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

[NewBie]

If you take a look, you'll note that even 5W zeners cannot take more than 20 Amps or so. Even in a short pulse. Unfortunately, adding a series resistor will only serve to limit the current.

Take a 60V 100 Amp pulse.

Zener kicks off at say 20V. So we have 40V remaining across the 10 ohm resistor. 40V/10ohms = 4 Amps. So, the pulse will easily have the ability to pass right by the zener stage, with little protection effect.

Example datasheet:
http://www.onsemi.com/pub/Collateral/1N5333B-D.PDF


It is quite a bit more likely you'll want to utilize one of the avalanche diode devices, such as an over voltage surge suppressor. These are specifically designed for an automotive environment, specifically for load dumps. They can take short pulses of 2500W, and non-repetitive currents of 600 Amps:
http://www.onsemi.com/pub/Collateral/MR2535L-D.PDF

If you want something more like a surface mount device, you can use a Transient Voltage Suppressor (TVS), which is a special large area zener like device. Pile three or four of these on and you should be okay. Something like the 18V device will clamp below 30V.
http://www.onsemi.com/pub/Collateral/1.5SMC6.8AT3-D.PDF

The advantage of using a decent series diode for reverse voltages, is that it will disconnect your circuit. But you suffer the Vf drop. Sometimes you can just put it across the lines, to just conduct if the voltage goes more negative than the Vf of the diode. But the circuit has to be able to handle this condition. A schottky diode will have a lower voltage drop in both scenarios. Using one rated for a much higher current than you actually need will typically result in a lower Vf drop across the diode, when in series, and will minimize your losses. It will also clamp better when across the lines.

If you poke around, you should also be able to find bi-directional voltage surge suppressors.

 

[andrewwynn]

i wouldn't worry all that much about the extremes.. might be nice to have an input cap or an output cap... the reality is in a worst-case scenario if the ldo chip blows it just won't turn on anymore.. and if the FET blows.. it usually becomes a 4 to 10 ohm resistor.. trying to put 18A through 4ohm means 36V drop.. obviously there is not going to be over-current at the emitters.

the FET and the LDO combined cost less than a fancy surge suppressor.. i would do some basics to 'help' more than 'be fool proof'.. and give it a go.. trial n error.. and.. if you want to be cautious.. start without the engine running 'til things are working nicely.

-awr

 

[evan9162]

Why you should isolate luxeon slugs when running in series:
http://candlepowerforums.com/vb/showthread.php?t=91700

Description of Vf and how it relates to current:
http://candlepowerforums.com/vb/showthread.php?t=77221
http://candlepowerforums.com/vb/showthread.php?t=72528

Vf shift in luxeons (with a ton of actual test data)
http://candlepowerforums.com/vb/showthread.php?t=60217

Achieving proper current balancing between parallel strings depends on a lot of factors. The easiest way is to use a regulator for each string.

Adding resistance helps, but the larger the potential Vf discrepancy, the larger the amount of resistance you will need.

The M vf range is 3.99-4.23V, and they are binned at 1A. If you are running at 1A, then the range of total Vf for a series string of 3 is 11.97V - 12.69V. So there's a potential to have an 0.72V mismatch, or up to 0.24V per LED. This could mean the difference between one strand running at 1A, and one at 1.5A. However, throw the potential of a large Vf drop occuring, and you could end up with a mismatch of 1.5V or more (0.5V/LED). That voltage difference can be disastrous for an LED.

 

[NewBie]

i wouldn't worry all that much about the extremes.. might be nice to have an input cap or an output cap... the reality is in a worst-case scenario if the ldo chip blows it just won't turn on anymore.. and if the FET blows.. it usually becomes a 4 to 10 ohm resistor.. trying to put 18A through 4ohm means 36V drop.. obviously there is not going to be over-current at the emitters.

the FET and the LDO combined cost less than a fancy surge suppressor.. i would do some basics to 'help' more than 'be fool proof'.. and give it a go.. trial n error.. and.. if you want to be cautious.. start without the engine running 'til things are working nicely.

-awr

The "fancy surge suppressor" that can handle the short 2500W 600A surges is only 1.06 dollars, and they offer free samples...
http://www.onsemi.com/PowerSolutions/product.do?id=MR2535L


You are spot on evan9162, as usual.

