The mean time between failures (MTBF) is a quick way to gauge the reliability of an individual part. If, for example, one MOSFET has a much higher MTBF than another, then the one with the higher rating is likely to last longer in the field without a problem.
But what if you want to compare one bill-of-materials (BOM) to another? Can you simply add together the MTBFs of each component to determine overall reliability? Unfortunately, no. This is because the MTBF, which represents hours-per-failure, doesn’t work in additive functions. However, the failure rate, a different number which represents failures-per-hour, is additive. The good news is that converting the MTBF to a failure rate is a simple matter of fractions. Here’s the formula:

Once you’ve converted the MTBF for each device to a failure rate, you can start evaluating the BOM for its overall reliability.
We looked at the BOMs for various LED drivers. For our example, we used low-power A19 designs, with a 10W load, and assumed similar components when applicable across five topologies. Table 1 lists each component and its associated failures in time, or FIT rating. The optocoupler has the highest failure rate. (Note: the FIT ratings don’t include acceleration factors).

Next, we applied these values to the LED driver BOMs of five common topologies. Table 2 gives the component counts and the final FIT values.

Table 2 shows that more components results in a higher failure rate, and that adding an electrolytic capacitor decreases reliability.
Table 2 does not reflect the impact of wear-out from solder joints. If we assumed two solder joints per component, the FIT rates would rise, but since the solder joints are associated with the number of components, the relative ordering of Table 2 would not change.
Table 2 also doesn’t take into account how factors like temperature, voltage, current and other environmental stresses will affect overall reliability. High temperatures can be particularly damaging to LED components, and can be a real problem in replacement bulbs. This is especially true for residential down lights, where the fixtures once used by incandescent bulbs don’t provide adequate air flow for the replacement LED bulbs.
Figure 3 gives the layout of the topology with the lowest FIT rating – a non-isolated buck topology, with no electrolytic capacitors, using our FL7701 driver controller IC.

Figure 1. FL7701 in a non-isolated buck topology

The FL7701 has integrated power factor control (PFC), and can be powered from the rectified off-line input using a small ceramic capacitor for hold-up. This eliminates the need for electrolytic capacitors at the input and for the IC’s self-bias rail.
Electrolytic capacitors on the output can be eliminated where the requirements for current ripple through the LEDs are fairly loose. The presence of ripple current can decrease the lifetime of an LED, but not having electrolytic capacitors may mean a better reliability figure for the fixture as a whole. However, as shown in Table 2, with electrolytic capacitors included, the FL7701-based design still has a relatively lower FIT rate.
Conclusion
The simple calculations we used aren’t meant as final predictors for LED driver reliability. It is important to add acceleration factors, and to perform life testing, which show how the drivers will respond in actual environments. Nevertheless, our calculations can be useful in understanding the tradeoffs for each topology. If the ambient temperature around the LED driver is managed properly, the non-isolated buck topology uses the fewest components and, as a result, offers the best FIT ratings.

This post is also available in: Chinese (Simplified)