Re: [time-nuts] MTBF (was Rubidium standard)
As I have learned in school from a department head, mean time between
failures (MTBF) means anything only if you are being mean. If you are
not being mean, it means nothing.
MS
At 01:14 PM 11/18/2009, Michael Sokolov wrote...
As I have learned in school from a department head, mean time between
failures (MTBF) means anything only if you are being mean. If you are
not being mean, it means nothing.
There is at least one practical use for MTBF, at least the real-world
statistical form.
If one has a reasonably large population of devices to maintain, the
number of spare devices needed to stay operational is a function of the
MTBF.
If you've got 10,000 cell sites, each with an Rb timebase, MTBF figures
can provide a pretty accurate estimate of how often one will fail and
need to be replaced, at least during the normal lifetime (between the
infant mortality and wearout stages). Coupled with MTTR statistics, one
can figure out how many spares should be stocked.
Manufacturers invest a lot of effort into using such statistical
measures to determine how many spare parts should be kept at various
points in the supply chain. Unnecessary parts on shelves = wasted
capital. Manufacturers also use MTBF statistics to price service
contracts on equipment.
This only works when one is dealing with a large population of devices,
so MTBF is meaningless at the individual unit level.
Sorry Mike , unless, as someone else said, the figures are derived from
field failures over at least a good porton of the expected like the MTBF
tells you absolutely nothing!! The statistics used on the usual 1000hour
test will only tell you the probability of failure in the first 1000hours of
use!! It cannot tell you anything mathematically about the extrapolated
life....this has become another urban myth. If it works it is more by luck
that by mathematical probability.
Alan G3NYK
----- Original Message -----
From: "Mike S" mikes@flatsurface.com
To: "Discussion of precise time and frequency measurement"
time-nuts@febo.com
Sent: Wednesday, November 18, 2009 7:06 PM
Subject: Re: [time-nuts] MTBF (was Rubidium standard)
At 01:14 PM 11/18/2009, Michael Sokolov wrote...
As I have learned in school from a department head, mean time between
failures (MTBF) means anything only if you are being mean. If you are
not being mean, it means nothing.
There is at least one practical use for MTBF, at least the real-world
statistical form.
If one has a reasonably large population of devices to maintain, the
number of spare devices needed to stay operational is a function of the
MTBF.
If you've got 10,000 cell sites, each with an Rb timebase, MTBF figures
can provide a pretty accurate estimate of how often one will fail and
need to be replaced, at least during the normal lifetime (between the
infant mortality and wearout stages). Coupled with MTTR statistics, one
can figure out how many spares should be stocked.
Manufacturers invest a lot of effort into using such statistical
measures to determine how many spare parts should be kept at various
points in the supply chain. Unnecessary parts on shelves = wasted
capital. Manufacturers also use MTBF statistics to price service
contracts on equipment.
This only works when one is dealing with a large population of devices,
so MTBF is meaningless at the individual unit level.
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
and follow the instructions there.
At 04:21 PM 11/18/2009, Alan Melia wrote...
Sorry Mike , unless, as someone else said, the figures are derived
from
field failures over at least a good porton of the expected like the
MTBF
tells you absolutely nothing!!
That is exactly what I meant by "the real-world statistical form" -
data gathered from in-service operation, not lab tests (or worse, a
purely mathematical combination of individual component MTBFs, like one
resistor contributes X, an IC of a certain type contributes Y, etc.).
But, things have to start from somewhere (e.g. a need to deploy spares
for a new product) so manufacturers will use shortcuts (lab testing,
etc.) for initial numbers, then refine based on experience. That
doesn't mean the shortcut methods are useless, only that they're less
accurate. Something is better than "nothing," and a simple device can
be expected to have a higher MTBF than a complex one (assuming similar
technologies, manufacturing processes, etc.). So I don't agree with
your "absolutely nothing" statement. Having an HP5071 doesn't make a
Timex worthless.