IMHO, in the scenario as it is made out to be so far, the small 0.1 ohm resistor is *far* from adequate.

Things may be able to be made to work at first, but it is doubtful that it will be all that reliable over time.

 

[Whitelitee]

Gryloc, damn man I cant belive you are doing this impossible task for your friend, you must be one good friend, or he is paying you a good amount of money. Exspecailly him not being a led nut, you should of just recomended him Hids. Or altleast for all you time typing, thinking and hard work you should be making them for you self being your the light nut. But whatever man I wish you good luck and you are one hell of a good friend... 🙂

 

[andrewwynn]

Somebody makes an active load balancer, i think i saw it on linear or national's site.. oh.. i remember now.. it was for a charging ckt i think but it would probably work for balancing load to emitters also.

nice on the free sample.. can't beat that.. i'd get one of those.. very good idea.. why re-invent the wheel.. buy a pre-fab unit for sure.

since i've discovered that the *starting* vF is below 13V.. there is plenty of overhead to use 1/4 ohm 1W resistors for balancing. combine that with initial matching of Vf and taking the logical assumption that as the vF drops they are *all* going to drop, and not make the ridiculous assumption that they could spread farther apart.. since as they drop.. the differences will become less in general.

I would set it up with a conservative 1.3A/string and monitor the vDrop on the balance resistor to measure the current in each string.

it would be *absolutely ok* if one string had 1300 and another had 1500.. this is NOT a case of a PDA display where if the corner is dim it's annoying.. the light will all merge together there is no way in hell you'll notice a string is outputting low output. As long as you aren't pushign things too hard and the HIGHEST string is 1.5A or less you'll be ok..

Since this will be a case of not being with it all the time.. just make sure the max output is a little conservative.. once they wear in and the Vf has stabilized.. then maybe you can push it up toward the full 1.5A. I would maybe start with 1.3.. it was amazing on the tester strand watching how much the voltage changes just in 1 minute while running current limit at 1.5A.

the thing evan didn't put in the equations was that to pull off resistance-matching you need to mirco-bin your emitters.. you will get within 0.1V almost certainly if you do this.. with 18 emitters it shouldn't be hard to find groups of three with a total vF within 0.05V which will do the trick to get the current within 100mA in each strand.

-awr

 

[SemiMan]

As Newbie suggests, put in the avalanche diode. It is well worth the time saved of debugging a toasted circuit.

If you insist on series/parallel, goes with something larger for the resistor, say at least 1/2 ohm or more (more is better). Odds are, you are going to have some LEDs where you do not have as good a thermal connection and this is going to create further differences in the Vf to add to the aging issues with vary from LED to LED. Again, this is the easy path and you have enough hard things to deal with. If you are planning to run at 1A, you can likely get away with not so tight matching, but if you plan to run at 1.5A for all LEDS, you need to be careful. Again, an LDO per LED string would solve the problem.

A reverse battery diode again is one of those things that makes life easier in automotive. Car batteries put out hundreds of amps and the fuse will not blow instantly. More likely your circuit will go up in smoke (literally maybe). If you can't handle the voltage drop, you can use a FET for reverse bias. If interested, I can email you a PDF showing how to do it. It takes a few more parts, but you can get the drop below 0.1V.

Are you planning to run with 3 or 2 LEDS in series? Realistically you can not run with any more than 2 unless you use a boost circuit. 3 LEDS running at 1.5A will have a Vf in the 13.5 range so already higher than the battery voltage with the car off. 2 LEDs will be 9V already, add in the drop for reverse battery, regulator, and not to mention drop in wiring in the car, and you quickly run out of head room.

FYI, went back to the specs. For low beam, to meet the spec, you need only about 320 lumens in the target area. For high beam, that only goes up to 390 or so. Of course, that is light in the target area and no one has yet developed a perfect optical system, hence why you need to input on the order of 1,000 lumens to get the needed light in the target area.

Good luck!

 

[andrewwynn]

gryloc already mentioned shoestring budget this is obviously a 'labor of love' and also it's not about logical or practical it's about innovative thinking outside of the box as a technology demonstrator.

