High Power LED Driver

[Lighty]

Hi all

Firstly, Dave what an awesome Blog!!! Thanks!

So I've decided to make my 1st post a somewhat complex one (well so I think) 

I'm wanting to building a LED lighting unit for my marine fish tank due to the costs of running Metals Halides.
 
My project objectives:

4x 50W Blue LED chips
2x 100W White LED chips
PWM Dimming
Thermal Protection
Arduino Controlled

The main problem is getting this project started is driving the LED chips due to their high forward voltages and currents.

50W Blue = Forward Voltage of 32-36V and Forward Current of 1750mA
100W White = Forward Voltage of 32-36V and Forward Current of 3000 - 3300mA

I did some searching on the net and found this handy IC by TI, the LM3464, a 4-channel high voltage current regulator with Dynamic Headroom Control (DHC), PWM and Thermal control (1 sensor input). This is very handy as it can control up to 4 strings (chips), and with its DHC it will assist in increasing the units efficiency. But the problem is that I don't know how to interface the DHC of the IC with a power supply! I was simply hoping I would be able to interface it with a 200W 36V SMPS (for the 2x 100w chips), but it doesn't appear to be the case?

What would be the best way to drive these chips, bearing in mind I need to be able to control them via PWM independently, or at least the Blues and Whites separately.

Can a SMPS be modified to work with the LM3464? If not, can they be modified to work as a CC driver, and accept a PWM signal?

I'm not totally concerned about the thermal monitoring being done by the LM3464, as I could just use the Arduino to monitor each LED chips temperature (which would be better) and let it adjust the Duty cycle for that channel.

I would prefer to go with a SMPS as to help keep the efficiency up.

Thanks in advance!

Regards

Justin

[Corporate666]

A few thoughts...

-I took a quick look at the chip, the Dynamic Headroom Control is a neat feature... it may be (probably would be) possible to hack a power supply to have this feature of your driver chip work, but I really think it appears to be designed to work with a power supply you either design, or which is intended to support this feature.  The power supply may not work (or not work well) at the voltages the DHC tries to adjust to, so that's why I am not sure modifying an off the shelf SMPS is such a great idea.

-You say you like the idea of an SMPS for efficiency and I presume you like the DHC of that chip because of the efficiency gains, but then you talk about using 2 x 100W white LED's.  I presume you are talking about those giant chips you can get from places like Deal Extreme?  Those are massively inefficient, which, IMO, cancels out any good you might achieve with a complicated power solution.  Also, unless you are buying quality LED's from a reputable supplier (like Cree, Philips, Osram, Nichia), those high wattage chips from no-name Chinese suppliers are laughable in their specs.  I have a 30W RGB chip that is an absolute joke at the light it outputs.  So I hooked it up to a power meter and the sum total of all channels at max is less than 10 watts - from the wall... so no way it is 30 watts to the LED - in reality, less than 10 after power conversion inefficiency.

I'm a regular at Reefnet and some other aquarium forums so I keep up to date with the posts about aquarium lighting... may I ask why you are looking to use so few LED's?  From what I have seen, most guys go with arrays of Rebels or Cree XML/XBD type LED's.  That would be much more efficient, but maybe more importantly, isn't such a small number of such high power LED's going to give you huge problems with hot spots in the tank?  Or is this some kind of tall cylinder tank of small diameter?

[Lighty]

Hi Corporate666

Thanks for your prompt reply!

As I thought, modifying a SMPS wouldn't be the best.

Yip, its those Ebay jobs, I do know that they are not the most efficient, but for the price you pay for them you can't expect the best right. I know I sound a bit contradicting about chasing efficiencies on the drivers and yet not using the most efficient LED, but this is more of a learning curve for me, forcing me to try and learn new aspects of the hobby, it would be easy to drive them with a LM317 etc, but wouldn't teach me much. Who knows, it might turn out to make a nice garden light 

[Corporate666]

100 watts of constant current is a lot.  That is a pretty beefy driver... and it's going to be a bit difficult to be chasing efficiency because you are naturally dealing with pretty beefy components by necessity.  You can buy 100W LED drivers for <$20 on eBay... I don't think you are going to really build something yourself and save money. 

