Designs requiring space-grade components must take into account the variances in component values that can occur over time. This is especially true for resistors, which can exhibit significant changes over their lifetime. Unfortunately, while most public guidelines for estimating Resistor aging are reasonably conservative, the issue is designers are following typical data and advertising hype and not the actual limits that vendors are allowed to deliver to.
Given there aren’t too many suppliers of space-grade resistors, and that MIL-PRF-55342 establishes specifications for resistors, the variance that different companies use in the aging tolerances and guidelines for resistors is quite surprising. Over the years, in the WCCA we have performed, we have seen everything from 0.1% to 4% for the aging/combined environments tolerance term, often referred to as just aging. Reconsider using 0.5%, or even 1%, for MIL-PRF-55342 resistor end-of-life effects; it is probably not supportable.
The initial and temperature tolerances (beginning-of-life, or BOL) are always well defined in the datasheet. The radiation tolerance is zero for passive components. This leaves only the aging tolerance to be defined as the end-of-life (EOL) variance. This is where programs and analysts tend to get creative.
Fig 1: This photo shows what 18 years of resistor aging can look like in air. (Image: AEi Systems)
Currently, our customers are using anywhere from 0.24% to 1.25% for Class A space missions with many critical programs opting for 0.5% aging for a 10-year mission. This is apart from the initial and temperature tolerances and just covering the end-of-life variance. These variations in expectation are normal and understandable since the materials utilized for the resistive film vary substantially and have correspondingly different properties. Therefore, any single data source selected would have limited applicability.
Let’s evaluate whether these numbers seem reasonable, or at least reasonably-conservative for a worst-case circuit analysis (WCCA).
The part tolerance database, often referred to as the PVDB (parts variability database) is at the heart of the worst case analysis. Touch the PVDB after analysis is started, or make a mistake, and the entire analysis could be impacted. This would certainly be the case for any resistor tolerance changes. This is one of the main reasons why the PVDB is largely developed at the beginning of the WCCA, and why customer/program approval is crucial. Without sign-off from all parties reviewing the WCCA, calculations should not begin.
I discussed the level of rigor needed for a WCCA in my blog WCCA: Lack of rigor will cost you, but clearly the most impactful EOL tolerances are that of the resistors. In my recent papers on the test vs. analysis budget ratio (References 1 and 2), I discussed total BOL vs. total EOL variance ratios of various components. Resistors are certainly the most impactful and can have the largest EOL change percentage-wise, as shown in the table below.
Fig. 2: Don’t pare down to the bone unless you know where the bone is (tolerance-wise). (Image: AEi Systems)
This table shows the BOL to EOL tolerance stack-up variance for several different types of parts.
The extreme value tolerance variation for each part used in the WCCA is the algebraically summed combination of the initial, temperature, combined environments/aging, and radiation tolerances. The aging tolerance is usually extrapolated from Arrhenius equations based on burn-in, or short- or long- term life test data (Reference 3). A sample calculation is shown in Fig. 3. If test data is not available, then public or proprietary guidelines are used as assumptions (Fig. 4).
Fig. 3: This sample calculation of the resistor aging is based on burn-in/life test data for an 84°C 10-year mission. For a 70°C, 10,000-hour, 2% life test limit (as per the military specification [Reference 10]), the aging is between 4.67% and 4.99% using Ea as 0.28 or 0.43eV. 0.28eV is suggested by ESA (Reference 4). It should be noted that the activation energy Ea, a critical element of the calculation, is not actually known with certainty. (Image: AEi Systems)