Case Study: Incredibly short life cycle of LEDs, requesting advice

[billbyrd1945]

I bought these cheap flashlights from Amazon (a ten pack for $19 (with batteries)). I bought them purposely so I could use their innards to make a 3d printed light source that would be powered by USB. The 3-pack of alkaline batteries measures 4.9v. My USB adapter outputs 4.8 volts. The flashlights have an array of 9 LEDs in a circle with one in the middle. After only about 48 hrs of use, all but two LEDs have burned out. Are these simply inferior LEDs, or, is there something about my design that's killing them? What can I do to extend the usability of my light source? Do I need a resistor? What value? I appreciate any help.

[TimFox]

If your USB voltage source is "stiff" (relatively low source resistance), there is nothing to limit the current in the LEDs, which will increase with temperature at a constant applied voltage.
The internal resistance of alkaline batteries is perhaps sufficient to not kill the LEDs, but is not a particularly elegant design.
See  https://www.rohm.com/electronics-basics/leds/led-characteristics  They state that the forward voltage drops with increasing temperature at a constant current:  that means that the current at constant voltage increases.  Normally, LEDs are never driven from a constant voltage.

[james_s]

A lot of these very cheap flashlights do rely on the internal resistance of the batteries to limit the current. The solution is to buy better quality flashlights that use a single LED and a proper driver circuit.

[mikerj]

They are simply garbage. I've had a had a couple of the same design flashlights (though aluminium case) and the LEDs are low quality and very overdriven, both of mine had several failed LEDs after very little use.  Even whilst the LEDs were working the grounding between the PCB and the aluminium case was very poor and caused flickering on mine.

[BrokenYugo]

A lot of these very cheap flashlights do rely on the internal resistance of the batteries to limit the current. The solution is to buy better quality flashlights that use a single LED and a proper driver circuit.

This, even if there is a resistor it will be sized with the droopy, probably carbon-zinc battery power in mind.

I'll hazard a guess that these cheap flashlights are the same blueish white LEDs that you buy from China for like $0.01/each in a bag of 100 or 1000 or whatever, probably rejects/seconds, unsuitable for anything serious.

[billbyrd1945]

Everyone seems to agree that it's a quality issue. If anyone can link me up to a proper source, I'd appreciate it. Thanks for all the responses.

[tunk]

Everyone seems to agree that it's a quality issue.

Or that you're driving them to hard. They could look like standard 5mm
LEDs with a maximum current of 20mA. It would have been interesting to
see how much they drew when all 9 were working. You could add one or
two diodes in series to lower the voltage.

[langwadt]

Everyone seems to agree that it's a quality issue.

Or that you're driving them to hard. They could look like standard 5mm
LEDs with a maximum current of 20mA. It would have been interesting to
see how much they drew when all 9 were working. You could add one or
two diodes in series to lower the voltage.

no use a resistor, try something like 20R

[amyk]

How much current were you pushing through them?

[thm_w]

Everyone seems to agree that it's a quality issue. If anyone can link me up to a proper source, I'd appreciate it. Thanks for all the responses.

https://www.digikey.com
You can get LED tape by the foot, LED stars, or round COB LED plates. Although they tend to be 9-12V (3 or 4 LEDs in series). If its on a 3D printer you have access to 12V though.

Personally I like the aliexpress COB stuff, some is good some is bad:
Comes with resistor, will get a bit warm: https://www.aliexpress.com/item/1005001424797203.html
Needs a external resistor: https://www.aliexpress.com/item/32910848200.html https://www.aliexpress.com/item/4001227120671.html

[billbyrd1945]

Ah! That's what I was hoping to hear. Not only is it a possible solution, but it also teaches me something about electricity. I didn't know you could overdrive something as long as the voltage was the same. I have 9 more flashlights, so I can certainly give your suggestion a try. And I have enough knowledge and equipment (I think) to report back with values. Thank you. On another subject, can you tell me how to be notified by email when someone responds to my posts. I've looked at 'account' and 'profile' and can't find an option for it.

[billbyrd1945]

Tunk suggested diodes, you suggest a resistor. That gives me two things to try. Thank you!

[billbyrd1945]

I'm about to find out. Will report back.

[billbyrd1945]

Thanks for the links.

[billbyrd1945]

OK guys. This is what I learned: With no modifications and using a USB wall adapter as power supply (4.8v), very bright light with .297 amps of draw.

Putting in one 20r resistor changed draw to .077amps with significant reduction in brightness of the flashlight.

