twist?
Dry-reed relays are constructed using a cantilevered beam switch enclosed in a hermetically sealed glass ampule. The ampule is placed inside a coil, attachments are made to the lead frame, and the entire unit is potted in plastic. The reed switch is built using a ferromagnetic material, usually a nickel/iron alloy. The cantilevered switch is designed to come together in the presence of a magnetic field of a specified magnetic force.
The Twist test allows CR measurements to be made over various portions of the contact surface. In qualitative terms, the switch is held loosely closed by reducing the coil current to just above the dropout point.
While the switch is held lightly closed, the current through the contact is varied in both magnitude and direction. The countering forces from the magnetic field produced by the coil and the magnetic field from the current passing through the reed contacts cause the reed blades to rotate slightly. A series of CR measurements is made, and these numbers are compared to the CR measurements that are taken with nominal coil current.
A Quantitative Look at Twist
The internal magnetic forces are calculated for a standard Form A reed switch in a SIP. The physical dimensions are 1 inch long and 0.3 inch wide. The coil that surrounds the switch has 2,000 turns and is 0.8 inch long. The field strength at the center of this coil can be calculated in amps/meter by the following equation:
where: H = magnetic force in amps/meter
N = number of turns
I = current in amps
L = coil length in meters
The dropout current is measured for each relay under test. The magnetic force produced at this point is equal to the mechanical spring tension of the contacts.
For the device in this study, the dropout current was approximately 9 mA. After this value was measured, the Twist range was established by adding a fixed current to the dropout current. This value, typically determined by experimentation, varies for each type of reed relay. In this example, the value was 3 mA. The incremental force from this Twist current is calculated by the following equation:
The magnetic field strength from the coil is calculated as follows:
The second factor in the Twist measurement is the magnetic field produced from the current flowing through the closed contacts. A typical value for the current used in this measurement is 100 mA. The equation for calculating the magnetic field at a point distance r from the center of a wire is given by the following:
where: H = magnetic force in amps/meter
I = current in amps
r = distance from center of wire
For a contact beam diameter of 0.56 mm, the magnetic force for r = 0.28 mm is calculated as follows:
From these calculations, it can be seen that the magnetic force from the Twist current in the coil is 316 A/m, and the magnetic force for the current flowing through the contacts is 56.8 A/m. The magnetic force from the contacts is approximately 18% of the force holding the contacts closed.
CR measurements are made at incremental points while the twist current is varied by �30%. The ratio of magnetic force between the contact and the coil is modulated from approximately 13% to 23%. CR measurements are obtained from different areas of the contact surface as the reed blades are twisted.
Why Does the Twist Measurement Work?
The Twist measurement is performed at low contact forces and looks for contact surface irregularities and particle contamination. These conditions result in nonuniform CR readings across the contact surface.
Life test studies have demonstrated that contacts with high Twist readings have a reduced life. The Twist measurement will pick out potential problem relays, even when all other electrical parameters are well within the specification ranges.
For the parts in this study, the mean for the contact Twist measurement was 70 m? with a standard deviation of 21 m?. Typically, 1% of a tested lot is rejected for the Twist test, with a limit set at 155 m?, 4 sigma over the mean.
What Is RDEL, and Why Does It Work?
The RDEL test looks at the stability of the CR measurement. A single measurement of CR will provide a resistance value, but it will not tell you if it will change with repeated closures due to contact self-heating or contamination. The RDEL test takes repeated CR measurements with multiple cycling of the contacts between each measurement.
Typically, the contacts are cycled five times prior to a measurement. This is repeated 10 times so the contacts are cycled a total of 50 times, and 10 CR measurements are recorded.
The maximum and minimum readings are compared, and the difference is RDEL. High-quality contacts will show very repeatable measurements, typically around 1 to 2 m? for a nominal CR of 50 m?. For the parts in this study, the mean was 1.5 m? with a standard deviation of 1.0 m?. Typically, 0.1% of a tested lot is rejected for the RDEL test with a limit set at 6 m?, 5 sigma over the mean.
Life Test Studies
Life test studies were performed with populations of Twist and RDEL rejects along with a population that passed all tests.3 All parts in this study passed the manufacturer tests: CR, coil resistance, operate and release times, dielectric withstanding, and insulation resistance.4 Each sample population had 14 relays.
