Would the 12V 10W high power LEDs for flood lights, such as the one linked below, require a current limiting driver or something?
Can I drive it directly from a 12V power rail, via a MOSFET for PWM?
No, you cannot. You definitely need a constant current driver, or a pretty beefy resistor (regarding power dissipation).
Absolutely does require current limiting.
IME those normally reach the maximum rated current with forward voltage of like 10v (it's 3 strings of LEDs paralleled, each string being three LEDs in series - you can sort of see the 3x3 array of dies if you look closely. Note that those are 1W dies, and there are 9 of them, and yet it's rated for 10W; this is typical chinese vendor math)
You can buy the drivers on ebay/aliexpress, or build your own drivers. Here's just one example (it takes AC or DC input - if you know you're using DC, could improve efficiency a bit by shorting the appropriate diodes in the bridge) http://www.ebay.com/itm/10pcs-AC-DC-Driver-12V-24V-High-Power-Supply-for-10W-LED-Lamp-Light-to-DC-9-12V/261266778014 (no specific endorsement) - but this wouldn't let you dim the output (most drivers don't have that functionality.
A pair of AMC7140 (would need two in parallel) could be used in place of the MOSFET and would work fine, since you'd be dropping just 2v or so, each one would be dissipating a watt, drop one of those little heatsinks on each one and you'd be golden. Otherwise, you're probably looking at a switching regulator, which if you're building your own board makes the design a little trickier.
I have driven LEDs like that with a carefully adjusted constant voltage supply (relying on the dynamic resistance of the LEDs to limit current - this works considerably better with arrays of 1W dies than it does with smaller LEDs - the IV curve isn't as steep in the region of interest) - I don't exactly advise this, but it can be made to work. Key points are that you need to slowly ramp up the voltage while measuring current when you're setting it up to find what the actual voltage you need is (may not be the same for each unit), and it must be well heatsunk, because the forward voltage that will draw a given current falls as the temperature increases and you need to avoid that producing positive feedback, and you should target maybe 2/3rds of it's rated current at most, to give room for error without exceeding the specs. Obviously, it's much better to use an appropriate constant current driver, though.
Thanks, that's very useful information!
I've been testing one of these common anode LEDs by connecting it in series with beefy resistors (between cathode and ground, if it matters), to a 12V 1.25A wall wart. Using one beefy 1 Ohm (50W) resistor, I measured up to 0.78V across it, which should equate to roughly as many amps, or just over 9W at 12V. I realize now that this would potentially damage the LED or overpower the wall wart. I think I was lucky, nothing seems damaged.
I don't have any high load constant current drivers at home, but I do have many unused TLC5940. While these would not be able to handle much heat on their own, or current per output, they should be possible to daisy-chain (either multiple outputs or even multiple chips) and also allow for dimming (either via greyscale PWM or via dot correction output current restriction).
I've tried to wrap my head around how much heat load this LED would actually put on one TLC5940 at close to full brightness (~250mA per color). To calculate the heat load, I assume I have to know the voltage drop from each LED color channel, which multiplied with current yields the wattage which the TLC5940 will need to dissipate.
With a large resistor as a current limiter, I measured approximate forward voltages of 5.5V for red, 7V for green, and 8V for blue. 6.50.25+50.25+4*0.25 = 3.875W, and the DIP-28 TLC5940 is rated at maximum 2.456W.
This means there will be far too much heat to dissipate in one TLC5940 on its own.
Maybe I can use MOSFETs to take some of the load?
www.ti.com/lit/an/slva280/slva280.pdf
Using MOSFETs and a 3.3V VCC for their Gate, the voltage at the TLC5940 output can be greatly reduced. I would also need outputs connected in parallel (for more than 120 mA per channel).
Would I need one MOSFET per channel, or can I use one MOSFET with 4-5 channels sharing the current load?
OH! You're using the RGB ones? The picture showed a white one.
For the RGB ones, you need to be more careful, because the applied voltage for each channel will be different...
I don't think the TLC5940 is an appropriate driver for these - you might be able to swing it with multiple in parallel, but I really don't like that proposal.
You could drive each channel with an AMC7135 (Put a zener + resistor on the supply pin to keep the supply voltage in spec, and drive the low side with a MOSFET - I've had success PWMing setups like this, even though it's using a somewhat liberal interpretation of the specs of the AMC7135), or use any of a number of other constant current regulators, or even beefy resistors for current limiting (separate resistor for each channel)
Ah, yes, the RGB ones. I see what you mean by the picture showing the white. Sorry about that.
AMC7135 seem to be limited to 6V input. How would that work with the blue LED at 8V forward voltage?
Do I hook the LED anode up directly to the 12V rail, and each cathode to an individual AMC7135? Instead of a zener, would a 5V rail, from the 12V via a linear regulator, suffice?
I then also assume I would need to PWM this to almost 50% just to stay within 250 mA, or is there any other way to limit AMC7135 to less than 350 mA?
logan893:
Ah, yes, the RGB ones. I see what you mean by the picture showing the white. Sorry about that.AMC7135 seem to be limited to 6V input. How would that work with the blue LED at 8V forward voltage?
Do I hook the LED anode up directly to the 12V rail, and each cathode to an individual AMC7135? Instead of a zener, would a 5V rail, from the 12V via a linear regulator, suffice?
I then also assume I would need to PWM this to almost 50% just to stay within 250 mA, or is there any other way to limit AMC7135 to less than 350 mA?
Aw damn, those are speced for only 250mA per channel? Well that's inconvenient - usually they're spec'ed around 350 (and that's what the 7135 is made for driving - only just one in a chain; the zener trick lets us drive more easily); that won't be such a good match if the max is 250mA, as you'd be pulsing them at 350, which is out of spec. The 7135's are not adjustable for a different current. Note that there are probably other analogous products with different (or adjustable) setpoints.
The way I've used the 7135's is with a MOSFET between their gnd pin, and the ground rail. This means that you can't use an external 5v rail - with ground disconnected, the 5v rail could be lower than what ground would have, causing current to flow through the protection diodes (which I suspect are there - haven't tested; if they're not, it would probably just damage the chip) and potentially lift the 5v rail above 5v and trash everything else on that line! No good. By using the zener + resistor trick (connected to it's ground pin, before the mosfet) you ensure that the voltage is always in safe territory.
What about AMC7140's, set for 250mA? That's well within their spec (put a heatsink on the red channel one for good measure), and they've got a dimmer pin that can be PWM'ed so you don't need an external mosfet. Of course, they're more expensive than the 7135.
You can drive a 10watt RGB LED from a regulated 12volt supply the cheap way (with a bit of heat).
Use three TIP120 darlington transistors (using mosfets means hotter resistors).
Emitters to ground. Bases via three 1k resistors to three PWM outputs.
LED anode to 12volt. LED cathodes via three current limiting resistors to the three collectors.
The three CL resistor values should be 12ohm/3watt for red, and 3.9ohm/1watt for green and blue.
This is assuming a 0.75volt saturation of the TIP120.
The 12ohm resistor can be replaced with a series string of three 3.9ohm/1watt resistors.
The resistors and the TIP120 will get warm. The LED will get hot, and needs a heatsink.
Leo..