MTBF, FIT and Failure Rate Converter

Three ways of saying the same thing, and everyone's datasheet picks a different one. Enter whichever you have — MTBF, FIT, or failure rate λ — and get the other two, plus the reliability R(t) of surviving a mission of your choosing. Runs entirely in your browser; nothing leaves your device.

MTBF (hours)

Mean time between failures, in hours.

Mission time t (hours)

For R(t). 8,760 = one year of continuous operation.

MTBF

—

FIT (failures per 10⁹ h)

—

λ (per hour)

—

R(t) — probability of surviving the mission

—

F(t) — probability of failing within it

—

The conversions

λ    = 1 / MTBF          (per hour)
FIT  = λ × 10⁹           (failures per 10⁹ hours)
MTBF = 1 / λ = 10⁹ / FIT (hours)

R(t) = e^(−λt)           F(t) = 1 − R(t)

A FIT is one failure per billion component-hours. It exists because λ for a decent electronic part is a number like 0.000000023, and nobody wants to read that on a datasheet — 23 FIT is the same fact, legibly.

Where MTBF = 1/λ is true, and where it is a lie

Every conversion on this page assumes a constant failure rate — the exponential distribution, the flat middle of the bathtub curve. That assumption is doing a great deal of work, and it is worth knowing when it fails:

Infant mortality. Early life has a falling failure rate. Burn-in exists precisely to get past it, and a constant-rate model over that period is optimistic.
Wear-out. Bearings, electrolytic capacitors, relays and anything mechanical have a rising failure rate at end of life. Here MTBF is not just imprecise — it is the wrong model, and you want Weibull.
MTBF is not a lifetime. An MTBF of 100,000 hours does not mean a part lasts 11 years. It means that in the constant-rate region, the rate is 1 per 100,000 hours. R(t) below makes this concrete: at t = MTBF, R = e⁻¹ ≈ 0.368. Roughly 63% have already failed by the MTBF, not half, and nothing like none.

The one that catches people Surviving to your own MTBF is a 37% proposition, not a 50/50. If a specification quotes MTBF and means "how long it lasts", the specification is wrong and worth querying before you build on it.

What to do with λ once you have it

λ is the input the rest of the analysis wants. Feed it to the λ to PFD calculator for a SIL band under IEC 61508, or use it as a basic-event rate in a fault tree — our failure rate reference lists typical values by component class if you are still sourcing numbers.

If your parts are redundant, resist multiplying probabilities and calling it done: common-cause failure usually dominates the result. The beta-factor model shows how much, and it is normally more than people expect.

Build the fault tree free → More tools