LED-Array, portable: impossible to drive?

[krutzbeuazen]

Hi!

(still) i plan to build a lamp/flashlight with many LEDs.
Since they are getting cheap, i thought about 100 or perhaps more LEDs. why not luxeon? would like 395nm N-UV.

okay.. how to power them.. nimh-cells seem to power them easily, up to 4A per AA cell. 4 or 8 of them should work nicely.

since the voltage drop on such high load will be quite strong, i need some kind of regulation (instead of resistors).

step-up-converter chips or whole circuits seem to be impossible to get over here in germany.

a voltage regulator would be really hard to build for 2, 3 or more Amp and would dim the output after minutes (as far as i know).

a current regulator would be possible to build, with a FET. nicest way so far.
(a big thank you to burnt_retinas for his hints on FET!)
problem with them is, if i hook several LEDs parallel to a constant current supply, some LEDs have slightly different specs, drain more power, heat up slightly more, lower their resistance, drain more power, burn. then the other LEDs get all the current for less LEDs, same thing happens, chainreaction, catastrophic failure of 100% of the LEDs in milliseconds. i guess the last few LEDs will almost explode, at 2A..


what to do? cant be that impossible to drive 100 LEDs with cells. portable. not necessarily in a nice small housing, but portable.

krutz, frustrated

 

[Harrkev]

That is why you use a resistor in series with each LED chain. Even a LED of only a few ohms should help, and not waste too much in efficienty.

You can also have one linear regulator for each LED string or two. This would also mean that each regulator would only carry a little current, so you distribute your heat over several regulators. The LM317 in current mode would be perfect for this.

Just a thought.

 

[krutzbeuazen]

about what regulation do you speak, current or voltage?

would it make any sense to use a constant-current-supply with all LEDs parallel and a small resistor for each LED which makes problems?

imho it wouldnt help anything, since its the too high current, or better, the unregular current "distribution" to the single LEDs?
i would need a little bit more current out of my FET, and the voltage wouldnt change at all, since its no voltage regulation?

just my guess, could anyone say something about what really would happen?

thanks!

krutz

 

[LED_ASAP]

If you don't over-drive the LED in the design, you probably don't need to worry too much about them exploding. LEDs can take some serious abuse---many successful commercial lights over-drive them at 60mA. At that current level the Vf is closer to 4V, instead of the 3.6V at rated 20mA. You probably won't see such a Vf difference (0.4V) from a single batch of LEDs anyway. So, here is a simple solution: design the array to drive each LED at 20mA. In reality, some will get higher current like 40mA, some will get less current like 10mA. But those are still within the safety margin. One word of precaution, though---in large LED arrays the LEDs in the middle can get much warmer than those near the outter circle. Make sure there is sufficient heatsinking for the central guys or you will really have thermo runaways.

 

[MrAl]

Hello there,

One idea is to divide the 100 up into sets of 20 each,
so you have five sets of 20 LEDs. You can check
their current and voltage with a set current
(whatever you intend to run them at normally) and
try to form four or five catagories of voltage range.
This gives you four or five sets of 20 leds that within each
set the measured voltage only varies by a small
percentage of the nominal operating voltage.

The best test would be to connect the 20 from each group
to twenty 1 ohm resistors, one resistor for each LED,
then apply 100ma and make sure one led isnt getting
too much current. Then increase current to 200ma,
then 300ma, finally 400ma (for 20ma normal per led).
If none of the leds currents are too high (say 20 percent
greater then normal) then that group will probably run
ok. In the actual application, you can regulate the
current to that group alone.
If you are really feeling paranoid, stick the 20 in
the freezer for 20 mins and test each one again, then
stick in the oven set for say 50 degrees C and test again.
If no resistor gets more then about 30% more current
then the normal for one led you're ok :-)

Although it's usually better to keep some series resistance,
the 1 ohm resistors are there so you can measure the
current through each LED, and can be removed for
the actual application as long as the group of 20
that were tested together STAY together.


