Preventing overcurrent in parallel circuits?

Mash

Enlightened
Joined
Dec 18, 2006
Messages
378
I was looking at the xitanium constant current PSU data sheets.
On bottom of page 2 (please look!) they have the reccommended wiring schematics, where they utilise parallel and series combined. This way you can power lots of LEDs from a single PSU, which is cool.

Now lets take the following arrangement as our example:
They have models (not shown on the above datasheet) that pump out 4A, which lets say we use to power 24 crees, in 4 parrallel legs of 6 leds in series per leg (as per data sheet diagrams). This way each leg recieves 1A.

All pretty standard so far, however whats got me thinking is this, hope I can explain clearly: The circuit is up and running as per above. If one LED in any leg fails (or a connection anywhere becomes loose), that whole leg is taken out of the equation, and suddenly the other legs are getting 1.3A each. The remaining three legs are now recieving 30% more current and are in effect overdriven (if we designed the fixture, sinking etc for 1A operation) and this will eventually cause another LED somewhere to blow soonish. Then the two remaining legs are pushed at 2A and surely will fail very very soon, if not instantly. So in effect failure of just one LED can lead to a cascading blowout of many more LEDs; which can end up expensive and total darkness to boot too! This is not what I would assume is a desirable thing!:grin2:
So question is how do we insure ourselves against such a possibility?
Would resistors work as a current limiter? In my thinking and comprehension I only see resistors used when the supply voltage is above the LED Vf, ie to set up the circuit, not as an insurance, especially when the dynamics of the circuit change as in this case. Also if we put a big enough resistor in there to prevent a blow out, what does this resitor do to the circuit during normal operation?
I can also think of maybe having a 1A fuse in each leg, but this would mean that if an LED fails, then 3 fuses will surely blow, which is better than instaflashing 18 LEDs though!
I can also think of a LM317 (in current regualtor type thing on every leg, but as with the resistors I wonder if its desirable to have it (in terms of heat and wasted energy during normal running).
I have given the above numbers only as an example here; whether crees can take 1.3A or 2A etc is not the question, finding the correct concept is the aim of this thread.

So what is the trick for preventing cascading failures in parrallel circuits?

PS: Also I just realised, that the same question applies to setting up and first time testing of your circuit as well. If you have a bad connector somewhere, or wired even one LED wrong, just powering up the circuit for the first time can have the same effect! How would you test and power up your setup safely for the first time?
 
Last edited:
I'm sure there are more complex protection circuits, but fuses are cheap. I'd use lots of them. Perhaps one for each emitter.

=Mark
 
That would be overdoing a bit, since fusing each leg would protect the LEDs in that leg, no need for individual fuses.
Anymore ideas from others? There must be a proper way.....
 
Seems there is a way as demonstrated herre. Only applies to an LM317T current controlled circuit though.
I dont really understand the circuit, but it seems to be addressing exactly this issue.

Im still open to input from knowledgable CPFers though, particularly reffering use with a PSU driver, as outlined in the opening post.
 
Last edited:
fuses would prevent over current hazards, but doesn't regulate the amount of current passing through that line to a specific level...
:anyone: thinking of wire gauges?
 
Cool drivers.. too bad they don't have one that quite suites my needs.. I have to do 9 LEDs @ 700mA.. I already have plans to build my own drivers.. but some pre made units would have been nice..

I get the same efficiency they do with my design.. so I imagine it's very similar. Except I don't understand why they charge so much more for dimmable! I wasn't even planning to bother with that for this specific project.. no need really to dim lights in the kitchen... but the drivers I'm making based on the LM3404 chip have a PWM input.. all you'd have to do is use a little 555 timer to create a PWM pulse to do the dimming.. so cheap to do... unless.. hmm.. if they are dimmable by a standard wall dimmer... then I can understand... that would be difficult to achieve properly... if not impossible.

I'll end up paying around $40-50 in materials to construct a 9x700mA driver myself.. The same driver though could go up to 12x1000mA with my design.
 
thats nice Im sure, but do you have any ideas ragarding the question of this thread?
 
fuses would prevent over current hazards, but doesn't regulate the amount of current passing through that line to a specific level...
:anyone: thinking of wire gauges?

