Searching \ for '[EE]:: Dead LEDs' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/displays.htm?key=leds
Search entire site for: ': Dead LEDs'.

Exact match. Not showing close matches.
PICList Thread
'[EE]:: Dead LEDs'
2008\07\16@093630 by Apptech

face
flavicon
face
Getting White LEDs with long lifetimes is easy. Getting them
at a low cost is another matter.

I've been testing various white LEDs for longevity and been
getting some rather bad results from Chinese sourced
products. Reports from others testing LEDs for the same
purpose are similar. This is not a total surprise as the
general received wisdom is that this is the case. It is
however a somewhat surprise as it is not obvious why the
Chinese products should tend to be so bad. A Chinese
supplier (name will not be stated) even made some changes to
try and meet my spec and the results were no better.

By Chinese I mean companies that are Chinese based - NOT
known internationals who are domiciled elsewhere but may use
Chinese manufacturing (eg Avago, Cree, Nichia, ...) - such
companies are demonstrably more liable to get it right.

The universal claim is that white LEDs last 100,000 hours. I
can assure you that many don't come anywhere close (by 2+
orders of magnitude in some cases).

My questions are:

- What mechanism makes Chinese LEDs so bad?

- Why is this allowed to be? ie why don't they do whatever
it takes to fix it.

If anyone feels that my statements are a generalisation and
that some Chinese white LEDs do have the sort of lifetimes
one would expect then *PLEASE* do tell me the brands!. I'd
be extremely happy to be wrong and to be able to source LEDs
at non-market-leader prices.

For white phosphor LEDs (blue radiator and yellow phosphor
re-radiator) the degradation mechanism seems to be actual
LED die output level. A possible mechanism in some cases MAY
be die over-temperature due to excessive over-rating of die
current capabilities. Phosphor death does not seem to be an
issue in what I have seen. (it is in some other cases). Die
bonding adhesive to the LED structure cup is claimed to be a
problem in some cases but when this was changed in the LEDs
I was getting to a Japanese sourced bonding product of good
parentage it made about zero difference.

Interestingly, and not directly related, some name brand
LEDs give atrocious spectral results at very low currents
while others are about as good across a wide range of
currents.



           Russell



2008\07\16@100426 by Wouter van Ooijen

face picon face
> The universal claim is that white LEDs last 100,000 hours. I
> can assure you that many don't come anywhere close (by 2+
> orders of magnitude in some cases).

Just out of interest: how do you test this? 100.000 minus 2 orders of
magnitude is 41 days, that's testable (with some patience), but how do
you test 410 days? Or are there known relations between lifetimes at
various currents?

--

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu

2008\07\16@100852 by Ariel Rocholl

flavicon
face
Mind you to ellaborate a bit on how you test LEDs for longevity?

How do you accelerate the process to understand 100K hours is wrong or what
variables/instrument do you use to measure performance?

2008/7/16 Apptech <spam_OUTapptechTakeThisOuTspamparadise.net.nz>:

{Quote hidden}

> -

2008\07\16@102547 by Joe Bento

face
flavicon
face
I am most curious, and perhaps someone can answer.  You mention that the
white LED might fail within a 2x magnitude of 100k hours.  100k hours is
about 11.4 years.  How is MTBF estimated or calculated?  I don't think
they light an LED at its recommended current and wait - white LEDs
haven't been around that many years.

I've bought many of those 100 LED collections off ebay, some with a
clear lens, others with the phosphor.  In my case they are primarily for
hobbyist use so gradual failure isn't as big a concern for me.

Joe



Apptech wrote:
> Getting White LEDs with long lifetimes is easy. Getting them
> at a low cost is another matter.

2008\07\16@103551 by Shawn Tan

flavicon
face
On Wednesday 16 July 2008 15:21:58 Joe Bento wrote:
> I am most curious, and perhaps someone can answer.  You mention that the
> white LED might fail within a 2x magnitude of 100k hours.  100k hours is
> about 11.4 years.  How is MTBF estimated or calculated?  I don't think
> they light an LED at its recommended current and wait - white LEDs
> haven't been around that many years.

Voodoo?? <grins/>

--
with metta,
Shawn Tan

Aeste Works (M) Sdn Bhd - Engineering Elegance
http://www.aeste.net

2008\07\16@105201 by Rich

picon face
I have spent some time in Chinese semiconductor manufacturing.  I found that
some had very low yields. I found that keeping contamination out of the
clean rooms was in some cases a challenge.  Running the clean room below
"clean" can result in below speck product.


{Original Message removed}

2008\07\16@110621 by Spehro Pefhany

picon face
Quoting Apptech <.....apptechKILLspamspam@spam@paradise.net.nz>:

{Quote hidden}

Probably some technology they are not using. ;-)

> - Why is this allowed to be? ie why don't they do whatever
> it takes to fix it.

