Off-line LED drivers

[Adrian Jansen]

Many sources now for these devices, but the few I have inspected all
have relatively large electro caps on the input side, to make DC from
which the current controller PWM draws its supply for the LED drive.

I strikes me that with the published figures for LED lifetimes of 50,000
hours, the electro caps in these are probably going to fail long before
the LEDs and other associated circuitry.

Has anyone investigated sensible ways of running the PWM circuit direct
off the ( full-wave ) rectified line, and accepting the fact that for a
few msec, the input voltage will fall below the 10 volts or so needed to
drive the LEDs ?

Seems to me that would extend the potential lifetime of the driver
considerably.

Any thoughts ?

 

[mike]

Many sources now for these devices, but the few I have inspected all
have relatively large electro caps on the input side, to make DC from
which the current controller PWM draws its supply for the LED drive.

I strikes me that with the published figures for LED lifetimes of 50,000
hours, the electro caps in these are probably going to fail long before
the LEDs and other associated circuitry.

Has anyone investigated sensible ways of running the PWM circuit direct
off the ( full-wave ) rectified line, and accepting the fact that for a
few msec, the input voltage will fall below the 10 volts or so needed to
drive the LEDs ?

And accepting the fact that line transients can exceed the limits of the
device.

 

[Phil Allison]

"Adrian Jansen"

Many sources now for these devices, but the few I have inspected all have
relatively large electro caps on the input side, to make DC from which the
current controller PWM draws its supply for the LED drive.

I strikes me that with the published figures for LED lifetimes of 50,000
hours, the electro caps in these are probably going to fail long before
the LEDs and other associated circuitry.

** Why is that ?

It is common for electros to be fine after 50,000 hours of continuous use.

One example is DC plug pack supplies left on 24/7.

Temperature is the issue and long as the electro is not subjected to more
than say 45C it can have a very long life.



..... Phil

 

[Phil Allison]

"Phil Allison"

It is common for electros to be fine after 50,000 hours of continuous use.

One example is DC plug pack supplies left on 24/7.

** The kind that have iron transformers .


..... Phil

 

[Fabio_78]

Many sources now for these devices, but the few I have inspected all have

relatively large electro caps on the input side, to make DC from which the
current controller PWM draws its supply for the LED drive.

I strikes me that with the published figures for LED lifetimes of 50,000
hours, the electro caps in these are probably going to fail long before the
LEDs and other associated circuitry.

Hello, if the working temperature is low enough and
the capacitors good enough (pricy?) they should
last enough
And also the optocouplers tend to age.

There are many design examples that lack a large
electro cap on the input. AFAIK this is done mainly for
high power factor, but as a side effect the large high
voltage electro disappears. The single stage High PF flyback
is simple to do for low power requirements, and moves the
cap from rectified mains to led output side.
For higher power one could use the usual PFC+flyback
(or PFC+resonant) and use high voltage film capacitors
as storage.
Check appnotes for ST L6561/2, powerint linkswich-PH,
if I remeber correctly also fairchild and ON should have
appnotes on the subject.

 

[Jamie]

Many sources now for these devices, but the few I have inspected all
have relatively large electro caps on the input side, to make DC from
which the current controller PWM draws its supply for the LED drive.

I strikes me that with the published figures for LED lifetimes of 50,000
hours, the electro caps in these are probably going to fail long before
the LEDs and other associated circuitry.

Has anyone investigated sensible ways of running the PWM circuit direct
off the ( full-wave ) rectified line, and accepting the fact that for a
few msec, the input voltage will fall below the 10 volts or so needed to
drive the LEDs ?

Seems to me that would extend the potential lifetime of the driver
considerably.

Any thoughts ?

Sure, a linear current sink will work just fine,
if you don't mind a small amount of energy loss in
thermal on the transistor?

You could do it with phase firing and that won't lose
too much, but that involves the use of a cap but not a large one,
one that most likely would last much longer...

Jamie

 

[E]

I don't quite understand why they don't use good old inductive ballast (or
current-limiting transformer) in those led things?
Cheap shit ballast is way more reliable than any cheap shit switchmode psu.