Take care,
Al

 

[HarryN]

This is kind of a wild idea, but since you are desperate - maybe it is interesting. It might be possible to drive them with a boosted voltage - say 48 volts in a pulsed DC arrangement. Input a pulsed DC from your cells into a step up transformer, out comes pulsed 48 volts.

Nominally, 48 volts / 4 volts = 12 leds in a series. For a cheap regulator for each string, add a small resistor. I am not sure but you might need a diode on each string as well to manage reverse current.

If you are into more sophisticated electronics, you might find a PWM current regulator, but maybe 8 x current regulators are expensive.

48 volts or similar is a good target, as there is a fair amount of telecom gear in this voltage range, and it still falls into the low voltage catagory.

I think that the shortest wavelength Luxeon offered is around 450nm with some distribution, maybe 430nm - need to check their site again, so you will probably have to use a 5mm based LED if you really want to go below 400nm.

Please consider eye protection as you are really getting into some higher energy wavelengths here.

I am not an EE, but this seems like a potential, but perhaps somewhat crude approach.

Good Luck

 

[krutzbeuazen]

Hi HarryN,

thanks for your idea!
actually this sounds like the savest way to drive many LEDs.
even if something goes horribly wrong, all that could happen is that one LED of each string burns, then the rest would be cut off.
i´m not sure how small these step-up converters can be built, but with that high current (almost 10w) i even had problems (= failed) finding a converter to output just 4V from 2,4V input.
one converter, then several strings of LEDs, each on one FET. sounds like the best way, definitely.

well, i planned differently now, will drive them in groups of 20, all parallel, on one FET for each group. a one R Resistor for each LED, and a fuse for 10.
between the FET circuit i´ll put a PWM, but i´m really not sure if my (nicd) cells can handle the "pulsed" drain.

parts are ordered, i´ll open a new thread when i can show results, be it positive or negative ;-)

thanks again for your help!

manuel

 

[Burnt_Retinas]

Personally, if I wanted to go this route, I'd go a voltage regulator with sufficient current and have strings of LED's hanging off that, each string with it's own resistor to balance the current through each string. The voltage regulator is to set the voltage just above the voltage needs of the LED strings so the resistors to each string can be minimal resistance and therefore result in minimal loss in efficiency.

Chris

 

[gibbled]

Do LEDs usualy fail open or short?

I haven't smoked one for a while and don't remember any trend.

 

[Ocelot]

LEDs usually fail open. That has been my experience over the years.

Scott

 

[Jack_Crow]

K,
Perhaps I need an education here.

Im diging up a memmory from nearly 20 years ago.
Let's set up a test stand and check the LED's.

If you run a fixed voltage, let's say regulated to .005% and carefully measure the current, you will work out a number for each LED's internal voltage drop.

Unfortunatly that will take long string DMM's like the HP 3456A or Fluke 8505A to get reliable numbers. Here is why.

For a hobbie project, measuring the voltage drop across each led will tell you if it's working should be enough. If the numbers are within a percent or two, that should be enough.

Most comsumer DMM will give you 3 and 1/2 digits. Even less on some ranges. Plus the inability to read numbers in the LSB (Least Significat Bit). Remembers errors are always additive in instruments.

So if your trying to read a voltage drop .72 volts across a diode, that last digit of the reading can be off by as much as 5 or 6 counts plus the error value of the instrument. The meter it's self will be on the 1.9 volt (1900) count range and your value will be half way to maximum (and most accurate reading).

A fancy meter will give you a reading like .7224 volts with very little error issues. As my wife is prone to saying. "Too much information"

I suspect your trying too hard to make something simple.

Build it one string at a time. Or build a 'pratice' one with visible light LED. Look in the "National Semiconductor Linear Regulator" guide for CC and CV regulator chips you can mail order. Pick out ones that look promising.

Do a web search for the P/N's . ON the web you can find data on almost any chip made for free with a little diging.

I use www.dogpile.com as my search engine. Most people use Google.

I hope this makes some sense.

Be well
Jack Crow in Iraq