The life test was run switching a 10-V and 10-mA resistive load at a 250-Hz operating frequency. The test was run to 200 million cycles, and 23 parts failed out of the total population of 42 relays. The Twist population had nine total failures, with eight occurring before 100 million cycles.
The RDEL population had 10 failures, with nine occurring before 100 million cycles. The Accept population with nominal Twist and RDEL values, had no failures by 100 million cycles and four failures from 100 to 200 million cycles.
The initial parametric data is shown in Tables 1, 2, and 3 for the three populations of relays. All failed parts were removed from the test at the point of failure, and the CR and RDEL readings for these parts were recorded at that time.
Table 1. Parametric Summary for Accepts
Note: Shaded rows indicate failures during life test
Table 2. Parametric Summary for Twist Rejects
Note: Shaded rows indicate failures during life test
Table 3. Parametric Summary for RDEL Rejects
Note: Shaded rows indicate failures during life test
All failures were due to high CR resulting in the test system recording the failure as a missed closure. From the data, it can be seen that the CR for failing parts ranged from 0.3 ? to as high as 10 ?.
All failing devices also show a high degree of contact instability in the RDEL readings, with readings ranging from 37 m? to several ohms. Both the RDEL and the Twist tests predicted which contacts would degrade and result in early life failures with better than 50% accuracy: eight out of 14 for the Twist test and nine out of 14 for the RDEL test.
The life test results are shown in the Weibull plot in Figure 1. B1 and B10 are defined as the number of cycles for a 1% and 10% failure rate, respectively. From this plot, a dramatic difference can be observed for the B1 and B10 failure rates between the Accept population and the Twist and RDEL reject populations.
The B1 for the Accept population was 60 million cycles vs. 100 cycles for the Twist population and approximately one cycle for the RDEL population. The B10 for the Accept population was 140 million cycles vs. 200k cycles for the Twist population and 7k cycles for the RDEL population. In an application requiring 100 million cycles, more than 50% of the Twist and RDEL populations would fail vs. only 4% for the Accept population.
What Do These Tests Mean to Field Returns?
From the life test study, early life failures occurring before 100 million cycles would account for half of the Twist and RDEL rejects. Since the reject rate for these two tests typically is 1.1%, approximately 0.55% of a given random population would be expected to fail.
More conservatively, if we assume that it will take 10 years to reach 100 million cycles, the AFR would equal 0.055%. Again, assuming an annual usage of 100k relays, a total of 55 devices would fail per year, and a system using 10k relays would experience five relay failures per year. The failure rate with either cycle rate is too high and would result in costly down time and repair time in the field.
Conclusion
Two very simple parametric tests called Twist and RDEL can accurately predict parts that will fail prematurely. These tests find reed-relay contacts that are poorly aligned, contaminated, or have rough surfaces.
The Twist test is performed at low contact forces and looks for contact surface irregularities and the presence of particle contamination. This test generally will reject 1% of a population, and half of these can be expected to experience early failures.
The RDEL test is performed at nominal coil voltage and looks for instabilities related to chemical contamination in addition to poor alignment and rough contact surfaces. The RDEL test typically will reject 0.1% of a population, and over half of these can be expected to experience early failures.
When combined, the Twist and RDEL tests provide a very effective measure of contact quality. The life tests performed in this study show strong early-life failures for the Twist and RDEL rejects. A significant improvement in the operating life can be achieved by the simple use of these tests.
References
1. Sutherland, E.F., �Measuring the Quality of Relay Contacts With the RTS-201,� MICRO ELECTRO, October 1987.
2. Mitani, S., Kamoshita, G., Ono, K., and Tanii, T., �An Analysis of Life Limiting Factors for Medium-Sized Reed Switches,� 16th Relay Conference, 1968.
3. Roettjer, P.G., �Life Testing and Reliability Predictions for Electromechanical Relays,� Evaluation Engineering, June 2004.
4. Sutherland, E.F., �Quality and Reliability Considerations for Users of Dry Reed Relays,� MICRO ELECTRO, May 1987.
About the Author
Phil Roettjer is president of Relay Testing Services and has more than 20 years of experience with electromechanical devices. Prior to RTS, he was the director of quality for the Server Products Division at Maxtor. Relay Testing Services, 89 Hartford Ave. East, Mendon, MA 01756, 508-473-5005, e-mail: [email protected]
FOR MORE INFORMATION
on relay reliability testing
www.rsleads.com/512ee-191
December 2005