Driver interface (fast and cheap) for a high power led

[djsoftlayer]

Hi guys, im very new at this, so, i guess this question will be easy to answer. I have a picaxe who outputs a very simple rectangular wave from one out pin. This wave have this times:

Ton: 0,01 s.
Toff: 0,3 s.

As you can see, its a wave for strobo applications. I have a power led that i want to connect to this output, but the led needs 3.5volts and 350mA.

My Vcc is 5V and can manage that level of current, i live in a very poor country, so the materials for making this driver must be very common, like an 2N2222 for example.

Other thinks to take care, is the times, is a relative fast wave, so the driver component must react faster than this times. Thanks for you help!

 

[JezWeston]

Hi there,

For a 350 mA LED, you will need a constant current driver. I've been playing recently with the MAX16836 chips. They only require two extra capacitors, although the chips themselves aren't that easy to come by. They can handle a 10 kHz PWM signal, so I expect they can turn on pretty fast.

There are lots of other LED drivers available, you can find a table of some at:
http://www.element-14.com/community/docs/DOC-3986

 

[djsoftlayer]

Thanks but no.

Thanks for your answer but i need a solution with common components like discret semiconductors like transistors.

 

[Technical]

Why not just use a ULN2803A darlington array like on most of the PICAXE project boards. This will work for you and is cheap!

 

[Jaguarjoe]

With a Vces of up to 1.6 volts at 350 ma for the darlingtons in a ULN2803A and a Vf of 3.5 volts for the LED, there is no room for a current limiting resistor if the OP sticks with his 5 volt power source.
A 2N2222 isn't a lot better at 5 volts.
A logic level MOSFET may be the next best bet but can the OP get them where he lives? Or, can the OP jack up his power source to a higher voltage?

 

[matherp]

You can use a LM317 regulator to create the constant current source if you have a voltage of over 7V (3v dropout + 3.5V forward voltage on the diode) available before you create your 5V supply (see the LM317 datasheet for the wiring). Then you would just need a npn transistor to turn the current flow on and off.

 

[lbenson]

Perhaps more experienced hands could make clear why this would not work in the long run.

Because the duty cycle is so low (1/30th--10ms on and 300 ms off), I thought it might be possible to overdrive the LED with respect to both current and voltage. I had a 1-watt LED that had been in my parts box for nearly 3 years, and I was willing to sacrifice it.

I set up the switching with a logic level mosfet--an IRL540, but any should do--and provided current limiting with a 10-Ohm, 5-Watt resistor, which running from 3 AAs at a measured 4.65 volts should provide .465mA. A 2-Watt resistor should be fine, and with the duty cycle, probably a lot less. I ran it for about 10 minutes. With no heat-sinking, none of the components got hot--all appeared to be at ambient temperature. I doubled the current by replacing the 10-ohm resistor with a 5-ohm--still no problem.

Here is the program.

#picaxe 14M

main:
  high 5
  pause 10
  low 5
  pause 300
  goto main

I didn't have one, but a small, inexpensive logic level mosfet which would probably do the job is the IRLML6401 (sot-23-3, but probably manageable for most hobbiests, and $.43 each in quantities of 10 from Digikey in the U.S.).

I don't know how fast the transitions are--if they needed to be sharp, perhaps a mosfet driver would be in order.

If those with experience of low duty-cycle use of high-powered LEDs can say that this would not work in the long run, I would be happy to hear about it. Otherwise it might provide an inexpensive and simple solution.

 

[hippy]

Over-driving LED's and 7-segment displays is commonly done, on the basis that if it has a particular current all the time it will, on average, be the same as twice the current for half the time. This is the principle used in multiplexing LED's and 7-segment with no current limiting resistors.

Whilst it does work, and often with no appreciable adverse effects, running anything above spec can cause long term or permanent damage. How to quantify the risk without a lot of detail about the component is almost impossible. Apparently one issue is 'thermal shock'; too much and you are repeatedly heating internal junctions which may cause failure. Heating effect is almost certainly proportional to time, so the less time the overdrive occurs the more likely it is not to heat up.

It's a bit like running a marathon. You can do it at a constant speed. You can run in bursts at twice the speed with the same length of pause, run ten times as fast and take nine times the pause. But it's hard to say what any individual's particular limits are, what someone could cope with and what they can't. Which tactics will have them finish the marathon, which will have them collapsed before they reach the end.

 

[Dippy]

Personally, if I were a newbie , I'd just use a resistor.

A linear based cc supply is (arguably) just as inefficient and a lot more bulky and needs the extra V headroom.

As stated above, the killer is heat which results in unpleasant elevated temperatures, which, eventually, may kill the device.

The best way would be to modulate a S/mode design OR just go and get one of the those LED driver chips made by Resistors Inc (ex Zetex) or National and modulate that. FAR more efficient, but difficult if you are a novice.

