Menu

Component derating

The Derating Table That Makes the Product Last

Notes · Testing, safety and certificationNovember 19, 202619 min readSenova Systems Research

Your board was designed at the ratings, and now the warranty claims are coming in, slow drift, not sudden death. Here is the derating table that makes the product last: the 50% rule, NASA's table, the tantalum numbers, and what MTBF actually means.

Picture the field. A thousand sensor hubs out in the world, each one on a wall somewhere, each one quietly doing its job. Then the warranty claims start coming in. Not all at once, one a week, then two. The failures are not sudden death. They are slow drift: a hub that reads a little high, then reads wrong, then stops answering. The board was designed at the ratings. Every part is inside its maximum. And the product is still dying in the field.

This article is about the table that would have prevented that. Derating is the practice of running every part below its published maximum, deliberately, by a written amount, for a reason you can defend. It is the cheapest reliability engineering there is, because it costs a few cents per part and it is decided on the schematic, before a single board is built. And it is the most skipped, because the datasheet says the part can take the full rating, and the full rating looks like a promise.

It is not a promise. It is a limit.

What derating is

Derating is the proportional reduction of electrical or thermal stress applied to a component, relative to its maximum published rating, to improve reliability and extend operational life. A common goal is to maintain the component's internal stress levels below 80% of its rated capacity. (Flyriver)

Say it more simply: a rating is the most a part is allowed to take, not the amount it should take every day. Run it at the rating and you are running it at the edge of what it can survive. Run it below the rating and you are buying life, the physical wear inside the part slows down, and the part lasts longer.

"Every electrical component experiences some degree of stress during normal operation. Heat, voltage transients, current surges, and mechanical vibration all accelerate the physical degradation mechanisms inside a component. By deliberately operating below the rated maximum, you slow those mechanisms and extend the functional lifespan of the part."

(ElectronLabs)

There are two numbers you will see for the same part, and the difference between them is the first thing to get straight. The common high-reliability rule, the 50% rule, says: resistors at no more than 50% of rated power, electrolytic capacitors at no more than 50% of rated voltage, ceramic capacitors at 70–80% of rated voltage. NASA's table, the KSC Preferred Practices, taken from MIL-STD-975, the NASA Standard Parts List, says: capacitors at a maximum of 60% of rated voltage, resistors at a maximum of 60% of rated power.

The two disagree, and the disagreement is not a mistake. NASA's 60% is a floor for space flight, where every gram and every millimetre is fought over; the 50% rule is the conservative practice for a product that has to survive in the field for years without a service visit. This article uses both, and says which is which every time.

If your design is at 60%, it meets NASA's table and it fails the 50% rule, and that is a decision to make with your eyes open, not an accident to discover in the warranty data.

The table

Here is the table, from NASA's KSC Preferred Practices (MIL-STD-975, Table 1, "Typical Part Derating Guidelines"), with the common high-reliability rule alongside it. Read it as a specification, not a suggestion: each row is a maximum, and anything above it is outside the table.

Part typeNASA (MIL-STD-975)Common high-reliability rule
CapacitorsMax 60% of rated voltageElectrolytic: max 50% of rated voltage. Ceramic: 70–80% of rated voltage
ResistorsMax 60% of rated powerMax 50% of rated power
Semiconductor devicesMax 50% of rated power; max 75% of rated voltage; max junction temperature 110 °CJunction temperature at least 20 °C below the maximum (Flyriver); for high-reliability work, Tj ≤ 110 °C or Tj,max − 40 °C (Reliatools)
MicrocircuitsMax supply voltage 80% of rated; max 75% of rated power; max junction temperature 100 °CSupply voltage 80–90% of rated (TrueGeometry)
Inductive devicesMax 50% of rated voltage; max 60% of rated temperatureRMS current ≤ 75% of rated (Flyriver)
Relays and connectorsMax 50% of rated currentContact current ≤ 60% of rated for inductive loads (Flyriver)

Two footnotes from the NASA document itself. First, the junction-temperature limits "should not be exceeded at any time or during any ground, test, or flight exposure", not on average, not usually, not except during the test. Second, "EEE parts derating can be established as either design policies or from reliability requirements.

