Hello,

If you are using the 3 Watt powerleds, keep the cooling of the leds in mind.
Without cooling the leds the lifetime will be very short.
They are mounted on an aluminium frame that can be mounted on a cooling frame.
When using a LM317 on 12 Volts for 1 led of 3.8 Volts (estimated for a white led),
the LM317 will dissipate about (12 - 3.8 - 1.2)*0.7 = 4.9 Watts, so this needs cooling also.

Greetings,
Bertus

 

hello bert,
I've mounted LED on an Old CPU heatsink (maybe thats an overkill for LEDs) so heats not problem.

Letting the LM317 drop out 4V makes no sense... Its wasting more energy than the LED itself.

I am on a lookout for Lower Dropout Chip that allows 3 LEDs in Series on 12v.

3X 3.4v= 10.2
Spare voltage left = 12-10.2 = 1.8

Any controller that works in 1.2~1.8 range?

 

he is using a linear CCS to drive the chip. and he can actually do better than that. But if the LEDs are running off at 2.4v@350ma, and powered by 12v, he is dropping 9.6v on the lm317. at 350ma, that's 3.5w of heat to dissipate.

what he can do is to put a few LEDs in serial. say that the lowest you will see is 10v. the LM317 needs 3 - 4v to work comfortably. that means you have about 6v left. That can drive 2 2.4v LED.

if he did that, his LM317 will see 12-2.4v*2=7.2v, and his power dissipation would have been lowered to 7w, a 20% improvement.

the basic wire-up comes from the datasheet: you just put the LEDs from the adj pin to ground.

 

BTW, that smps also has a 3.3V output, Shall i connect the LED directly to that?

that's a terrible idea. the purpose of having a resistor is to help you define the current: the resistor provides a negative feedback to help stablize the current.

if you don't have it, and the slightest change in the voltages, either from the power supply or the diodes' forward voltage as they heat-up or age, will cause tremendous amount of current fluctuation through your diode.

a disaster waiting to happen.

 

Letting the LM317 drop out 4V makes no sense... Its wasting more energy than the LED itself.

take a look at their datasheet. the minimum voltage differential is 3v, and can be as high as 3.5v (for 2amp out), or 2.5v (for 0.5amp or lower).

I am on a lookout for Lower Dropout Chip that allows 3 LEDs in Series on 12v.

3X 3.4v= 10.2
Spare voltage left = 12-10.2 = 1.8

Any controller that works in 1.2~1.8 range?

for that work work, you need a regulator that can deliver that much current, as well as being absolutely rock solid at its bottom range.

LT is usually very good in that area.

 

here is something simpler than what you had, and more robust but less efficient.

it is a typical CCS used in audio amplifiers, Nelson Pass uses a lot of it in his amp.

the output current is defined by 0.7v/R2, in this case, 2amp.

you can put your LEDs in serial or in parrallel. and if you do put them in parrellel, as I did here, you need those current sharing resistors (R3..R5, 0.22 - 1ohm).

Q1 can be any small signal BJT, and M1 can be any medium power mosfet (irf510 or bigger).

R1 can be anything between 1k - 100k. 10k may be a good compromise.

the mosfet should dissipate at least 5v for the current to be reasonably stable. so you are wasting a lot of power.

this should be a lot easier and simpler to put together than a lm317 based solution.

 

the minimum power supply voltage, for this circuit to work, is around 5v + Vf, where Vf is the LED's forward voltage drop.

you can also use a BJT (like 2n3055), which will be more efficient and work at lower supply voltage. however, those transistors usually have low beta so you need to scale back R1, and use a medium power transistor (tip for example) for Q1.

 

Thanks for the truck load of inputs.
I'd better work up my way around the market and see whats available. I live in a small town, so sourcing odball componnets is No go.

I've heard about the LM 1086 thats same ast LM 317 but works only on 1.8v, Is that true?

if that works... That Linear chip is wasting only about -
1.8V X 700mA = 1.26W...
When you're driving 9 watts worth of LEDs on it....not bad...is'nt it?

 

And please,
Don't forget that i am NOT trying to drive a single LED on 12V... Thats inneffcicent by the very common sense.

 

one big factor in doing anything electronics is that the circuit should perform with some certainty and it should depend less on active device parameters (which vary a lot) and more on physics / passive devices.

here is a demonstration of that example: two circuits, different mosfets, (vastly) different LEDs. the current going through the diodes is just 4ma different, out of 750ma.

