Light bulb reliability

[peteuk]

Dear all,
I have 6 Megaman bulbs in my kitchen. (CFL - Reflector - R80 - code 616897)

http://www.megamanuk.com/products/cfl/8/reflector/32/r80/93/616897

They were bought in April 2009 from B&Q and a 4th bulb has now failed.
They are meant to have a 15,000 hour life so I think 4 failures in this time is very poor.
B&Q replaced 2 of the bulbs. I replaced a 3rd (B&Q had stoped selling them for a while) and now the 4th needs replacing. B&Q have started selling the bulbs again but refuse to replace the 4th and have referred me to Megaman.
I have e-mailed Megaman asking them to replace the 4th but I am not hopeful they will do so - let's hope I get a pleasant surprise.

Assuming Megaman don't send a new bulb I will have to buy another. Can anyone recommend a more reliable energy saving bulb?

Thank you

 

[ban-all-sheds]

http://www.google.co.uk/search?sour...GB&ie=UTF-8&rlz=1T4GGHP_en__426GB426&q=genura

 

[plugwash]

how many hours per day are the lights used for?

 

[XtraRaptor]

Lamp life is an average, which is based on the number of starts and the length of operation, so if you turn lights on and off frequently they will not last as long (actual operating hours) as lamps that run continuously.

 

[peteuk]

Dear all,
Thanks for the quick replies.

Ban-all-sheds: I hadn't heard of the Genura lamps but I notice they are higher wattage and nearly 3 times the price of a Megaman. Do you have experience of the reliablity of these? A quick Google search of Genura found a complaint like mine - LOL

Plugwash / XtraRaptor: Obviously it's different on different days but if we say 5 hours (I think an overestimate), then this would be 1,825 hours a year. A 15,000 hour life would mean they should last 8+years so let's say 3 years lost for switching on/off I think the bulbs should last ~ 5 years
Even accounting for the fact that 15,000 hours is an average, 4 failures out of 6 within 2.5 years seem excessive to me (but then I am not a lightbulb expert else I would not be asking for guidance)

Thanks again,

Pete

PS Have just read the Megaman should cope with 600,000 switching cycles as well having a 15,000 hour life

 

[ban-all-sheds]

1) A 15,000 hour life does not mean it's the average - it means that by 15,000 hours half will have failed. On the face of it that might look the same, but it would still be a 15,000 hour life if the remaining 50% of the lamps failed at between 15,000 and 15,001 hours.

2) My experience of Genuras is that they are long-lived. I've never measured the life, but I had 4 in the kitchen and 2 in the landing & hallway for well over 20 years, and I was never conscious of having to replace them often.

3) When new they don't really have a slow-start problem, they get as near-as-makes no difference to maximum output within a few seconds, but when they get old they are a bit slow to get going. Like people really. Doesn't bother me - TBH I've never understood why some people get so bent out of shape when their lights don't do 100% instantly. I'm sure it's just an irrational rejection of anything different, and that if the same people had always had low-energy lights that took a few seconds to warm up, and you presented them with halogens they'd be complaining that these new-fangled lights were horrible because their eyes didn't have time to adjust. They put out enough light to see by the instant you turn them on, but if you are one of the people who cannot bear to wait for his full complement of lumens to arrive you might not like them when they've got a few years under their belt.

 

[echoes]

1) A 15,000 hour life does not mean it's the average - it means that by 15,000 hours half will have failed.

I wish to be a pedant: assuming a normal distribution of lamp failures, that's exactly what the average (mean) is. An 'average' lamp will last 15,000 hours. Half the lamps will be below average, and the other half above average.

 

[JohnD]

that's NOT how you calculate a Mean.

A Mean would be to add up the total hours of all the lamps, and divide them by the number of lamps.

So if you had 1,000 lamps, of which 499 failed at 1 hour, 1 failed 15,000 hours, and 500 failed at 15,001 hours, you would still meet the 50% failure at 15,000 hours measure.

But the arithmetic mean of lamp-hours would be

(499x1)+15,000+(500x15,001)/1,000 = (499+15,000+7500500)/1000 = 7516 hrs (approx)

You are thinking of the Median.

 

[echoes]

No I'm not. For a normal distribution, the mean is at the peak of the curve.

So if you had 1,000 lamps, of which 499 failed at 1 hour, 1 failed 15,000 hours, and 500 failed at 15,001 hours, you would still meet the 50% failure at 15,000 hours measure.

That is nothing like a normal distribution, and where the mean would be a pretty misleading statistic.

 

[JohnD]

But the "50% fail by X hours" is not a calculation of Mean Hours Life

It is an calculation of Median Hours Life.

There is no evidence that lifetime of Megaman lamps fits your Normal Distribution curve, and the Median Life claim does not require them to.

