LED Troubleshooting

[karlmarx]

Hi I got two LED multichips and both have failed within days of each other after a year of flawless service. One of them is still kinda of running here's the circumstances of events.

The colors started flashing like crazy and then went completely blue. Now only the blue inner and outer rows are lit. The center Blue and the two white don't light anymore. The colors occasionally will flash some purpler coloring.

I already ordered new multichips, my question is can it be reversed, and how can I prevent it from happening in the future. I have a 65c temp cutoff switch, was it a result of overheating?

 

[mcgyvr]

LED = dead..
Assuming you just stuck a 65deg C switch somewhere on the heatsink the actual LED junction temp could be much higher..

 

[karlmarx]

LED = dead..
Assuming you just stuck a 65deg C switch somewhere on the heatsink the actual LED junction temp could be much higher..

 

[mcgyvr]

even with it mounted there the temps can still be considerably higher at the junction..
add the fact that these have parallel strings and thermal runaway its a high possibility..

 

[karlmarx]

even with it mounted there the temps can still be considerably higher at the junction..
add the fact that these have parallel strings and thermal runaway its a high possibility..

Do the LEDs get more sensative over time to temperature, as the temperature outside heats up the AC has a bit of trouble keeping up. On both the occasions I found the room to be a bit warm, say maybe 74-76 F instead of the usual 71-72F. Normally when the room gets hot the lamps will just shut down for a few minutes and then turn back on.

I'm wondering if in the future I should replace the thermal cutoff with a 60 or 55C temp cutoff for bulbs.

 

[mcgyvr]

In general.. the LEDs will reach a "constant temp" in as little as 30 minutes to 2+ hours..
After that the temperature of everything will increase and decrease as the ambient temp fluxuates..
temp of the junction will be
current temp + (delta of the ambient)
so if the ambient goes up 2 deg all the measured temps follow..

so if your lights were switching off when it got to 74F.. that would say the sensor got to 65c.. so when the temps go down to 71 the sensor would now be just below its setpoint and the lights would come on..
So the temp on that plate was approx 60-65 deg C.. Junctions could easily be 40C above that..
Running LED's at what could be close to 100 deg C WOULD shorten their lifespan greatly..

 

[oreo57]

Normally when the room gets hot the lamps will just shut down for a few minutes and then turn back on.

Well that shows a problem right there.. What current are you driving them at?

 

[zachts]

Well that shows a problem right there.. What current are you driving them at?

Running way too much current. Likely your running them at the max current the vendor said they could handle. Bad Idea!!!!!! unless of course they are running in a freezer.

Buy two more so you have 4 and at the very least run two chips in parallel (or two stirngs) off the same driver(s) as the two you just burned up. Better still cut the current in half and get two more drivers as well as lower temp cut off switches. Those things should really never get anywhere near 65C or they are doomed!!!! 55C at the heatsink is probably pushing the limit even. You should be able to touch the heatsink without it burning you or being uncomfortably hot. Especially with those large multi chip arrays. The internal temp isgoing to be way, way, higher than the temp on the heatsink.

You should be able to run them with ambient of over 90F with no issue.........

 

[karlmarx]

Running way too much current. Likely your running them at the max current the vendor said they could handle. Bad Idea!!!!!! unless of course they are running in a freezer.

Buy two more so you have 4 and at the very least run two chips in parallel (or two stirngs) off the same driver(s) as the two you just burned up. Better still cut the current in half and get two more drivers as well as lower temp cut off switches. Those things should really never get anywhere near 65C or they are doomed!!!! 55C at the heatsink is probably pushing the limit even. You should be able to touch the heatsink without it burning you or being uncomfortably hot. Especially with those large multi chip arrays. The internal temp isgoing to be way, way, higher than the temp on the heatsink.

You should be able to run them with ambient of over 90F with no issue.........

This looks identical to my driver:
http://www.dx.com/p/50w-led-driver-circuit-input-85-265v-output-22-26v-2300ma-42805#.U0ohhPldXD4

so according to this it's, 22-26V, 2300mA. That does seem like it's very high current, would it be ok to run two 50w led chips with this driver? I know I should get meanwell's but that would probably be part of a bigger project.

 

[oreo57]

This looks identical to my driver:
http://www.dx.com/p/50w-led-driver-circuit-input-85-265v-output-22-26v-2300ma-42805#.U0ohhPldXD4

so according to this it's, 22-26V, 2300mA. That does seem like it's very high current, would it be ok to run two 50w led chips with this driver? I know I should get meanwell's but that would probably be part of a bigger project.

