Ran into severe LED problems with my headlight cluster. Please help!

First off, I am working on a LED headlight replacement for my friends truck. I cover the project here:
https://www.candlepowerforums.com/threads/117661

I just got the LEDs today (36 TWOM K2's) along withg the optics. I could not resist getting a mock up ready for just one headlight. I have 18 of the K2's hooked up in 6 parallel banks of 3 series LEDs. This would work well. I was impatient and I was carefully driving the whole cluster with a little battery pack to see how it would light up. Everything looked awesome, untill a couple banks of LEDs turned off! Eeek!

I just discovered that the slugs on the bottom is very well connected to the negative leads! I never had problems making small clusters with LuxI's and LuxIII's using the same aluminum heatsink. I can hook up a continuity tester (built into my multimeter) to the slug and negative and it will beep at me for a K2, and not the older LED's (which I already knew about and dismissed a while back). I didnt know the K2's were built that much differently.

I hooked the whole cluster to the battery pack (two old standard 6V lantern batteries) and I made sure the current didnt go up too high (I measured current using the multimeter). It should run at 12.0V to 12.7V @9A max (1.5A x 6). It read about 2.5A or so with this pack.

I wanted to try the 9A a little to get these things going at full brightness. I used a lead acid jumper battery pack (for jump starting your car, of course). It was a small pack inside not rated for a whole lot of current. With this pack configuration, I got the whole cluster running at 7.3A (1.22A per LED bank) and the light was killer!

I was checking things out making sure the LED's made good thermal contact. I used thermal compound temporarily on an aluminum heatsink and I held them down with some strong tape. I have done things in the past this way with no problem. I wanted to save the two-part thermal epoxy for the nice copper heatsinks I need to build yet (waiting for the copper in the mail yet). None of the LED's really got hot. They were cool and I felt the giant heatsink warm up a little. All light was bright and uniform.

Anyway, when I had the LED's hooked to the lead acid battery (not fully charged), I had one LED stop working, turning off one bank. I freaked out a bit, but I found the issue with the one LED and the other two in series were fine. I had a spare of a slightly different tint K2 and I replaced that. It lit with no problem again. Bright! I have been getting pictures of the killer cluster and I decided to take it outside. It has no optics yet and it lights all the trees up outside for 50 or more feet. It lit up a flood with the same intensity all around as the beam of my 4XLuxIII 2DMAG mod with the 4 SO20 reflectors (really tight beam when untrimmed and used with joker Luxeon LED's)!!! Holy crap!

Here is some pics when indoors (my messy room). Here is a comparison of the ambient light (the Quad LuxeonIII flashlight shining up at the ceiling and my monitor) and with the K2 headlight cluster on a 7.3A. I had the same speed and amperature setting on my camera, so here:



As I was saying, I was outdoors and that is when the nail-biting crap happened. I saw after having it lit for about 10 seconds that another bank went out as another got dimmer! I quickly cut the power. The LEDs were still not hot or anything. I am left with a dead K2 in one bank with the others working fine and another bank with a dimm K2 with the others fine. The dim one was visibly damaged when you light it very dim. Part of the die doesnt light. Here is a picture of the cluster lit at a lesser intensity. You can see the problems clearly:



Okay, this is where I need help. How can I hook these up so this doesnt happen? Here is a diagram of how each light is wired:


I found when hooking the negative from a 6V battery pack to the negative of the cluster, then touch the positive of the battery pack to the alumunum heatsink, a certain few LEDs light up dimly. I believe that is a real problem when hooking the full voltage to the LED's for use in the car. Here is a small picture of how the few LEDs that lit up:

Notice how just the one LED at the end of each bank (the last one that is hooked to the main negative) are lit. The one labeled "1" was the one that was first replaced. LED "2" is currently not working but the others in the bank are.

How can I successfully isolate each LED (if I need to do this) while still providing good thermal contact? Would a thin sheet of anodized aluminim work? What about thinly painting a small spot underneath the slug of each LED? Would a thin sheet of mica work to isolate the LEDs? Hmmm.

Any help here would be appreciated. I am lost and this is holding me and the project back. Is there something I am missing? Thanks all. This is why I love CPF forums!

-Tony

im not a modder but I think i ran into this problem before...

Hooking up too many strands of christmas lights in a linear/parallel configuration...new bulbs but a few strands wouldn't light...

have you considered running a quantity of diodes on one power supply and another batch on another power supply?

Gryloc, idleprocess posted this in your project description thread:

"If you're using individual emitters, prepare to go to extraordinary measures to isolate each slug from your common heatsink"

And andrewwynn posted this in the same thread:

"If i designed this system i would put a zener diode of about 4.5V across each emitter in case of an 'open' failure.. "

Sorry for restating those posts but they seem to be applicable to your situation.

7.5 amps at 12 volts? Thats 90 watts, a quite bit more than a single halogen car lamp draws on low beam.

One assumption seems to be that the current will divide up so that 1/6 of it will go through each string of 3 LED's. This is not true (for a given voltage, the current will vary, Luxeon lottery).

And yes it appears you will have to electrically isolate the ground from your heatsink if you want to wire the circuit this way.

Not2bright, where did you get those quotes from (idleprocess and andrewwynn). I checked my own thread and I didnt see those replies. Thanks for fowarding them to me. Nice of you.

Cobb, I know. These things wont be as efficient as a nice, hot halogen headlight, unfortunately. Its in the name of science. Each headlight actually runs at 9A (full, but may be reduced to reach the 1500 lumen goal), so that is a max of of abut 120W . Each headlight will produce about 1800 lumens each, though. That is better than those nice aftermarket headlights (HID reaches the 2000 to 3000 lumen mark though at 40W or so). Its not terribly inefficient. This would be a better project in about 5 years when LEDs get brighter and better. K2's arent the most efficient, but they are bright (perfect for me).

Oh, as for brightness, these arent just low beams, these are high and low beams (check my thread). Most halogen headlights produce about 1000 lumens for low beams and 1500 lumens for your good 55-65W lights. This is per headlight, too. I should be okay.

Coherence, the LEDs I got are from the same batch. They are the same binned lights (all 36 of them). They have the same forward voltage (+/- 0.1V I am sure). In the parallel/series configuration, they work well when it comes to even amperage and brightness. When three of any of these matching LEDs them are in series, they should have the same Vf as the next bank of three. I was given the suggestion of using 0.1 ohm resistors in line with each bank to even the current. I will most likely use these.

So thanks for the responses. I went to Walmart tonight and found something when I went to the hardware and paints section. I needed electrical tape and I saw this new tape (new to me atleast) that is called "Aluminum Foil Tape". It was cheap and I thought about using this to isolate the slugs of the LED's. I was looking at the stuff and it uses thick gauge aluminum (your heavy duty foil like used on the grill) and a unique adhesive that is strong and not too thick. Perfect. I may add 1-3 degrees C per watt if resistance, but they seem to conduct heat pretty well. I stuck it to my hand and was touching objects cooled by the ouside air (lower 70's) and I felt the coolness immediatly after touching the object. I felt it quicker than with my bare hand. This doesnt tell me much and it isnt scientific, but It will work for me! I will cut little squares and stick them to the heatsink under each LED base. I will then check the continuity between each aluminum pad and the heatsink to be safe. I think this should work, hopefully.

About the zener diodes as suggested by andrewwynn, will that work? I dont understand the properties of zener diodes and I dont know how they will protect each LED that may short out or burn out. Can normal rectifying dioes work? Just throwing out ideas.

Alright, So even if I have each LED isolated, is there anything else I need to worry about?

I was also wondering, each LED has a forward voltage of X, and if you attach a higher voltage supply to it, the LED will just soak up more power and the current used will increase. So X is sort of a minimum, right? It regulates it's own voltage almost. Any less and the LED brightness drops, along with how much current it will use. I know this isnt exactly very technical, but this is just from my experience so far. If I have batteries that are low in a battery pack when testing an LED, I can usually add another cell in series to increase the power going to it. I usually do this until the appropriate current is reached. Thing is, how many volts over this forward voltage per LED can you go safely before you start doing some damage? I was wondering because I dont want to accidently add too large of a source voltage and fry these jewels.

Thanks again for your posts!

-Tony

how many volts over this forward voltage per LED can you go safely before you start doing some damage?

There is no one answer to this question. The Vf from the manufacturer is a 'nominal' value, that is an average part at Vf the LED will consume say 1.0 watt.
At Vf+2.0volts, maybe it will consume 2.0 watts, maybe 1.5 watts, or some other number. It depends on the individual part.

When the car is running, the alternator will raise the elec. system voltage above 13 or 14 volts, etc. to charge the battery, so you do not have a fixed voltage supply anyway.

To control the power used, you MUST control the current. LED's do not obey ohm's law, think in terms of varying the current, not the voltage.

The easiest way to do this is to wire all the emitters in series, then they will all get the same current through them. But if one unit fails as an open circuit, then they all go dark.

Another way is to drive each emitter individually, then your problem of the heat sink connected to negative terminal is eliminated. Just tie the heat sink to ground and all the neg. terminals to the heat sink.

one of the ways to isolate the slugs is to have a anodized aluminium heatsink.

the layer will very well isolate the slug at they are not electrical conductive.

thermal conductivity will stay.

not sure if the thermal conductivity will be affect in anyway.check with other members who know more.

Gryloc said:

Coherence, the LEDs I got are from the same batch. They are the same binned lights (all 36 of them). They have the same forward voltage (+/- 0.1V I am sure). In the parallel/series configuration, they work well when it comes to even amperage and brightness. When three of any of these matching LEDs them are in series, they should have the same Vf as the next bank of three. I was given the suggestion of using 0.1 ohm resistors in line with each bank to even the current. I will most likely use these.

Alright, So even if I have each LED isolated, is there anything else I need to worry about?

-Tony

Click to expand...

Your LEDs will not stay matched, with time, their Vf will change. Also, the Vf drops with temperature, causing that string to continue to pull more current, which makes it heat up even more, which causes the Vf to drop further, so on and etc. As the temperature changes, the Vf ratio's between the LEDs will certainly change. They are *NOT* on the same slope.

It is not uncommon to see one string pulling easily 2x the current of another string, and it can get much worse with time, heat, and age.

You will find, over the first month, some of the LEDs will drop by 0.3V Vf, where other ones will not, in the same batch. Search for posts by evan9162, he documents and characterizes this.

This is the whole reason why folks utilize constant current sources with LEDs. With high power LEDs, it becomes even more essential.

Electrical isolation. Use a razorblade to make a somewhat transparent (partially see thru) layer of thermal epoxy where you intend to mount your LEDs. Let it cure. Somewhat low thermal resistance insulator. Watch out with your aluminum tape, it typically has an issue with shorting at the edges, especially over time.

Great points there Newbie....as usual.

:wave:

Thanks Coherence. Well, the reason I wanted to know was because I was hooking it to a fully charged 13V lead acid battery and the LED forward voltage is anywhere from 12.0V to 12.7V (going by 3 of the M-bin Vf LEDs -still need to test the true Vf with my meter). This is only temporary and for bench testing.

When this cluster is used in the truck, it will be using Andrewwynn's Hotdriver. I still dont have it, so I am just direct driving. It will be regulated for sure, so those funky 15V surges from the starter motor or whatever doesnt do any damage. This regulator will also has current limiting, so the LED's wont get over 1.5A of current.

Oh, the heatsink I have isnt anodized. All I have to connect the LEDs with is some Arctic Alumina ceramic thermal paste inbetween. I dont know if that isolates the LEDs or not. I will transfer them later to a nice copper heatsink. The heatsink will be raw copper, so I will have to find a way to isolate them again. This tape I am using is only reated for 150 degrees F, so it will fail in the real world. Plus, epoxie will hold the LED to the tape, but the tape adhesive will melt and it will fall off.

How do I isolate the LEDs with copper? Is there any material that can coat the heatsink permanantly while the LED can still be epoxied? Is there any good two part thermal epoxy that is not electrically conductive? I have some "Arctic Alumina Thermal Adhesive". Will this stuff work? Im afraid I may squeeze too much of the adhesive out from between the slug and HS when I put pressure on the emitters, so I would get electrical contact. I dont know too well. I used the stuff with my previous projects, but the slug is already isolated from the negative with the LuxIIIs I had (I believe).

Thanks...

-Tony

gryloc.. the AA paste is 'iffy' regarding isolation.. i'm betting you have been getting some conduction paths to ground through the paste.. fortunately when you really use the AA adhesive it is MUCH better at electrical isolation!