Oh.. gryloc.. funny thought i just had when i read that mention of LDO per string..

the nano driver would be absolutely perfect for such a task!.. having a hotdriver for V regulation to the whole set and a bunch of nano drivers to hold the current stable at a max would be an ideal solution.. for 'version two' i would definitely work toward that end. The chips are not expensive .. you could build up a circuit with all the ldo's on one board.

semi.. you are not correct about the 3 in series.. it takes less than 13V for 3xK2 UxxM to reach 1.5A It is true that with the car not running it won't be at full power but that's just fine.. just like incan that dim when the car's not running these will do just the same (in high beam).

interesting on the spec.. i'd be interested about the target area.

-awr

 

[NewBie]

Andrew,

Please pardon my french here, but it would be good to note that a UxxM, is actually binned at only 1.00 Amps.
http://www.lumileds.com/pdfs/DS51.pdf


The M designating the Vf bin, means at 1.00A the Vf will range from 3.99 to 4.23.
http://www.lumileds.com/pdfs/AB21.pdf


The Vf continues to increase at the 1.5A that they are talking about here. The production parts I ordered, and have, I'm looking at typically 4.3-4.6 Vf @ 1.5A.


The interesting thing I have noticed, is even off the same strip, these K2 LEDs all change their Vf over time, some very little, and some a surprising amount. The variation is quite amazing. The other issue, is the thermal problem. As the Vf changes, the amount of heat dissipated in each LED also changes. This causes a further change in the Vf, and can lead to pretty decent and surprising variations on series-parallel strings.

 

[andrewwynn]

i was basing my numbers on actual K2s i strung up and pushed 1 to 1.5A through with my current limiting bench supply.

Once again, newb.. let some people that have actually built something build. we love the theoretical database you bring to the table but I actually build lights.

-awr

 

[NewBie]

i was basing my numbers on actual K2s i strung up and pushed 1 to 1.5A through with my current limiting bench supply.

Once again, newb.. let some people that have actually built something build. we love the theoretical database you bring to the table but I actually build lights.

-awr

Andrew,

I'm not sure if you are aware of it, but I have many products in production that use LEDs, specifically in lighting related applications, and have for many years now. I actually do have hundreds of thousands of LEDs in products I have designed, in use, all over the world. In my field, it is important that things are reliable.

I would encourage you to refrain from this tac in the future.

Since you brought it up, this is not just conjecture, it is based on actual reality, as many times my comments are.

I've tested quite a number of K2 devices now, and ran plots on them, just like the one you see a few posts back. This isn't guesswork, nor working off the datasheet, but reality. These are the sorts of things I normally do, before I design things into a product. It is often important, because it is not uncommon that you learn more about the devices, than what is on the datasheet.

However, it is possible, that my setup has more agressive heatsinking, which would cause my Vf values to be higher than yours would be, if you were not as agressive in the heatsinking as I was.

Here is an example of one of my orders I have made, so I could characterize the various K2 devices (this bag is from my first order of the production K2 devices):

 

[SemiMan]

Andrew,

While Gryloc has stated he is running on a budget, at a minimum, he is looking at probably $100 of parts per light for the LEDS, optics, and power by the time he blows some of them out, tries different optics, etc. It will be well worth the extra $5 or $10 tops per light to make sure that they do not fail prematurely. In the long run, that will be be best way to protect the shoestring budget. Nothing is worse than getting close to completion and having something fail and have to tear it all apart to repair it.

Gryloc, load dump and other nasties are reality in automotive. I suggest you learn how to protect your circuit. It will be a useful thing to know for the future. You are also going to have significant Vf differences in your LEDS in actual operation. Again it is reality, so it is best to be prepared for it.

Semiman

 

[andrewwynn]

semi-man.. i was only referring to the budget meaning.. i'm pretty sure that it's a non-paying job.. for a friend.. i agree about the not skipping out $10 and already had him upgrade to at least the high-temp circuit vs the very most basic.. i would have personally run each light from a separate driver.

Good advice about the protect the circuit, i like the advice from newbie on the protection ckt, it's absolutely worth the couple bucks if it's actually available.. a 'price quote' without a 'place to buy' (a single unit and not 1000) is not necessarily helpful... fortunately you can probably sign up as a customer and get a free sample and that's even better!

Newb, you have to know you are about the most contentious member of cpf, nit-picking everybody's eveything to death.. it is helpful to make us double-and-triple-check our work but you are not always right, and you do *not* *always* have to do everything *the best* way, and some good old trial-and-error with real-world building does find solutions that are outside the box of the data specsheet.