There's a thread on CPF about these LED chips.

http://www.candlepowerforums.com/vb/showthread.php?344819-100W-LED-Chips-Any-Experiences

400 watts of LED light is a lot.  I mean a LOT.  Even for low efficiency LEDs like those.  Have you considered the difficulty of spreading that light around?  You'll have a small number of point sources which really isn't what you want unless your tank is very small but very deep.

Due to the low LED efficiency, you are also going to have a ton of heat to get rid of.. a majority of all that wattage you are pumping into those LED's is going to turn into heat.  You're looking at substantial head sinks with attached fans... like high-end CPU cooler level heat sinking.

[Jeff1946]

Not sure if these would be useful, Phillips makes 60 W led light bulbs that are blue leds surrounded by yellowish plastic that converts some of the blue to yellow and red to give white light.  They can be dimmed with a standard ac dimmer.  You could take the plastic off and just get blue.  They also make wifi drivable bulbs that have separate red, green and blue outputs.

[poorchava]

drivers for <$20 on eBay... I don't think you are going to really build something yourself and save money. 

I wouldn't expect them to last very long. All the components are most likely overdriven and shitty quality anyway. My father used to say: "you buy cheap, you buy twice"

I would either use some dedicated chip (like the LM3464 mentioned above) or make the driver on my own. There are some microcontrollers with special dedicated SMPS peripherals. I mean your typical constant current led drive is nothing more than a normal buck converter with the little twist, that you don't regulate around output voltage divider, but rather a shunt resistor in series.

Generally integrated "all-in-one" solutions are not optimal in my opinion. You can quite often notice, that some chips from some manufacturers (notably: Allegro, Infineon, Macroblobk, ST) with which you simply cannot shrug off the feeling, that this chip was designed specially for some customer and then released to the open market. Kind of seems that way with LM3464.

[Lighty]

Thanks for the replies everyone.

If I chose to go with a something like the LM3464, I'm a little confused as to how the buck system works on it as there is no inductor in the circuit  Is that not one of the main part of a buck or boost system?

I know they are using a shunt resister as discribed by poochava, are they using the secondary of the SMPS as the inductor?

Also, if I power it from say 40V, is the excess power disapated as heat like that of a linear power supply? Can I use either a SMPS or Toroidal?

Anyone have any schematics of a decent CC power supply suitable for this application with PWM support?

Thanks in advance

[Lighty]

Hi All

Been thinking about this and wondering why why I haven't concidered going with a linear power supply.

After some testing, I've found the 100W Chip to run very nicely at 2.5A (approx 30V). So if I had to setup a LM338 as a constant current regulator to run at 2.5A it would require a shunt resister of 0R5, 3.125W.


Now if I can have the power supply run at about 34-35V, this would mean that the LM338 would have a volt drop across it of about 2.75V - 3.75V, This would equate to about 6.875W - 9.375W.

This means the regulation system would disapate a maximum of 12.5W for an LED chip driven at approx 75W, if Im correct this would equate to about 85% efficiency.

I hope my calculations are correct?

Any comments.

[Corporate666]

Hi All

Been thinking about this and wondering why why I haven't concidered going with a linear power supply.

After some testing, I've found the 100W Chip to run very nicely at 2.5A (approx 30V). So if I had to setup a LM338 as a constant current regulator to run at 2.5A it would require a shunt resister of 0R5, 3.125W.


Now if I can have the power supply run at about 34-35V, this would mean that the LM338 would have a volt drop across it of about 2.75V - 3.75V, This would equate to about 6.875W - 9.375W.

This means the regulation system would disapate a maximum of 12.5W for an LED chip driven at approx 75W, if Im correct this would equate to about 85% efficiency.

I hope my calculations are correct?

Any comments.

You could do that, but honestly, those LED's you are using might be 20% efficient, whereas a higher end LED might be 40% efficient.  I don't really understand the point of chasing efficiency on the power supply side and then using a really inefficient LED.  It seems like hardly using the gas when taking off from a red light in town, but then zipping along the highway at 100mph.   But whatever.

If you are going to supply the LM338 with 34-35V from another supply, well, that supply will already be dropping 120/240 down to 34-35.  Why not just have that supply go a little lower and/or operate in CC mode and supply the LED?