Because it costs more to develop, buy or license the technology than it's
worth to them. In which case you may find (as I have in some cases) that
you're better off to buy the name brand products that have been forced  
down in price by the shoddy stuff.

{Quote hidden}

There are more than one type of phosphor used in the blue-LED-plus-yellow
phosphor configuration. Here is one commercial supplier:

http://www.mt-berlin.com/frames_cryst/descriptions/led_phosphors.htm


> Interestingly, and not directly related, some name brand
> LEDs give atrocious spectral results at very low currents
> while others are about as good across a wide range of
> currents.
>
>
>
>             Russell

Out of curiosity, have you tried blue LEDs from the same suppliers?
I'd be interested if they show a similarly short life.

Blue (and white) LEDs have a relatively high voltage drop compared to
other types such as superbright red, so the power dissipation at the
junction can be twice as high at the same If.


>
> -

2008\07\16@110830 by Spehro Pefhany

picon face
Quoting Rich <rgrazia1spamKILLspamrochester.rr.com>:

> I have spent some time in Chinese semiconductor manufacturing.  I found that
> some had very low yields. I found that keeping contamination out of the
> clean rooms was in some cases a challenge.  Running the clean room below
> "clean" can result in below speck product.

It's perhaps worth noting that with LEDs they sometimes have two  
grades-- one for
export to first-world countries, and another for local toy manufacturers and
sale to India, Bangladesh etc. where a 20% failure rate and wide variation is
tolerated. Naturally the price of the third-world-rated floor sweepings is
considerably lower.

Best regards,
Spehro Pefhany
--
"it's the network..."                          "The Journey is the reward"
.....s...KILLspamspam.....interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

2008\07\16@115611 by Bob Axtell

face picon face
Russell,

The problem with the white LEDS is similar to problems with other Chinese
goods (i.e.
Li-Ion cells, many other items). The causes seems to be poorly-made
chemicals. The
raw chemicals used in many processes don't seem to be chemically pure enough
for
the task at hand. The labor is fine, the business practices are OK (but not
to Western standards), but critical chemicals don't seem to be available.
That's why a battery maker is able to make decent batteries for months then
make a batch of really shoddy goods one
week. That's what I was told when _I_ asked about it...

--Bob

On Wed, Jul 16, 2008 at 6:33 AM, Apptech <EraseMEapptechspam_OUTspamTakeThisOuTparadise.net.nz> wrote:

{Quote hidden}

> -

2008\07\16@123311 by Funny NYPD

picon face
Material science and industry wasn't build in one week, it took years to accumulate experience and process practice. Eventually they will get there. It might take some time.

Funny N.
Au Group Electronics, New Bedford, MA, http://www.AuElectronics.com



{Original Message removed}

2008\07\16@133604 by Dr Skip

picon face
Wouldn't that result in an _above_ speck product? :) and below spec... (sorry,
I couldn't resist).

I can also add that since I run a lot of multi-white-LED items, that there is
some percentage of individual LEDs that have died in a few hours of runtime
rather than decades. No numbers or analysis after that, but enough to note and
wonder why...


Rich wrote:
Running the clean room below
> "clean" can result in below speck product.

2008\07\16@163307 by William \Chops\ Westfield

face picon face

On Jul 16, 2008, at 10:35 AM, Dr Skip wrote:

> some percentage of individual LEDs that have died in a few hours of  
> runtime
> rather than decades.

A fair number of the LED traffic lights around here have developed  
blotches of dead LEDs.  Especially the white LEDs used for "Walk"  
signals, but also even RED lights.  It's sorta sad; I bet someone in  
the local DoT is unhappy that the believed the "saves replacement  
costs" spiel..

BillW

2008\07\16@172943 by Nate Duehr

face
flavicon
face
William "Chops" Westfield wrote:

> A fair number of the LED traffic lights around here have developed  
> blotches of dead LEDs.  Especially the white LEDs used for "Walk"  
> signals, but also even RED lights.  It's sorta sad; I bet someone in  
> the local DoT is unhappy that the believed the "saves replacement  
> costs" spiel..

They haven't replaced them though -- or you wouldn't have seen them.  (GRIN)

I suppose there's probably some rule about how many of them have to fail
before the assembly has to be replaced.

But agreed... everyone touted CFL's for home lighting so I've tried a
few... all of which have failed in slightly longer lifespans than
incandescent bulbs.

(Continuing this on [EE] because I'm interested in the LIFESPAN
engineering of these  electrically engineered products... LEDs, CFLs,
all the "stuff" that's supposed to last longer.  Any discussion of CFL's
as the saviors of the planet or whatever, should go to [OT] of course.)