Thats true, however, the aim here is to prevent an overcurrent in case a section of the circuit fails, ie we need a failsafe mechanism.
In normal woking status, the current passing through each leg is determined by the PSU and the number of parallel legs.
 
thats nice Im sure, but do you have any ideas ragarding the question of this thread?


You answered yourself already... fuses.. the obvious solution.

Trying to rig up a transistor kill switch with a seperate current sensing resistor on each leg would require you opening up your xitanium driver and modifying some SMT components.. so I doubt that's on the table.

I'd just test it first.. get fast acting fuses and short 1 LED in a string to make sure all the fuses burn and there's no more power... It would be nice to have the 1 faulty string shut down like the link pictured with the LM317, but it's just not worth it.

Sorry if I sounded selfish.. I hadn't seen quality drivers like this on the market before.. I was excited to compare them to the ones I'm building. Do you know of other high quality AC drivers? ..Their life is rated at 50,000hours.. so pretty good, but the LEDs will probably outlast the them.
 
I dont think you have got the gist of this question, please read the thread and post relevant information.
There are many threads regarding many AC drivers, right here on CPF.

The question is, once again: What is the best way of protecting parrallel legs from a cascading failure?
(apart from fuses on each leg, as suggested by myself, which will ALL blow, in case of a failure).
The transistor diagram which I reffered to, is for use when the whole circuit is being POWERED by an LM317 (I mentioned it to demonstrate that other people have come across this situation). In our case we will be using an AC constant current driver, so we need a solution specific to that setup.
 
Last edited:
Fuses won`t help, the current is constant, if an LED fails the current remains constant but the voltage may vary.

Use the correct driver for the application is the real answer, if your using 1W LEDs use a 1W driver and put them in series. Allow enough voltage overhead to drive a whole string.Driver cost has gone through the floor in last 2 years.

Paralelling LEDs can work if they`re matched and are going to be running in similar conditions, big clusters were a common culprit ones in the middle get hot and current hungry.

Xitanium is quite an old driver relatively and some of the wiring suggested would suggest is not thought of now as best practice.Exactly for the reasons you point out.

HTH
Adam
 
Fuses won`t help, the current is constant, if an LED fails the current remains constant but the voltage may vary.

Use the correct driver for the application is the real answer, if your using 1W LEDs use a 1W driver and put them in series. Allow enough voltage overhead to drive a whole string.Driver cost has gone through the floor in last 2 years.

Paralelling LEDs can work if they`re matched and are going to be running in similar conditions, big clusters were a common culprit ones in the middle get hot and current hungry.

Xitanium is quite an old driver relatively and some of the wiring suggested would suggest is not thought of now as best practice.Exactly for the reasons you point out.

HTH
Adam
 
Thank you Adam for your input. The reason I like the Xitaniums in this type of circuit is that you can drive lots of leds with one driver.
On other drivers with series wiring only, I have found that they can drive a max of 16 or 8 LEDs ( 350, 700mA) generally. So a xitanium would be able to replace 6 or more other drivers (although this type of xitanium is about 4-5 times the price! so maybe you come out close to even!)
 
Last edited:
I dont think you have got the gist of this question, please read the thread and post relevant information.
There are many threads regarding many AC drivers, right here on CPF.

The question is, once again: What is the best way of protecting parrallel legs from a cascading failure?
(apart from fuses on each leg, as suggested by myself, which will ALL blow, in case of a failure).
The transistor diagram which I reffered to, is for use when the whole circuit is being POWERED by an LM317 (I mentioned it to demonstrate that other people have come across this situation). In our case we will be using an AC constant current driver, so we need a solution specific to that setup.

Hmm.. I'm not so sure you understand your own issue. You don't have an understanding of the fundamental electronics at work here.

The LED driver in the xitanium will have a current sensing resistor just like the LM317 does... but you won't be able to split up that resistor for each string and add a transistor for the safe shutdown unless you want to modify the SMT components...

...so once again, fuses are the best option.. You can't have one LED go out and the just turn off that string, You CAN'T keep any of the LEDs on when one fails.. not without modifying the current sensing resistor value..

...so put some fuses in and stop being rude?


LED-FX:

The current is constant in a series.. if you start making parallel series arrays with these drivers you half your current... so if one string fails the current will double over the other string (if you had 2 in total) ...so yes a fuse will help... and again... is the best solution.
 
Last edited:
A few thoughts regarding the situation of 4 parallel strings of 6 LEDs in series each, and a few driver and motivation/objective thoughts as well:

1. If you had a cascading blowout only one LED per series string would blow, leaving 20 working but unpowered LEDs, although 15 of those many have been over-current abused (with 1.3A, 2A, and briefly 4A).

2. Small value resistors are good (or at least sometimes useful) in parallel string setups because they help balance current between strings. I've seen Cree operating voltages (@ 1A) from 3.3V to 3.6V or so. If you happen to have more of the lower ones in one string and more of the higher ones in another string you can get your strings unbalanced very quickly. Driven in parallel with a 2A supply, a string of lower voltage LEDs might draw 1.5 A while the higher voltage string would only draw 0.5A. (Example only.) Adding a resistor to each string would help even out the difference. It would be best to have total operating voltage matched as close as possible to start with.

3. The idea of using a fast-blow fuse for each string isn't actually too bad for protecting your LEDs. The chance that an LED will blow or a string will go open circuit should be very low (assuming your soldering is good), so the likelihood that the fuses would be needed isn't high, so the situation arguably doesn't deserve high-cost protection. Using fuses thus seems very good. Perhaps you could wire a small LED in parallel with the whole lot, wired with a large value resistor so it would provide emergency lighting when your fuses blow. That way you wouldn't be left in the dark when all the fuses blow - and remember, if one fuse goes, all do. :tinfoil:

4. From memory the LM317T has a sense voltage of 1.25V, but the dropout voltage is from 2 to 2.3 volts, which at 1A would mean 2 to 2.3 watts dissipated by each one. (Multiply by the number of strings you have.) That could be worse (say, 9-11% waste for driving 6 LEDs per string) but would be good protection for the LEDs. I can think of other ways of electronically limiting the current that might be less wasteful, but they would depend on the power supply not increasing the voltage much to try to overcome the loss in current drawn. The LM317T can cope with 40V, but it would be better to run them with a constant voltage supply, not a constant current one.

5. The link you gave, Mash, using transistors with a single LM317 is a very clever way of using only one LM317 for multiple parallel strings of LEDs. An LM317T can typically cope with 2.2A so in the example you gave you'd need a couple. It's an improvement - only half as wasteful as using one LM317 per string.

In conclusion, whether you choose to do 3, 4, or 5 (or something else) will depend on how electronically adventurous you want to be, how much money and time you want to put into it - your call on both of those - and also how your driver works.

If you had four parallel strings each drawing a current controlled 1A, drop a string to only three strings, would your driver try to increase its voltage to compensate (according to V=IR). Is there a limit to how much it would increase its output voltage? Would that be bad for it? Would the increased voltage (and therefore wattage to be dissipated) for bad for the LM317s? Would burning them out be any less expensive than burning out an LED per string?

If you had a high current regulated (or even semi-regulated) voltage supply that you could use instead, you could just use some LM317s or something else that does a similar job, perhaps even more efficiently. Adam (LED-FX) was wrong about the fuses not helping, but right about the cost of drivers recently.

Fuses seems like the easiest, cheapest, and simplest - which normally goes hand in hand with reliability. The downside is that it's all or nothing. Is having the light more or less important than protecting the LEDs? If having light is the most important thing then don't run the LEDs in parallel - use another driver that controls the current for only one string at a time.
 
Thank you TorchBoy for your informed and insightful comments.
Regarding some of your questions:
If it fails and goes dark, its not a big deal!
The xitaniums I think supply a max of 24Vs. Thats why you see a max of 6 LEDs in series per leg. So usually you are operating at max voltage anyway. Basically, at maximum, the driver is putting out 24V and 4A.
I see what you are saying regarding the LM317s and using them with constant voltage, as a power source for the whole circuit.

