Characterizing Global Temperature and Humidity Environmental Severity With Relation to Published Standards

Size: px
Start display at page:

Download "Characterizing Global Temperature and Humidity Environmental Severity With Relation to Published Standards"

Transcription

1 Characterizing Global Temperature and Humidity Environmental Severity With Relation to Published Standards Dustin S. Aldridge, Raytheon Missile Systems Keywords: Acceleration Factor, Diurnal Cycle, Natural Environment, Peck Model, Temperature-Humidity Test, SUMMARY & CONCLUSIONS Severe yet common product environmental tests are 1000 hours at 85 C/85%RH or 95 C/95%RH for products, circuit card assemblies and electronic components. Such environments never occur naturally, however they attempt to simulate the corrosion damage that could be expected in service. To what natural environment do these tests correlate? Further, how do tests based upon Mil-Hdbk-310, Mil-Std-810 and STANAG 2895 B3 daily environments compare to these standard test environments? The analysis employs the physics-of-failure Peck power law temperature-humidity model with common conservative values for the Arrhenius activation energy and the relative humidity exponent, based on aluminum corrosion, coupled with climatological data from 49 United States weather stations and 19 international locations. The monthly average temperature and humidity extremes were transformed into an hourly diurnal cycle assumed to occur every day of each month. Using the power law model the equivalent time at the test condition was calculated for each day, and summed for each month and location. The 85 C/85%RH test can be correlated with about 10 years in a hot and moist natural environment, such as Singapore, with 1000 hours of 95 C/95%RH exposure equivalent to 25 years. Long term tests based upon the worst case diurnal temperature and humidity cycles in military standards are on the order of 1 in 1,000,000 probability of occurrence in nature. 1 INTRODUCTION Product reliability and durability testing evaluating electronic product degradation due to corrosion with time in a hot moist environment is common. For some products the service life can be a typical controlled office environment, however for many products, the service or storage life may be dominated by the natural environment. Electronic components are commonly tested with a 1000 hour 85 C/85% Relative Humidity (RH) Temperature Humidity Bias Test such as JEDEC JESD22-A101, JEITA EIAJ ED- 4701/100 methods 102 and 103. [1-3] Common questions include: What do these test simulate? How does this test relate to the natural environment? What do long term test based upon Mil-Std daily diurnal cycles relate to in severity to the natural environment? These questions are answered based upon the Peck Power Law Model with standard conservative assumptions for the model exponents. The comparison to the 1000 hour 85 C/85%RH evaluated for world locations indicate that this test is equivalent to approximately 10 years in a severe hot and humid natural environment. For defined Mil-Std daily diurnal cycles applied for 10 years, the severity ranges from negligible impact to extreme. 2 METHODS A commonly used PoF model for combined temperature and humidity exposure based upon aluminum corrosion is the Peck power law temperature humidity model, [4, 5] % RH test AF = e % RH field n E a 1 1 K Tfield T test where %RH Test,Field are the percent relative humidity of the test and field, T Test,Field are the absolute temperatures in K of the test and field, E a is the activation energy in ev, n is the humidity exponent, and k is the Boltzman constant Commonly used values for n and E a are 2.7 and 0.7 respectively. The typical range for the activation energy for temperature humidity testing of electronic components is between ev/ K, although a wider range can be found in literature for various mechanisms, and for the humidity exponent n, between 2.5 and 3. The lower each of these numbers, the less test acceleration is provided with increasing stresses of temperature or humidity. The chosen values in this analysis are purposely conservative. The variation between the most conservative values and the least conservative values amounts to a difference in acceleration factor of approximately a factor 3.5X. As is typical in inverse power law acceleration factor equations the exponents have a significant impact on damage equivalence. There are many sources for climatological temperature and humidity data that have been recorded for significant periods of time commonly available as monthly averages. While even finer daily averages are available for some sites, it was determined that sufficient fidelity would be achieved with the monthly averages of daily high and low temperatures and humidity s. [6-9] The locations chosen were required to (1) /17/$ IEEE