In general, NASA has taken the approach of establishing derating policies that cover all applications of the various part types in space flight equipment. These policies are available in MIL-STD-975. If derating is to be determined from a reliability requirement, the reference document is MIL-HDBK-217."

MIL-HDBK-217, "Reliability Prediction of Electronic Equipment", is the handbook that turns the table into a number. It is the reference for quantitatively estimating the effects of stress levels on reliability: given a part, a stress level, and a temperature, it gives you a failure rate. The table tells you what to do; the handbook tells you what it buys you.

Other standards exist for the same job, MIL-STD-1547 (Department of Defense), NAVSEA SD-18 (U.S. Navy), ECSS-Q-30-11A (European Space Agency), MSFC-STD-3012 (NASA Marshall), and EDN's warning applies to all of them: "For the same component, the derating parameters in different standards may vary." Pick one standard, write it down, and apply it consistently. A design that is half NASA and half 50%-rule is not derated.

It is unexamined.

Why it works: the physics underneath the percentages

The percentages are not arbitrary. Each one is a shorthand for a physical law, and the law is what makes the table work.

Heat kills by chemistry. For electrolytic capacitors, the rule of thumb, rooted in the Arrhenius equation, which governs how fast chemical reactions run, is that every 10-degree Celsius increase in operating temperature roughly halves the capacitor's expected lifespan. ElectronLabs' example: "A capacitor rated for 2,000 hours at 85 degrees Celsius that operates continuously at 95 degrees Celsius may deliver only 1,000 hours of service before its electrolyte degrades to the point of failure." The electrolyte dries out. It does not fail suddenly; it dries out slowly, the capacitance falls, the equivalent series resistance rises, the ripple on the rail grows, and one day the regulator that depended on that capacitor stops regulating. That is the slow drift the warranty claims describe. The demo below runs the numbers: the same capacitor, rated 5,000 hours at 105 °C, lasts 0.6 years at its rating and 9.1 years at 65 °C.

Voltage kills by field strength. A capacitor's dielectric is an insulator with a voltage across it, and the higher the voltage, the harder the electric field pushes on it. For tantalum capacitors the effect is so strong that the failure mode is catastrophic, the part does not degrade gracefully, it shorts, and a shorted tantalum on a power rail can take the board with it. Vishay's white paper states it flatly: "Solid tantalum capacitors should never be operated at or above their rated voltage due to their susceptibility to damage and their catastrophic failure mode." The numbers are in the tantalum section below, and they are the most dramatic in this article: the same part, derated or not, differs by a factor of a thousand.

Temperature kills semiconductors by doubling. "For many semiconductor technologies, a 10–15 °C increase in junction temperature can approximately double the failure rate." (Reliatools) The junction is the temperature that matters, not the ambient, and the junction is hotter than the ambient by the power the part dissipates times its thermal resistance. ElectronLabs' example: "A voltage regulator with a junction-to-ambient thermal resistance of 50 degrees Celsius per watt, dissipating 1 watt in a 40-degree ambient environment, reaches a junction temperature of 90 degrees Celsius, still within the rated maximum of 125 degrees, perhaps, but already operating in territory where long-term reliability is measurably compromised." The datasheet's maximum is 125 °C. NASA's limit is 110 °C. At 90 °C you are inside both, and still running hotter than you should, because the failure rate at 90 °C is roughly double what it is at 75 °C.

And the ratings themselves are measured at room temperature. "Most datasheets publish their maximum ratings at an ambient temperature of 25 degrees Celsius, a comfortable room temperature that describes almost no real operating environment." (ElectronLabs) The derating curve buried in the back of most datasheets quantifies exactly this: it shows how the maximum allowable operating parameter, power, voltage, current, must be reduced as temperature rises. A resistor rated for 0.25 W at 25 °C is not a 0.25 W resistor at 85 °C. If you sized it from the headline number, you sized it from a temperature your product will never see.

MTBF and FIT: what the numbers mean, and what they do not

The handbook's output is a failure rate, and the failure rate is expressed in two units you need to be able to read.

FIT, failures in time, is one failure per billion hours of operation. A part rated at 10 FIT will, on average, fail once in a hundred million hours of operation. MTBF, mean time between failures, is the reciprocal: MTBF in hours = 1,000,000,000 ÷ FIT. A part at 10 FIT has an MTBF of 100 million hours, or about 11,415 years. A system MTBF is the reciprocal of the sum of its parts' failure rates: add up every part's FIT, divide a billion by the total, and that is the system's MTBF, the MIL-HDBK-217 method, in one line.