Nate

2008\07\16@174445 by Paul Hutchinson

picon face
> -----Original Message-----
> From: piclist-bouncesspamspam_OUTmit.edu On Behalf Of William "Chops" Westfield
> Sent: Wednesday, July 16, 2008 4:33 PM
>
> A fair number of the LED traffic lights around here have developed
> blotches of dead LEDs.  Especially the white LEDs used for "Walk"
> signals, but also even RED lights.  It's sorta sad; I bet someone in
> the local DoT is unhappy that the believed the "saves replacement
> costs" spiel..
>
> BillW

I saw this happening soon after the LED lamps were introduced in SE
Worchester County MA back about five years ago. Within a year a number of
the lamps had been replaced and since then many have been replaced with the
old incandescent type. A new traffic light was added on my commute a month
ago and that town put up incandescent instead of LED, I'm guessing that town
has given up on LED lights.

>From a drivers perspective I'd like them to put them all back to
incandescent because the LED units with multiple dead LEDs are very
difficult to see in bright sunshine. I suspect the decreased light output
when hot contributes to my difficulty in seeing some of them on hot commutes
home. The town's around here seem to wait until they have more than 66% dead
LEDs before they replace them.

My guess as to why they are such a problem up here is that they failed to
take into account the over 55 degC ambient temperatures inside a traffic
light on a hot southern New England summer day. Add in all the heat from the
LEDs and it wouldn't surprise me if on some days it's well over 80 degC in
the light fixture.

Paul Hutch


2008\07\16@175903 by Richard Prosser

picon face
Nate,

I've had good results with CFLs - especially on my garage door opener.
This is a high vibration application and standard incandescent bulbs
last about a month. "Rough Service" rated bulbs last about 6 months.
I've had one CFL bulb fail at 2 weeks (internal dry joint) but one has
lasted about 3 years so far and the other about 2 and it's replacement
(so far) ~1year.

In non "rough service" areas I've not had to replace any IIRC so
better than 3 years and running. This is about 7 bulbs in addition to
the ones in the garage.
We still use incandescent where we have dimmers and in some less
common fittings.

Richard P

2008/7/17 Nate Duehr <@spam@nateKILLspamspamnatetech.com>:
{Quote hidden}

> -

2008\07\16@184918 by Apptech

face
flavicon
face
> Dr Skip wrote:
>> some percentage of individual LEDs that have died in a
>> few hours of
>> runtime
>> rather than decades.

> BillW
> A fair number of the LED traffic lights around here have
> developed
> blotches of dead LEDs.  Especially the white LEDs used for
> "Walk"
> signals, but also even RED lights.  It's sorta sad; I bet
> someone in
> the local DoT is unhappy that the believed the "saves
> replacement
> costs" spiel..

Where's the warranty document ... .

Quite apart from actual LED longevity, poor driving can lead
to early failures. Driving multiple LEDs while maintaining
efficiency is a modest challenge that many do not rise to. I
have seen many consumer products with little or no attempts
to provide anything like constant current drive. Most 3-cell
battery consumer torches place LEDs in parallel and across
the battery. Some use a single shared series resistor. I
have a 30? LED lantern powered by a 6V SLA with LEDs
connected in pairs across the battery with no resistors.
Works amazingly well. When you have many LEDs and a highish
supply voltage available then multiple series parallel
strings are possible. You'd HOPE that that was what eg
traffic lights did. Using series resistors you get a degree
of mismatch between strings and some losses in the
resistors. For mains powered signalling equipment this is
usually tolerable. For lighting applications the lumens/Watt
suffer accordingly. When mains powered the voltage can vary
substantially and if brightness and current are to be
maintained at reasonably stable levels you need 'quite a lot
of drop' [tm] in the series element.

For battery powered equipment with multiple LEDs, if best
possible efficiency is required then all LEDs need to be
operated in a single series string with a (usually boost)
converter used to provide appropriate voltage for desired
current. With multiple series strings you need an indecently
variable converter per string. Anything less gives
suboptimum efficiency (assuming that all converters have the
same efficiency). I have never yet seen equipment with
multiple series strings and a converter per string.

More on practical lifetime measuring later today ... .



       Russell

2008\07\16@184919 by Apptech

face
flavicon
face
> My guess as to why they are such a problem up here is that
> they failed to
> take into account the over 55 degC ambient temperatures
> inside a traffic
> light on a hot southern New England summer day. Add in all
> the heat from the
> LEDs and it wouldn't surprise me if on some days it's well
> over 80 degC in
> the light fixture.

Temperature is a great killer of LEDs. 10C increases can
decrease lifetimes by 2 to 10 times depending on other
factors. It doesn't take many 2X reductions to turn 100,000
hours into disaster.