 

[[email protected]]

Copper is expensive, shipping weight too, efficiency, size...

If you try to put some 50 Hz inductances into some E27 lamps, the
thing would be quite heavy and existing luminaries could not handle
it.

You throw away the converter with the lamp,
at least one I have has it buld in.
The lamp will only work for a few thousand hours (LED life).
So makes no sense to make the converter last longer.

Please read the LED specs carefully.

If the LED is marketed by (absolute) maximum current of say 1000 mA
and hence marketed as "3 W" LED, the life time (after LED and phosphor
degradation) is a few thousand hours.

However if respectable manufacturers only specify the lumen output
(with 1000 mA abs max) at 350 mA (1 W) and the 30.000-50.000 hours
might be believable.

 

[MrTallyman]

Copper is expensive, shipping weight too, efficiency, size...



You throw away the converter with the lamp,
at least one I have has it buld in.
The lamp will only work for a few thousand hours (LED life).
So makes no sense to make the converter last longer.

The right move for the entire industry would be to place the DC source
conversion IN THE LAMP and make the "light bulbs" just the LEDs and
current limit elements and attachment socket stub.

 

[Adrian Jansen]

The right move for the entire industry would be to place the DC source
conversion IN THE LAMP and make the "light bulbs" just the LEDs and
current limit elements and attachment socket stub.

We already have that, and they have much the same problems of CCL lamps
- the temps generated by the lamp itself impact on the electronics.
Much better to separate the LED and its heat from the electronics.

With the long life ( quoted ) on the LEDs, they essentially become
fixtures in the building, rather than replacable elements like the
tungsten lamps we have all used. So it makes more sense to have the
entire lamp installation permanent. That is, if the rest of the driver
can also be made long life.

I rather doubt that inductive ballasts can supply the relatively
stringent constant current requirements of LEDs, although it is
tempting. Certainly the simplicity is attractive. And weight is not
really an issue in permanent fittings. Copper and iron cost maybe.

 

[Tim Williams]

With the long life ( quoted ) on the LEDs, they essentially become
fixtures in the building, rather than replacable elements like the
tungsten lamps we have all used. So it makes more sense to have the
entire lamp installation permanent. That is, if the rest of the driver
can also be made long life.

I've got it!

We'll build a high-reliability circuit, conservatively rated transistors,
big transformer/coils, lots of heatsinking, and leave the inevitable
electrolytic hanging out the side, mounted in a screw base! Patent
pending!

Ya know, the scary thing is the plausibility. They need to hurry up and
make high voltage aluminum polymers work. Then this'll (mostly) go away.

Tim

 

[josephkk]

Copper is expensive, shipping weight too, efficiency, size...



You throw away the converter with the lamp,
at least one I have has it buld in.
The lamp will only work for a few thousand hours (LED life).
So makes no sense to make the converter last longer.

Gosh, i am seeing LED lamp and luminaire ratings of 50,000 hours and up,
even claims of over 100,000 hours. You would need damn good PSUs for
those.

?-)

 

[Adrian Jansen]

Or you need some imagination. That's around 35 years at 8 hours a day, so
there's little risk to making the claim and being off by 4:1 or so.

I too am very sceptical of those 50,000+ hour ratings. One can only
presume they do some sort of high temp accelerated test to base the
figure on. Unless its just purely marketing guesswork.

However even if they are optimistic by a factor of 5, that still makes
the LED regardable as a fixture, rather than a replaceable item, in the
average domestic/business installation.

 

[[email protected]]

Or you need some imagination. That's around 35 years at 8 hours a day, so
there's little risk to making the claim and being off by 4:1 or so.

you never know, I just read of a guy making a fuss with safety
agencies
and the manufacturer because his PH lamp fell down when the plastic
gizmo
it was hanging in broke. It had been hanging there for 27 years!


-Lasse

 

[E]

I too am very sceptical of those 50,000+ hour ratings. One can only
presume they do some sort of high temp accelerated test to base the figure
on. Unless its just purely marketing guesswork.

