Certainly, if MTTF = 10000 hrs =~ 400+ days, then half the units can be expected to fail at less than that period - and some proportion of that 50% within a one-year warranty (if such exists). I don't have the formulae to hand to calculate the exact numbers.
I suspect the hapless coder producing the web doc was given some hallucinated specs to work form and '10000' looks good. best regards, 웃 Victor Churchill, Netley Abbey, Southampton On Wed, 10 Jun 2026 at 19:34, Peter Merchant <[email protected]> wrote: > 4 days is roughly 100 hours, so divide your 10000 by 100 = 100. > 100 * 4 days is 400 days which is close to your figure. Did they lose a > couple of 00's in printing the figure which might give you 40000 days. What > sort of figure would you expect? > > I remember at uni having to calculate the MTBF of a system by calculating > on the MTBF of each component. I couldn't do it now. > > Peter > > On 10/06/2026 14:42, Ralph Corderoy wrote: > > Hi, > > > > I was idly looking at this power convertor, and saw the ‘MTBF 10000h’. > > https://www.olimex.com/Products/Power-Supply/Modules/ACDC-85-255-5V0.7A/ > > > > That's quite a short time for an always-on device. > > > > $ units -1 10000hour time > > 1 year + 51 day + 10 hr + 11 min + 14.025322 sec > > > > Just long enough for a one-year warranty. :-) > > Not that MTBF works like that. > > > > Skimminghttps://en.wikipedia.org/wiki/Mean_time_between_failures, > > it is perhaps more of a MTTF value since I don't suppose they get > > repaired. And the odds of it lasting a year are about 42%. > > > > R(t) = e^(−λt) = e^(−t / MTBF) > > > > $ bc -l > > e(-(24*365)/10000) > > .41644536602038009593 > > > > Where have I gone wrong? > > > -- > Next meeting: Online, Jitsi, Tuesday, 2026-07-07 20:00 > Check to whom you are replying > Meetings, mailing list, IRC, ... https://dorset.lug.org.uk > New thread, don't hijack: mailto:[email protected] > -- Next meeting: Online, Jitsi, Tuesday, 2026-07-07 20:00 Check to whom you are replying Meetings, mailing list, IRC, ... https://dorset.lug.org.uk New thread, don't hijack: mailto:[email protected]