Three things MTBF is not. First, it is not a lifetime. A board with an MTBF of 100,000 hours will not last 100,000 hours.

MTBF is the average time between failures across a population, some boards fail early, some late, and the average is the number. Second, it is not a guarantee. A 1,000-hour MTBF does not mean your board will fail at 1,000 hours; it means that in a large fleet, one failure per thousand board-hours is the expected rate.

Third, it only applies in the flat middle of the bathtub curve, the useful-life period, where the failure rate is roughly constant. Early life has infant mortality (manufacturing defects, caught by burn-in and by the test coverage in Article #32). Late life has wear-out (electrolyte drying, the Arrhenius rule above).

MTBF describes the middle. A capacitor that dries out in two years is not an MTBF failure. It is a wear-out failure, and no amount of derating the voltage will fix it if the temperature is the problem.

The failure-rate levels have names, from the Vishay presentation: M = 1% per 1,000 hours; P = 0.1%; R = 0.01%; S = 0.001%. A commercial solid tantalum capacitor sits at M-level, 1% per 1,000 hours, at or near its rated voltage. The 50% derating takes it to 5–15 FIT, which is roughly 0.0000005% per 1,000 hours. That is the scale of the difference the table buys you, and the demo makes it concrete.

The tantalum case: the part that fails catastrophically

Tantalum capacitors deserve their own section, because they are the part where derating is not a refinement but the difference between a product and a fire.

A solid tantalum capacitor with a manganese dioxide (MnO2) cathode, run at or near its rated voltage, "typically experience[s] a 1% failure rate (FR) per 1000 h of operation." (Vishay) That is the M-level rate: one in a hundred parts fails every thousand hours. And the failure is not a graceful fade.

The part shorts. A short on a power rail draws the full current the rail can supply, and the localised heating can ignite the part and the board around it. "Solid tantalum capacitors should never be operated at or above their rated voltage due to their susceptibility to damage and their catastrophic failure mode."

The fix is the 50% rule, and the handbook quantifies it: "Commercial solid tantalum capacitors have typical failure rate of 1% per 1000 hours. However, for high reliability applications that require a much lower failure rate, the MIL-HDBK-217 standard was established. This included a specification for the derating of capacitor voltage that was a cornerstone of increasing the operating life.

This standard dictates the 50% voltage derating to achieve a 5 FIT to 15 FIT (failures in time) rate. 1 FIT is one failure per billion hours." (Vishay white paper, document 40292, revised 19 July 2024)

Read that again: 1% per 1,000 hours is 10,000 FIT. The 50% derating is 5–15 FIT. The same part, run at half its voltage instead of at its rating, fails roughly a thousand times less often.

The Vishay CMSE presentation puts it in field terms: "Derating of applied voltage from full rated voltage down to 50%–66% dropped field FRs to approximately 5 FIT–15 FIT from hundreds of thousands of FIT." And the presentation's worked example: a life test of 100 samples for 1,000 hours at rated voltage predicts a failure rate of 2.3% per 1,000 hours; derating by 50% for MnO2 capacitors brings the predicted failure rate to approximately 0.1% per 1,000 hours and "increases MTTF from five years to over 100 years."

The practical rule, from Vishay's recommended derating guidelines (below 85 °C, for MnO2 solid tantalum, using EIA-809 as well as the NASA and MIL-STD guidelines): the working voltage should usually be between 0.5 and 0.6 times the rated voltage. For rated voltages of 35 V and above, Vishay recommends an even higher derating ratio, with a minimum of 50% derating. A 5 V rail wants a 10 V tantalum.

A 12 V rail wants a 25 V tantalum. A 3.3 V rail wants a 6.3 V tantalum, and a 3.3 V rail on a 3.3 V-rated part is not a design. It is a failure waiting for a warm day.

One distinction to keep: polymer tantalum capacitors need less derating. Polymer cathodes self-heal, the localised heating of a potential failure site makes the polymer non-conductive, insulating the spot, so the same 0.1% per 1,000 hours is reached at 80% derating for polymer where MnO2 needs 50%. Polymer tantalum now has its own military standard, MIL-PRF-32700, and Vishay's FIT calculator follows MIL-HDBK-217 revision F. If you are specifying a tantalum today and the board can afford the polymer part, the derating table is shorter, but it is still a table, and the part still has one.