For short-term tests you can get away with using thin tape to isolate the alm .. just make sure you still use the AA paste.

You can use the AA thermal adhesive to bond with electrical isolation from the copper heat sink.. that's exactly it's job!

Make sure the feet are bent up away from the HS of course!

You should put a zener diode from Bat+ to Bat - on the hotdriver of maybe 14-15V to make sure there are no nasty spikes that get to the ckt from the ignition etc.

Oh.. since we have discovered that you have more voltage overhead than we thought.. absolutely put a 0.1ohm resistor in series with each string of emitters.. This is the best way to get balanced output between a bunch of strings.. if one string tries to have more current (because of the combined Vf is lower).. it will have more Vdrop on that resistor.. it will help the balance.

Also.. match your Vfs of each emitter.. 'microbin'.. what i do for this is to hook them all up in series and crank up the bench supply 'til they start to light.. they will come on in groups and i can match them.. if i have for example.. 6 that light at 2.9V.. 6 that light at 3.0V and 6 that light at 3.1V.. i can put one of each in each string and get balanced total Vf.

Since they are series wired.. they will all get the same current in the end and everything balances out nicely.

I'm concerned about the 'dim' branch.. you need to check the Vf of that branch.. each emitter while on..

I've had similar weirdness of lighting up one of a few that are in a string.. this happens i believe from the internal diodes inside the emitter that go between the leads and the heat sink slug.. not sure where the current path is to make it happen if you have all the emitters on tape, but it's definitely a quirky thing about series emitters.. I've seen it before.

I would test each emitter's lead to ground with your diode tester on your multimeter once gluing with AA.. i've done hunreds of these and only 1 case i think of a 'spread too thin' issue shorting out.

-awr

The Vf of the K2 emitters will shift with time, and the amount of shift will vary from part to part. My test data on one part:




0.1 ohms is definitely not enough to current balance your strings. The technique is dicey at best. LumiLEDs specifically advised against doing this in one of their seminars I attended in 2002.

In an automotive environment, don't forget your series diode, many cars/trucks have a reverse voltage transient during load dump. It can easily amount to hundreds of Joules, and can exceed thousands, and it is common to see bond wires blown when it happens. Without a series resistor, even large stud mounted zeners have been blown during positive transients. It varies widely from model to model of car/truck, and even varies alot within a single model, depending on the condition of the various electrical system components.

During starting, the starter current on one series of vehicle varied between 100 and 380 Amps. When you open the solenoid to stop the flow of current, the energy that was built up by 100 Amps at 12 Volts, reverses its flow when the field collapses, and produces a voltage and currrent in the opposite direction. This occurs quicker than the solenoid can fully dis-engage and prevent a arc-over.

One item that causes these load dumps is the starter. Another situation that can be *much* worse, is when a vehicle is jump started.

Gryloc said:

Not2bright, where did you get those quotes from (idleprocess and andrewwynn). I checked my own thread and I didnt see those replies. Thanks for fowarding them to me. Nice of you.

Click to expand...

Just for reference:

idleprocess's post #2 and it is his fourth bullet point.

andrewwynn's post #13 and is the 9th line of text below the sweet CAD mock-up.

http://www.candlepowerforums.com/vb...ad.php?t=117661

I agree with Newbie. 0.1ohm is not nearly a large enough value to balance the strings over time. You could use a low cost low drop out linear regulator set up as a current source in place of the resistors. This will be much easier to do, but of course when changing currents, you would need to control each one. Alternately, you could PWM the LEDS.

Plan for up to 16V from the alternator under typical operating conditions at least for short (hundreds of milleseconds) of time. Typical voltage is likely in the 14-14.6V range as the car is running. Also note that you will get spikes (load dump) up to 40 or 60V depending on what car you are putting this into. If you have some basic current limit on the LEDS, they will take this pretty easy. On the other hand, your regulators probably will not and will fail either unexpectedly or over time.

In terms of ease of circuit implementation, a boost circuit and driving all the LEDs in series will be the simplest. The Zener diode as mentioned takes care of single failures. The Zener has a higher conduction voltage then the LED and is wired in parallel with the LED. If the LED is conducting, essentially the Zener does nothing. If the LED fails, the Zener will start to conduct and complete the chain. Keep in mind that Zener will draw even more power than the LED and you need to get rid of that heat.

