High-Power LED Photoflash

What is the ACTUAL minimum "on" (emitting) time that an LED can be driven to?

I've got a couple of 10W LED's on their way, should be arriving any day. Let's assume I have reasonably appointed drive circuitry, which allows for what would be PWM, except for my purposes, it's only going to be a single pulse. I'll be using it for up-close stop-motion photography as a prototype.

Unless I misunderstand, LED's rise time from power on to emitting is virtually zero, and the transition back to dark is instantaneous for all purposes also.. but what's the ACTUAL minimum "on" time that would actually produce output? There are a number of commercial photography flash units and panels based upon LED's now, either single emitter units or many discrete LED's. These mostly cater to driving huge amounts of light with little heat and faster cycle time, but what about actual minimum flash duration?

Another consideration is overdriving/cooling. As I understand it, the main monster is thermal breakdown.. but a single pulse is much less an issue in this respect. CREE and other makers have "out of spec" documents discussing driving their emitters many times their rated values, pulsed.. I wonder just short/powerful a pulse I could really get out of a 10W emitter?

I either don't understand what I am looking for, or can't find it.. thoughts?

You need to look at the data sheet of the LED, this will tell you. It is not instant, a lot will depend on the capacitance of the device as to how quickly you can get the voltage rise into it, as well as the impedance of the power supply

Okay.. so it's not really "instant". I'll do some practical testing.. taking photos is a good way to clock it, actually. Reality sometimes trumps theory. I figure getting a pulse duration well under a millisecond should be easily done, and it looks like that's well within reasonable limits of the emitter.

As the power increases on these and price drops, I wonder how much of the photo market is heading this way. More to come, I'll post results of some testing once it gets here and I twiddle with it..

I figure getting a pulse duration well under a millisecond should be easily done

Yes I think so too, but I haven't done this with power LEDs only little ones.
I hope you are using a constant current driver rather than just using a resistor to limit the current.

Resistor? I was just going to use a car battery and an old relay.... is that bad? :grin:

Actually, that is a question. I do have a constant current driver, it runs on 120vAC. It does NOT however have a PWM input, it's from a commercially sold LED fixture I purchased (I posted on it a while back-- remember Heather?). My plan is to switch the output from that using a simple NPN transistor (maybe darlington). There's going to be some rise/fall delay with the transistor, I know.. but I'll play around and see what I actually get out of it in terms of performance.

Though less output than a Xenon strobe of course, I'm interested in seeing if the duration of the flash might be a real boon to using LED.. stopping a moving object (photographically) is a lot of fun, but the faster the event, the shorter the flash has to be to "stop" it...

I have not considered a one-shot high powered LED before, but maybe constant current is not so important as the diode will have plenty of time to cool...

Please look at Avalanche transistor - Wikipedia and consider it rather then your darlington. I'll bet what you want is the "Capacitor discharge avalanche pulser" described there with the addition of a current limiting resistor. WRT selecting the resistor value, you may want to over-drive the LED a bit since the pulse will be such a short duration -- at least look up your LED's pulsed, not constant current rating.

That circuit at least gets you a fast rise time. Now how to create a fast fall time? Maybe some kind of uP controlled short to ground using another avalanche transistor?

Arduino compatible constant-current LED driver: www.toastedcircuits.com

This idea is being expanded upon, but as it does, I need a bit of clarification:

An LM317 adjustable voltage regulator can be used as a constant-current source. I understand that for durability, it's preferable to use a switching regulator circuit, but that they also tend to be more expensive. I also am aware that using an LM317 as a constant-current driver isn't very efficient, a lot of power gets wasted as heat. The thing is, they ARE cheap, and setting different current values only involves changing a resistor.

What I need clarification on is if I as a hobbyist should worry about it, as LM317's are cheap and plentiful, as are heatsinks. Because I really don't understand why the regulator is working this way, I also don't understand what kind of load rating the regulator should have in relation to the load on it. This 10w LED is a good example.. what rating (current) LM317 do I need, if we assume I take efforts to get the feed voltage fairly close to the LED's required voltage?