For example.. there is no way in hell the hotdriver 100's should work based only on the specsheet... but i spent a month fine-tuning the exact way for them to start and they work like the bomb.. My point is... put your money where your mouth is so to speak.. i have little to go by other than 'your word' about these 100's of 1000s of items 'out there in the world'... and they really don't apply to batching up 18 K2s in banks of 3..

'the whole world' has been using balancing resistors since the dawn of the white LED to arrange this sort of thing.. true it's been for 30mA.. the concept even though lumelid's frown upon it.. logically enough because the balance might suck.. with simply under-driving the average enough so that the swing in current will make the highest close to the max allowed.. it's a very doable thing.. I'm working with gryloc to figure out a solution that specifically does NOT have a regulator of any kind in each bank of LEDs... so.. working within the confines of that.. now come up with something helpful to the project.. That means.. pretty much.. you are limited to the only known solution.. balance resistors.. maybe they'll suck.. maybe they'll work just fine.

My point of the comment that isn't 1/10th as contentious as most your posts on my threads was based on this quote:

Andrew,

Please pardon my french here, but it would be good to note that a UxxM, is actually binned at only 1.00 Amps.
http://www.lumileds.com/pdfs/DS51.pdf


The M designating the Vf bin, means at 1.00A the Vf will range from 3.99 to 4.23.
http://www.lumileds.com/pdfs/AB21.pdf


The Vf continues to increase at the 1.5A that they are talking about here. The production parts I ordered, and have, I'm looking at typically 4.3-4.6 Vf @ 1.5A.

I built up a hotdriver for gryloc.. dialed in 13.0V and with THREE in-series K2 U--M emitters.. it went over 1.5A within 20 seconds... i had to dial the voltage down to below 12.95 to keep the current at 1.5.. that means 4.31V/emitter.. FAR below 4.4 estimated in the datasheet.. My point was.. you were referencing the datasheet with the numbers.. I was referencing a live load with the same bin K2 emitters going to be used ... and they matched *my* measurements and NOT your hypothetical results.. this is not the first time it's happened nor expected to be the last.. Your tone comes off saying 'it can't be done'.. My tone is.. "it just WAS done i JUST did it"..

A while back somebody told me that my hypothetical solution for the hotdriver just absolutely couldn't be done.. my reply.. was to say.. tell the prototype that's been working for 4 months 'cause i think you just hurt it's feelings.

similarly.. it is possible.. the numbers ARE working out very positive.. that the total vF is less than 13V, and will drop.. which gives MORE headroom meaning LARGER balancing resistors can be used, and I have extreme confidence that this solution will be absolutely fine.

I personally recommend using two drivers.. and setting the max voltage for an initial current of about 1.5A before tweaking the current limit to hold the current to 1.3 per row maximum... and after a few months of service.. see how much the vF has changed and measure the current on each row.. it's a very doable thing, it just takes work to balance the vF by hand if you want to do it w/o a regulator on each bank of LEDs.

-awr

 

[NewBie]

Newb, you have to know you are about the most contentious member of cpf, nit-picking everybody's eveything to death.. it is helpful to make us double-and-triple-check our work but you are not always right, and you do *not* *always* have to do everything *the best* way, and some good old trial-and-error with real-world building does find solutions that are outside the box of the data specsheet.
-awr

By CPF rules, please attack the post, not the poster- Thank you.

'the whole world' has been using balancing resistors since the dawn of the white LED to arrange this sort of thing.. true it's been for 30mA.. the concept even though lumelid's frown upon it.. logically enough because the balance might suck.. with simply under-driving the average enough so that the swing in current will make the highest close to the max allowed.. it's a very doable thing.. I'm working with gryloc to figure out a solution that specifically does NOT have a regulator of any kind in each bank of LEDs... so.. working within the confines of that.. now come up with something helpful to the project.. That means.. pretty much.. you are limited to the only known solution.. balance resistors.. maybe they'll suck.. maybe they'll work just fine.
-awr

The reason LumiLEDs greatly frowns upon the balancing resistors, is that they have experience in traffic lights, where balancing resistors were used, and lead to long term failures. They did present how to utilize balancing resistors if you absolutely have to, making them large enough such that they can take up the slack, and minimize the problem, much like SemiMan said- with his larger resistor value.