Philips / Luxeon / Lumiled have some excellent material
available on lumen maintenance wrt temperature and mA drawn
(essentially independent variables wrt effect despite
usually being linked in practice by practical realities). eg
you can get all combinations of IxTy and with differing
results.



       Russell

2008\07\16@184919 by Apptech

face
flavicon
face
> But agreed... everyone touted CFL's for home lighting so
> I've tried a
> few... all of which have failed in slightly longer
> lifespans than
> incandescent bulbs.

I date label (almost) all CFLs on installation.

Some small wattage ones which are always on have run for
literally 2 or 3 years. This exceeds the manufacturers'
claims handsomely.

While some others die early I find that they generally last
long enough to more than justify the capital cost :
operating cost tradeoff involved.

But, then you have to see using them.


           Russell

2008\07\16@193335 by Cedric Chang

flavicon
face
{Quote hidden}

or in my case , shorter
cc



Nate


2008\07\16@204911 by mrresp1

flavicon
face
" everyone touted CFL's for home lighting so I've tried a
> few... all of which have failed in slightly longer lifespans than
> incandescent bulbs "

we switched over to CFLs all over the house a couple years ago.. we actually
got a check back from the power company after about 3 months because they
had estimated our usage based on the year before.  nice.

however, after about 6~7 months or so, 95% of them died out... strangely,
the ones located outside that are subjected to northeast winters have
survived to this day.  the indoors surviving members go into a 'flicker
fits' all the time until they warm up after about 20-30mins.  .there is no
doubt in my mind that this is a waaaayy shorter life span than what i had
with incandescents for which i'm absolutely positive i still have original
bulbs installed & working fine from 17 years ago....

these CFLs (all different brands) seem to have had self-destructive thermal
problems..  certainly there's no doubt that the manufacturers knew all about
it and gambled on the usual consumer laziness not to claim their warranty
coverage ...

after all told, the manufacturers made their quick buck and we ended up
wasting money, filling up the landfills with electronic garbage, and thus
increasing our carbon footprint...

way to go world....

resp~
.


2008\07\16@205333 by Apptech

face
flavicon
face
> > The problem with the white LEDS is similar to problems
> > with other Chinese goods (i.e.Li-Ion cells, many other
> > items).
>> The causes seems to be poorly-made
> > chemicals. The raw chemicals used in many
>> processes don't seem to be chemically pure
>> enough for the task at hand.
>> The labor is fine, the business practices
>> are OK ...
...

A friend says:

The problem goes much deeper than that.

All companies can have problems with the quality of their
input materials
from time to time.

Reliable companies have rigid quality control policies and
procedures that
detect problems early in the production process and do
something about them
such that the quality of the finished goods they ship to
their customers is
not in turn adversely affected.

In the case of many Chinese companies (and similar companies
in other
countries) quality control tends to be very poor at all
levels and a totally
foreign philosophy often exists along the lines of "lets
ship anyway and see
if the customer notices".

Some disreputable companies deliberately set out to
islead  - providing
good quality product meeting spec initially and then cutting
corners and
(their) costs once the production volume orders arrive.  I
have seen
examples where inferior types of plastics have been
substituted (or
reinforcing additives deleted from the mix) or alloy
formulations changed.
In a couple of cases the substitutions have been so blatant
that you have to
wonder how the supplier thought they would ever get away
with it  - yet they
try it on  - and feign surprise and concern when they are
caught out.

To some extent the problem is one of ethics and morality
although plain
carelessness and poor management are probably just as big a
problem with
many Chinese companies.  The old adage that you get what you
pay for applies
just as much to the Chinese as to anyone else.  As their
input costs
escalate over time (as they inevitably will) and they can no
longer compete
entirely on price, they will be forced to address issues of
quality and
reliability of the supply chain  - and the situation will
improve.  Until
then, eternal vigilence, combined with an open up-front
zero-tolerance
attitude to quality issues when receiving goods, will help
disuade most
suppliers from any thoughts of "cheating".

Regards,

Ken Mardle

Applied Digital Research Ltd.  Tel    : +64 9 415-2514
P.O. Box 6480                  Fax    : +64 9 415-3514
Wellesley St                   Mobile : +64 21 879-648
Auckland
NEW ZEALAND                    WWW    :
http://www.acqura.com

2008\07\16@205335 by Apptech

face
flavicon
face
A number of people have asked how LED lifetimes can be
inferred from limited duration test data.
This is a representative query:

Joe Bento said:
> I am most curious, and perhaps someone can answer.  You
> mention that the
> white LED might fail within a 2x magnitude of 100k hours.
> 100k hours is
> about 11.4 years.  How is MTBF estimated or calculated?  I
> don't think
> they light an LED at its recommended current and wait -
> white LEDs
> haven't been around that many years.