You should easily be able to calculate the heat dissipated in the LED and resistor and any transistor used for switching.
You have Ivy watts and I-squaredR and you have the duty. Easy peesy.
I , too , would use a MOSFET.

Whether or not you need a proper driver depends on the MOSFET chosen and speed of switching required.
300mS on and 10mS off is not a fast frequency but do you want the switching transitions to be super-quick? (Why?)
For low current (<1A) MOSFETs most will have very low Gate / Input capacitance.
And THAT is one of the things that defines whether you need a pukka driver or not.

Designing a circuit is a holistic thing.
(Horrible word but it gets the message across).

 

[SAborn]

I have had no real problems running a 5v picaxe down to 3.5 volts, so as matherp said why not just use a LM317 vreg set to 3.5 volt, for the whole circuit and drive the led with a simple transistor.

No current limiting required, Lm317 vregs are everywhere in junk so should be able to salvage one of them.

Then its just a matter of setting up the vreg and adding a drive transistor to the picaxe output.

 

[Dippy]

Which is easier?

1. A regulator plus 2 x res and caps to produce a regulator.
OR
2. A single 4.7 Ohm 1 watt resistor (standard cheapy half watt 4R7 would be fine at that duty).
R= (Vsupply - VfLED - VfTr) / I

Good luck, I'm sure you'll get this easily sorted.

 

[matherp]

In my experience it is better to drive these high power LEDs properly with a constant current source rather than just using a resistor. I've experimented driving a 10W one with my bench supply and the voltage changes (reduces) significantly at a constant current as the device fully warms up and heat soaks. With a resistor that limits to the "warm" current requirement you may find that it doesn't ever get to full brightness. With a lower resistor that provides the full current cold you are going to be significantly overdriving when it is warm. I use a PT4115 to provide the constant current (just needs an inductor, a resistor, a capacitor, and a schottky diode to complete the circuit) It has a PWM input which I control with a 08M to give a clean dimmable light. In the diagram I use a nominal 12V supply on J1. The 5.1V zener creates the supply for the 08M and the LED connects to J2.

let dirs=%00000101
do
readadc 1,b0
w1=b0 ' scale 0-255
w1=b0 * 256 ' scale 0 - 65280
w2=w1/164 +2
pwmout 2 , 99, w2
pause 10
loop

 

[BeanieBots]

...My Vcc is 5V and can manage that level of current, i live in a very poor country, so the materials for making this driver must be very common, like an 2N2222 for example.

Other thinks to take care, is the times, is a relative fast wave, so the driver component must react faster than this times. Thanks for you help!

With this criteria, I think a simple series resistor is a blatently obvious choice.
matherp's comments are correct but by simply sacrificing a little on maximum brightness a simple resistor (as per Dippy's calculation) will do nicely.
The 2N2222 transistor should also be OK for 350mA.

 

[matherp]

djsoftlayer

The circuit I posted above will of course completely meet your original requirement if you can source the PT4115 as the program to do your flash pattern is very simple.

Best regards

Peter

 

[gengis]

The classic simple two transistor linear current regulator might meet the needs (common readily available transistors) Q1 and 2 are not critical, so long as Q1 can dissipate the power required and carry 350 milliamps of current.

http://en.wikipedia.org/wiki/Current_limiting

For 350 ma Rsens would be ~2 ohms (chosen to drop ~.7 volts to turn on Q2 which effectively removes the drive signal by shorting it. Rsens would dissipate 1/4 watt so I'd use a 1/2 watt resistor. The dissipation of Q1 will depend on how much voltage it has to stand off when limiting current. A five volt supply, with a LED having 3.2 volts across it is about 630 milliwatts so it may need some heat sink.

They show the circuit drawn with the load (your led) connected to ground - that isn't ideal and you'd be better off putting the load between Vcc and the collector of Q1. (Q1's emitter to ground)

The drive circuit must have a current limiting resistor since the current limiting transistor Q2 shorts the drive voltage to ground. It is needed to protect both the axe (or external +V drive signal) and the transistor. A 470 ohm 1/4 watt should do it - connected between the external drive and Q1's base.

When playing with expensive high current LEDs and setting up the current regulators etc. a string of 4 or 5 series connected rectifier diodes makes a good dummy load until you verify that the current is being limited and nothing is getting unduly warm. Another caveat - if you are running this in a warm climate, pay attention to the current derating specifications for the LED you are using, and choose the current limit point with temperature in mind.

 

[BeanieBots]

djsoftlayer

The circuit I posted above will of course completely meet your original requirement if you can source the PT4115 as the program to do your flash pattern is very simple.

Best regards

Peter

Except it's more expensive, harder to obtain and won't run off 5v.
Other than those minor details, it's fine