Pros and cons of buck vs. CCR to drive high-power LED?

I'm looking to drive a high-power 3W RGB LED (350mA per channel) and I've come across two options for doing so that I'm not sure how to choose between. I see a lot of people put together buck drivers using inductors and other parts, but I've also come across these nifty CCRs from ONSemi that appear to do the same thing in a much smaller package. I was planning to put a CCR in series with each LED channel on the low side, then control each CCR through a ULN2003 so that an Arduino can do PWM control on it.

My question is, what are the advantages of using a traditional buck driver over this new CCR from ON Semi? Am I overlooking any important details by choosing to go with these CCRs?

Those devices are high power "constant current diodes" or something like that, a depletion-mode
FET with gate wired to the source. They have several volts drop out and waste a lot of power as
a result. a CC buck regulator will be more efficient. The ONsemi devices need a constant
voltage supply anyway, so I see little point in going that route unless you are happy with the
power loss.

Hm, I didn't know about the dropout. The datasheet says something about an overhead voltage of 1.8V at 70% of the normal current output. I will be giving my LEDs a stable 5V, and the biggest voltage drop I see will be 3.2V-3.8V., which I think means that the 1.8V dropout will be fine (it will still leave enough voltage for the LED). Am I understanding that correctly?

While a CC buck regulator may be more efficient, it also has a much bigger footprint and parts count. As long as these CCRs allow safe and consistent operation of the LEDs I'm OK with them running a little hot (I'll just make sure to add more thermal footprint). I need something that is better than straight power resistors (which is what I've been doing up until now), so maybe the CCR is a nice "middle of the road" option between resisters and a full buck regulator?

Why?, Do pray tell...
The losses are the same with an NPN power transistor sensing emitter current which means the same power loss as heating thus it requires the same thermal footprint as does the CC diode and as an SMD part is considerably more difficult to deal with.. Small 2 to 5 watt sensing resistors are available not much more than a metal ribbon with solderable leads welded to the resistive element.
The "Nice" thing about the BJT current limiter is that the transistor doesn't occupy as much board area if externally mounted as the CC diode will.
There are other considerations that are two fold in nature...

  1. the CC diode should have 2 Oz. Copper for proper heatsinking and i'd Not use it with Pb free solder for the second reason which is premature solder joint failure on inadequate PCB layouts..
    If you can visualize a 5W heatsink unfolded you can approximate the required 2 OZ PCB area required and soldering the diode to an area of that size is a PITA... Been there, done that. Long ago..
    One more thing to consider is that for PCB integrity and long term reliability the heatsink area should Really should be thermally isolated from the control circuitry.
  2. Were I You... I'd contemplate an Op-Amp like an LMC662 and a .05 to .1 ohm sensing resistor. total dissipated power in the sensing resistor should be about 1/2 to 1 W... most of the resistors I mentioned above start at 1W..
    Using the inverting input a driver base current limiting resistor and a little feedback it should be trivial. There is another advantage in that you can vary the gain to vary the current to the led.

Doc

On second thought there is another method especially for low power... ~1W led's that might work better.
Use a buck mode switcher and set it's output ~ 1 V higher than the CC limiter requires.
This would relieve most of the thermal issues and make a PCB much more realizable because the ground plane is now the heatsink..
Same number of wires too. One Vsupply and as many returns as there are led's.

Doc

Sorry, I'm going to need you to dumb that down a bit for me, lol. I have no idea what is meant by "NPN power transistor sensing emitter current", for example :stuck_out_tongue:

If the CCR has so many downsides, why does it exist? What types of scenarios would it actually be good for? The datasheet seems to suggest that it is meant to efficiently drive LEDs, and since it can output 350mA, this seems to match a high-power LED perfectly.

Why would one need a 5W thermal footprint? If I am driving one channel of a 3W RGB LED, shouldn't there only be 1W of power in the system to dissipate?

I've never heard of your method of using an op-amp to drive high-power LEDs, is there a name for that kind of circuit I can Google?

zenwebb:
Why would one need a 5W thermal footprint? If I am driving one channel of a 3W RGB LED, shouldn't there only be 1W of power in the system to dissipate?

1W in the LED, you can dissipate an infinite amount in the power supply (if you want).

zenwebb:
which I think means that the 1.8V dropout will be fine (it will still leave enough voltage for the LED). Am I understanding that correctly?

Look at the datasheet Fig 2, if you have about 1.8V at VAK, then you have only about 250mA, you need about 7V overhead to get close to 350mA.

Dang, it looks like you're right. On the very first page of the datasheet it states "Ireg(SS) = 350 mA
@ Vak = 7.5 V", which means that at 5V this thing won't do what I need it to.

For my project, buck drivers are just too large and expensive (especially considering that 3W/5W RGB LEDs can be bought for under $2 each), so I'm going to go back to working with resistors for now :stuck_out_tongue: