High Power LED Driver Circuits

High-power LED's: the future of lighting!

but... how do you use them? where do you get them?

1-watt and 3-watt Power LED's are now widely available in the $3 to $5 range, so i've been working on a bunch of projects lately that use them. in the process it was bugging me that the only options anyone talks about for driving the LED's are: (1) a resistor, or (2) a really expensive electronic gizmo. now that the LED's cost $3, it feels wrong to be paying $20 for the device to drive them!

So I went back to my "Analog Circuits 101" book, and figured out a couple of simple circuits for driving power LED's that only cost $1 or $2.

This instructable will give you a blow-by-blow of all the different types of circuits for powering Big LED's, everything from resistors to switching supplies, with some tips on all of them, and of course will give much detail on my new simple Power LED driver circuits and when/how to use them (and i've got 3 other instructables so far that use these circuits). Some of this information ends up being pretty useful for small LED's too

here's my other power-LED instructables, check those out for other notes & ideas

This article is brought to you by MonkeyLectric and the Monkey Light bike light.

Step 1 Overview / Parts

There are several common methods out there for powering LED's. Why all the fuss? It boils down to this:
1) LED's are very sensitive to the voltage used to power them (ie, the current changes a lot with a small change in voltage)
2) The required voltage changes a bit when the LED is put in hot or cold air, and also depending on the color of the LED, and manufacturing details.

so there's several common ways that LED's are usually powered, and i'll go over each one in the following steps.

---

Parts

This project shows several circuits for driving power LED's. for each of the circuits i've noted at the relevant step the parts that are needed including part numbers that you can find at www.digikey.com . in order to avoid much duplicated content this project only discusses specific circuits and their pros and cons. to learn more about assembly techniques and to find out LED part numbers and where you can get them (and other topics), please refer to one of my other power LED projects.

Comments

[Spuzzum]

This works great! I just finished my first box, consisting of 7 drivers, and will be making 2 more boxes, for 21 drivers in total.

I found that anything above 500mA and I needed a heatsink on the fets, so I bulked them up. I then found that the current actually drops after 900mA.. wasn't the driver, it was my tester diodes needing a fan on the tester heatsink. :p Now it runs steady at 1000mA, and can be dimmed to as low as 100mA. I modified with a 10ohm 1/2w trimpot, and a 10ohm 1/2w resistor paralleled with the trimpot.. making a 5ohm 1w variable resistor. Results.. an "adjustable" constant current driver. :)

Works out to about $8 a driver.

I also found that splurging for 24v sources is actually cheaper than using 12v from an atx psu... less drivers needed in the long run.

[profpat]

great tutorial, will try it out got lots of LED BULBS for projects

[sprocketme2]

I understand that R3 is calculated as R=(Saturation voltage of Q1)/(desired current through LEDs) If I am wrong please correct me. Now for circuit 5 specifically, is R1 still functioning as a pull up (100kohm) like the other circuits, or is it functioning as gate resistance for Q2? In which case it would be in the order of 25-100ohm.

[dante.ragnarok]

i need to kwon how calculate the set current, because i have a 18v and 3,5 A power source and i want conect 4 red led 3w and a blue one 3w at 700mA each one

[peterlonz]

This seems to me to be a very imprecise way of driving an LED.
Any key component with a set point that may vary by 100% is surely utterly useless.
You will not know untill you receive the PTC exactly what its trip point is.
The example shown raises more Q's than A's:
You may end up driving at the desired 500mA or the unwanted value of 250mA.
The circuit copied by the poster possibly employed this method but I'll bet the PTC 's were carefully sped'd & tested to be sure they were within an acceptable range, not something the hobbyist can easilty do!

[dan]

yes, it is imprecise but very cheap and simple. for some applications that can be useful.

[burton2663]

what do you suggest to drive 1 watt luxeon leds from an lm3915 output? tnx

[ggodinho]

Hello, Sir Dan. I've liked a lot your circuit and I've been making some updatings in order to fit with my project. So, I'd like to share my experiences with all and I'd like your technical remarks and sujestions.