However even if they are optimistic by a factor of 5, that still makes the
LED regardable as a fixture, rather than a replaceable item, in the
average domestic/business installation.

For domestic that may be considered fixture, but for business use 50000
hours
is just 2.5 times fluorescent tube lifetime ie. about 5 years. Lack of
replaceable
*standardized* bulbs is big drawback for led technology. Leds are damn
expensive to begin with and the need to replace the whole luminaire every
five years means much more installing work (expensive).

Also at 50000 h they typically quote something like 70% light output
remaining.
It means that to achieve some minimum lighting level at end of life, you
must
design with 30% extra power initially so most of lifetime you are making
more
light than necessary. Fluorescents drop only by 10% so not much a problem.
Of course if you have feedback on lighting level, then it is not any
problem.

Energy saving of leds compared to modern well designed fluorescent lighting
is anyway questionable at best. Of course if you compare leds to some old
non-designed system with 18W halophospate tubes with D-class ballast,
plastic diffuser and no reflector you get impressive numbers to put in ads.

Even if you spend big bucks for led lighting you do not necessarily save any
energy over fl. lighting. But using conventional light sources, natural
lightning and good design does not give you any "Green PR" like leds do.

 

[josephkk]

For domestic that may be considered fixture, but for business use 50000
hours
is just 2.5 times fluorescent tube lifetime ie. about 5 years. Lack of
replaceable
*standardized* bulbs is big drawback for led technology. Leds are damn
expensive to begin with and the need to replace the whole luminaire every
five years means much more installing work (expensive).

Also at 50000 h they typically quote something like 70% light output
remaining.
It means that to achieve some minimum lighting level at end of life, you
must
design with 30% extra power initially so most of lifetime you are making
more
light than necessary. Fluorescents drop only by 10% so not much a problem.
Of course if you have feedback on lighting level, then it is not any
problem.

Energy saving of leds compared to modern well designed fluorescent lighting
is anyway questionable at best. Of course if you compare leds to some old
non-designed system with 18W halophospate tubes with D-class ballast,
plastic diffuser and no reflector you get impressive numbers to put in ads.

Even if you spend big bucks for led lighting you do not necessarily saveany
energy over fl. lighting. But using conventional light sources, natural
lightning and good design does not give you any "Green PR" like leds do.

Some Sites/pages to consider:

http://www.lrc.rpi.edu/programs/nlpip/lightinganswers/t8/05-t8-lamp-life.asp

http://www.lrc.rpi.edu/programs/futures/LF-LampLife/

http://en.wikipedia.org/wiki/Electric_light

http://www.lithonia.com/micro_webs/electronicballast/source.pdf

http://www.lrc.rpi.edu/programs/nlpip/lightinganswers/hwcfl/HWCFL-life.asp

It seems that fluorescent life varies a bit over the range of 10,000 -
20,000 hours, LEDs range from 50,000 - 100,000 hours. I thus challenge
your only 2.5 times the life.

Depending on usage LEDs can deliver more lumens per watt, near equivalent
CRI, dimmability (something fluorescent does not handle well at all) and
some other useful trade offs. For some applications LEDs rock, for others
there are plenty of lamp technologies.

Induction lighting in particular is still giving all other technologies
fits in several applications.

?-)

 

[[email protected]]

I too am very sceptical of those 50,000+ hour ratings. One can only
presume they do some sort of high temp accelerated test to base the
figure on. Unless its just purely marketing guesswork.

This is just a design parameter.

The lifetime is defined when the light output is reduced to a specific
amount (say to 70 %). The rate of reduction can be measured with
relative short (1000 h) tests and extrapolated from there.

Those LEDs that dim in a few thousand hours are usually driven with
Imax, i.e. a LED with Imax=350 mA ("1 W") at 350 mA. This causes
severe thermal problems (high junction temperatures), heating the PSU.
Running at such high current also reduces the efficiency [lm/W].