Example Labs is a made-up company we use for worked examples.

Example Labs' sensor hub board, derated

Example Labs is the fictional 60-person company this series uses. The board is the field sensor hub board from Article #32: 100 × 60 mm, four-layer, 120 nets, 114 test points. It runs from the 48 V bench pack (13S1P NMC, 54.6 V full, from Articles #20, #23, and #42) through a buck converter to a 5 V rail and a 3.3 V rail for the STM32H7. Here is the board, derated, and then not derated, with every figure computed in Python before writing.

Step 1, the tantalum on the 5 V rail. The board carries ten solid tantalum capacitors with MnO2 cathodes. As drawn, they are 6.3 V parts on a 5 V rail, 79% of rated voltage. That is inside NASA's 60%? No, it is above it. It is above the 70% commercial ceiling, above NASA's 60%, and far above the 50% rule. At or near rated voltage the failure rate is 1% per 1,000 hours, which is 10,000 FIT per part. Ten of them give a combined failure rate of 100,000 FIT, an MTBF of 10,000 hours, 1.1 years, and across a fleet of 1,000 hubs running 8,760 hours a year, 876 tantalum failures a year. Replace them with 10 V parts, 50% derating, and each one falls to 10 FIT (the midpoint of the 5–15 FIT range). Ten of them give 100 FIT, an MTBF of 10,000,000 hours, 1,141 years, and 0.88 failures a year across the whole fleet. Same board, same circuit, same capacitance. The only change is the voltage rating printed on the part. The warranty arithmetic: at $150 per field replacement (the article's own assumption, parts, shipping, and labour), the rated-voltage board costs $131,400 a year in warranty, and the derated board costs $132 a year. The extra bill of materials, ten parts at roughly $0.10 more each, $1,000 across the whole fleet, pays for itself 131 times over in the first year.

Step 2, the electrolytic on the 12 V rail. The buck converter's input capacitor is an electrolytic rated 5,000 hours at 105 °C. Inside the enclosure, on a warm day, it runs at 85 °C: 5,000 × 2^((105−85)/10) = 20,000 hours, or 2.3 years. At 95 °C it is 10,000 hours, 1.1 years. At 65 °C it is 80,000 hours, 9.1 years. At 55 °C it is 160,000 hours, 18.3 years. Every 10 °C is a doubling or a halving, and the temperature is the design variable, the enclosure, the airflow, and the heatsink from Article #35 decide it, not the capacitor. The voltage matters too: as drawn, the part is a 16 V capacitor on a 12 V rail, 75% of rated voltage, the ElectronLabs example of a designer reasoning that a 4 V margin is adequate headroom. It is above the 50% long-life target and at the 70% general-commercial ceiling. The 50% rule wants a 25 V part on a 12 V rail. The 16 V part is not wrong by NASA's 60% table, 75% is above it, and it is not wrong by the commercial 70% ceiling by much. It is a decision to make with the temperature in hand: at 65 °C and 75% voltage, the part lasts; at 85 °C and 75% voltage, it does not.

Step 3, the resistor that is running at 80%. A 0.25 W resistor on the board dissipates 0.2 W, 80% of its rating. It passes NASA's 60%? No, 80% is above 60%. It fails the 50% rule by a wide margin. The fix is a 0.5 W part in the same footprint, which runs at 40%. The cost is a few cents and a slightly larger package. The failure it prevents is the slow one: the resistor runs hot, its value drifts with temperature, the circuit it sets, a current limit, a divider, a pull-up, drifts with it, and the hub reads a little high, then wrong, then stops answering. That is the warranty claim from the opening paragraph.

