Most engineers don't grasp the assumptions and subtleties behind a calculation of mean time between failures for multi-component electronic assemblies.
Other bloggers have written here arguing that MTBF is not a very interesting measure for LED luminaires, and that, in fact, we should talk about MTTF (mean time to failure) instead.
Three techniques are typically used to establish MTBF for an electronic product:
- Parts-count method
- Parts-stress method
- Demonstrated MTBF (most relevant to a mature product with an extensive documented field history)
The parts count method is the dominant choice of commercial product manufacturers: power supplies, modular LED drivers, LED luminaire systems, electronic control modules, etc. When all is said and done, it probably is as good a guideline as the other two methods and a much more practical tool for establishing an MTBF fairly quickly.
The stress method, though it has a reasonable degree of legitimacy, is rarely used for non-military products, because of the tedium involved in verifying the real stress on components during a multitude of possible operating conditions. It is a method that seeks to answer the question "How conservatively are the components rated?"
The demonstrated MTBF method is viewed as the ideal way in that it establishes the "real" field reliability. However, for such a number to be valid there must be:
- An absolute consistency in the way the product is being used
- A substantial and documented history in the field
- An unchanging design and manufacturing process
- A legitimate way of obtaining feedback on field performance
These criteria are rarely all encountered in electronic products. As a result, demonstrated MTBF figures can be ambiguous and prone to marketing manipulation. That is, they must be taken with a grain of salt.
It is common in the computer and telecom industries (not so common in LED products) for a customer to ask a power supply vendor, "What is your MTBF?" If one is familiar with deriving an MTBF, he or she has learned which items most influence the number: electrolytic capacitors, optocouplers, ICs, and, to a far lesser degree, transistors and diodes. The more complex the IC, the worse the number. Temperature is a dominant determinant. If one has no ICs, optocouplers, or electrolytics, the MTBF can be very high -- well above 100,000 hours, even at elevated temperature.
Charts or software programs based on MIL-217 or Telcordia-332 make no provision for white LEDs other than just treating them as any other ordinary diode -- no surprise, since there is very little long-term field data on products using phosphor-based LEDs. Contrast that with MTBF or life expectancy numbers for a dual-ball-bearing brushless DC fan, where tens of millions have been operating 24/7 for years in networking equipment. You can be sure Apple and Cisco have information as to how many returns or complaints they have had in last 10 years due to power supply fan failures.
It is common for most firms, when establishing their datasheet MTBF number, to use a 25°C ambient in the calculation. This is OK for chokes and transformers, but the MTBF for an electrolytic capacitor, microcontroller, or optocoupler can drop radically at high temperature. Also, for things like MOSFETs or LEDs, one must crank into the equations, not the ambient temperature, but the junction temperature.
You begin to see that, to calculate an MTBF, one sometimes must actually know how things are designed and made. These nuances simply aren't incorporated into MTBF software programs.
Finally, we should note that these MTBF methods have a variety of "multiplier factors" and categories for each type of component. It can be totally arbitrary which of these you employ; there is no single right or wrong way. Without knowing all the assumptions used in an MTBF determination, a customer has absolutely no way of knowing what the MTBF figure means.
Not one engineer in 50 has ever done an MTBF calculation or has a clue about what is in the previous several paragraphs. Consequently, vendors, if asked for an MTBF figure, typically (and permissibly) derive the number for the most optimistic condition possible, potentially inflating it by a factor of 3, 5, or even 10 times. A vendor trying to be completely helpful by providing a worst-case MTBF -- based on the toughest operating environment -- could be shooting itself in the foot from a sales standpoint.