2 have average temperature and humidity extremes provided and were intended to represent the expected severity ranges for cold, hot, dry and damp environments. The locations used in the analysis were 50 US cities such as Las Vegas, Phoenix, Detroit, Boston, Atlanta, Houston, New Orleans, Miami, and a number of international locations for worldwide extremes Riyadh and Madinah, Saudi Arabia; Neuquen, Argentina; Kuwait City, Kuwait; Acapulco and Zacatecas, Mexico; Halifax, Nova Scotia, Canada; Singapore; Hong Kong, China; Bluefields, Nicaragua, Haller Station in Antarctica, and Nord Greenland. For each location, the monthly high and low temperatures and humidity s were used to create a sinusoidal diurnal cycle for each day. The lowest temperature and the highest humidity occurred concurrently at 6AM and the highest temperature and lowest humidity occurred concurrently at 4 PM. 3 RESULTS The Peck equation was evaluated for each hour s temperature and humidity for an equivalent time at the 85 C/85%RH test condition with n=2.7 and E a =0.7 ev/ K. These were summed for a daily equivalent then multiplied by the number of days in each month; summing each month s contribution for an equivalence for each city. (See Table 1). Table 1: 85 C/85%RH Field Exposure Equivalence If one looks at the analysis by hour and location, the temperature severity across locations can be reasonably modeled using a logistic distribution. (Figure 1) The humidity environment can be divided into very dry environments with the moist environments requiring two lognormal distributions to represent the humidity environment as seen in Figure 2. To understand severity in this context, a 1% severity indicates that 99% of expected conditions would be higher than the noted level, and 1% would be below this level. Similarly for the 99% severity, 1% of expected conditions would be higher than the noted level, and 99% would be less than this level. Figure 1: Example US Cities 2PM Temperature Severity Distribution Figure 2: Example US Cities 2PM Dry and Wet Humidity Severity Distribution Now considering the 10 year equivalent duration at 85 C/85% RH for US cities based upon the Peck equation we can model the severity distribution as seen in Figure 3. The shape is dependent upon the selected cities where a number between Buffalo and Detroit are in a similar latitude. There are three regions dry, temperate, and humid. The end effects are likely due to the amount of data in the analysis and that it is not a true random sample across the country. Here we see that Las Vegas is the lowest combined severity from a Peck model perspective, with Miami being the highest of considered cities. This dataset indicates that the percentile environment would be about 1300 hours 3.1. Expansion to World Environment When we calculate the equivalent test time under 85 C/85%RH test conditions for multiple worldwide locations, adding extreme cold and dry locations, Antarctica and Greenland; hot and moist locations such as in Equitorial Guinea, Marshall Islands, and Nicaragua; and similarly modeling worldwide severity prediction as seen in Figure 4, there is clear asymptotic behavior to estimate the highest likely global environment. In an attempt to give equal weighting to dry, temperate, and moist environments, an analysis with an equal number of locations for each region was made with a random sample