Step 4, the regulator's junction temperature. The 5 V regulator dissipates 1 W, with a junction-to-ambient thermal resistance of 50 °C/W, in a 40 °C ambient, ElectronLabs' example, and Example Labs' board is the same shape. The datasheet maximum is 125 °C. NASA's limit for semiconductor devices is 110 °C. At 90 °C the part is inside both, and the margin to NASA's limit is only 20 °C, and the failure rate at 90 °C is roughly double what it is at 75 °C, because a 10–15 °C rise approximately doubles it. The fix is the same as in Article #35: cut the dissipation (a more efficient converter, a lower drop) or cut the thermal resistance (copper, a heatsink, airflow). Halve the dissipation to 0.5 W and the junction falls to 65 °C, and the failure rate roughly halves again. And from Article #35, the MOSFET's on-resistance approximately doubles from 25 °C to 125 °C, so the hotter it gets, the more it dissipates, the hotter it gets. Derating the junction is what breaks that loop.

Tj=40+1×50=90 °C\text{Tj} = 40 + 1 \times 50 = \text{90 °C}

Step 5, the board's MTBF, as a sum. Add the parts up, MIL-HDBK-217 style. Ten tantalums at 10 FIT each (50% derated) = 100 FIT. Twenty resistors at 1 FIT each (the article's own assumption, labelled) = 20 FIT. Five ICs at 10 FIT each (the article's own assumption, labelled) = 50 FIT. Total: 170 FIT, an MTBF of 5,882,353 hours, 672 years. The same board with the tantalums at rated voltage: 10 × 10,000 FIT = 100,000 FIT, plus the same 70 FIT of everything else, gives 100,070 FIT and an MTBF of 9,993 hours, 1.14 years. One part type, derated or not, moves the board's MTBF from about a year to about seven centuries. The other parts barely matter, because the tantalum dominates the sum, which is the point of the sum. Find the part with the biggest FIT, fix it first, and the MTBF follows.

Step 6, what the watchdog sees. The failures derating prevents are the ones Article #37's watchdog is built to catch: the brownout, the hung regulator, the rail that sags and never comes back. The watchdog's job is to recover from a failure. Derating's job is to make the failure not happen. A watchdog with a 1,000 ms nominal timeout, actually 681 ms at the fast end of the LSI tolerance and 1,882 ms at the slow end, is the last line of defence. The derating table is the first.

The decision test

Five questions. If you can answer all five from your own schematic, your board is derated. If you cannot, it is designed at the ratings.

  1. What is the derating standard for this board, and is it written down? NASA's MIL-STD-975 table, or the 50% rule, or your own table. Pick one, write it in the design notes, and apply it to every part. A design that is half one standard and half another is not derated.
  2. For each capacitor, what is the working voltage as a percentage of the rating? Electrolytic at 50% for long life, 70% for general commercial. Ceramic at 70–80%. Tantalum MnO2 at 50–60%, and never at or above the rating, because the failure mode is catastrophic.
  3. For each resistor, what is the dissipation as a percentage of the rating, at the hottest temperature the board will see? 50% at room temperature is not 50% at 85 °C, because the rating itself falls as the temperature rises. Size from the derating curve, not the headline number.
  4. For each semiconductor, what is the junction temperature, not the ambient? Tj = Ta + P × θJA. Compare it to NASA's 110 °C for discrete semiconductors and 100 °C for microcircuits, and to the datasheet maximum minus 20 °C. If you do not know the junction temperature, you do not know whether the part is derated.
  5. What is the board's MTBF, and which part dominates it? Add the FITs. The part with the biggest number is the one to fix first. If the answer is a tantalum at rated voltage, the fix costs a few cents and the MTBF moves by a factor of a thousand.

Objections

"The datasheet says the part can take the full rating. Why would I use a bigger part?" Because the datasheet's maximum is a limit, not a recommendation, and it is measured at 25 °C ambient, a temperature your product will almost never see. The part can take the full rating once, in a lab, at room temperature. Your product has to take its actual stress every day for years, in an enclosure, on a warm day. The table is the difference between those two situations.

"Derating everything will make the board bigger and more expensive." Derating the tantalum costs $0.10 a part. Derating the resistor costs a larger package on one footprint. The board does not get meaningfully bigger, and the bill of materials does not meaningfully rise, the demo's $1,000 across a thousand hubs saves $131,268 a year in warranty. The expensive version is the one that comes back.

"Our product is a consumer product, not a spacecraft. NASA's table is overkill." NASA's table is the floor for space flight, where weight is fought over gram by gram. The 50% rule is the conservative practice for a product that has to survive without a service visit. If anything, a consumer product that cannot be serviced needs the 50% rule more than a spacecraft does, the spacecraft has a ground team watching it. Your hub has a customer who will return it.