First of all, I couldn't understand why did you insert the R2... Since I needed more input power, I just verified my NFET datasheet (IRF260) and Q1 (BC548) components and I noticed Q2 supports until 200V. So, I connected all the circuit at 180V, (an local 110V plus half-wave retific and filtered voltage). This power in drives a string of 40 5mm white leds (40x3.1=124V).

I know I could put more leds until reach near 180V, but I preferred put 3 paralel strings of 40 LEDs (a resistor was added in series on each LED string in order to fit all the strings to the same current). Of course, NFET will waste the extra voltage, but it doesn't matter to me, the important is to sustain the same 60mA constant current (3 strings of 20mA LEDS) and for this, the circuit is perfect.

So far, so good but, I have questions to ask for the group: what would happen with my 120 LEDs if Q1 openned its base or other else problems with this component ? There is a way to introduce a protection for them?

and more: How this circuit would behave if I strobe the LEDs feed with any flip-flop? it seems that the voltage delay a bit to fall when I just insert the leds... so, in a pulsed condition I belive the leds power will rise a bit.

Thanks

[johnjuan]

Not sure if I understand right - does the Zener in fig. 3 make so the circuit will output more than 20v, or just intake more than 20v? I want to connect to 24v supply. Will Zener be enough, or do I need a better NFET? If I need better NFET, can you suggest one capable of 24v or more?

[toyds]

i want to make a simple high power led driver for this one

Input supply : 12V-13V motorcycle battery
# of LED : 2 LED in series at around 350mA


does this driver be possible to use?please reply
thank you very much

[rguzmán]

Hi, i just tested it with a IRF520...

works fine, too cool really

the IRF520 (at 800mA calculated R3) was "cool", no heat at all

so I tested the consumption of the three 3W leds I use... only 540mA with R3= 0,66 ohms

so I changed R3 with a New Voltage Drop value... to get close to 800mA. R3= 0,45 ohms. But Consumtion was only 570mA.

could it be that with IRF520 or any mosfet I would need to use a really low R3?

or should I test another NPN (I'm using MPSA18 which has much more gain than 2N3904)

[rguzmán]

the Voltage at GATE from IRF520 was 10,2 , but at SOURCE onle 0.29V...

maybe here is the problem,.... as I read that on BASE from the NPN there should be more than 0.5V for it to work right.

[kmantesla]

Great instructable, I just threw this together to get the hands on experience of a current source. I designed single led desk lamp driven at just under it's max rating 1.0A. I am very pleased with the results, and I wanted to chime in on everyone's rants on efficiency, because most people don't understand this circuit and real world efficiency. Granted, it's not the most efficient circuit in the world, but it does EXACTLY what it's designed to do, change your input to an exact output, keeping the leds happy and healthy.. but at a cost.. all the "changing" is converted directly to heat, not an ounce is saved.. My circuit uses a switching wallwart; 5.2v @ 1.0A.. we can safely assume this type of adapter has an efficiency of 85%, if not more. My circuit "inputs" 5.3v @ .90A and my load, the leds get 3.0v @ .88A. Yay, this is exactly what it's designed to do, my led is nice and BRIGHT and happy, but all that stepping down and regulation doesn't magically convert, not with this circuit, it's converted to HEAT. So, my "regulator" circuit has a weak efficiency of about 56%, coupled with the wall adapter efficiency, the total circuit is around 50% efficient. Yikes, thats a scary number, right? But this is why we use leds, because an amateur can still be "green" even with a lossy, dirty regulator like this.. My lamp is SUPER bright, at less than 6W. I guarantee you 10 of these would EASILY rival a singular 60W bulb, in brightness and cool factor no doubt. Plus, I built mine out of spare parts, you could easily increase efficiency ALOT by designing a circuit with specific values on the adapter and fet.. and my led was an old rebel, newer ones are even more efficient.

