Genesis: On your question on confidence limits.
MTBF figures are published for 60% confidence with an application temperature generally of 55°C. Where the test is accelerated, then the manufacturer provides traceability back to a non-accelerated number. We do not accept unsupported accelerated figures.
Manufacturers publish the numbers of components used in their the test, so one can calculate where the 90% confidence limit quite easily - see the foot of
Quality - Reliability Programs for example. The numbers of components in the tests are high, but the failure rate is so low that companies generally give the 60% MTBF figure based on the duration of their test, and with no failures. That is, if 1000 components are tested for 1000 hours, then the manufacture will claim a 60% MTBF of 1 million hours (put roughly and in simple terms). Of course, it could be higher than that. Getting 80% confidence generally means twice as many hours, 90% twice as many again, etc.
The documents indexed by
Quality - Metrics: Reliability gives the background to how a typical component is tested. The detailed component reports provide the data over time, that is the test ran every 8 weeks, so one then has a larger population still to work with. Using that data, the 90% limits are close to the 60% limits.
The electronics MTBCF is 2.9 billion hours, with 60% confidence. It is still well over 1 billion hours with 90% confidence. With the twin scrubber unit, the number is squared.
On your point regarding the front of the bathtub curve, the units are soaked in, and the manufacturer has a 10 year record of 0 ppm failures except for 1 year, with 1ppm, which was traced and rectified. This is on millions of populated circuit boards a year, used in temperatures from -40C to +120C (automotive - the latter is how hot cars get left in a dessert).
In addition to the manufacturer's testing, all components are tested by us for suitability. The next few months should see some of the reports on that published.
On the mechanics, we are doing a lot of testing at the moment and will publish those figures later. The biggest challenge is the hoses.
Alex