Hey all,
I've been building LED lighting projects for a long time and really enjoy it. The one limiting factor is my electronics knowledge. I've always used linear current regulators, which are very simple but also very inefficient. I want to change that! I know that I can buy plenty of these online, but I want to understand it fully and build one. My motivation is mostly for education.

My latest project requires driving 2 parallel strings of 3 LEDs each from a 14V source. Each string requires 2.9A. LED voltage is approximate 3.3V each at this current. A quick and dirty calc shows that efficiency with a linear regulator would be ~70%. I am aiming for more like 90% to 95%, with the ability to change LED configurations (see below).

I've been researching extensively and understand the basic principle of an RLC circuit, and that I most likely want a synchronous buck type driver since my voltage will always be higher than the series LED voltage. An asynchronous type looks like it will also work, but with lesser efficiency.

Here are some basic requirements:
Form factor - rectangular up to 1.5" x 2" x 1" this is very flexible
Input voltage - 14V standard, up to 30V spike (possible dirty voltage)
LED configurations - flexible 2s2p; 3s1p, 3s2p, 3s3p, 3s4p
Output current - constant 2.9A per string, so 2.9A, 5.8A, 8.7A, 11.6A total - if this means duplicating a single circuit capable of 2.9A that is fine. I don't need any fancy dimming, flashing, or anything like that. Simple and durable is most important.
Is it possible to use thru-hole components entirely?


I realize this is quite a bit of an undertaking, but I am willing to put in the effort. If someone could help by pointing me in the direction of a specific controller with data sheet so I can start reading up on it, that would be awesome. It's overwhelming with so many different options out there. By searching here, I found the PT4115, but the user required an external mosfet to actually drive the LEDs due to the current limitation.

Edit: Here is the post I was referring to: http://forum.allaboutcircuits.com/threads/buck-current-regulator-help-needed.80383/ See the bottom of the page.

Thanks a bunch to anyone willing to point me in the right direction.

 

I have a little one I worked on a bit for a low power one. I might be able to scale it up.
This particular one regulates the ground side, but if all you hang on the output is LEDs it should be ok.
If you try to make one size fits all it will be more expensive and probably quite a bit bigger.
Let me check on a few components.

 

Silly me.

http://www.ebay.com/itm/DC-8A-Buck-...176?pt=LH_DefaultDomain_0&hash=item3cd6f286b0

 

I actually have one of those linked above - very similar, except it is rated for 10A. It looks nice, but I want to understand it more and know how to select components.

Here are two others that I have. The left one is rated for 3-18V 3A CC and the right one is rated for 3-16V 2.3A CC. They both work as advertised. The one on the right actually achieved 91% efficiency from a 14V source powering 3 series LEDs. It achieved 86% efficiency when powering 2S LEDs from the same source.

I have used the WEBENCH tool by TI, which is great, but the schematics it's giving me have 30+ components. How come they need so many, but these small circuits pictured above can get away with 7-12 components?

The only tip I've received is that I should avoid too high-frequency of a regulator to avoid noise issues.

Thanks again.

 

Here is a very simple one. I just used parts that I had in my spice library. We might be able to find some better ones if you decide you want to build one. It is not a hi def circuit, but should work. It comes in at about 88%.

 

Thanks for that. I can't really understand the theory of operation. A comparator circuit with U1 as the brain and two fets for switching? Some sense resistors for setting current. Caps and inductors for voltage/current smoothing. Is U1 an LM339?

I did a few hours of research today and found this option that looks pretty neat. Especially valuable to me is the datasheet with PCB design guidelines and they also offer a pretty advanced calculator for sizing components.
http://www.allegromicro.com/~/media/Files/Datasheets/A6213-Datasheet.ashx

The only thing is I don't see any of these available for purchase, so something similar would be nice.

 

I'll try.

Lets start with the BAT54. It is just a little schottky diode so it has a voltage drop of about .27 volts. It is used as the reference voltage for the 339 comparator. It is a weak point as it's voltage changes with temperature so so will the LED current. But not to bad for home brew and easy.
The - of the comparator turns off the big fet if the voltage across the sense resistor is higher than .27 volts.
The 4.7 meg adds some hysteresis so the voltage has to go above and below the reference otherwise it would oscillate around the reference. This lowers the switching frequency.
The 2.7k "pulls up" the output of the 339 when it is off.
The 7002 allows for more drive current so the FET can switch faster than it could if it were driven by the 339.
The 470 ohm pulls up the voltage to shut off the FET. This is also a power waster, but easy.
M1 is the main power switch.
L1 stores the current and continues to supply it to the LED after the switch is off.
D2 keeps the voltage from going below ground when the switch is off so the current will continue to flow in the inductor.
C1 filters the voltage.
This one runs at a low frequency so it needs a large inductor for storage. The ones you buy in IC's usually run at a high frequency so you can use a smaller inductor which is cheaper and more efficient but usually requires a good PCB. This one might run on a wired breadboard if you were careful with the layout.
I have not shown decoupling capacitors, but there should be one on the supply and probably a small one on the 339 and the reference voltage.

I am not very good at explaining things. Maybe someone else can help.