In parallel....not series.. Better yet 3.......
2300/3
Well lets get real.. Common "specs"

For the above chip you want a constant current driver AT 1700mA MAX... Then you need a PS w/ 34V plus drive loss Volts out.. Say 36-38V

Not sure of your exact chip nor exact driver or exact power supply (I assume AC/DC driver so PS is not present) but yes you seem to have a mismatch

As my own general recommendation.. never run them at rated max current unless you have SUPER DUPER COOLING..
Even then it is questionable w/ such large clusters of LEDs in one chip.. It's own internal heat management is questionable..regardless of external heatsink/fans
Applies always to cheap chinese chips.. rule of thumb to good chips..

LEd's really get more efficient at lower currents (L/W).. You can make up for the gross loss w/ more chips.. Running chips full out is a bit counter productive (heat degradation/loss of efficiency)

 

[karlmarx]

OK I have completely revamped the LED system with new multichips and new Meanwell drivers on a 5up board.

I have a couple of questions about running the LEDs in parallel. I have gotten them to run in series but they were too dim IMO. So I tried to use the calculator here: http://led.linear1.org/led.wiz

I entered the values:
48v source
36v FV
500mA
2x leds

However the first resistor quickly burned up. The resistors were attached to the negative leads of the multichips. I probably need a bigger resistor but I could also be doing everything wrong. If anyone could offer some help it would be much appreciated.

 

[zachts]

OK I have completely revamped the LED system with new multichips and new Meanwell drivers on a 5up board.

I have a couple of questions about running the LEDs in parallel. I have gotten them to run in series but they were too dim IMO. So I tried to use the calculator here: http://led.linear1.org/led.wiz

I entered the values:
48v source
36v FV
500mA
2x leds

However the first resistor quickly burned up. The resistors were attached to the negative leads of the multichips. I probably need a bigger resistor but I could also be doing everything wrong. If anyone could offer some help it would be much appreciated.

You need a 10 to 20 watt resistor, preferably an adjustable one to run things this way. Audio resistors might be available that big in the correct value at an affordable price but you'll not pay much more to buy an LDD to drive the multichip leds much more effectively. At those specs a resistor is going to be wasting about 7 watts each and get darn hot as you already know.

Edit, I think your already using an LED driver looking back, it is just too powerfull, you need to wire two LEDs in parallel not in series to reduce the current to each chip off your drivers.

Resistors won't work well if you are already using a constant current led driver to run them.

 

[karlmarx]

You need a 10 to 20 watt resistor, preferably an adjustable one to run things this way. Audio resistors might be available that big in the correct value at an affordable price but you'll not pay much more to buy an LDD to drive the multichip leds much more effectively. At those specs a resistor is going to be wasting about 7 watts each and get darn hot as you already know.

Edit, I think your already using an LED driver looking back, it is just too powerfull, you need to wire two LEDs in parallel not in series to reduce the current to each chip off your drivers.

Resistors won't work well if you are already using a constant current led driver to run them.

Im running this driver:

The chips won't both run in parallel. Only the last chip lights in the chain.

 

[oreo57]

Im running this driver:

The chips won't both run in parallel. Only the last chip lights in the chain.

36V 3400mA ... that is what you'd need in a driver/ps to run those 2 in parallel.. if they spec'd as I assumed above..

1 1000mA driver per chip w/ your ps should give you bright and stable light..

Those chips are composed of 10 led in series w/ 5 rows.. and your doubling that..

The resistor in the calculations assume constant voltage... which you won't
have, nor in this case want...........

Bottom line your driver is WAY under the ability to supply enough current for the 2 chips together...

As a test just power one w/ the LDD ..see what it looks like..
It should light and be fairly bright..

 

[zachts]

36V 3400mA ... that is what you'd need in a driver/ps to run those 2 in parallel.. if they spec'd as I assumed above..

1 1000mA driver per chip w/ your ps should give you bright and stable light..

Those chips are composed of 10 led in series w/ 5 rows.. and your doubling that..

The resistor in the calculations assume constant voltage... which you won't
have, nor in this case want...........

Bottom line your driver is WAY under the ability to supply enough current for the 2 chips together...

As a test just power one w/ the LDD ..see what it looks like..
It should light and be fairly bright..

It sounds like karlmarx was trying to run the two chips in series not in parallel. Even on a 500mA driver the two chips should light up quite bright if the LDD is connected to the 48v power supply listed previously. If you did have them wired in parallel and they did not both light up then there are some wiring faults somewhere that need corrected.

