Pi-FAN for RPi4 with 4 (instead of 3) cables?

remove "electronic" and it makes sense.

No. Again, MTBF characterizes failures during normal service life.

Take a fan which usually fails after 5 years of operation due to wearout, which happens on all of those fans after about the same amount of time - those 5 years are *not* parts of the MTBF. MTBF characterizes the statistical mean time between failures *during* those 5 years due to *other* reasons.

So you can have a MTBF of 50 years, and a lifetime of 5 years. If you take

The important thing is to keep MTBF and lifetime separate - both will lead to failures, but are separate mechanisms.

cu Michael

It does not really matter if you repair or replace - what is important is that a technician needs to drive to the installed equipment to get it going again, and MTBF helps calculating how often that will be necessary during normal operation.

That assumes that properly repaired equipment is as reliable as comparable, non-failed equipment of the same age, otherwise, the numbers will be skewed.

cu Michael

Thanks Michael, very instructive. That clears up my misunderstanding. So MTBF is a random type, rather than a wear type thing. You just have to assume bits are replaced before they get old.

I read MTBF for people is about 700 years.

Well yes and no,

Given f = failure rate (e.g probability of nuclear decay)

half life = ln(2)/f. MTBF = 1/f.

Therefore:

half life = ln(2) * MTBF = ~0.69 * MTBF

essentially they are the same, apart from the 0.69 conversion factor.

For a stingy old man like me, yes. I wait till things break. So MTBF is not particularly useful. I want long average life, not reliability until they get old.

MTBF is the type of metric used by someone who will replace bits before they get old.

If I were managing a data centre and I knew human intervention was very costly (I suspect it is), I might have a policy of replacing parts before the failure rate became high due to old age.

I know the companies I worked for hated investigating failure, or fixing stuff. They much preferred replacement. Essentially components were cheap, human intervention was not.

Obviously if the equipment is expensive, or you have cheap slave labour, the equation may change.

I read MTBF for people is about 700 years.

Well yes and no,

Given f = failure rate (e.g probability of nuclear decay)

half life = ln(2)/f. MTBF = 1/f.

Therefore:

half life = ln(2) * MTBF = ~0.69 * MTBF

essentially they are the same, apart from the 0.69 conversion factor.

For a stingy old man like me, yes. I wait till things break. So MTBF is not particularly useful. I want long average life, not reliability until they get old.

MTBF is the type of metric used by someone who will replace bits before they get old.

If I were managing a data centre and I knew human intervention was very costly (I suspect it is), I might have a policy of replacing parts before the failure rate became high due to old age.

I know the companies I worked for hated investigating failure, or fixing stuff. They much preferred replacement. Essentially components were cheap, human intervention was not.

Obviously if the equipment is expensive, or you have cheap slave labour, the equation may change.