reliability
A torch that lights every time you need it for years is reliable; one that works in the shop but dies in your bag a month later is not, no matter how clever it looked. Reliability is the probability that a circuit keeps doing its job, correctly, for as long as it should, under the real conditions it will face — heat, cold, vibration, ageing, and abuse. It is the difference between a design that merely works once on the bench and one that keeps working in thousands of hands across years.
Failures over a product's life famously trace a bathtub curve. Early on, the failure rate is high as the weak, mis-made units die quickly — infant mortality, which makers weed out with burn-in, running units hot for a while before shipping. Then comes a long, flat valley of low, random failures, the useful life. Finally the rate climbs again as parts wear out: electrolytic capacitors dry up, solder joints fatigue, contacts corrode. Engineers sometimes quote an MTBF (mean time between failures), but read it honestly — a million-hour MTBF does not mean each unit lasts a million hours; it is a statistical rate during the flat valley, not a promised lifetime.
You earn reliability mainly by reducing stress, because heat and electrical stress drive most failures: derate parts, sink heat away, keep electrolytics cool, protect inputs against surges and ESD, and avoid fragile clever tricks in favour of simple robust ones. Honesty is central here. Reliability cannot be tested in at the end; it must be designed in from the start, and it is statistical — you cannot promise a single unit never fails, only shift the odds. A good designer also designs for failure: choosing how the circuit fails safe, and respecting that the parts that wear out (heat, capacitors, connectors, anything mechanical) deserve the most margin.
Two boards meet the spec on day one. Board A puts its electrolytic capacitor next to a hot regulator; Board B puts it across the board and derates its voltage. A year later, A's cap has dried out and the board fails, while B runs on — same schematic, very different reliability, decided by stress and heat.
Same schematic, different lifetimes — reliability is designed in through stress and heat, not added at the end.
MTBF is a rate during the flat part of the bathtub curve, not a guaranteed lifetime. A million-hour MTBF does not mean a unit lasts a million hours — it means a low random failure rate while it is in its useful life.