For DC/DCs sold to CERN a Run In during 8-12h at around 35-40°C was performed including an electrical test. 4 weeks however also looks good, but is rather not realistic for industry.
Temperature cycling should be a good screening test, especially for ceramic capacitor damage and soldering issues (dry solder joint). From experience this should not significantly reduce the lifetime of a well produced device.
A ceramic capacitor damage could take 1-2 months in operation to show a failure
For the design, it should always be avoided to place ceramic capacitors close to the PCB edges, where they are exposed to high mechanical stress when the boards are cut or as well to friction on the surface.
Critical components: If well designed, semiconductors and magnetics should not be a problem. For a good design, electrolytic capacitors should be the only wear out issue.
Radiation degradation to be considered additionally however!
For MOSFETs, BJTs and diodes this is only true if not frequently power cycled
Experience from power converter failures with ~3500 components inside shows that generally only cables and connectors are affected, as well as dust sometimes is an issue. Semiconductors are practically never an issue, if well designed and radiation excluded.
Diagnostics: Measuring the output V_ripple would be an option to draw conclusions about the capacitors ESR/degradation. (difficult implementation)
For el. capacitors Hitachi has been chosen in the past as a good manufacturer. Best ratings are to be used (highest lifetime rating, low ESR, …). The price should not be an evaluation criteria.
Power cycles are also a big problem for el. capacitor degradation.
Many usually fail after repowering after a LS.
It has been assessed in EPC to place a resistor in parallel to the circuit breaker in order to keep the capacitor always supplied.
Input from a manufacturer was to better use a 450V rating for a capacitor which is always regulated at 400V rather than a 500V or 550V rating. Too much margin may be an issue.
Maintenance strategy should be to enable an easy desoldering of the capacitors in order to replace them before the estimated lifetime.
Annotation: In specific applications additionally to be cross-checked with radiation tolerance and dose exposure of the RaToPUS.
The same applies to fans
El. capacitor lifetime equation to be checked for correlation between temperature and ESR in order to find the best trade-off. It would be best to contact the manufacturers to get this input and their recommendation.
ALT testing is very difficult to implement and to interpret the data correctly. It was only performed by EPC on POPS capacitors, regarding charging cycles.
It is not as easy as for instance for ball bearings, especially because high stresses on el. capacitors can cause various mechanisms which are not present during normal operation, e.g. saturation on the MOSFET gate.
SMD resistors of EPC designs are derated, for example down to 0.15 Watts for 1206 (0.25W rated)
To measure diagnostics, an idea would be to measure the discharge curve of the DC/DC after turning it off with a resistor in parallel (RC time constant). For used el. capacitors, the discharge time should be shorter than for new ones. But this has not been done in EPC yet.
Annotation: Difficult to implement for the RaToPUS, because the el. capacitors are enclosed on the 48V bus between the AC/DC and DC/DC
Additional power cycling stress to be taken into account! This could lead to many failing; Spares should be ready
Therefore it may be a strategy to uninstall and test (and inspect if e.g. blown up) some modules manually after some years of operation close to an estimated EoL
Leakage current change is probably more complex to measure.
EPC’s focus is a lot on production issues. This is beforehand to analyse the process, what could go wrong, and afterwards performing e.g. temperature cycles
Load sharing other than ~50:50 does not sound feasible.
Probably no major reliability problems to be expected for ramping up from 50% to 100% of one RaToPUS after a redundancy failure. The remaining RaToPUS would pass from an already operational, hot state, to an even hotter state.
E.g. Pulse is a good commercial manufacturer, but showed failures due to dust in the past.
Fans should be equipped with filters
Design advice to put a film capacitor in parallel to an electrolytic to take care of the pulsed current. (space constraints)
Major Outcomes & Actions
Strong focus on designing for high reliability, in particular on derating, using fans (+filters), avoiding electrolytic capacitors and purchasing highest quality (contact manufacturers)
El. Capacitor lifetime formula to be checked for best trade-off between lifetime and ESR rating (manufacturer to be consulted for feedback)
Same strong focus on high quality during the production process and following screening tests
During operation, power cycles should be avoided, easy desoldering and replacement of electrolytic capacitors should be enabled, fans regularily replaced for EoL, protection against dust/dirt in place and the discharge time of capacitors may be measured for degradation diagnostics by uninstalling and testing operational modules after some years