Great instructable! It reignited my desire to custom build drivers over store bought ones, so I can start to really push leds, like the newer XML's to 3amps..
Following this build, I immediately bought some high quality, low dropout, high EFF adjustable voltage regulators with built in buck/boost options all in a T0220 package, got them for around 2$ apiece. I currently use luxdrive's drivers for all my projects, but now I'm going to start playing around again and see if I can come up with something just as rock steady and functional as my beloved buckpuck.

[yohanc777]

Dear Sir.Your right and your wrong also. The right part you have figured out and said yourself.The wrong part is that when your using 5 v @1a input to suit ur 1 led need that entirely is your choice,but you shouldn't calculate the heat loss or efficiency in this instance as you have not yet used the fet to its minimum potential also.
I say this because when u run 3 or 4 led of 700mA each on a 13 v dc 3 amp supply.Dan's simple but polite circuit doesn't even heat up and it remains cool and efficient for even more that 36 hours at a stretch (my bench test) . It only heat up to waste voltage and current when you use 1 or 2 leds or 300mA or 700mA.
Dan's circuit is superb in almost all ways.Sometimes the most simplest things are the best things in life.

[kmantesla]

Well yes, but in the above post I was mentioning the "excess voltage"... 3 or 4 leds is going to be pretty much RIGHT at 13v so you're not going to have to drop much if any, hence your driver is probably doing just about nothing and you'd be better with one series resistor. In my case, if I hooked up just the led to my 5.3v source the led would be instantly fried, and this circuit is used to drop the excess, which gets converted to HEAT :)

[kmantesla]

-I wanted to add; people trying to calculate efficiency using any AC settings on an ordinary multimeter (less than 100$) are probably not getting accurate results... at all... Use DC for all measurements, and estimate the mains to circuit supply adapter efficiency.. Depending on the type of adapter and load, assume it can vary as much as 65-95%.. Good AC meters and true rms meters are hundreds of dollars. If you understand them, generic meters are basically worthless for AC measurements outside of simple residential mains checking.

[GrumpyOldGoat]

Without going through the laborious process of reading all 256 replies dating back several years apparently...

Would a recycled computer power supply converted to a benchtop power supply do the job?

You have a complete set of tightly controlled AC and DC voltages that could be used.

I raided the dead computer repair shop and collected quite a few and tested them. Most still performed flawlessly.

[Spuzzum]

This is exactly what I have just done.. a 350w antec, and will be spreading my led's over the 5v, 12v1, and 12v2 rails. If just using palin resistors, and splitting up over 3.3v, 5v, 12v1, and 12v2.. I calculate 92% efficiency. That's for 127 Cree's at 350mA. But I want PWM capabilities.. so went with this driver. Can't run it on the 3.3v rail, but 5v and the 2 12v rails have plenty of amperage to offer. Just don't go over the maximum 10 amps the psu's 18 awg wire's good for. But at 4 or 5 lines per rail, that's still quite a lot of power to be tapped. :D

I mentioned in my last post that it wasn't working on my 5v line.. it is.. must've been a ghost. And it's working with just the 100K R1 resistor as well, didn't need the 22K.. nor did it need the zener. At least not with the 5v and 12v sources the psu gives.

Just need to look at the specs on the psu (usually on the side, or on the site). Make sure you don't go over the maximum amps per voltage rail, and note that 3.3v and 5v rails are "combined wattage" maximums. And some have "combined maximums" on the 12v1 and 12v2 rails as well.

As for the 10ohm 10watt "dummy resistor" needed for the bench supply mod.. I found using an led and 470ohm resistor on both the 5v and 5vsb lines for the "power on" and "standby" led's. Uses about 1/2w. If you look on your motherboard, you'll see (or should see) a green led.. that's your pc's "dummy load". That's also connected to the 5vsb (5v standby).

For 127 led's at 700mA, I calculate about 15 drivers over the 5v and 2 12v lines.