I have some 100mA LM317LZ .. little plastic buggers, but they sell for 30 cents delivered in two days and are dang useful. What kind of load could I expect to drive with a 100mA rated regulator? 100mA? or is that the amount (max) that I can dissipate with the regulator itself? (which would make the potential load considerably higher)... or do I need to go and get a pack of LM317T rated for 1.5A ? I do like the idea of using adjustable voltage regulators as constant-current sources (even if inefficient) because of course it allows you to kill two engineering needs with one stone..

In addition, I'm shopping around for cheap N-channel logic-level MOSFET.. something rated not too crazy in voltage, but able to handle up to say, 40v.. but at decent currents like 5A or more. What I'm looking for is what I'd call a "hobby level" MOSFET, cheap.. 5v Arduino-easy.. and can handle loads like large wattage LED's, stepper motors, etc. I still don't quite fully "get" the MOSFET thing, still trying to get my head around it. Not understanding doesn't have to stop me, I'm going to go ahead and try to use one and see what kind of trouble I'll need to figure my way out of...what's confusing me is that they seem to refer to a switching voltage of 10v.. I need a part that takes 5v or below, yes?

In fact, since I'm doing an order with them, which would you get for general purposes to have around:


I'm also probably getting some TIP31, as general purpose NPN power transistors to have around.. good choice?

What kind of load could I expect to drive with a 100mA rated regulator? 100mA?

Yes that is the current rating, probably limited by the bond out wires.

or is that the amount (max) that I can dissipate with the regulator itself?

No that will be another value, it will be in the data sheet.

do I need to go and get a pack of LM317T rated for 1.5A ?

Probably better.

As for the FET look at the data sheet for it saying "logic level" or specifying the on resistance Ron at 4.5 or 5V. Ignore the gate threshold figure.

All right, since the parts arrived this morning, I decided to slap together the 10 watt LED with it's heat sinking and Constant Current driver made from an LM317 Voltage regulator. Note that this is a VERY inefficient method of providing regulated current, but it does work, and is cheap...

Ghetto LM317 - based Constant Current Driver with 10w Power LED... this dang thing is BRIGHT... rated 900-1000 Lumens... somewhere in the range of a 65watt incandescant bulb's output- except the LED emits that over 160 degrees rather than 360, so the apparent intensity is even higher. By eyeball, I'd compare it to a 75 watt incandescant floodlight, maybe a bit more, except for the very cool light color- the LED is 6500K. For reference, Xenon flashtubes operating at their theoretical ideal of 6600K can produce approximately 90 Lumens per Watt, but that of course is only in extremely short pulses. If we can get 900-1000 Lumens (rated output) from the LED, it will be performing as efficiently (if not more so) than a standard short-arc strobe tube. Once the switching circuitry is in place, I can then start experimenting with pulse width and actual minimum illumination time. At least at low power levels, it would appear that these power LED's will be giving Xenon a run for it's money as a light source... but only if the flash duration can be manipulated well.

Cost of build so far:
10 watt LED - $1.00 on eBay auction from Hong Kong - I got a deal, but normally like $7 now
LM317T 1.5A adjustable Voltage Regulator $ .36 (mail order.. 20 for 6.99)
1 Ohm and .5 Ohm 2 watt resistors $ Free Salvaged from a printer circuit board, if bought, maybe $2 total
CPU Heatsink $ Free Salvaged. Just about any chunk of metal (Al or Cu) will do

So, even paying full prices of $7 for LED and $1 each for the resistors (outlandish!), for under $10 you can assemble a BLINDING regulated Power LED, set up for typical voltages found in an automobile or some wall warts. I've managed to cobble this together so far for a grand total of under two dollars :wink:

Metered out the resistors to 1.5 Ohm, so that provides for (1.25/1.5) = 833mA regulated. This should be right in the slot in terms of keeping the LED/Regulator relatively cool while giving relatively close to rated output. There's likely to be a little resistance drop with heating, so I'm okay with this power level given the crudeness of the circuit! I may eventually reduce the resistance to 1 Ohm, which would give 1.25A (25% over rated max) when using it as a photo flash.. as the short duration of use will take away most of the heat dissipation concerns. I've used no heatsink compound (none in the bins), but the component's heatsink tabs are screwed tightly against the Aluminum CPU heatsink and this seems to be enough. A loop of uninsulated copper wire also holds the two resistors firmly against the heatsink to help keep them cool also. After ten minutes of operation at 13.5v from a car battery, the heatsink was only mildly warm to the touch and the components themselves were comfortable to hold a finger on.