"We have a watchdog. It recovers from failures." The watchdog recovers from a failure after it happens. Derating prevents the failure from happening. Both belong on the board, Article #37 is the last line of defence, and this article is the first. A product that recovers from a failure every week is still a product that fails every week.

"MTBF is a made-up number. Nobody believes a 672-year MTBF." Nobody should, and the article says so. MTBF is an average across a population, not a lifetime, and it only applies in the flat middle of the bathtub curve. Its value is comparative: it tells you which part dominates the failure rate, and how much the number moves when you fix it. A board at 1.14 years and a board at 672 years are not two predictions. They are two designs, and one of them is better by a factor of six hundred.

Sources

  1. Flyriver, "Derating is the proportional reduction of electrical or thermal stress applied to a component, relative to its maximum published rating, to improve reliability and extend operational life. A common goal is to maintain the component's internal stress levels below 80% of its rated capacity."
    What we used

    "For passive components like resistors, derating is typically applied to the power dissipation rating. A standard industry practice, particularly in high-reliability applications (e.g., military or aerospace), is the 50% rule for power derating. If a resistor is rated for 1 Watt, it should be operated at 0.5 Watts or less." "For capacitors, voltage derating is paramount, especially for electrolytic and ceramic types. Operating a capacitor near its rated voltage accelerates the degradation of the dielectric layer, leading to increased leakage current and eventual short-circuit failure. High-reliability standards often mandate that the applied DC voltage (VDC) should not exceed 50% of the rated voltage (Vrated) for electrolytic capacitors, and 70% to 80% for ceramic capacitors." Table 1: "Resistor (Power) Pactual ≤ 50% Prated; Electrolytic Capacitor Vactual ≤ 50% Vrated; Semiconductors (TJ) TJ ≤ TJmax − 20 °C; Inductor/Transformer IRMS ≤ 75% Irated; Relays/Switches IC ≤ 60% Irated (for inductive loads)."

  2. ElectronLabs, "Rated to Fail"
    What we used

    "Derating is the practice of operating a component at a fraction of its published maximum rating." "Every electrical component experiences some degree of stress during normal operation. Heat, voltage transients, current surges, and mechanical vibration all accelerate the physical degradation mechanisms inside a component. By deliberately operating below the rated maximum, you slow those mechanisms and extend the functional lifespan of the part." "Most professional design standards, including those published by NASA, the military under MIL-HDBK-338, and major commercial aerospace bodies, mandate specific derating percentages for different component classes. A general-purpose guideline used across the industry recommends operating resistors at no more than 50 percent of their rated power dissipation, capacitors at no more than 50 to 70 percent of their voltage rating, and semiconductors within similarly conservative thermal and electrical boundaries." "Most datasheets publish their maximum ratings at an ambient temperature of 25 degrees Celsius, a comfortable room temperature that describes almost no real operating environment." "For electrolytic capacitors, a widely cited rule of thumb, rooted in the Arrhenius equation governing chemical reaction rates, holds that every 10-degree Celsius increase in operating temperature roughly halves the capacitor's expected lifespan. A capacitor rated for 2,000 hours at 85 degrees Celsius that operates continuously at 95 degrees Celsius may deliver only 1,000 hours of service before its electrolyte degrades to the point of failure." "A voltage regulator with a junction-to-ambient thermal resistance of 50 degrees Celsius per watt, dissipating 1 watt in a 40-degree ambient environment, reaches a junction temperature of 90 degrees Celsius, still within the rated maximum of 125 degrees, perhaps, but already operating in territory where long-term reliability is measurably compromised." "A designer selects a 16-volt electrolytic capacitor for a 12-volt power supply rail, reasoning that a 4-volt margin provides adequate headroom." "For capacitors, target an operating voltage at or below 50 percent of the rated voltage for long-life applications, and no more than 70 percent for general commercial designs. For semiconductors, verify junction temperatures under worst-case thermal conditions using the thermal resistance data in the datasheet, not the ambient temperature alone."

