Hi,

In a few particular sockets in my home, our light bulbs die unusually quickly (they last about 2 months) compared to other sockets. What could be the cause?

 

Just a thought, they could be getting hot. That or the bulb itself is seeing a large draft, creating a large temperature differential.

Old sockets can have high resistance area, which means they also generate heat. This is an unsafe condition. Do you see any discoloration or signs something like this could be happening?

 

Hi,

In a few particular sockets in my home, our light bulbs die unusually quickly (they last about 2 months) compared to other sockets. What could be the cause?

Are the lights in the basement??? Or on a level with another floor above??

 

Just a thought, they could be getting hot. That or the bulb itself is seeing a large draft, creating a large temperature differential.

Yes, they burn out quite quickly. I thought bulbs were designed to operate at high temperatures though?

Old sockets can have high resistance area, which means they also generate heat. This is an unsafe condition. Do you see any discoloration or signs something like this could be happening?

Hmm... my house is rather old (about 15 years, I think). I don't see any discolourations on the bulbs though. Is there another way to (safely) test if the socket does have a high resistance area? Also, why is it unsafe?

Are the lights in the basement??? Or on a level with another floor above??

They're at the ground floor; my home has no stairs up or down.


Thanks for your responses guys; we're getting quite frustrated at having to replace our bulbs regularly!

 

You don't say how the light bulbs fail, so we're guessing a bit. Assuming it isn't a catastrophic failure like the envelope breaking, the usual failure mode is the filament breaking. Tungsten filaments experience grain growth over time while they are at incandescent temperatures and when a grain boundary essentially extends over the width of the filament, they become mechanically weak: thermal or mechanical stresses can easily break them. I believe Langmuir and others at GE studied this stuff in detail in the first half of the 1900's.

Knowing this tidbit, then, some strategies to lengthen the lives of bulbs are

1. Soft start them (ramp the current slowly to reduce the thermal shock).
2. Minimize mechanical shock on the filaments.
3. Run them at reduced power to reduce the filament's temperature.

For example, a mad teenager stomping around in her bedroom upstairs might provide too much mechanical vibrations to a lamp fixture in the ceiling of the floor below, leading to an early demise of the light bulbs. The last strategy does have the annoying feature of reducing light output though...

The failure time distribution of light bulbs has classically been reported to follow the exponential distribution (see, e.g., Nelson's book on life data analysis (pub. by Wiley)), but the experiments I've seen that measure this just turn the light bulb on and leave it on. This isn't realistic in terms of typical household use, especially with a teenager who always is turning lights on with an energy-conscious father following her around the house turning lights off. I'd be curious to know if anyone has some hard distribution data concerning the latter case. At any rate, the mean and standard deviation of the exponential failure times are equal, meaning you'll see quite a bit of variation in your small samples. In other words, you might want to write down the bulb installation times and failure times and collect data for a while; you might be looking at normal random variations and interpreting them as non-random.

Here's a sorted list of 100 pseudorandom exponential deviates with a mean of 1000:

14.6   135.   193.   298.   447.   670.   916.   1100.  1220.  1830.
29.2   138.   216.   303.   470.   670.   953.   1110.  1290.  1840.
38.4   147.   238.   317.   471.   673.   1000.  1130.  1300.  2100.
61.4   154.   251.   321.   493.   705.   1010.  1140.  1370.  2260.
64.9   160.   252.   328.   553.   745.   1010.  1140.  1420.  2490.
67.3   161.   275.   342.   564.   808.   1020.  1150.  1450.  2610.
74.1   161.   283.   365.   572.   815.   1030.  1160.  1490.  2610.
83.2   171.   283.   385.   591.   815.   1050.  1180.  1580.  2630.
132.   174.   287.   403.   592.   861.   1050.  1180.  1770.  3100.
133.   182.   294.   419.   656.   916.   1090.  1210.  1830.  3810.

Folks who aren't trained in statistics and familiar with the behavior of various distributions will often draw the wrong conclusions from such data, especially with small samples. Look at some of the statistics from this sample:

Mean         831.
Std dev      744.
Sd/mean      0.896
Count        100
Minimum      14.6
Maximum      3810.
Range        3800.
Median       663.
Geom. mean   518.
Skewness     1.47
Kurtosis     2.40

Estimated Percentiles (%, value):
    0        14.6
    5        67.2
    10       135.
    15       161.
    20       191.
    25       269.
    30       297.
    35       337.
    40       436.
    45       559.
    50       663.
    55       773.
    60       916.
    65       1010.
    70       1090.
    75       1140.
    80       1210.
    85       1420.
    90       1830.
    95       2500.
    100      3810.

 

Being from the USA I don't know Austrailian wiring. Do you use a center tapped feed at the the service entrance 220v-0-220v, similar to that used here 120-0-120, with a common neutral?

Ken