Questions about power LEDs

Discussion in 'General Electronics Chat' started by takao21203, Dec 5, 2012.

  1. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    I have made a power LED assembly today.

    4x 10W LEDs are mounted on a Pentium II heatsink using a carrier PCB.
    These cheap proto PCBs are good for this actually- they can be worked easily.

    I made one test assembly already with 1x blue, 1x red in series.
    That runs directly from my laptop PSU!
    Current is too low, however. It is only about 500mA.

    -I noticed the red LED wasn't very bright at first, but brightness has increased over the past 2 days. Is this normal? Should I expect more increase when I run it at full current?

    -The new assembly has 2x LEDs in series. I want to use a large toroid which has a no-load voltage of more than 50V. Rated for 36V x 2.
    I want to use a DC/DC converter that is well capable of high voltage.

    The problem is when I have larger load, eventually voltage will drop to less than 44V. That is what I will need for all the LEDs in series.
    Applying nearly 110V DC to my DC/DC converter PCB is also something I never attempted. It has 12V supply for the TL494.

    -Is it better to use all the 4 LEDs in series, or 2x in parallel each?

    The transformer isn't optimal for that (48V would be better).

    If I have a large voltage difference, I think the DC/DC losses will be higher. All 4 LEDs in series only needs 1A current.

    My DC/DC circuit is capable of 10Amps permanently, probably more, and much higher inrush current.

    But I am not sure what choice I should make. Maybe I will try with all LEDs in series.

    I tried the laptop PSU with 2 red LEDs in series- only very low current. It has just a few volts less than needed!

    Actually today I looked at some power LED curves.
    The LED forward voltage will increase a lot with current.
    Since there are chips in series already, and two or 4 LEDs in series, there are several volts working area. Even a sudden spike of +1 volts won't neccessarily destroy the LEDs with so many in series.

    -Question 3. The forward voltage should be 7 to 9V (3 red chips in series). It is even a bit higher than for the blue LEDs! So it is definitively incorrect. Maybe these are white LEDs with a phoshor for red?

    I will wire up my PSU soon, inline the Amp meter, and see what voltage is required. And yes the PSU circuit has a current preset also, I made a mistake when I built the circuit, so it actually does not work at all. It won't be so hard to make it working maybe.
  2. Markd77

    Senior Member

    Sep 7, 2009
    I'd recommend a constant current circuit. Power LEDs have problems with thermal runaway without it - their temperature rises, so their forward voltage drops, the current rises and they get warmer until they fail. It might seem OK today but if the room temperature rises they could die.
    Something as simple as this will work OK (with recalculated values), replace the red LED with a zener if you are using higher voltages:
  3. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    I need 1 Amps, 44V, and already have a dc/dc circuit.
    The sink needs a cooling fan at 40W I highly guess so.

    Do you have any datasheet that shows thermal runaway for power LEDs?

    Mainly large power LEDs are interesting right now. I am also getting some 30W LEDs soon.
  4. Markd77

    Senior Member

    Sep 7, 2009
    It was surprisingly hard to find, it isn't mentioned in a lot of datasheets.
    If you look at page 9 of this one you will see the forward voltage drops around 0.2V with a temperature rise from 25 - 90 Celcius. That can be enough to double the current if you give it a constant voltage supply.
  5. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    1. I never allow 90C.
    2. It is adjusted to match the current in hot condition. Means I don't measure the current when the LED is cold.
  6. mcgyvr

    AAC Fanatic!

    Oct 15, 2009
    Stop feeding LED's from constant voltage power supplies.. Its NOT correct and will cause all sorts of problems. I thought that was already discussed/beaten into your head in your piranha build post... guess not..

    You've got 4 x 10W LED's.. I'd suspect max forward voltage for those is 11V maybe..and they are rated 1000mA each.
    Run all 4 in series and you need a 48V constant "CURRENT" power supply that is rated for close to 1A output.
    Something like a meanwell HLG-40H-48B or similar will work perfect.
  7. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    You misunderstand the purpose and nature of this thread. It is not to discuss the way of power supply. I wrote three questions. They are marked like that


    The rest of the post is only explanation to help understand the setup.

    I welcome your concerns. However the questions I raised remain the same no matter if constant current or not. So you are actually not talking about the raised questions. I wrote them, because I do not have information about the topics. I write them again:

    -LED burn-in (the red LEDs were not very bright, and still don't have full brightness). Is it required to do a burn-in at rated current? Or even higher currents? Has this been missed maybe during manufacturing?

    I have one assembly running for a few days, current is 500mA (10W LEDs). The blue LED is very bright, while the red LED still does not output much light. Even if it increased. Nominal current is about 1A. Having a cooling fan, this assembly is totally cold.

    -LEDs in series (what is better, highest possible voltage, or parallel LEDs)

    -Forward voltage incorrect, it should be 7 to 9V, but is much higher (3 chips).

    The new 40W assembly has 4 red LEDs. I have not yet powered it up. I will try to run it at rated current.
  8. mcgyvr

    AAC Fanatic!