If they were wired in series though and do indeed spec out to have Vf of 36v then two in series would need to have a 72v power supply and that would exceed the LDDs input/output voltage capability. Likely with two chips in series it was trying to illuminate them but only able to pass a few mA at the available 48 volts with them connected in series and the one that illuminated likely had a slightly lower Vf.

Vf = forward voltage across the LED

 

[zachts]

Im running this driver:

The chips won't both run in parallel. Only the last chip lights in the chain.

Just for Clarification See below for "Series" and "Parallel" wiring. I think you have it misinterpreted.

 

[oreo57]

and the one that illuminated likely had a slightly lower Vf.

I've had a similar thing happen w/ LDD's when trying to drive single LED strings in parallel where V(f) is different..

In electronics, a constant current system is a system that can vary the voltage across an electronic circuit to maintain a constant electric current

Once the v(f) is satisfied for one chip and IF that v at that mA is lower than the V(f) of the next chip.. it doesn't light the other chip..

AFAICT..

to my understanding that is why only one lights at low mA.. it produces a low V(f).. at least from the standpoint of one chip..

could be wrong, and I know it is a bit more complicated than this.. but I "think" it generally applies..

I=r/v.... so a driver like the LDD s nothing more than a precision self adjusting voltage regulator..

At least that is the definition that makes me happy..

Driving multiple LED light sources with one driver is generally done with the LEDs arranged in series strings to avoid uneven light levels resulting from voltage variations.

very uneven...

 

[zachts]

I've had a similar thing happen w/ LDD's when trying to drive single LED strings in parallel where V(f) is different..........very uneven...

Once the v(f) is satisfied for one chip and IF that v at that mA is lower than the V(f) of the next chip.. it doesn't light the other chip......

Pretty much.

Must be a substantial difference in Vf then for only one to light on a LDD in parallel. Never had issues running strings of single LEDs in parallel, though they can be off by several hundred mA in some cases so that may be the issue here, just not enough total current being supplied to over come the imbalance in Vf and light both chips. If you upped things to an LDD1000 then both probably will light up but one will be substantially brighter.

Each could be measured ran by itself to determine if that is the case and if they are very different you could either mix and match the ones you have to balance them out, or you could use a very basic current mirror to correct (err, work around) the problem if running them in parallel is still the desired option. However one LDD per multichip LED is probably the best choice.

There is a process termed "balancing" that is discussed on several of the older parallel LED build threads that was used to "fix" this imbalance problem. But it's darn labor intensive on large arrays as many of those old builds were. (also not very helpful for multi chip builds unless you have many chips to work with and pair up those with matching Vf) Thus it has largely been abandoned for running one driver per LED string, which is simpler for most DIYers to accomplish.

 

[zachts]

FWIW:

http://www.reefcentral.com/forums/showthread.php?t=2097086&highlight=current+mirror

A pair of BD139 on a very small heatsink works pretty well for total driver currents up to at least 1.3 amps and voltage at or below 48v (been running some for about 2yrs now this way on an ELN-60 to run 2 or 3 string arrays)

 

[oreo57]

Must be a substantial difference in Vf then for only one to light on a LDD in parallel. Never had issues running strings of single LEDs in parallel, though they can be off by several hundred mA in some cases so that may be the issue here, just not enough total current to light them both due to the imbalance.

Each could be measured ran by itself to determine if that is the case and if they are very different you could either mix and match the ones you have to balance them out, or you could use a very basic current mirror to correct (err, work around) the problem if running them in parallel is still the desired option. However one LDD per multichip LED is probably the best choice.

In my case V(f) was substantially different.. (different colors w/ much diff V(f) but equal chip count)
IF my thinking is correct.. a higher mA driver would solve that problem since those chip differentials would not be as extreme as in my case.. and the voltage "boost" @ higher mA would overcome it ..

However one LDD per multichip LED is probably the best choice.

agreed...
Each multichip is composed strings in parallel, w/ different characteristics per LED per string internally.. Matching them is a bit problematic..at least at borderline supplies/currents (both upper and lower).

Then again it could just be a bad chip ect.. Which is why I suggested testing each chip w/ the LDD separately. Then work on the series/parallel problem..
I can't find any complete specs on multichips just general info..
but considering the generality of LEDs..This tells me a few things..

As to my understanding trying to run a red and a blue in parallel wouldn't work since the voltage @ mA "latch" wouldn't permit a v(f) high enough to light a blue (say the driver is 250mA constant current) @ 25C ..
and of course the lower the mA the duller.. I assume you could extrapolate to close to zero???

sorry, just babbling..