An excellent paper on the subject is here

       http://www.lumileds.com/pdfs/WP12.pdf

Note section 7. where they discuss the use of the Weibull
distribution to infer long period results from shorter
period data. Alas, there are 3 factors which are "derived
from experimental data" so this is most suited to large
suppliers who have the resources to do the extensive testing
needed to establish the parameters for their product ranges
AND to understand how these are affected by various other
factors.

However, from their work lumileds have provided some
excellent rule of thumb guidelines which can be applied with
due care to other manufacturers' products. See the above
paper for information - some of it quite sobering.

Also, individual manufacturers do provide long term test
results, with data occasionally updated as tests proceed. I
have some Nichia data (may not have been meant for on
copying - I could ask) which shows real results out to
around 10,000 hours under various conditions of temperature
and humidity.

In my case I have found that operating continually for
several weeks is usually a good indicator of whether a LED
is either very bad or MAY be OK. Bad LEDs have very
measurable degradation after say 300 to 500 hours. I have
some samples which have now run for almost 6 months and a
light meter is not needed to see the result. Degradations of
up to 50& are often not detectable with certainty in
isolation to the untrained eye. "This light looks a bit
dull" may or may not be the conclusion. Even in side by side
comparisons 10% to 20% is hard to see (for me anyway).

I run batches of LEDs in series at constant current AND have
a reference LED that is in series but is usually
short-circuited. When measuring I enable the reference LED.
This gives me a zero (almost) operating time comparison and
provides a sanity check for fluctuations in ambient
temperature, current. light meter health etc.

In my tests the best Chinese LEDs operated at or below rated
current and in modest ambient conditions (say 20C mean or
less) would be extremely lucky to reach 1,000 hours at above
50% initial output!!! Hence my 2 orders of magnitude
comments. Not very hard to measure in short periods and Mr
Weibull need not be invoked. (For precision testing control
of ambient temperatures is necessary. In my case the results
are clear enough that no such control is needed. Also, heat
sinking to a level similar to that to be experienced in
practice is desirable as temperature rise can play a major
part in longevity).

Note when looking at test data that in many cases LED output
RISES after a while and then begins to decline. This is said
to be due to "annealing".Whatever the cause it is often a
good indicator of a LEDs long term performance. Those that
peak super early, or not at all, can be expected to have a
poor lifetime.



           Russell

Luxeon reliability tool

       http://www.luxeon.com/technology/reliabilitytool.cfm

Summary versions of above paper

       http://www.lumileds.com/pdfs/AB07.PDF

       http://www.luxeon.com/technology/lumenmaintenance.cfm

K2 Star
70% at 50,000 hours

       http://www.luxeonstar.com/luxeon-k2-star-red-75-lumens-700ma-p-131.php

Discussion of practical applications

       http://www.electronicsworld.co.uk/asset/361/leds.pdf








2008\07\16@211347 by peter green

flavicon
face

> these CFLs (all different brands) seem to have had self-destructive thermal
> problems
What type of fittings do you tend to use

In my experiance the worst situation for a CFL is a small enclosed
fitting with the lampholder at the top. The heat rises to the top of the
fitting where it is trapped cooking the electronics.

2008\07\16@213155 by mrresp1

flavicon
face
"worst situation for a CFL is a small enclosed
fitting with the lampholder at the top"

yup, "hi-hats"... as we call them...

but let's face it, they're probably one of the most popular fixtures out there and after all, these were your standard CFL R20, R30, & R40 floods ... so where exactly did they 'think' these bulbs were going to be used???... obviously not in a table top lamp....

~


{Original Message removed}

2008\07\17@033214 by Dr Skip

picon face

Apptech wrote:
>
>
> Where's the warranty document ... .

Mail to some site, my expense, possibly pay for return postage, possibly accept
a reconditioned unit in return (not my personal one), wait 10-12 weeks, their
discretion as to repair or not, blah blah blah...

And unless they changed their process, design, or sourcing, it will happen again.

>
> Quite apart from actual LED longevity, poor driving can lead
> to early failures. Driving multiple LEDs while maintaining
> efficiency is a modest challenge that many do not rise to. I
> have seen many consumer products with little or no attempts
> to provide anything like constant current drive. Most 3-cell
> battery consumer torches place LEDs in parallel and across
> the battery. Some use a single shared series resistor.

I've actually had better luck with the 3 cell units. The ones I talked about
have a surface mount controller chip and handful of components, has flash and
multiple brightnesses, etc. Certainly more development than throwing a resistor
at it...


2008\07\17@042522 by Alan B. Pearce

face picon face
> Interestingly, and not directly related, some name brand
> LEDs give atrocious spectral results at very low currents
> while others are about as good across a wide range of
> currents.