[yohanc777]

3 leds work best on the 12v lines.(3 volts or less go for the consumption of the cc regulator) So dont bother with the 5 volt or lesser voltage lines. depending on the ampere required play around with R3.

[Spuzzum]

Cree XP-E Reds use 2.3V a piece.. 2 LEDs per 5V line wastes less than 5 LEDs on the 12V line. The Cree XP-E Orange-Red uses 2.1V. Osram Golden Dragon Plus "Hyper Red" uses 2.15V (@ 400mA).

So if you "only" use the 12V line, then you have wasted voltage. You'd still have waste using the 5V line, but less. And although it's only 2 LEDs in series on the 5V line.. you can wire up more in a series/parallel set-up. There's quite a bit of Amperage available, so even though each series is only using 4.2-4.6V per 5V line, you can still fit up to 9Amps worth. Each voltage line is good to 10Amps, but use less just to be safe.

Solution 0: 2 x 3 array uses 6 LEDs exactly
+----|>|----|>|---/\/\/----+ R = 1 ohms
+----|>|----|>|---/\/\/----+ R = 1 ohms
+----|>|----|>|---/\/\/----+ R = 1 ohms


Also.. don't forget.. you need a "minimum load" on every rail.. 3.3V, 5V, and the 12V. It'll work without it, but it's safer, and better for the psu, if you ensure the minimum load for each rail.

[GrumpyOldGoat]

I think you just raised my IQ by at least 2 points with that response.

THANKS My Friend. Good information to keep handy!

[Spuzzum]

You're very welcome. :)

[rguzmán]

Hi,

i'm trying to calculate the dropout voltaje of a JFET or MOSFET...

any help to build this?

[csepzol]

Can anyone help me to calculate the R1, R2, R3, R4, Q1, C1 with an lm317?

[nrao2]

I need to rig up aLED white cold light source of approx 100w, what best circuit you would recommend & the White LED part.

[bau05]

Hi Dan

I use a group of 32 Cree MC-E leds(4 in series x8 in parallel), mounting in a XLarge heatsink with fans.
An IRF3710(Vd=100V,Rds=23mΩ, Ιd=57Α) drives these leds.
All work fine, except that leds open/close "unevenly" during first/last steps(Sunrise/sunset mode)
(it isn't a PIC problem, cause i tested with four 5mm simple leds that worked "smoothly")

Q: Can i replace the two resistors(100R, 10K) with your circuit and drive safely the leds?


*Vtot=12,5Volts/Itot=3,9Amperes

*See IRF.sch below.

[spark light]

It sounds to me like you just completely paralleled the led's.
Here's the problem:
a constant current source works by changing voltage to stay at set current.
because of this, you have to make sure each led shares current evenly, as they each have different turn on voltages. you can do this by putting a small resistor in series with each one. (10 ohms) The ideal solution would be a separate driver for each, but that might be expensive.....

[bau05]

A friend gave me yesterday his ebay cheap Led Wireless Dimmer(with remote ctrl). It uses an smd ATMEL microchip inside.
After testing this i noticed that ATMEL doesn't minimizes the light at all, plus remote ctrl step to step cliks are different.
So with a digital voltmeter, i deside to measure the PWM output steps.
1)My PIC diy dimmer steps: 0.00>0.01>0.03>0.05>0.07>.....>5.02 Volts
2)ebay Atmel dimmer steps: 0.02>0.04>0.08>........>4.98 Volts

That's the answer:
the different voltage PWM steps in each microcontroller's software.
A smooth opening light mode should have 0.00>0.005>0.01>0.015>0.02>0.25>....voltage steps(maybe this will be done with an analog potentiometer).
I guess that there is no solution, except someone re-programs the .hex file. Cause i dont "speak" the PIC-language, my circuit will stay as is.
Anyway thanks for answers.

------
*To help others:
A 10x10cm IRF heatsink gone useless. IRF+heatsink were TOO HOT!
I cooled the burning IRF, mounting it to a cheap 5x5cm PC VGA heatsink fan!
IRF now has the room temperature!