 

I've heard about the LM 1086 thats same ast LM 317 but works only on 1.8v, Is that true?

there are a lot of ldos out on the market. the lt1086 series is a good one, also much more expensive. it also has a different Vref so you will need to recalculate.

as to its drop-out, check the datasheet.

if that works... That Linear chip is wasting only about -
1.8V X 700mA = 1.26W...
When you're driving 9 watts worth of LEDs on it....not bad...is'nt it?

that's not how the math works.

 

nice...
The major challenge in my area is What is available.
Even a Lm317 is godsend if i get it here. And its still better than running them off directly...

If if ask all the components you mentioned, what will is get is=
a bland, long stare on Shopkeeper's faces.

 

obviously, you have to play with the cars you are dealt with.

But if all you have is lm317, why bother looking for other LDO regulators that you don't have?

 

most (>90%?) of the LED drivers out there are made with boost switching mode regulators, either with internal switches or external switches for high powered one.

the advantage of using a boost dc/dc converter is that 1) you can configure the chip into a constant current source - that's good for the LEDs; 2) you can switch far more efficiently than you can with a linear regulator and 3) it is typically very simple and inexpensive.

here is an example of using MC34063 to drive 5 - 8 LEDs. the current is defined by Vfb / R3 where Vfb for this particular chip is 1.25v. With R3=63ohm, we have a drive current of 20ma.

the power source here is a 5v rectified dc that swings from 3v to 7v, at 100hz. as you can see from the two charts, the current going through the LEDs are rock solid, and the controller adjusts its output voltage to account for different LEDs it has to drive.

 

boost converters require that your output voltage be higher than your input voltage. In cases where you have a high voltage power source, and you want to drive just a couple LEDs, you should use a buck dc/dc converter.

here is an example where you can configure a mc34063 into a buck converter. Again, the drive current is 20ma, regardless of the number of LEDs driven (as long as the total forward voltage drop is lower than the input voltage).

 

there are tons of such chips and they are incredibly cheap: mc34063 is like $1 each from digikey and you can get even cheaper ones if you are willing to use newer dedicated chips. or you can go with TO220 chips if you want to deliver lots of power.

something like this offers lots of flexibility and it is easy to diy. all you need to do is to figure out a) which topology you want? boost or buck? and b) what's the feedback voltage on a given chip. after that, it is pretty much automatic.

 

dimming with such a chip is very easy: you just need to fool the chip into pumping more or less current, by manipulating the current on the Vfb pin. This can be done analogly or digitally.

here is an example of analog dimming. a voltage source higher than the chip's Vfb(th) will fool the chip into less current and a voltage source lower than the chip's Vfb will yield more current.

in this example, a 0.5v source caused the current to go from 20ma to 35ma - you can work out the math as to why.

 

here is how you do digital dimming.

Vdim is a pwm signal coming from a mcu, or a 555 if you will.

its duty cycle controls the dimming. the pulse train goes from a pi filter formed by R3/C3, and is then applied to the feedback pin. in this particular example 5v 10% duty cycle equates roughly 500mv, so we should be getting around 35ma, as we did before. Instead, we got 1ma, due to the loading on the capacitor - you can alleviate that by increasing R2 and R4.

the pwm pulse train doesn't need to be fast: anything faster than 200hz would do.

and the pi filter isn't necessary if your pulse train is faster than 200hz - but some chips may not function too well if the Vfb voltage is too stable.

 

any of the method would beat out linear CCS, or worse yet a linear driver that do not provide a constant current source to the LEDs: such a driver will severely shorten the life of those LEDs.

 

I knew I had it somewhere in my junkbox, .

here is a similar design using a slightly different chip: ncp3065. the topology is basically the same: step-up dc/dc converter.

I put on four batteries (5.4v power in), and put radioshack led (2.2v Vfwd, 20ma) in serial with a 1k resistor as the load. the multi-meter measures the voltage across the 1k resistor (23v) so the current is 23ma. that's equivalent to driving 11 such LEDs in serial.

the filter caps are 22u/50v in and out, and timing capacitor is 4.7nf (a little bit on the low side). the inductor is hand-wound with hook-up wires over a recycled torroid. I would estimate the inductance at around 33-47uh.

it is completely silent and the chip is cool to the touch.