So, your testing shows you get what you pay for - pay more get better
performance ??? ;))

2008\07\17@064235 by Apptech

face
flavicon
face
>> Interestingly, and not directly related, some name brand
>> LEDs give atrocious spectral results at very low currents
>> while others are about as good across a wide range of
>> currents.

> So, your testing shows you get what you pay for - pay more
> get better performance ??? ;))

No.
As noted, "name brand LEDs were used in both cases. Cost
in volume is uncertain but both would be comparable.

And it was hardly testing - more just observation from
simple experiment.
And most people do not want to run LEDs in the manner
concerned. What was notable was that with one supplier you
could do so and with the other you couldn't, if you wanted
white light.

Using parts from two name brand / 'top shelf' suppliers,
Nichia & Avago, the results were very different. Note that
the result would not be a problem in any "normal"
application. Note also that vendors expressly note that when
dimming LEDs they should be PWM'd so that peak PWM on
current  ~= rated current and NOT DC fed at lower levels
expressly due to the colour shift. Caveat Emptor.

When the 20 mA spec LEDs were operated in the 1 mA to 5 mA
range the Nichia LEDs still produced essentially white
light, whereas the Avago LEDs from the same batch produced
EITHER extremely yellow or extremely blue light. When a row
of the Avago LEDs were compared it appeared that they were
of two colour types. The result appeared to be a mix of two
different colours.

Few people want to run 20 mA nominal LEDs at about 1 mA. It
happens that I have a client who does. Total current is up
to about 5 mA and this was shared amongst from 1 to about 5
LEDs during development testing. Notionally the amount of
light is similar with 1 to 5 LEDs - you just get less light
per LED as you add more. But you also get colour shift.

The results are due to the LED proper being blue and the
phosphor converting a proportion of it to white so the two
are added to produce "white". Clearly different things
happen with Nichia and Avago at low current levels.

I have no business or other links to Nichia.
I have become increasingly impressed with Nichia LEDs and
their technical approach to doing business. I don't know too
much about their business side of business or even pricing.
>From what I've seen so far I'd rate them as generally too
dear for products that can get away with not using the very
best :-). There may be others, but so far Nichia are the
only company that I have seen put a chromaticity shift with
colour graph in a standard data sheet. Also from what I've
seen so far, their chromaticity binning range seems to be
tighter for comparable products than at least one other
market leader and maker of very fine products. (Name not
mentioned and not Avago - they are good in other respects so
I don't want to be seen to be criticising them here for
something which they make clear and which may well be
acceptable in some applications. .


       Russell







2008\07\17@072752 by Ariel Rocholl

flavicon
face
But still you didn't explain how do you measure LED life time to understand
is not 100k hours but 2 orders of magnitude less.

Some mentioned CFL lasted 2-3 years. Well, that means probably lasted before
failure (or refused to start) but that doesn't mean they were providing the
same lumens for 2-3 years. A CFL radiation curve is typically 60% less after
12 months in a few datasheets. For LEDs there are interesting curves as
well, most will have only a fraction of initial radiation after 2-3k hours.
They still work, but that doesn't mean they keep the properties at the level
one specific project may need. And I saw these curves for well documented
products such as Luxeon, but cheap chinnese brands not even provide that. So
how do you know where are you on the curve and what is the expected lifetime
with your own tests?

2008/7/17 Apptech <KILLspamapptechKILLspamspamparadise.net.nz>:

{Quote hidden}

> -

2008\07\17@080130 by cdb

flavicon
face
Sometime ago I wandered across the schematic for a standard LED
traffic light unit, it was nothing more than 1 series resistor and a
string of LED's upto 240vac, If I recall correctly (I probably still
have the diagram somewhere) the LEDs where used as rectifiers, by
having so many on one string, pointing one way, and so many on the
other the other.

I'll see if I can find the circuit.

Colin

:: I saw this happening soon after the LED lamps were introduced in SE
:: Worchester County MA back about five years ago. Within a year a
:: number of
:: the lamps had been replaced and since then many have been replaced
:: with the
:: old incandescent type
--
cdb, RemoveMEcolinTakeThisOuTspambtech-online.co.uk on 17/07/2008

Web presence: http://www.btech-online.co.uk  

Hosted by:  http://www.1and1.co.uk/?k_id=7988359







2008\07\17@083747 by olin piclist

face picon face
Apptech wrote:
>> A fair number of the LED traffic lights around here have
>> developed
>> blotches of dead LEDs.  Especially the white LEDs used for
>> "Walk"
>> signals, but also even RED lights.  It's sorta sad; I bet
>> someone in
>> the local DoT is unhappy that the believed the "saves
>> replacement
>> costs" spiel..
>
> Where's the warranty document ... .
>
> Quite apart from actual LED longevity, poor driving can lead
> to early failures.