  3. NASA KSC Preferred Practices, "EEE Parts Derating" (MIL-STD-975, Table 1)
    What we used

    "EEE parts derating can be established as either design policies or from reliability requirements. In general, NASA has taken the approach of establishing derating policies that cover all applications of the various part types in space flight equipment. These policies are available in MIL-STD-975, 'NASA Standard Parts List.' Table 1 provides typical derating factors from that document. If derating is to be determined from a reliability requirement, the reference document is MIL-HDBK-217." Table 1: "Capacitors Max. of 60% of rated voltage; Resistors Max. of 60% of rated power; Semiconductor Devices Max. of 50% of rated power, Max. of 75% of rated voltage, Max. junction temperature of 110 °C; Microcircuits Max. supply voltage of 80% of rated voltage, Max. of 75% of rated power, Max. junction temperature of 100 °C; Inductive Devices Max. of 50% of rated voltage, Max. of 60% of rated temperature; Relays and Connectors Max. of 50% of rated current. NOTE: Maximum junction temperature levels should not be exceeded at any time or during any ground, test, or flight exposure."

  4. EDN, "Electronic Components Derating, Made Easy"
    What we used

    "Electronic components derating (or de-rating) is the operation of a component at less than its rated maximum capability in order to prolong its life." Standards in common use: "MIL-STD-975, published by NASA; MIL-STD-1547, published by the Department of Defense; NAVSEA SD-18, published by the U.S. Navy; ECSS-Q-30-11A, prepared and maintained under the authority of the Space Components Steering Board in partnership with the European Space Agency; MSFC-STD-3012, prepared by NASA's Marshall Space Flight Center." "Note: For the same component, the derating parameters in different standards may vary."

  5. NAVSEA SD-18, "Thick Film Resistor: Current/Voltage, 80%, Power, 50%."
  6. Reliatools, "Component Derating in Electronics"
    What we used

    "For many semiconductor technologies, a 10–15 °C increase in junction temperature can approximately double the failure rate." "For electrically stressed components such as capacitors and dielectrics, lifetime often follows an inverse power law." "Semiconductors: Focus on junction temperature, not ambient. For high-reliability applications, targets such as TJ ≤ 110 °C or TJ,max − 40 °C are common." "IEEE Std 2818-2024 / VITA 51.4-2024, Reliability Component Stress Analysis and Derating."

  7. Vishay Sprague white paper, "Derating Tantalum Capacitors, Depends On The Cathode System" (document 40292, revision 19 July 2024)
    What we used

    "Solid tantalum capacitors should never be operated at or above their rated voltage due to their susceptibility to damage and their catastrophic failure mode." "Reliability is usually expressed as a percentage of failures per 1000 h of operation. Tantalum capacitors used at or near their rated voltage typically experience a 1% failure rate (FR) per 1000 h of operation." "Commercial solid tantalum capacitors have typical failure rate of 1% per 1000 hours. However, for high reliability applications that require a much lower failure rate, the MIL-HDBK-217 standard was established. This included a specification for the derating of capacitor voltage that was a cornerstone of increasing the operating life. This standard dictates the 50% voltage derating to achieve a 5 FIT to 15 FIT (failures in time) rate. 1 FIT is one failure per billion hours." "For solid tantalum capacitors with MnO2 electrolyte, Vishay uses the EIA-809 standard as well as the NASA and MIL-STD guidelines." "RECOMMENDED VOLTAGE DERATING GUIDELINES (below 85 °C)."

  8. Vishay / Electronic Design, "Derating Guidelines for Tantalum Capacitors"
    What we used

    "Tantalum capacitors used near their RV usually have a 0.1% to 1% failure rate per 1,000 hours." "Commercial solid tantalum capacitors have a standard failure rate of 1% per 1,000 hours ('M-level'). For high reliability applications, the MIL-STD-217 standard was established, which includes a 50% voltage derating to achieve a 5 to 15 failures in time (FIT) rate, where 1 FIT equals one failure per billion hours." "Usually, the application (working) voltage should be between 0.5 and 0.6 V of the RV. For RV ≥ 35 V, Vishay recommends an even higher derating ratio, with a minimum of 50% derating." "Polymer tantalum has recently been provided a MIL-STD, MIL-PRF-32700, to guide users."