However my way of thinking was (with a xitanium power source), that if each leg is protected by one LM317, set at eg 1.2A, then if lets say two legs fail, we still have 24V and 4A incoming total, however then the LM would only let 1.2A through to each remaining leg, instead of the available 2A, therefore protecting it. And also this would help with Vf differences -unbalanced legs- in the legs, ie only allowing max 1.2A through regardless of how low the total Vf for that leg is.
I once managed to blow TWO crees in series (yes both of them), by feeding them 10V by accident, still dont how it happened; I agree with the principle that you would/should only blow one LED in each leg in case of overcurrent/voltage, but I have managed better as in the above example!
And thanks for the clarification regarding the wastage involved in having LMs in circuit during normal operation.
I am trying to get my head around the best way to run multiple LEDs; CC driver-->parallel strings, is one (therefore this question); and I think the next best option would be something like a 24V CV source (with high wattage), feeding into strings of 6 LEDs fed with one LM per leg. Do you think there are better ways?
The reason I started playing around with LMs was actually because of your previous comments on CPF. At first I was vary of learning another circuit etc, but when I saw how simple it seemed, pulled out something similar (its not even an 317 but works the same!!!) from a broken amplifier and PRESTO, it WORKED!!!
BTW as a side question, if for the R1 in the sense pole of the LM we use a potentiometer, this would/could work as a crude dimming, correct?
Thanks again for your input, it is much appreciated.
 
Last edited:
I once managed to blow TWO crees in series (yes both of them), by feeding them 10V by accident, still dont how it happened; I agree with the principle that you would/should only blow one LED in each leg in case of overcurrent/voltage, but I have managed better as in the above example!
Wow. I've learned something there. I didn't know they could blow so quickly. (I guess I'm glad I didn't have more than one Cree in series when I connected my multimeter to it the wrong way in my car - 12V not good.)

The LM317 works quite nicely and simply. The only reason I don't use them any more is because of the big dropout voltage. That isn't so much a problem when using a supply like you are doing, although you'd have to have enough voltage head room from the driver. If the LEDs each need an average 3.6V, that makes 21.6V total and if the Itanium does 24V maximum without overheating because it's trying to balance the impossible (trying to push 4A but only 3A being drawn), that should be OK, even ideal.

if each leg is protected by one LM317, set at eg 1.2A, then ...
... you'll need very good heatsinking on the LEDs, and good heatsinking on the LM317T wouldn't go amiss.

BTW as a side question, if for the R1 in the sense pole of the LM we use a potentiometer, this would/could work as a crude dimming, correct?
Yep, and I wouldn't even call it crude. Just make sure you have enough resistance there not to exceed the maximum setting you're aiming for. It's not a problem to put a resistor in series with the pot so when the pot is turned to zero ohms you have the resistance from the other there still.
 
Blowing the leds: Yeah it was very very quick. Was playing around, seeing what voltages give what brightness, current etc, so connected two in series, and wanted to check 5V, BUT using a pc PSU, must have either given it 10 or 12V. Both very very bright for a split second, then nothing, and they were both gone!
Regarding heatsinking:
Now Im playing with some undercabinet lighting; I have two rebels in parallel at 400mA each, running off a nokia charger, epoxied to an old hard disk case thin lid (hey its aluminium!), this barely gets warm.
Also three crees at 800mA to a smallish heatsink, with the LM epoxied on to sink as well. This setup needs a fan though, otherwise things can get quite hot!
 
Update:
Just got me an LM317 to play around properly!
TorchBoy, you are quite right about the voltage drop across the LM. With a 12V psu, and 3 crees, its quite difficult to hit high currents even with 0 R1 value. And when you add an R1 to ensure max current protection, the resistor cuts into the circuit voltage as well, further dropping the V and A values. However I like the peace of mind and easy controllability it gives you.
TorchBoy as you mentioned you are not using LM317 anymore, may I ask what you are using?
 
You should never use 0 ohms for the current sensing resistor on an LM317, they are rated at 1.5amp, therefore you shouldn't use anything less than 1ohm really.. or more to be safe... in CC mode the LM317 will drop about 3 volts.. so you can't run 3 CREE off 12v reliably. A 14.4v drill battery would work perfectly.

Remember to check how many watts you'll be dissipating.. at 1A running 2 CREE off 12v you'll be dropping 5v, 5v X 1A = 5watt.. that's a pretty decent amount of heat from your LM... And your resistor as well will be dissipating some heat.. 1.25v X 1000mA = 1.25w, so you can see efficiency really goes out the window with these.. but they're great for prototyping LED designs. If it will be for constant use it can be tricky trying to find the right supply voltage for a linear driver so you can do what you want, but not make too much heat..
 
Last edited:

Latest posts

Top