In terms of a way of isolating the LEDS, you may be able to your hands on some high quality base metal core board material, which is essentially an aluminum or copper backing, a very thin layer of electrically isolating and thermally conductive material and topped with typical circuit board copper. If you are creative, you can likely find a way to get your circuit drawn on top of the copper and etch it like a normal circuit board.

If you go with the anodized aluminum, make sure it is hard anodized. Even in this case, be careful not to scratch it.

I have also had some good luck with Arctic Silver, silver epoxy. Even though it is silver filled, it is only slightly conductive, and that should have no issues with the protection diodes conducting to the slug of the K2.

Another way of doing the high/low beam is to turn on/off different sets of LEDS. Some would be used for low beam, and some would add in the high beam functionality. That would allow you to control direction better.

I see the hotdriver control chip can take up to 38V on it's supply.
In one of the hotdriver variants, the MOSFET is only rated for 30V. A bit above this and the MOSFET breaks down, acting as a short from source to drain. At this point, the luxeons would directly be connected to the transient source.

I've actually measured transients on vehicles of 70V, just from a load switching on and off, such as a starter.

One may also want to look at the transient response time of the hotdriver, a wise man would consider this...

yup the FET is a 30V model... like i *JUST* posted.. i recommend a zener to protect the HD against spikes.

Considering btw.. if the hotdriver is running.. there is 12-13V on the source.. that means 42-43Vbat before that becomes a problem to the FET.. however.. w/o the zener a spike of 40V would shoot right through the FET because it won't feedback to the circuit 'til it drops on the load.. it is pretty fast to update but not fast enough for 'spikes'.. (microseconds typically).

in regard to balancing strings.. that of course is down-stream of using a current source.. and the point is this: *better than NOT using them*.. not that it's ideal. it could be a bit larger value.. based on the fact this circuit likely has 1/2V or more to spare.. maximize that balance resistor.. maybe it can be 1/4ohm.. you'd want to use a 1W 1/4ohm reisistor (it will drop just over 1/2W continuous at 1.5A).

-awr

Ok, so there is so much information just rolling out right not. I am hearing so many things right now. Considering strictly my scenario (not just generally), should I use the .1 or even .25 ohm resistors in series with each bank? This is considering that I am using a current limited Hotdriver circuit.

Tell me exactly how to use these zener diodes. I just heard to use them with each LED and one for the entire circuit. That doesnt help me much. So I need one in parallel with EACH K2? (I am worried about cost and complexity). Do I have one also before or after the Hotdriver? Now how are these supposed to be rated? What is the min/max voltage handling of these? Is there a certain voltage I have to look for that these can suppress? Are they supposed to handle so much current? How are they supposed to be hooked up, in series ir in parallel with the circuit? I am not very knowledgable with zeners. You can tell me all day to use them, but I will need a bit of help. If these are across each LED, and an LED fails, then the current will pass through the zener diode and the other two LEDs will stay lit? What if the Vf of that zener is real low?! Then more power will go to the remaining two and they will get hot and fail, too. I would much rather just have the whole bank/strand turn off than risk more damage. The headlight will still be lit with the other banks and it would be safe to drive home, even thought they would get <1.5A of current distributed amongst the other 5 banks (<0.3A extra each -not too dangerous, right?).

Ok, so if I use a zener before the Hotdriver, can it handle what the starter motor and alternator throws at it? They have fully regulated and protected automotive switching power supplies out there, but they are too expensive for our budget. Keep this discussion up! It is fascinating!.

With the forward voltage drop, should I just set the Hotdriver down 0.75 of a volt to play it safe for when the LED's VF drops over a month? I get 0.75V because according to the chart, the K2 at 1000mA has a Vf drop of 0.25V in a month. There are three in series, so 0.75V. I know this varies with every bin of K2's. I have a M-bin K2's. The reason I say this is because I will be off for college this fall (September), and I would sure like a set it and forget it type project. There will be a few days of constant testing (for heat, change in current, etc), and I will address issues then. Right now I am concerned about getting all the parts and putting them together. I will then set up the Hotdriver and watch it for a few days. Then, the project will be in the hands of my friend, who will be using the truck. He is not as knowledgable with LEDs, so he will call me for help, but only if he notices that the LED's dim too much, go off, or change color dramatically. I will be 2 hours away and wont be able to run over right away to help. He would have to wait a month or so.