I built up a hotdriver for gryloc.. dialed in 13.0V and with THREE in-series K2 U--M emitters.. it went over 1.5A within 20 seconds... i had to dial the voltage down to below 12.95 to keep the current at 1.5.. that means 4.31V/emitter.. FAR below 4.4 estimated in the datasheet.. My point was.. you were referencing the datasheet with the numbers.. I was referencing a live load with the same bin K2 emitters going to be used
-awr

No, I was not solely referencing the datasheet. You will notice I said 4.3V - 4.6V, based on my actual measurements. Your 4.31V does agree with my measurements that I stated. I also offered up the fact that more or less heatsinking affects the forward voltage.

similarly.. it is possible.. the numbers ARE working out very positive.. that the total vF is less than 13V, and will drop.. which gives MORE headroom meaning LARGER balancing resistors can be used, and I have extreme confidence that this solution will be absolutely fine.
-awr

If you look up at my chart above, you will notice that the majority of the shift appears to happen in the first 60 hours. One might consider burning in the unit, for 3 days, and then matching at that point.

A good reference paper, for dealing with Vf variations, and other items, was done by LumiLEDs for tail lights. You can rescale things accordingly. They also mention the possibility of -300V spikes, and even up to -600V spikes in large trucks, and offer up the series silicon diode option I mentioned earlier:
http://www.lumileds.com/pdfs/AB20-3.pdf

I'll poke around, and see if I can find the Luxeon presentation dealing with the same type of matching issues for you.

 

[andrewwynn]

The problem i had with your first post i quoted was that once again you implied i don't know what i'm doing which is incorrect.

Apologies for being blunt, there are member far more blunt than myself... i was not 'attacking' you, stating something out-loud that me and you have talked about before regarding how "direct" you are.. no offense is meant, none should be taken.

I agree upon further examination that it's a very good idea that the balancing resistors are more than 1/10th ohm!

it's clear that using the balancing act.. presents a higher maintenance solution.. I will be willing to bet that those traffic lights did NOT have hand-matched vF per string and NOT checked-up upon a week or month later for fine-tuning.

It's clear that it's *better* to not use them.. but better than nothing to have them, and my contention is that it's quite possible, quite doable and nothing wrong with it.. just 'more work'.

The silicon diode protection is a very good idea, and why we love having you around to fine-tune our hacks :-D

-=awr

 

[dano]

This is a warning: Stop the attacks, and keep the posts relevent to the topic. Personal atatcks or public nit-picking b/w members will not be tolerated.

--dan

 

[Gryloc]

Jeesh. Both of you need to stop. I have been looking in on this thread for some time and I was never sure if it was a good time to post or not. CPF is a poor place to discuss your differences like that (Especially in my thread). I like when all of you work together on CPF to solve problems. That is what these forums are for! It is good to stay positive here, or you may scare away newer members that may think this is a dark, hostile place. I invited someone new I found while shopping on eBay who designs and builds simple boost curcuits for Luxeons. I bought his nicely priced driver and told him about CPF. I told him to look for my threads and make himself at home. Lets not scare him away. Anyway, I am sure we can can get back to buisness here.

Actually my main question at the beginning (why my LEDs are failing?) was already answered by a thread by evan9162. In this thread he discussed his findings when the slug of an emitter is not isolated properly. After reading how the Luxeons use the ESD diodes and what happens when the LEDs go into "failure mode", all I can say is "ooooh, ahhh!".

So, I have a question. Can we move this back to my main project thread in the general light discussion section? I think the topic of my next few questions about power actually strays from the purpose of this thread. It is located here:
http://www.candlepowerforums.com/vb/showthread.php?t=117661

Please, everyone, lets all move over there and lets get a fresh start. I hope there is less bickering there (I dont like taking arguements home). I have some questions that concern using balancing resistors and noise supression....

Thanks for everyones concern, though. I am glad that some support one of a kind projects. I hope we get all of this figured out soon.

 

[Empath]

With the author of the thread satisfied with the completeness of this thread, we'll honor his request to move to: http://www.candlepowerforums.com/vb/showthread.php?t=117661

We'll close this one as complete.