 

[echoes]

But the "50% fail by X hours" is not a calculation of Mean Hours Life

It is an calculation of Median Hours Life.

For a normal distribution, the mean and median are the same. I don't know if lamp failures are normally distributed. Probably not far off though.

 

[echoes]

It seems intuitive that the 'average' lamp life is interpreted as 'I expect most* lamps to survive this long'. When one realises that the figure quoted is the median of a distribution (may be identical to the mean), that interpretion may be far from realistic.

* most being a clear majority, not just 50%

Being cyncical, I'm inclined to wonder whether a typical distribution is such that the mean is less than the median, permitting a longer life to be quoted.

 

[ericmark]

Now I know why I failed "A" level statistics! However it is still true that having a life of 10,000 hours does not mean it will last that long. However the number of starts, the local voltage, and temperature of the room can all effect the life. I fitted 18 bulbs together and after one year three had failed. The replacements were slightly different in colour so since one room had 8 and other 10. I put all old bulbs in room with 10 and all new in room with 8 keeping spare old bulbs as stock. Huge difference in price old were around £6 each and new £1 each.

Since then not a single bulb has failed. Either old or new so old have lasted 3 years new 2 years. Living room and dinning room open plan house so they run at least 5 hours a day. So I consider it's just a case of luck.

My landing lamp is an 18W HF lamp bought second hand in 1992 and this year I have changed first tube. But kitchen with 4' and 5' tubes have not lasted any where near as long. However they are not HF fittings. I would like to change kitchen lamps to HF type.

I did find in industry that if one didn't change a HF lamps tube quickly once failed the ballast tended to fail as well. But one doesn't tend to leave failed lamps without changing in the home.

As far as life is concerned I see little difference between cheap and expensive tubes or bulbs. But I have found with cheap bulbs the fuses seem to be missing and when they fail it can be a pig to find which one has failed. Also if more than 6A MCB is used it can weld the contacts to a BA22d bulb holder. But how to tell difference between a cheap bulb sold at high profit or an expensive bulb is another thing.

So all bulbs I buy are cheap and I avoid Ikea after finding theirs did not have built in fuse and I had to change a fitting. My fault really should not have been on 16A MCB.

 

[JohnW2]

Maybe timne for a little 'Statistics 101'?

BAS, as I'm sure you know, 'average' is just a general term which covers a whole host of summary statistics (those which Statisticians call 'measures of central location') - the two (most common) which have been mentioned, arithmetic means and medians, are just two specific examples of ‘averages’.

Echoes, it is extremely unusual for survival times (or times to failure) to be anything like Normally distributed (exponential distributions, or ‘fancy’ ones like Weibull and other distributions, are much more common), either in engineering or biology. If they were Normally distributed, then, of course, the arithmetic mean and median would be identical.

Thinking of just the arithmetic mean and the median, which is more appropriate depends upon the nature of one's interest. If, for example, cost of replacing individual failed items (over a long period of time) is the interest, then one needs the arithmetic mean, regardless of the distribution. On the other hand, if one has, say, a large bank of LEDs, the entirety of which has to be replaced when 50% of the LEDs have failed, then you would want the median. However, without knowledge about the distribution, medians can be very misleading - since a satisfyingly long median survival time could, for example, be ‘hiding’ the fact that nearly 50% of the items fail extremely early.

In engineering circles, by far the most common index of 'average time to failure' for manufactured products is the Mean Time Between Failures (MTBF) or Mean Time To Failure' (MTTF) - median failure time is very rarely used but, if required, can (with knowledge of the distribution) be derived from the MTBF.

I'm not sure what 'average lifespan' is quoted in relation to consumer products (like lamps) - does anyone know? It wouldn't surprise me if it was the same MTBF/MTTF that electrical engineers are used to using. The marketing people would probably like that, too - since the 'long upper tails' of most failure distributions (i.e. a few of the items tend to last 'for ever'!) is likely to result in mean failure times being appreciably longer than median ones!

Kind Regards, John

 

[JohnW2]

So if you had 1,000 lamps, of which 499 failed at 1 hour, 1 failed 15,000 hours, and 500 failed at 15,001 hours, you would still meet the 50% failure at 15,000 hours measure.

That is nothing like a normal distribution, and where the mean would be a pretty misleading statistic.

It certainly isn't anything like a Normal distribution - but see what I've just posted. For a start, Normally distributed times to failure are very unusual. Secondly, whether or not an arithmetic mean (or any other type of 'average') is useful or misleading depends upon one's interest - if it were in terms of the cost of replacements, then the arithmetic mean would be precisely what one wanted, regardless of distribution. (and, in the sort of example given, a median could be almost as misleading as an arithmetic mean).

Kind Regards, John.