Failures accelerated by temperature are usually converted from test time to service time with the Arrhenius equation. Under the same conditions, changing Ea alone swings the conclusion a lot.
AF = exp[ (Ea / k) · (1/Tuse − 1/Tstress) ] · k = 8.617×10−5 eV/K · T = °C + 273.15 · equivalent years = test hours × AF / 8,760
Equivalent service life by activation energy
Conditions
Absolute temperature, test → use [K]
Acceleration factor AF
Equivalent service hours
Equivalent years
Check against the text ()
Test 125 °C, use 55 °C, 1,000 hours, 1 year = 8,760 hours. The "Text" column holds the values from the Arrhenius paragraph and Ea table of Session 6, Section 7, and from misconception 9 in Section 10.
Item
Computed
Text
What this calculation leaves out
Ea differs from one failure mechanism to another. That is why a single test cannot tell you "this chip will last 9 years."
JESD47I also cautions that this period may be shorter than actual use requirements and depends on the failure mechanism and the use environment, and it says long-term wear-out mechanisms should be covered by wafer-level reliability data.
The equation applies to temperature-accelerated failures. JEDEC JEP122 collects the acceleration models for each failure mechanism.
JESD47I, in a note to Table 1, uses the same conditions (0.7 eV, 125 °C → 55 °C) as an example and gives an acceleration factor of 78.6, about 9 years. The last digits differ because of constants and rounding.