    Oct 15, 2009
    No. LED's do not require a burn in period. But due to your lack of proper current regulation what you are experiencing could be related to a change in the amount of current that device is seeing. It could also be related to the ability of the human eye to see/effectively judge brightness for different colors of certain wavelengths.. It could be a manufacturing defect or numerous other issues.

    better in relation to what? safety? cost? manufacturability?
    Anything over 60VDC is considered a greater danger/hazard to a human than operating voltages under. Notice how most constant current drivers stay under 60VDC..(well you would when you stop playing with that CV supply and do it right) Paralleling of LED's is not recommended unless properly managed. Thermal runaway does occur. A current mirror circuit could be constructed to balance the current levels in the even of a shorted LED or change in forward voltage,etc.. In general it is my feelings that one should #1 stay under 60VDC and #2 avoid parallel LED circuits like the plague.

    These are from China.. Since when were published specs from China ever accurate? Not sure how you are measuring this either or what you consider "higher".. When you add LED's in series the strings forward voltage is the sum of the forward voltages of each LED. So 3 in series with a Vf of 9V is a string forward voltage of 27 Volts.

    And I 100% understood the nature of this thread and saw your questions.. BUT crap in = crap out
    Do it properly..then perform all the testing you want to validate the design. Right now its just a failure waiting to happen.
  9. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    I have no idea why the current should change when the voltage doesn't.
    I wrote already that it is 500mA. The laptop PSU can not produce more voltage.

    At first the red LED had a light output of less than 1W. It has increased a lot, but it still far less than what I expect from 10W. Or 5W in that case. The blue LED is very bright, compareable to my white 10W LED.

    This is a test assembly to see if the heatsinks are good, what voltage is needed, how much light it produces.

    then I made another, with 4 red LEDs.

    I tried a LED constant current chip. It gets so hot the LED actually starts blinking. The coil isn't right, wrong material for 500 KHz. And it needs a PCB with via's, mounted on a heatsink. I have these chips here.

    The brightness did not increase much. So I run the two LEDs in series from 19.5V (electronic PSU not linear supply). These are 660nm red LEDs.

    There is not any current increase or variation. Effectively it is a 500mA constant current supply.

    Just better in any aspect. I don't really want to deal with 100V DC, even if I wired the transformer that way some months ago. It is a 500VA toroid, so maybe it can still give off the 44 to 46 volts that are needed.

    If I adjust a TL494 to some voltage, it will maintain it to 0.05 volts.
    There is also the option for current limit. All I need to do is to change a few wires.

    Why not parallel? On the Piranha PCB all the strings are parallel. Brightness is exactly the same. I will see how many 1000 hours they live.

    I also soldered 3mm LEDs, 2 different brands. One brand has trouble, some strings go off, turn on again. The other brand LEDs all remain on.

    I don't think these are regular red LEDs. The forward voltage must be 11V or 12V. When I connect two in series to 19.5V, they just turn on. So the forward voltage is higher than for blue LEDs.

    Failure to happen- I don't understand what you write. If it works like I want then it is not crap.

    So actually I have LED driving chips here, but not the right coils. And the chips are very small. I need larger chips.

    What do you expect, someone never used power LEDs before, but will know everything in advance: the right Watts, the right cooling, the right voltage regulation, the right way to make a lamp assembly?

    I don't know everything. Things will be tried, research will be done, and then what works best will be replicated as many times as needed.

    I still use a 10W LED wired to 12V inside my computer, using a diode to drop off voltage so the current is exactly 1A. No brightness change whatshowever. It is screwed to a small VGA cooling fan.

    I have no idea how many watts are possible for these cooling fans. But that is part of the research I am doing now. And the most important aspect!

    Actually the laptop PSU has an overload protection. If I short the wires, it will turn off for a second. So if I work the laptop PSU right on the limited (it's rated output power), I will have a protection against overcurrent. The LEDs effectively can not be destroyed in any circumstances.

    Like a current regulation, but one that never kicks in.
  10. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    The 3mm LEDs which have trouble actually seem to be almost home made. They were cheap, no brand information, and the epoxy formation is different from LED to LED.

    I made two strings with all the remains from that batch, mixed in a few green LEDs, and these strings never turn off.

    The seller is no longer active on eBay for two years, so I will never have these LEDs again. The other brand comes from a 1000pcs bulk bag, also seems to be rather 640nm.

    And these have no trouble with direct supply, 10 LEDs, no dropper diode. I will run the panel for some months, and will observe any detoriation or failure.

    But I am not sure if I should build a larger panel from 5mm LEDs only. I am thinking about it, but it is a lot of soldering work. Maybe it is cheaper than to use large power LEDs? I have many many questions, so if you have any ideas or comments, this is all interesting.

    If you know some good constant current chips, I may also buy a few samples and see how they work. What I have here are PT4115, for 1.2A, but not the right coils. So I would need 2A chips if I don't make a PCB for them. For the cooling sinks I need a PCB anyway to hold the LEDs in place.
  11. Pantaz

    New Member

    Nov 10, 2011
    As a "power" LED warms up, its light output decreases. (Data sheets should have a chart to show the temperature/lumens correlation.)

    Also, LED lumen output decreases over its lifetime. This deprecation can be accelerated significantly when the LED is driven above it's rated forward current.

    Another common cause of reduced output is excessive junction temperature. This can radically reduce lumen output in a very short time. Thermal issues can cause much grief even when driving the LED well within its electrical specs.

    Cree has a pretty good selection of literature on lumen maintenance and thermal management for its products, but much of the information is applicable universally.