High power LED driver

[dcoultas]

I recently took up the endevor to create an LED driver for the seoul p7 900 lumen led. The problem is that I would like to make a driver that can support more than just one. I would like it to be a ten amp driver using the on semiconductor NCP3066PG. This driver ic only supports 1.5 amps through the ic so it has to be used to control external higher power transistors. The website offers schematics to their predesigned boards including the 3 amp version. I like to make a similar version to the 3 amp model they provide however Iwould like it to regulate at 10 amps rather than three. My question is, is it possible to modify this design to support ten amps and what components would it require?

 

[dcoultas]

http://www.onsemi.com/PowerSolutions/evalBoard.do?id=NCP3066S3BCKGEVB
here is a link to the 3A board

 

[Mitchekj]

What type of input voltage are you using? May be much easier to run the LEDs in series, that way you'd only need 350mA. (Or whatever you'd planned to run the LEDs at, I believe 1400mA is the recommended from Seoul.) The max duty cycle on that IC looks like it's in the 90-95% range, so you should be able to get a fairly small difference between input and output voltage, but it's probably not ideal to run the switcher at such high duty cycles.

(Edit: By that I mean you can run three LEDs in series, at a decent current, with an output in the range of 9-10V. You should be able to do that with a 12V input, no problem. That way you don't need to up the current rating.)

Are you familiar with DC-DC converters? They're not easy to design, even for many professional engineers. Changing the max current out may sound trivial, but there are a whole slew of things you have to consider, but I believe there are members here who design such things for a living, who may be willing to help there.

I'll try not to go through the lecture, but these LEDs will get HOT! if you run them anywhere remotely close to the 2.8A figure that they give for the P7. Almost 12W? I hope you have thermal management considered. Liquid nitrogen perhaps? Seriously though, these will get to surface of the sun temperatures and die a horrible death quick-like if not heat sunk properly.

By the bye, I hope you're not looking for a nice white light... I've noticed that all the Seoul LEDs I've worked with have this slight sickly green tint, they look horrible... to me anyhow.

 

[dcoultas]

Heat won't be an issue. The issue is cost. I want to be able to run as many as possile off one driver and get as much light as I can at the lowest cost possible. So a ten amp driver will be running at least 5 of these LEDs and the power supply will be whatever is nessesary to run them in series. I really just need help determining what components I would need to make that circuit 10 amp instaed of 3.

 

[Mitchekj]

You can run a billion LEDs in series at 1400mA, provided you have the voltage. You don't need more than 3A. Set the output to 1.4A and you're good. You will need a ~20-24v input to run five.

 

[(*steve*)]

Heat won't be an issue.

Incidentally, this is why manufacturers of LEDs are often really careful who they sell their parts to.

Even experienced designers sometimes believe this.

 

[dcoultas]

I relize I can run as many in series as voltage allows however as I said I need as much light as Incan get as cheap as possible. Since LEDs are current driven if I run more than the current to each will divide between each led. I would like to run each led at at least half power and not need one driver for each led.

 

[Mitchekj]

I'll try again:

Connect five LEDs in series.
Apply a constant current of 1.4A, which the 3A demo board should do, with one resistor change.
Don't connect the LEDs in parallel.
Done.

5 LEDs, each LED will be running at 1.4A. Heat, not being an issue, will not kill the LEDs, so you're good to go.

IN SERIES current is constant through all components.
IN PARALLEL voltage is constant over all components.

 

[Mitchekj]

And to pre-empt, to get 1.4A constant current out of that demo board:

Add another 0.68 Ohm resistor as R18.
If R17 (the 0.33 Ohm) is not installed, install it.
You want two 0.68 and one 0.33 resistors all in parallel down there.

Edit: As an input into the buck, you'll need about 20Vdc minimum, btw. I'd use 24Vdc. Turns out the chip has a max duty cycle of ~85%.

 

[(*steve*)]

Mitchekj is right. Run LEDs in series wherever possible.

A string of 5 LEDs in series, each drawing 20 mA draws a total of 20mA and ther eis no risk that the current will not be shared evenly.

A set of 5 LEDs in parallel, each drawing 20mA draws a total of 100mA and there are real risks of one LED drawing more than its fair share.

The benefits of series connection, and the problems of parallel connection grow as the LED current increases.

in both cases the brightness will be the same.

The only argument against series connection is where you want to deliberately vary the current through individual LEDs.

 

[dcoultas]

Ah thanks I'm really not sure what i was thinking. For some reason I was thinking that current was not constant in series. This solves my issue completely.