Should I also have a nice rectifier diode (that can handle the current) in series to protect the Hotdriver and LED's from reverse voltages? It would have to be a beefy one. I had three diodes in parallel one day and I was messing with a 50W or so load. All three got burning hot in seconds. They were about a half inch in diameter each (dont remember the specs). They can be rather pricy, too.

Hmmmm, should I consider a higher voltage mosfet? On that can handle atleast 40-60V to protect from spikes if the zener fails or something?

Just so you know, the foil tape I am using is working. It isolates the slug alright, but it doesnt conduct heat very well. After about 10 seconds of being on, you can feel the tops of the LEDs get a little warm. At the same time, the heatsink does warm up a bit, too. I am glad the thermal adhesive would work. The tape is only temporary. I was able to have the heatsink on for 20 seconds at a time for dozens of times so far with no failures, so my problem is nearly solved, I hope. Good!

The output is still awesome (I will post more pictures later). Even the flood is so overwhelming. I wont hook up the narrow optics I got untill I recieve the copper. This is because I am not sure if I want to "heatsteak" every Fraen optic like it shows in the pdf's. I can also trim the legs and have the optics held down by a sheet of quality plexiglass (or acrylic? where could I get sheets of this? hmmm), which is bolted to the heatsink in each corner and maybe in the middle by some standoffs.

I have set on each K2 some various optics and reflectors I had laying around in some boxes. I wanted to see what the beam would roughly look like since I wont use all those new Fraen optics till I get the copper. I have some narrow optics, some 20mm and 27mm reflectors, and some misc relectors (D and minimag maglite reflectors set on top). The beam is rather wide and not even very centered with all these optics, and not focused, but damn it is still fricken bright. I have to set the amperature on my camera to 8.0 and the speed to 1/1000 and it still shows up as a nice, medium spot! Its awesome. I am tempted to start trimming legs, but I would probably regret doing this.

Im done rambling. Keep me updated, please. This willingness to help by all of you is awesome. Thanks!

-Tony

Hello, you are looking for a very difficult solution for a very easy problem, that happen to me in many opportunities, none LED has the parameter exact with respect to the other, I recommend to you, first, individualize each LED thermally to avoid short circuit among them and not to use the configuration, parallel/series for a same circuit, uses single serial using a resistance of 1.5 ohm 3 watts to have a voltage of 4,2 and 1200 ma in each LED, or 1.8 ohm 3 watts to 1000 ma... Let me know if this help you... Good Luck...

gryloc..

i would absolutely use the balancing resistors, especially considering that you'll be using a current-limiting driver.

zener diodes are designed to work in a circuit backwards from normal diodes.. All diodes work in this way.. they normally allow current to flow one way not the other (like a one-way or 'check' valve).. but.. they also have a 'breakdown' voltage where they will allow current from the 'backwards' direction.

zener's are tuned to have a precise breakdown voltage.. so.. for example.. if you put a 4.5V zener diode in parallel with each K2 emitter.. should that emitter fail open for any reason.. or.. should voltage ever get over 4.5V.. it will start to conduct and maintain no more than 4.5V.. The main rating on zener's is the power rating.. calculating what current/voltage to expect if it ever kicks in.. in this particular case.. i'm not sure about a practical solution.. because it would have to handle the same power the K2 has.. like 6W.. forgetabout that..

however.. for the spike protection ckt.. we are talking about microsecond or millisecond pulses.. so a typical zener can be used.. it's not a bad idea to have some series resistance like 10ohm or something.. but what happens.. is if there is an attempt to put 40V across the hotdriver.. the zener shunts the current to ground.. It would simply be wired up in parallel to the Bat+ and Bat- of the hotdriver.. with the 'pos' side and 'neg' side opposite.

the zener would also protect from revere voltages.. becaues it wold conduct at a very low voltgae.. that said.. i'm not sure what if any reverse voltage the hotdriver can take.. a full bat of reverse will blow the LDO chip every time.

I wouldn't worry about a different FET unless you have any problems with the monster one on there.. i put a triple size D2PAK on there.. i think it's a 30V vs the 20V i normally use.

Ok.. time to rest.. had a very long day with nanos being a very big pain in my backside.

-awr