I do however have a question.. when I meter the LED, I see only 9.3v across it when connected to a 12v wall wart that meters to 17v with no load and 13.6v with the led circuit connected. thoughts?

Now to add a logic-level switching circuit. I've got a few power transistors around, only need something rated for like 2A @ 20v, I'm sure I can find something in the bins that will work..

when I meter the LED, I see only 9.3v across it when connected to a 12v wall wart that meters to 17v with no load and 13.6v with the led circuit connected. thoughts?

So you are "loosing" 13.6 - 9.3 = 4.3V
This is made up of the volts drop across the current monitoring resistor and the volts drop across the regulator. Sounds about right to me.

Sure electronics sells lots of nice LED constant current modules, most can be controlled with a TTL voltage level PWM signal. I see 1,3,5, and 10watt models shown here:



Yep..seen the ones from Sure and others.. those are switching-type regulators, right? A lot more efficient.. but for a 10w one (and the one they sell is wrong voltage for this LED) they want at least $11 + shipping... about $13 total.. and that's just the driver. Add the LED (and their prices are a little higher than some) and you are over $25.

Even if you paid $1 per resistor and $1 for the regulator (robbery on both fronts), the circuit above is $2. Add a TIP120 (I have a couple around, so that's what I'm using) Darlington for $1 and you are up to $3 for the PWM-friendly driver. Add on the LED and you are totalling $10. Add on 50% if you don't shop around or want express shipping- still only $15.

Now, I wouldn't want to build a home lighting fixture with the LM317 circuit.. but for something that is only going to be "on" for very short periods of time, efficiency (and even LED life) aren't as much of a concern as just keeping it cheap. A lower-wattage setup might be useful like this however.. I have a pack of LM317L, which handle 100ma- which is a nice level for driving a handful of discrete LED's in series from 12v.

I'm glad I "invested" in the LM317T's, got 20 for $6.99. I already keep around 7805 and 7812 (5 & 12v regulators) but for some reason never really considered just having the adjustable ones on hand. The price is only a few cents more each for the adjustables, and they can do a wide variety of tasks rather than just being a one-trick pony. The 100ma version only cost something like 20 cents, the 1.5A version about 35 cents each.

I knew about the 3v regulator drop, I wasn't aware I was going to lose some to the resistors also. I know I'll be adding another 3-4v drop with the TIP120 (drawback of using darlington), so I'll need to keep that in mind. I think I may have a couple of regular bipolars ripped off a circuit board with a decent enough current rating, otherwise I only have 2222's rated at 800mA. Can you run transistors in paralell to spread the load, or would they have to be perfectly matched, etc to prevent thermal problems on one of the parallel transistors?

Can you run transistors in paralell to spread the load

No, not unless they are thermally linked like being on the same substrate, but you can do it easily with a FET.

So, if I understand right, the advantage in this case to using a FET would be that the voltage drop would be less and the overall current capacity would be higher.. in short, the transistor won't be "working" as hard and wasting as much as heat... yes?

Yes sort of.

I wouldn't say working as hard, it is just that the work differently to leave a constant voltage across them when they are on, where as a FET has a constant resistance across it when it is on. For a given current a FET will generate less heat because the on resistance is very much lower and so will not produce as much voltage drop.

A lot more efficient.. but for a 10w one (and the one they sell is wrong voltage for this LED)

Again with the voltage parameter ]:slight_smile:

A constant current driver cares little about what voltage you are using for it's input as long as it's higher then the forward voltage drop of the LEDs that it is driving. One in theory supply 50 volts dc to the input of the driver that would then 'force' a constant current, say 1 amp, through the led string. Bare forward Led voltage is a characteristic of the led not a operating requirement specification (other then the source voltage has to be higher).