3 from each zone. The intent here was to look at the confidence bounds at the highest severity. With this analysis the projection to the 99.99% level is approximately 1400 hours with the 90% bounds of 1200 to 1650 hours. Considering multiple analyses, the highest likely global level is estimated to be 1300 hours, which corresponds with the lower 90% bound at 99.99% severity of the highest fidelity model for expanded worldwide locations and within the bounds projected from the US only data. The desire is to define the simplest distribution that has a minimum equivalent duration of 13.6 hours representing the 0.01% severity, 254 hours representing the average duration, and 1300 hours representing the 99.99% severity. A 3 parameter Weibull with a slope of 1.53 and a characteristic life of 300 hours meets these criteria, and is contained within the 90% confidence bounds of the higher fidelity model up to about 500 hours (90% locations). Relative location severity in terms of combined temperature and humidity natural environments can be reasonably estimated from this simpler 3 parameter Weibull model. Now we can rescale the severity of each location with this model assumption for various locations as seen in Figure 5. From this, Singapore at 1022 hours of 85 C/85%RH exposure represents the 99.8% severity. Kansas City, Missouri, at hours is about the average environment. The least severe location with data is for Halley Research Station, Antarctica, at 13.6 hours under test, assumed to be at the.01% level Consideration of Mil-Std Temperature Humidity Diurnal Cycles Mil-Hdbk-310 [10] provided the worst case daily diurnal concurrent temperature and humidity in Table IV which corresponds with the STANAG Table B3 [11]. If one were to base a test upon this daily cycle for 10 years assuming the rescaled equivalence model shown in Figure 7, this would correspond to the percentile severity or on the order of 1 in a million. Table 2 provides the comparison to tests based upon various [10, 12] daily diurnal cycles applied for 10 years. It must be noted that for the extremes, it is highly unlikely that any of these cycles would be observed for every day of the year Relationship of 1000 Hour 85 C/85%RH Test to Natural Environments Table 3 provides examples of correlation of the 1000 hour 85 C/85%RH to various natural environments. There are also 1000 hour tests performed under the conditions of 95 C at 95% RH, which represent an additional acceleration factor of ~2.5. Hence to simulate 10 years of exposure in Singapore would only require hours exposure. Table 2: Mil-Hdbk/Std Diurnal Cycle Based 85/85 10 Year Test Duration & Severity (n=2.7, Ea=0.7) Table 3: Equivalent Natural Environments to the 1000 Hour 85 C/85%RH Test 4. DISCUSSION The purpose of this analysis is to put the various natural temperature and humidity environments and test environments into perspective. It is informative as to what these standard tests represent relative to real world needs. [13] It also provides information regarding the width of the temperature humidity world environment that can prove useful in reliability risk analysis. This analysis is limited by the number of sites with available data for an extreme temperature and humidity climate. Certainly with different model exponent values the effective times will be different but the relative severities will remain the same. The concept of severity is useful from a stress strength interference perspective, and can be leveraged in reliability analysis. [14] Considering the worldwide natural environment is useful to guide reasonable requirements choices and mitigate the fascination with the extreme, such as assuming every single day in the life of a product is a worst case combined environment day. While these conditions may occur for a day

4 or two a year, they do not represent a typical lifetime environment. More comprehensive daily temperature and humidity simulations could be performed, but would not significantly change the conclusions. It is reasonable to expect 3 regions of equivalence: one for extremely dry environments; one for temperate environments; and another for extremely moist environments. Statistical models can produce unrealistic values at extremes and caution is warranted. The extreme values of 13.6 hours for the dry environment 85/85 equivalence or 1300 hours for the moist environment are engineering choices within the confidence bounds. The fact that no location with data produced a 10 year accelerated test duration above 1158 hours favored the lower bound of the projection at 1300 hours. Some conservatism may be warranted, as this analysis does not consider rain events where the humidity can be 100% for an extended period of time. Rain events are generally associated with a lower temperature, so the increase in equivalent time compared to the 85/85 exposure due to a rain event is reduced. The worst case tropical monthly average high humidity s were above 95%, so this is not considered a significant error. While some specific areas of Antarctica may produce less equivalent hours, this also is not considered significant. Based upon the Peck model the standard 1000 hour 85 C/85%RH does have a reasonable correlation with 10 years in a hot and humid natural environment, such as Singapore or the Marshall Islands. The 1000 hour 95 C/95%RH is representative of about 25 years in Singapore. Engineering judgment is required to determine what requirement is considered reasonable and prudent. Figure 3: US City 85/85 10 Year Equivalent Test Duration Figure 4: 10 Year Equivalent Severity Distribution Hours at 85 C/85%RH Worldwide Locations