Running the LED at Imax/3 will slow down the degradation considerably,
making 50000+ hours believable. The junction temperature is much
cooler, even when running three Imax/3 LEDs on the same heatsink as
one at Imax (Rth j-c is in parallel). It might even be possible to
reach over 100 lm/W claimed efficiency. Of course, to get similar
initial light output as the one LED at Imax, you need nearly three of
those devices at Imax/3, which of course increases the cost
significantly.

For even longer lifetime, reduce the current even further and add more
LEDs, but of course this drives the costs up and is only justifiable,
if the replacement costs are high.

Look carefully at the LED specs, optical characteristics are more or
less useless at Imax, many manufacturers specify the optical
characteristics at some more usable fraction of Imax.

Those over 100 lm/W efficiency claims are usually measured below
Imax/10 and at 25 C junction temperature (just when turned on . In
addition the highest outputs only apply to "cool white" i.e. those
with a very strong blue spectral line.

 

[josephkk]

Fluorescent lifetime varies very much with use pattern and ignition method.
I have read that in the US they use these rapid/instant start systems that
reduce lamp lifetime somewhat. Around here we use glow starters nearly
allways if using ballast. I think most here haven't even heard of those
rapid
start things
Also if you look at two first links you find that typical lamp lifetime is
given
with many on/off cycles. In typical business use lamps are on for long time
(if not 24/7) so startup wear is not that much concern. If you look at graph
in second link you can find that under these conditions it is reasonableto
expect about 30k life for a tube.
And 2.5 times 30k hours is 75k hours. Of course if there is much on/off
cycles lifetime drops and that 10k figure might be realistic for home use
with instant on starter.

Modern fluorescent lamp with electronic ballast and good reflector give
system efficacy of about 90 lm/W (tube only is approx 110 lm/W) so
they are about equally good as good led lamps. Also modern control
gear is not that bad at dimming.

Yes, certainly there are many places where leds are very good choice.
It just irritates me when green weenies are pushing led lighting as saving
the world wonder thing. Every magazine and newspaper is nowadays
full "leds save 60% energy" articles with no hint as to where on earth
they are pulling those numbers from. And then city engineers have
hard time telling politicians why changing every lamp in city to leds
ain't that great idea.

Nice trim.

The 60% number is averaged over all other technologies without frequency
of use weighting. Thus incandescent bend the figure a lot.

?-)

 

[[email protected]]

The right move for the entire industry would be to place the DC source
conversion IN THE LAMP and make the "light bulbs" just the LEDs and
current limit elements and attachment socket stub.

A better approach for the industry would to standardize some
constant_currents_ (DC) such as 20/50/100/350/1000 mA so that any
light panel from any manufacturer rated for a specific constant
current could be plugged into a series string, just like christmas
tree lights.

In a system driven by a constant current, each load would have the
same current, but the voltage could be different, depending of the
power levels needed. This is very similar as the Pxxnn series tubes
used in old European TVs, in which the filaments were in series across
the 220 Vac mains (with some inrush current limiting).

For larger LED panels it would be mechanically feasible to use
multiple constant current generators with separate strings with two or
more of those standard currents through the sockets. The panels would
only have pins for the required current and the socket would short
circuit the unused current lines, maintaining loop continuity. Some
open/short circuit detection at each constant current source could
take out unused current loops.

 

[[email protected]]

There are many design examples that lack a large
electro cap on the input. AFAIK this is done mainly for
high power factor, but as a side effect the large high
voltage electro disappears. The single stage High PF flyback
is simple to do for low power requirements, and moves the
cap from rectified mains to led output side.
For higher power one could use the usual PFC+flyback
(or PFC+resonant) and use high voltage film capacitors
as storage.
Check appnotes for ST L6561/2, powerint linkswich-PH,
if I remeber correctly also fairchild and ON should have
appnotes on the subject.

Wouldn't replacing the output voltage feedback divider at pin 1 with a
high side load (LED string) and 100 ohm to ground and feeding the
voltage across the resistor to the pin 1 (feedback) produce a 20 mA
constant current source ? This should be able to drive a string of 100
LEDs.

Thus only one switching transistor and one inductor would be needed
for both PFC as well as constant current regulation.

Of course, with such highly integrated chips, this might cause havoc
e.g. in startup functions.