That would be rather extreme poor driving.  He wasn't saying cars were
crashing into the light poles, just that the LEDs were failing in the light
fixtures.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000.

2008\07\17@083814 by Apptech

face
flavicon
face
> But still you didn't explain how do you measure LED life
> time to understand
> is not 100k hours but 2 orders of magnitude less.

Yes, I did. You may have missed it.
See my post about 12 hours ago (the time is NZT - about GMT
+ 12)
which begins:

   From: "Apptech" <spamBeGoneapptechspamBeGonespamparadise.net.nz>
   Sent: Thursday, July 17, 2008 12:52 PM
   Subject: Re: [EE]:: Dead LEDs

   A number of people have asked how LED lifetimes can be
   inferred from limited duration test data.
   This is a representative query:

Real people use the Weibull distribution to infer long term
characteristics from shorter term data. See muy post and
Philips paper.

I have not needed this degree of complexity yet. ALL chinese
LEDs tested so far reveal themselves well enough in a month
of continuous operation.

> Some mentioned CFL lasted 2-3 years. Well, that means
> probably lasted before
> failure (or refused to start) but that doesn't mean they
> were providing the
> same lumens for 2-3 years. A CFL radiation curve is
> typically 60% less after
> 12 months in a few datasheets.

Depends on manufacturer and whether you mean continuous
running.
People like Philips define their terms well.
An 8000 hour life will probably mean something like 50% of
initial at 8000 hours.

As the philips Tornado CFLs are up to 2X as efficient as
some competitors they are as good as the worst at the end of
their life.

While I have had a number of CFL early fauilures (and have
destryed many more due to mishandling etc) I generally feel
that they are up to the claims made for them when ones from
repiutable manufacturers are used.  For general lighting use
at teh 20 Watt level I buy ONLY the best - Philips Tornado.
I buy them when they are on special at about $NZ3.33
typically (3 for $10) or about $US2.70. For 40W, 60W and 100
Watt ones (:-) :-) :-) :-) :-) :-) ) I buy the brightest on
the demo display stand. I do like the 100 Watt ones !!!!.

> For LEDs there are interesting curves as
> well, most will have only a fraction of initial radiation
> after 2-3k hours.

See the detailed lumen maintenance document from Philips
referenced in the post I mention above.

Reputable manufacturers provide guaranteed degradation data.
Typically 70% of initial at 10,000 or 50,000 or 100,000
hours depending on various parameters.

> They still work, but that doesn't mean they keep the
> properties at the level
> one specific project may need. And I saw these curves for
> well documented
> products such as Luxeon,

as above

> but cheap chinnese brands not even provide that. So

Indeed. Which is what started this thread.

> how do you know where are you on the curve and what is the
> expected lifetime
> with your own tests?

As per my prior posts.
Chinese LEDs are so far bad enough that weeks of testing is
enough to see the trend.

Operate N at constant current.
Short one with a switch.
Unshort this refence only for measuringh. Allow a minute or
so to thermally stabilise before measuring.

Use light meter of choice.

I take a white plastic kitchen funnel and cut it down so
large end fits light meter dome and small end is trimmed to
fit over 10mm lEDs and bed square on PCB. tape all outside
of cone withe black tape. This makes a crude but OK
integrating radiometer. Most LED light shines onto meter
dome (which is itself an integrator) and the rest gets
reflected off cone. As the radiation patterns I am testing
are much the same for all LEDs I'm testing the errors
introduced are (IMHO) lowish.

Measure all LEDs. Turn off reference LED.
Record levels.
Repeat occasionally.

168 hours / week.
730 hours in a month.

If it drops noticeably in a week bin it.
If it drops noticeably in a month ask why.
I have some that have been running about 6 months so far.
Not a pretty sight.




       Russell

2008\07\17@090214 by Apptech

face
flavicon
face
> Sometime ago I wandered across the schematic for a
> standard LED
> traffic light unit, it was nothing more than 1 series
> resistor and a
> string of LED's upto 240vac, If I recall correctly (I
> probably still
> have the diagram somewhere) the LEDs where used as
> rectifiers, by
> having so many on one string, pointing one way, and so
> many on the
> other the other.
>
> I'll see if I can find the circuit.

Would be interesting to see.
That's an entirely valid way to do things BUT it can be done
well or poorly.

Using AC means they are often below peak value and often off
during a cycle.
This tempts people to run them over spec at peak.
Colour shift with current not a major issue probably.
Current variation with mains voltage change not trivial.

Rectifying the mains and using DC and a basic current source
would give far better control.