[polilies]

Hi dan!
firstly thanks for your project!it's great, i have try it few days ago. and it works with one led.But the NFET heating too much . Do you know something about it? i've try it with 5v zener. and when i tried the circuit with three power led it wasn't heating but the leds were not bright enough.actually too dead.. ?

[afuller5]

I ran across this post recently as I kept buring high power LED's or had very hot resistors. So, I tried this method, I already had a fet connected in the manor shown in Circuit 5. So I disconnected the Source on my fet and put R3 (3x1ohm 1/2 watt resistors) and the Q1 parts in. For the FET @ Q2 I was already using a NTD5806N and had a 3904 transistor laying around to use for Q1. R1 from my ATTINY85V was 500ohm as anything more and the FET doesnt fully turn on. So, I rigged this up and tried it to limit the current to 1.6Amps on 13.8vdc,. I fried yet another LED. When I put my meter in the line I was hitting close to 8 amps of current! I know I did the calculations right.. Any thoughts and wisdom?

[suguimoto]

Hi Dan! Thank you very much for your article!
I´m building my own RGB LAMP, I bought a 9W (350mA per color) RGB 6 Pins (Chinese-Star) Led and i´m using an Arduino to control it.
My main problem was that I couldn´t find many of the IC described above, so I bought similar IC´s that are working quite fine:
Q1: BC 547
Q2: IRFZ44N
R1: 100k 1/4W
R3: 1ohm 1W

I really dont know how efficient it is, but it´s working fine and the components didn´t get so hot.

I hope this can help some other people that are willing to work with high power leds.

[maqs.a]

Hi,
Can u pl's share the circuit diagram of what you have done?

can it be used with 1w LED?

[adi1977]

My dear friend

I dont have enough knowledge in electronics so i'm asking you for help. I want to make a LED Tube on a PVC batton with 1watt or more wattage of an led when given a power supply of direct 220 - 240 Volt so that a single tube lightens up the entire room.


My dear friend can you please design a circuit for me and mail me along with all the parts to be bought and their specification . (i.e. resistance of how much wattage, capacitor, diodes etc.)

A little help from you will be highly appretiated.

Thanking You

Adit Pal

[erikp]

Can someone help me out regarding the proper heatsink to use for Q2? My LED (part number 475-2581-1-ND on digikey) draws 1.4A with a forward voltage of 2.5V. I am using a wall-wart that gives me about 10V. My LED runs smoothly, but the two heatsinks I've tried using with Q2 get *very* hot to the touch. I bolted both of them on using thermal compound. Both are "U" shaped. One is about 2cm tall by 1cm wide (about the height and width of Q2). The other is about 3cm tall by 2.5cm wide.

[burton2663]

i have a vu meter using 3915 with 10 small LEDs, im trying to find a way how i can convert those little LEDs to 3w LEDs using simple transistor drivers. any diagram is appreciated thank you very much

[drbogger]

Will a 1N4733A 5.1 volt zener diode with a Voltage Forward (Vf) of 1.2V @ 200mA work?

There is also one that doesn't specify a Vf... Which one should I get?

[apocello]

Maybe I'm just missing something but how do I determine the value of R2 if I'm not concerned about interfacing with a microcontroller but need an input voltage greater than 20v?

[letniq]

Hi, Great article but I didn't understand how to calculate the how many ohm should be R3. For example I'm having 36 3W power leds. 12x3 leds Source: 12V PC PSU LED Forward: 3.2V LED Current: 750mA I will have 9A through Q2 (IRF3205) Q1: 2N3904 R1: 110K It will be great if someone can help me. Thanks.

[wolf1728]

I see my website has been referenced. A month ago, the domain name 1728.com was stolen from me and so that link should no go to parallel resistor calculator

[pmj_pedro]

If i use one of this circuits like the 5, and d the supply font is 18V, and the 4 high power leds only need 13.6V, should i use a resistor or something to not burn the leds?