5 Figure 5: Re-Scaled Worldwide 10 Year 85 C/85RH Equivalence vs. Location The worst case diurnal cycles per the various military standards should be considered in that context. These represent a worst case day, but are not representative for a long term environmental requirement. Mil-Std-810G Change Notice 1 cites the basis for the various worst case natural humidity environment exposures to Majuro, Marshall Islands. Using available climatic data with the analysis methodology used in this paper indicates a 10 year 85%RH/85 C equivalence of hours, a 99.7 percentile severity for that location. The standard does not suggest any accelerated test method, rather it recommends between days of real time humidity test exposure, conceding the suggested tests are minimums. When using the Mil-Std-310 or 810 daily exposures to define a long term test requirement, the reliability demonstrated on such a test should consider the severity of the test requirement in relation to the natural environment. The demonstrated field reliability can be much higher than the demonstrated test reliability due to the severity of the test. Another consideration in terms of requirements is for enclosed systems. While this analysis covers the natural environment, the response in an enclosed system depends upon the initial atmosphere (humidity and temperature) and leakage rates based on pressure gradients experienced given the daily temperature variation. There will always be a lag in the interior environmental conditions. If there is any sealing of the system, the interior environment will require significant time at those conditions to achieve the natural environmental extremes. Some systems have essentially a one way leakage system, being able to draw in moisture but unable to expel it, leading to contained liquid water where the internal humidity is no longer correlated with the natural environment. Qualification tests need to balance multiple considerations in terms of environmental severity, exponent choices, and system design relative to closed system vacuum suction effects to assure satisfactory field reliability. REFERENCES 1. JEDEC, JESD22-A101D, Steady-State Temperature- Humidity Bias Life Test, July JEITA EIAJ ED-4701/100, Environmental and endurance test methods for semiconductor devices (Life test I), August 2001, pg C. Zorn, N. Kaminski, Acceleration of Temperature Humidity Bias (THB) Testing on IGBT Modules by High Bias Levels 2015 IEEE 27 th International Symposium on Power Semiconductor Devices & IC s 4. D. S. Peck, Comprehensive Model For Humidity Testing Correlation, IEEE International Reliability Physics Symposium, 1986, Pg O. Hallberg, D. S. Peck, Recent Humidity Accelerations, A Base for Testing Standards, Quality and Reliability Engineering International, Vol. 7, 1991, Pg NOAA: National Centers for Environmental Information. Available online: World Climate & Temperature, Available on: World Climates, Available online: World Climate Data- Temperature, Weather and Rainfall, Available online: US Department of Defense Mil-Hdbk-310, Global Climatic Data For Developing Military Products, 23 June NATO Standardization Agreement STANAG 2895 Extreme Climatic Conditions and Derived Conditions for use in Defining Design/Test Criteria for NATO Forces

6 Materiel, February US Department of Defense Mil-Std-810G Change Notice 1, Environmental Engineering Considerations and Laboratory Tests, 15 April Egbert, Herbert W. The History and Rationale of MIL- STD-810, 2 nd Edition, January 2010; Institute of Environmental Sciences and Technology, 1827 Walden Office Square, Suite 400, Schaumberg, Il A. Kleyner, Effect of Field Stress Variance on Test to Field Correlation in Accelerated Reliability Demonstration Testing, Quality and Reliability Engineering International, Vol. 31, 2015, pg Dustin S. Aldridge Raytheon Missile Systems 2751 E. Ozona Pl. Tucson, AZ USA BIOGRAPHY dustin.s.aldridge@raytheon.com Dustin Aldridge is a Senior Principal Systems Engineer in Reliability Engineering with Raytheon in Tucson, Arizona where he is a specialist in reliability risk analysis. He is the recipient of the 2011 IEST Reliability Test and Evaluation Award, an IEST Fellow, Director for the Product Reliability Division, an Editorial Advisor for the Journal of the IEST, and represents IEST on the RAMS Board of Directors