LED constant current driver

[Rachie5272]

I'm building an LED device, and need to make a controller for it.  These are pretty big LEDs though: 2-3.5 V, 13.5 A, 25% duty cycle.  I have six of them.  They're red PT-54 Phatlights, meant for projectors.

I've never designed a power supply beyond a simple linear chip, so I need some help starting.  I'm thinking about using a MAX16818, and driving them all in series from a big 12 V 18 A power supply.  I'm thinking about using beefy overkill FETs, too.

The device will be AC line powered, and efficiency isn't very important.  I have a fairly big budget for a one-off, too.  I'm most concerned about simplicity and longevity.

The desired mode of operation is fairly simple.  Normally I want it on but very dim, then ramp up to full power in a second or two on command.  I'm guessing this is fairly trivial to do with a microcontroller and PWM.

Is this a good driver chip to use?  Which SMPS configuration should I use?

Thanks!

[duskglow]

Wow, that is one honking big LED.

[_Steven_]

please post your results. We can learn of this. Normally I drive smaller leds with a current source..but this is over my head. That many amps for 1 single led? 13,5 Amps? I cannot think about any linear device that can handle this
But 13,5Amps??? are you sure??
What kind of rectifier are you using than...and what kind of transformer...

13amps is really alot.
Can you please post some pics or some extra info.

[NiHaoMike]

PC power supply, connect LEDs in strings of 3 in series with a resistor to allow operation from the 12V rail, then add a MOSFET to PWM the LEDs.

[mariush]

Sounds like a normal computer power supply would be a good start ... those can do 3.3v and 5v at about 20a each (but about 130w in total on both 3.3v and 5v so about 15a on each is reliable) + 12A at about 30-50 amps. For 50-60$ it's very hard to do better.

There are linear regulators like LT1083 and LT1084 and LT1085 that can do 7.5A and 5A and 3A respectively.

LT1083/LT1084/LT1085 datasheet : http://cds.linear.com/docs/en/datasheet/108345fg.pdf

Looking at the datasheets it seems that minimum forward voltages are 2v and 3.5v, but you probably need to look at higher voltages:

8.1A driven continuously :
min VFmin 2 3.5 3.5 V   (R G B )
typ VF 2.7 4.5 4.6 V  (R G B )
max VFmax 3.5 5.6 5.7 V (R G B)

13.5A with 25% 50% 25% duty cycle

min VFmin 2.2 3.8 4.0 V  (red green blue)
typ VF 3.0 5.0 5.1 V (red green blue)
max VFmax 3.8 5.9 6.0 V

LT1084 has a current limit set at about 6.5A so you could probably just pair two of them and set them to output 3.7-3.5v with a 5v input  and rely on the built in current limiting
just the same you could use 3xlt1085 with its 4A internal current limit to get the 4.6-6v for green and blue from 5v or 12v (3 lt1085 to dissipate the heat easier), with 25% and 50% pwm you're basically going to run these 3 3/4 of the time.

[Rachie5272]

They're huge LEDs alright.  Internally I think they use a dozen or so dice, but externally it acts like a single 2-3.5 V, 13.5 A LED.  I'm trying to drive six of them in series.

I'd rather not use a linear power supply, because of the huge power involved.  I know the losses aren't as high as they would be with higher voltage, but it seems so inelegant.

I have a nice big 12 V 18 A switching power brick meant for some sort of computer workstation, which will be a good starting point.  I'm hoping someone can tell me if the Maxim chip is any good for this application, and which SMPS topology I should use (buck-boost, SEPIC, or ?uk).

The LEDs are still in the mail, but I'll post pictures of my setup when they come.  I'll be using three big CPU heatsinks just to keep them cool.

[MagicSmoker]

The buck topology is the easiest to stabilize and lends itself well to constant current applications. Very interesting LEDs you are using, btw.

[Rufus]

Seems a waste to switch mode from mains to DC then switch mode again for current control.

Manson do some bench power supplies (re-badged by various companies) with digital and analog voltage and current remote control.

http://www.manson.com.hk/products/detail/156

I see the 32v 15A version listed at 179 euro.

[lgbeno]

Looks like a fun project, I would think that you could use some buck topology regulator and low side current sense with an opamp feeding the buck fb to trick the buck into regulating the voltage across the current sense resistor and hence LED current.  Another amp could be used to scale the current up with a uC pwm.  Doesn't Dave have a video on this from awhile back?

13A is a lot though so there will certainly be some challenges.

[Rachie5272]

It seems like a bigger waste to use a lovely bench supply in a dedicated application  Besides, I won't learn anything.

Buck alone won't work here, as my input source is right around the series forward voltage of the diodes.  It needs to be able to produce a slightly higher or lower voltage.

[IanB]

it acts like a single 2-3.5 V, 13.5 A LED

It's a minor detail, but read the data sheet carefully. Vf is 2-3.5 V when operated at 8.1 A continuous. Vf is 2.2-3.8 V when operated at 13.5 A 25% duty cycle. It also recommends the pulse frequency should be > 240 Hz when operated in pulsed current mode. It may perhaps be easier to design for the 8 A continuous case?

You will need a big heat sink of course, possibly with a fan to assist it. A re-purposed CPU heat sink could be an option. What did you have in mind?

Also I see there is a thermistor on board. Ideally you should monitor the thermistor in the driver circuit and reduce power in the event of the temperature being too high.

I can't help with the choice of driver, but ignore suggestions of a simple series resistor. That's not sensible for high power LEDs. Look for dedicated LED driver modules. You can typically put these in parallel to get the current you need, and often they have a PWM input for brightness control.

[Rufus]

It seems like a bigger waste to use a lovely bench supply in a dedicated application  Besides, I won't learn anything.

I have a fairly big budget for a one-off, too.  I'm most concerned about simplicity and longevity.

Make your mind up then.

[lgbeno]

It seems like a bigger waste to use a lovely bench supply in a dedicated application  Besides, I won't learn anything.

Buck alone won't work here, as my input source is right around the series forward voltage of the diodes.  It needs to be able to produce a slightly higher or lower voltage.

If simplicity is the goal, I'd suggest making the circuit buck mode only not buck/boost...  Maybe a 24 V vin supply?

[Rachie5272]

I can do 24 V input.  How does the efficiency compare to a 12 V buck/boost?

It looks like I'll be making two of these now.

[Marco]

Why not just pick one of the hundreds buck current supply ics specifically designed for this?

[Rachie5272]

I thought that's what the Maxim chip was.  Do you have a better suggestion?

[Marco]

There are specific ICs for non isolated LED drivers, although hundreds might be hyperbole.

http://www.fairchildsemi.com/pf/FL/FL7701.html for instance ... dunno if it can drive the great big MOSFET you will need directly though.

[IanB]

I thought that's what the Maxim chip was.  Do you have a better suggestion?

The MAX16818 does indeed look like it is designed for driving high current LEDs.