  9. Vishay CMSE 2023 presentation (Jon Rhan & Jerard Jose), "Derating of applied voltage from full rated voltage down to 50%–66% dropped field FRs to approximately 5 FIT–15 FIT from hundreds of thousands of FIT."
    What we used

    "M = 1% per thousand hours; P = 0.1%; R = 0.01%; S = 0.001%." "A life test of 100 samples for 1000 h at the specified temperature and rated voltage predicts an FR of 2.3%/1000 h." "Derating of 50% for MnO2 capacitors and 80% for polymer capacitors brings the predicted FR to approximately the same level of 0.1% and increases MTTF from five years to over 100 years!" "While 50% to 60% derating is recommended for MnO2 capacitors, the following is applicable for polymer capacitors." "The FR [λp] for MnO2 capacitors is defined in MIL-HDBK-217F Notice 2."

  10. Vishay Tantalum FIT Calculator, "Vishay's Tantalum FIT calculator program was developed to follow the calculations defined in MIL-HDBK-217 (revision F). This program enables users to predict the reliability of tantalum capacitors."
  11. Demo arithmetic verified in Python before writing: 1% per 1,000 h = 10,000 FIT; 5 FIT → MTBF 200,000,000 h (22
    What we used

    831 years); 10 FIT → 100,000,000 h (11,415 years); 15 FIT → 66,666,667 h (7,610 years); 10,000 FIT → 100,000 h (11.4 years); 100,000 FIT → 10,000 h (1.1 years). Ten tantalums at rated voltage: 100,000 FIT, MTBF 10,000 h (1.1 years), fleet failures 876/year across 1,000 hubs. Ten tantalums at 50% derating (10 FIT each): 100 FIT, MTBF 10,000,000 h (1,141 years), fleet failures 0.88/year. Warranty at $150 per field replacement (article's own assumption): $131,400/year vs. $132/year; extra bill of materials $0.10 × 10 parts × 1,000 hubs = $1,000. Electrolytic rated 5,000 h at 105 °C: 105 °C → 5,000 h (0.6 years); 95 °C → 10,000 h (1.1 years); 85 °C → 20,000 h (2.3 years); 75 °C → 40,000 h (4.6 years); 65 °C → 80,000 h (9.1 years); 55 °C → 160,000 h (18.3 years). 16 V capacitor on a 12 V rail = 75% of rated; 50% rule wants a 25 V part (24 V computed, rounded to the next standard value). Resistor: 0.2 W on a 0.25 W part = 80%; on a 0.5 W part = 40%. Regulator: Tj = 40 + 1 × 50 = 90 °C; NASA limit 110 °C; margin 20 °C. Board total: 10 tantalums at 10 FIT + 20 resistors at 1 FIT + 5 ICs at 10 FIT = 170 FIT → MTBF 5,882,353 h (672 years); same board with tantalums at rated voltage: 100,070 FIT → MTBF 9,993 h (1.14 years)

More in design notes

EMC/EMIEvery board is a radio station. The only question is whether it knows it is one. A 60 mm clock trace on a 2-layer board fails radiated emissions at its 5th harmonic by 1.7 dB; the same trace on a 4-layer board passes with 14.4 dB of margin. This piece shows the three ways a board radiates, the return path that decides most failures, and the stackup that fixes them.25 minPCB Design for Manufacture and TestThe average PCB project needs 2.9 respins at $46,000 each, and teams that simulate first need 3.0 spins to production against 4.6 for build-and-test teams. A board that works on a bench is not the same as a board a factory can build, test, and repair. The rules that close that gap, in plain words.14 minThe Profile That Doesn't TombstoneYour boards come back with tombstoned parts, cold joints, and voids, and the assembler says the profile is fine, because the oven's setpoints are fine and the board's temperatures are not. Here is the reflow profile as four zones, measured on the board, with the numbers.17 minAluminium, Steel, or PlasticAluminium, steel, or plastic? The material decides the cost, the weight, the heat, the EMI, and whether the product survives outdoors, and the wrong one fails in under two years. Three axes, four plastics, three metals, and the arithmetic for Example Labs' field sensor hub.20 minCNC Machining DFMA 30 mm deep pocket with R2 corners forces a 4 mm end mill that deflects 16 times more than the tool the job needs, and a drawing with ±0.01 mm on every dimension costs eight times the same part at ±0.5 mm. Eight design rules, each with a number, and a machined sensor hub worked from $1,370 a part to $430.25 min