Consider the result for the SDM reliability survey for the DA 16-50/F2.8 lens as of 18th Feb 2012:

Let's examine the 2009 figures:

Total | OK | Failed | Failed % |

75 | 41 | 34 | 45.3% |

What is the failure rate?

Let's simplify. Consider only lenses purchased in 2009, looking at them up to 2011. Let's say these 2009 lenses were all bought at the beginning of that year.

Say we had a 20%/yr constant failure rate. Here is the result of 100 lenses purchased in 2009.

Year | Start of Year | OK at End of Year | Failed | Failed % of Original Amount |

2009 | 100 | 80 | 20 | 20% |

2010 | 80 | 64 | 16 | 36% |

2011 | 64 | 51 | 13 | 49% |

But lenses probably won't have a constant failure rate per year. They're probably like other mechanical devices: the longer you use it, the more likely it is to wear out. So let's consider years of use. Say it fails at a rate of 18.3%/years-of-use. So now the failure numbers look like:

Year | Start of Year | OK at End of Year | Failed | Failed % |

2009 | 100 | 82 | 18 | 18% |

2010 | 82 | 63 | 37 | 37% |

2011 | 63 | 51 | 49 | 49% |

There are a number of other complications:

1. In reality, lenses were purchased throughout the year of 2009 so the avg. age of these lenses at the end of 2011 would be 2.5yrs.

2. Older purchasers are probably under-represented as 2009 purchasers are more likely then 2011 users to, by now, either have switched brands, or not use DSLRs at all, so they are less likely to come to this forum and participate in the survey

3. In this survey, more lenses are being purchased each year.

4. If we knew the year of failure of these 2009 purchases (2009, 20010, 2011), it would help to more accurately determine the failure rate and problem years of manufacture.

Here's what I've done. I've considered years-of-use. Since I don't know the year of failure, I'll assume they all failed today. I've started from 18th Feb 2012. This is the 49th day of 2012. The amount of years is 49/365/2 (divide here by 2 since some lenses were bought on 1st day of 2012 while others may have been bought on the 49th day, so the avg. is 24.5 days) = 0.07yrs.

Adding in 2011, we add 0.5 years (again since lenses could have been bought throughout that year) = 0.07 + 0.5 = 0.57yrs.

For 2010 and for earlier years we add the full year, so 2010 is 1.57yrs.

Here are the figures:

Year | Elapsed Yrs | OK | Fails | Total | Fail % | Fails-Yrs | Total-Yrs | Total Fail Rate % |

2012 | 0.07 | 14 | 4 | 18 | 22.2% | 0.3 | 1.2 | 22.2% |

2011 | 0.57 | 73 | 18 | 91 | 19.8% | 10.5 | 52.8 | 19.8% |

2010 | 1.57 | 55 | 23 | 78 | 29.5% | 46.5 | 175.1 | 26.6% |

2009 | 2.57 | 41 | 34 | 75 | 45.3% | 133.8 | 367.6 | 36.4% |

2008 | 3.57 | 30 | 29 | 59 | 49.2% | 237.2 | 578 | 41.0% |

2007 | 4.57 | 20 | 11 | 31 | 35.5% | 287.5 | 719.6 | 39.9% |

2006 | 5.57 | 2 | 0 | 2 | 0% | 287.5 | 730.8 | 39.3% |

2005 | 6.57 | 1 | 1 | 2 | 50% | 294.1 | 743.9 | 39.5% |

Here are 3 graphs:

The 60-250 curve is exponential as you'd expect for straight wear and tear failure. The failure rate is relatively low: only a total of 9% of lenses purchased within the last 2.5yrs have had SDM failures.

The other two lenses show high early failure rates and the rising part of the curve is not indicating a constant years-of-use failure figure.

Searching now for problem years of manufacture (YOM), let's look at the failure rate for years where the yearly total sample size is reasonably large (at least a total of 18 OKs + Fails in a year)

YOM | Lens | Fail % |

2007 | 16-50 | 35% |

2008 | | 49% |

2009 | | 45% |

2010 | | 29% |

2011 | | 20% |

2012 | | 22% |

| | |

2007 | 50-135 | 34% |

2008 | | 42% |

2009 | | 30% |

2010 | | 21% |

2011 | | 7% |

2012 | | 24% |

| | |

2009 | 60-250 | 13% |

2010 | | 8% |

2011 | | 2% |

Without knowing the year of failure, it is not possible to determine problem years of manufacture.

Dan.

*
Last edited by dosdan; 02-18-2012 at 01:16 PM.
*