Say VLED on is 2V1.
Say allow 10% for series resistor.
Say VLED_very_dim = 1V7 (make own assumptions)

For 230 VAC, Vpeak = 325 Volts.
Vleds = 325 * 90% = 293 V
LEDS = 293/2v1 = 139 = 140

Vdim = 325*1v7/2v1 =~~ 263 V
Phase angle past peak V = acos(263/325) = 36 degrees

ie LEDs go from full to very dim in +/- 36 degree area
either side of peak V or 36/90 = 40% of cycle.

This drives LEDs at far far less than max and temptation is
to use high I peak, which MAY be OK.

Running from DC rectified mains gives constant current,
constant brightness. Rectifier cost is low.
Capacitor cost may be less so. At 20 mA you need 2000/V uF
where V is ripple.
At 10V ripple need 200 uF At 400 V. Electrolytic OK. Not
tiny.

Say LEDs are $0.05 x 140 = $7 of LEDs.
Cap will be less than that but a significant part of cos.
Rough current source can cost little.
Rectifier is v cheap (1N 4007).


   Russell









2008\07\17@090908 by Apptech

face
flavicon
face
>> Quite apart from actual LED longevity, poor driving can
>> lead
>> to early failures.

> That would be rather extreme poor driving.  He wasn't
> saying cars were
> crashing into the light poles, just that the LEDs were
> failing in the light
> fixtures.

What have you done with Olin?


                   Russell


                                               :-)

2008\07\17@100625 by Ariel Rocholl

picon face
2008/7/17 Apptech <TakeThisOuTapptechEraseMEspamspam_OUTparadise.net.nz>:

> > But still you didn't explain how do you measure LED life
> > time to understand
> > is not 100k hours but 2 orders of magnitude less.
>
> Yes, I did. You may have missed it.


I definitely missed that. Probably my brain capacity is lower in the curve
than I expected...


>
> Real people use the Weibull distribution to infer long term
> characteristics from shorter term data. See muy post and
> Philips paper.
>

So this means you basically reuse Philips experience to measure other
brands. But I am not sure Philips provide *all* the required data to make
their test reproducible. It would be better IMHO to have ISO / ANSI / DIN
papers to measure parameters in a well stablished, fully reproducible,
neutral way. Many times manufacturers measures parameters in the most
convenient way *for them*.



>
> I have not needed this degree of complexity yet. ALL chinese
> LEDs tested so far reveal themselves well enough in a month
> of continuous operation.
>

So I am not being critic, I am just trying to learn how feasible is to
extrapolate what Philips documents, to measure other brands' lifetime. I
think is easier to predict degradation to a certain amount of radiation but
lifetime may be different. I think is really safe for you to say "lifetime
is shorter than brand X" but giving absolute numbers looks challenging to me
(like 2 orders of magnitude less, etc).



{Quote hidden}

Right. And depending on the application. For instance, I am ok with a CFL
being at 50% in the second year if it is for a kitchen.
If the application is a UVB CFL type for a reptile terrarium, you will kill
the animal by lack of D vitamin if it doesn't receive at least 75% UVB
radiation of specified for the species. So for that application, end of life
means 75% radiation. That is typically 6-8 months for best brands (8 hs/day,
25C). Not too much.



>
> typically (3 for $10) or about $US2.70. For 40W, 60W and 100
> Watt ones (:-) :-) :-) :-) :-) :-) ) I buy the brightest on
> the demo display stand. I do like the 100 Watt ones !!!!.
>


100W CFL? !!!
I've still to see that! With sun glasses anyway!


{Quote hidden}

Sounds reasonably reproducible anywhere. Did you validated Philips
parameters using this method in one of their Lumileds?

Very informative. Thx

--
Ariel Rocholl
Madrid, Spain

2008\07\18@061218 by cdb

flavicon
face


:::: Sometime ago I wandered across the schematic for a
:::: standard LED
:::: traffic light unit, it was nothing more than 1 series
:::: resistor and a
:::: string of LED's upto 240vac, If I recall correctly (I
:::: probably still
:::: have the diagram somewhere) the LEDs where used as
:::: rectifiers, by
:::: having so many on one string, pointing one way, and so
:::: many on the
:::: other the other.
::::
:::: I'll see if I can find the circuit.
::::
:: Would be interesting to see.

Wouldn't you know it, I can't find it on my CD backups, so I'll check
my paper folders. Even searched Gurgle, all I found was a number of
people have patented the 'a diode backwards across the LED for AC'
amazing what can be patented.

Colin
--
cdb, RemoveMEcolinspamTakeThisOuTbtech-online.co.uk on 18/07/2008

Web presence: http://www.btech-online.co.uk  

Hosted by:  http://www.1and1.co.uk/?k_id=7988359






More... (looser matching)
- Last day of these posts
- In 2008 , 2009 only
- Today
- New search...