Computer power supply good for LEDs ?

[cpf user]

After having used Chinese LED drivers with Chinese 3-watt stars for a while now, I'm now trying AC power adapters that convert 120/240 to 12vdc @ 600ma, in order to driver four 3-watt stars in series. Now, one of the stars always seems to die after about 5 to 7 days of operation. They are mounted on an aluminum plate which only reaches 115 F, which is actually lower temp than the LED drivers would reach.

Is it possible that the AC adapters are introducing spikes into the supply?

 

[LilKevin715]

When you mention computer power supplies I automatically think ATX PSU's, is that what you are referring to? Or a "wall wart" AC adaptor?

Have you measured the output of the power supply with a multimeter? Irreguardless of which one you are using when converting AC to DC there will always be some ripple introduced. If you are using a good power supply the amount of ripple introduced should be minimal and a non-factor.

 

[ianfield]

After having used Chinese LED drivers with Chinese 3-watt stars for a while now, I'm now trying AC power adapters that convert 120/240 to 12vdc @ 600ma, in order to driver four 3-watt stars in series. Now, one of the stars always seems to die after about 5 to 7 days of operation. They are mounted on an aluminum plate which only reaches 115 F, which is actually lower temp than the LED drivers would reach.

Is it possible that the AC adapters are introducing spikes into the supply?

Is the wall wart heavy or light? - if its heavy it has an old fashioned iron-cored transformer - unless it also has a 7812 regulator, its regulation will be pretty poor - that is the voltage will fall noticeably under load - it has significant internal resistance. You might be only just not getting away with it! If the wall wart is light, its more likely a switch mode with much better regulation - you're going to need an external current limiting resistor.

 

[DIWdiver]

Are you using a driver in addition to the 12V supply? Or have you measured the current going to the LEDs?

No power supply should generate significant spikes. Ripple, yes. Or they can let through spikes that come from the line. Or the voltage could be off, and you could be overdriving the LEDs.

We'll need more info in order to help you much more than making vague suggestions.

 

[cpf user]

The original LED drivers, and the power adapter (desktop style), are both switching. The LED drivers were rated 700ma, but the 600ma adapter makes more LED heat and LED brightness. The KillOwatt measures 8 watts from the wall.

As for a resistor, none of the ten Chinese suppliers said I would need one, but then I did not ask. They knew specifically though that it was for 4 stars in series. I did specifically request constant-current 600ma adapters, and this was a unique enough request that many of them had to special prepare the samples. Apparently power adapters are normally constant voltage.

An interesting pattern is emerging: In addition to the failure being about day 5 to 7, it is almost always the 3rd star in the series, as counted from neg to pos. This would seem to rule out the enapsulants, (electronic grade white silicone on the solder contacts, and the entire thing sealed in condensation-cure 2-part clear silicone) and heat (measures about 115 F on the aluminum plate).

 

[DIWdiver]

How are the stars attached to the plate? Sounds like you might have a thermal bonding issue.

 

[cpf user]

With these same thermal pads we've used for hundreds of others:
http://www.tglobaltechnology.com/LED-die-cuts/LP0001v01-Li2000-0.15.PDF

I also measure the LED temp itself before sealing in silicone, pointing the laser thermometer at it, and it's lower temp than the aluminum plate side is.

I know some of the units we did not use flux remover before sealing them; might there be a reaction of the flux with the silicone? The units that have failed have all been after sealing, never before (interestingly, the aluminum temp drops a few degrees after sealing, presumably because the silicone is conducting heat away)

 

[DIWdiver]

I'd be cautious about trusting the 'laser' thermometer unless you have proper surface prep. Stick a piece of black electrical tape on a bare aluminum plate, and the temperature of the tape reads much higher than the plate (if the plate is hot)!

 

[cpf user]

Actually the outside of the plate is painted satin black for cooling; so the temp is fairly correct. Plus, it's not too hot to touch. Neither are the actually LEDs.

Two more units went out today; exactly 5 days from encapsulation, within hours of each other. I'm thinking the tin-cured silicon, which releases alcohol during curing, is causing some type of reacting with the bare aluminum inside the compartment that kills the LEDs. When I open the compartment and tear the silicone, it still has a strong smell of uncured silicone. So I'm changing back to platinum-cured silicone, to see if it helps.

 

[DIWdiver]

Hmm...

Sounds like you have the thermal issues handled, so what's left?


I don't know about tin-cured vs. platinum-cured, but I do know that household silicones are corrosive while curing, and most made for electronics are not. But you are using electronic-grade stuff. Still, if you still have smell of uncured silicone after 5 days, something is not right. Any silicone that hasn't cured after 5 days isn't being used correctly. Could it be an interaction between the two silicones being used?


Have you verified that the LEDs are actually dead? Grasping at straws here.

 

[cpf user]

Yes I hand-tested the LEDs; either opens or shorts.

Since tin-cured silicone gives off alcohol as it cures, it might need more of an opening to vent; currently it only has a 1/4" pour hole which fills up. Am going to try filling 1/3 at a time, allowing a full day cure between each.

 

[cpf user]

I think I found the problem:
www.cree.com/products/pdf/xlamp_chemical_comp.pdf‎

 

[DIWdiver]

Uh, no.

This failure mechanism describes a reduction in output or change in spectrum. It does not include or explain shorted or open emitters. Keep looking.

 

[cpf user]

Yes it did say that failure could happen.