Putting one green 5mm LED in series (no resistor) with a new flashlight (9 LEDS) lit the LED and the flashlight LEDs but almost totally killed off the light coming from the flashlight. Draw was reduced to .54mA.

So, I'm guessing from this that I can use the resistor method to throttle back the current, but I don't know how far to go. Are there any rules of thumb for this so I can proceed with my project with confidence that the LEDs will hold up for, say, a year or so? If not, I guess I'll just let them burn for a few days on the bench with different resistors to see how they hold up. This was fun for this old man. Thanks!

[amyk]

OK guys. This is what I learned: With no modifications and using a USB wall adapter as power supply (4.8v), very bright light with .297 amps of draw.

If they're all in parallel, that's 33mA per LED, which is way over the 20mA limit that ones of that size are typically rated for. I would guess that they are all in parallel, given the number of them (and the other possible arrangements, 3 x 3 series or all 9 in series, wouldn't light from 5V) but you would do better to dissect the light and find out what the actual arrangement is.

[jonovid]

Resistance Wheel & a DMM adjust for minimum current at maximum perceived brightness

[billbyrd1945]

"If they're all in parallel, that's 33mA per LED, which is way over the 20mA limit that ones of that size are typically rated for. "

Yes, you're right. No matter which two pads I probe with my meter (in diode setting), it illuminates the only remaining functional LED.

[langwadt]

"If they're all in parallel, that's 33mA per LED, which is way over the 20mA limit that ones of that size are typically rated for. "

Yes, you're right. No matter which two pads I probe with my meter (in diode setting), it illuminates the only remaining functional LED.

looks like it actually has a resistor, I have similar light and it doesn't have resistor

[Siwastaja]

It still relies on the combined resistance of that resistor and the internal resistance of the battery cells. Even just "too good" batteries kill such a design. This idea works best on single-use key fob lights which come with button cells that can't be replaced by the user (and are small and crappy enough to have significant resistance).

33mA alone isn't problem, the problem is lack of current regulation which combined with the positive thermal coefficient of LED forward voltage causes thermal runaway: high current heats up the LED, which lowers the voltage drop over the LED, which, given very little resistance, increases the current through the LED, and the cycle continues until destruction. The simple solution is to experimentally find amount of resistance which gives safe current level even after testing it for extended time (say half an hour for thermal equilibrium), at highest possible ambient temperature. 10-15mA per LED would be on the safe side, because having them in parallel without per-LED resistors is also risky and worth extra derating.

[Gyro]

"If they're all in parallel, that's 33mA per LED, which is way over the 20mA limit that ones of that size are typically rated for. "

Yes, you're right. No matter which two pads I probe with my meter (in diode setting), it illuminates the only remaining functional LED.

Can you show the other side of the PCB? Are the LEDs mounted flush to the board?

[billbyrd1945]

Yes. It does.

[billbyrd1945]

What a fool I've been. All I ever needed to do was to figure nine LEDs at 20mA and then come up with my resistor using Ohm's law. Or am I still lost in the weeds? One thing's for sure, using various resistors (currently 10r) is doing the job. Just a matter now of finding the optimal one with overnight burnout testing.

[BrokenYugo]

You don't necessarily need 20mA, you just need enough to make however much light you want. 20mA would be the maximum, for an induvial, typical good LED, that comes with a datasheet with lifetime and current specifications.

10 ohms and 4.8V in is 20mA per LED (simple ohm's law problem). Except putting diodes in parallel like that and expecting them to split the current evenly rarely works out in reality, especially if they're not thermally connected, since forward voltage drops with heat and every diode is a little different, one ends up in a thermal runaway condition, taking the bulk of the current available, and then it blows and next lowest Vf diode starts taking the bulk of the load, until there are none left, as you've witnessed. Works alright for a throwaway flashlight that's expected to run maybe 100 hours at most though.

So more realistically, with this crude circuit, you want more like 100 ohms (20mA total), or perhaps even lower current/bigger resistor, you'll have to experiment as these are unspecified LEDs. 20mA is just a rule of thumb and the mystery parts in question may actually handle considerably less if you need them to go for years.

A better way to your goal might be to get some of the "1 watt" LEDs offered on ebay/ali pre mounted on a small circular PCB, and size a resistor for that, keeping in mind they probably need more than the included bit of PCB for a heatsink to run them anywhere near 1W, which is likely exaggerated (usually pretty safe to assume Chinese amps/watts are half size), but you could probably drive one at least as bright as those flashlights without issue, and you get some choice in color temperature as a bonus.