Cooling fan MTBF is measured hot and extrapolated cold. What really kills a fan, what a field swap costs on site, and why no moving parts change the sums.
A cooling fan datasheet usually carries one line buyers read as a promise: MTBF 70,000 hours. Taken literally that is eight years of continuous running. Taken properly it is a short test on a small sample in a hot chamber, converted by an acceleration factor into a figure for an ambient the fan was never tested at, describing a population rather than the unit in your enclosure.
Measuring efficiency in watts is too narrow a frame for that. A fan's power draw is a small charge you pay every hour. Its failure is a large charge you pay once, at a moment you do not choose, and the size of it has almost nothing to do with the price of the fan.
Cooling fan MTBF is a statistical mean time between failures for a population of fans, derived from accelerated life testing at elevated temperature and extrapolated back to a stated reference ambient, not a guaranteed service life for any individual fan. Two consequences follow immediately, and both are routinely missed at the point of specification.
First, the terminology is loose. A fan is not repairable, so the honest metric is mean time to failure. MTBF strictly belongs to repairable systems, and fan hazard rates are not constant, which makes the single-number summary awkward to compute and easy to misread [1]. If failures were exponentially distributed, roughly 63 per cent of a population would have failed by the time the mean is reached. It is not a warranty. It is a centre of gravity.
Second, the reference condition matters more than the value. SEPA quotes its MTBF data at 40 °C ambient unless otherwise stated, and observes that other manufacturers frequently publish considerably higher values taken at room temperature [2]. Two fans with the same headline figure can differ substantially in the same box.
The more useful metric is L10, the point at which 90 per cent of a sample are still running. Sanyo Denki defines expected life at 90 per cent survival, ending when rotational speed falls to 70 per cent of its initial value, and states plainly that the determining factor is ball bearing grease life rather than the electronics [3]. SEPA gives the population ratio L1 : L10 : MTBF at about 0.21 : 1 : 4 [2]. The number you were quoted is therefore roughly four times the number you should plan maintenance around.
| Metric | What it measures | Basis | How to read it |
|---|---|---|---|
| MTBF / MTTF | Population mean time to failure | Accelerated test, extrapolated to a reference ambient | Comparative only. Check the stated temperature first |
| L10 | Hours at which 90 per cent still run | Bearing and lubricant life | The planning number. Roughly a quarter of MTBF [2] |
| L1 | Hours at which 99 per cent still run | Bearing and lubricant life | Use where one failure is expensive. About 0.21 of L10 [2] |
| End-of-life criterion | Speed at 70 per cent of initial | Airflow and noise degradation | A fan can be "alive" and no longer cooling [3] |
Almost none of it is electrical. Fans die from the bearing outwards.
Lubricant degradation is the dominant mechanism, and it is thermally driven. The Booser relationship for grease life carries an acceleration factor of about 1.5 per 10 °C, so extrapolating from 80 °C down to 40 °C means multiplying by roughly 5.1 [1]. Vendors run life tests between 70 °C and 85 °C ambient to get answers in reasonable time [1]. Sanyo Denki's own example puts a fan rated 40,000 hours at 60 °C at approximately 70,000 hours at 40 °C [3]. The leverage cuts both ways: a fan specified at a benign ambient, then installed beside a hot power stage or behind a partly blocked intake, gives that multiplier back fast.
Extrapolation is itself a source of error. The same source notes that applying acceleration factors derived from bearings at 100 °C down to 25 °C "is probably questionable, but that is often how they are used" [1]. The model has wide error bars, and none of them are printed.
Then there is dust. Loading on the impeller unbalances it and loads the bearing; loading on the heatsink raises the local ambient, which accelerates the grease chemistry, shortens bearing life, reduces airflow and raises the local ambient again. Continuous duty removes the thermal recovery intermittent operation gives. None of this reaches the headline number, because the chamber is clean and the sample is new.
A 40 mm fan is cheap. Replacing it costs whatever the site costs.
Once intervention cost rather than component cost sets the number, "efficiency" stops being watts and becomes expected cost over the deployed life: the power integral, plus the probability of failure inside the service life multiplied by the cost of the visit, plus the redundancy you paid for because you did not trust the first term.
A dielectric barrier discharge actuator is roughly 200 micrometres thick and has no bearing, no impeller and no membrane. Solid-state cooling with ionic wind moves air by accelerating ions along a surface, dragging neutral air with them. Published work reports an ionic wind superimposed on a 0.3 m/s bulk flow doubling the average heat transfer coefficient and delivering roughly 20 K of additional convective cooling for under 100 mW [4].
The point is not the milliwatts. It is that the wear-out mechanism has moved. There is no grease to oxidise, so the temperature acceleration that governs bearing life does not apply in the same form, and there is no rotating mass to unbalance when dust arrives. The maintenance plan changes shape: from scheduled replacement of a consumable, to a component evaluated like any other solid-state part on the board. Plasma actuators and the Y-Flow thermal line are in development against exactly that specification pressure.
Solid-state does not mean immortal. It means a different failure mode, and an honest comparison has to say so.
Dielectrics age under continuous electrical stress. The drive is kilovolt-class AC, a constraint on the power supply and on creepage and clearance. Operation in air produces some ozone, dependent on waveform, geometry and duty cycle, and manageable by design; we treat that in ionic wind reliability and ozone. Electrical-to-fluid conversion efficiency for corona-driven ionic wind sits around 1 to 2 per cent [4], poor for moving bulk air, and winning only where the alternative in that space is no airflow at all.
Ionic wind also produces metres per second, not tens of metres per second. If the problem needs high volumetric flow, a fan remains correct, and the comparison is set out in ionic wind versus fans. Fan field data carries decades of history a newer component does not have. Piezoelectric coolers remove the bearing but introduce membrane fatigue, so "no rotating part" and "no wear-out" are not the same claim.
The defensible position is narrow: where a box is hard to reach, sealed, or running continuously in a warm ambient, and where the flow needed is local rather than bulk, the arithmetic that matters is the cost of the visit you never had to make. In fanless designs that term dominates.
No. MTBF is a population statistic derived from accelerated testing, not a service life for one unit. Under an exponential assumption, around 63 per cent of a population would have failed by the mean. Use L10, the hours at which 90 per cent are still running, for maintenance planning [1][2].
It depends on the vendor, which is why the figure is often misleading. SEPA states its MTBF data assumes 40 °C ambient unless otherwise noted, and that other manufacturers frequently publish higher values taken at room temperature [2]. Life testing itself is typically run between 70 °C and 85 °C and extrapolated back [1].
Bearing lubricant degradation, accelerated by temperature, is the dominant mechanism, with dust loading and imbalance contributing. Grease life follows an acceleration factor of roughly 1.5 per 10 °C [1], and manufacturers identify ball bearing grease life as the determining factor in expected life [3].
No. Sanyo Denki sets the end of expected life at the point where rotational speed drops to 70 per cent of its initial value [3]. A fan can still be turning while no longer delivering the airflow the thermal design assumed.
It changes the risk rather than removing it. There is no bearing or grease to wear out, but dielectrics age under continuous electrical stress and the drive is kilovolt-class AC. Ionic wind also produces only a few metres per second, so high volumetric flow applications still need a fan [4].