Set me straight here. High powered LED lamp? Is this feasible?

[beelzebozo]

Hi, beginner here,

A bit of a background. I'm an engineer by training. However, apart from the courses I had in the university I've never actually worked electrical engineering, I went into software development right after getting my degree. Cut to 15 years later I've been rediscovering my electrical roots lately with a project I've come across somewhat accidentally. After discovering that we have a local LED producer that was manufacturing 155lm/w LED strips I decided to build a 50000lm lighting fixture.

Basically it goes like this. I've picked the strips. 10 of them. 72 V, 0.4A each. Now how do i drive them? After searching a bunch,  I've decided that TI's LM3464 in cascade configuration will be easy enough and efficient enough. This however presupposes that I have a 350W power supply, that outputs like 70 to 80VDC,  that supports Dynamic Headroom Control (DHC). As of now I havent been able to find supply with such a feature, as far as I can tell one must build one themselves.

So I'm hoping I can find some reference design, after finding out what DHC is, and calculate it to my specs, then find someone more experienced to build it, after I've made all the design and calculations. The same goes for the actual LED drivers (LM3464). I've no practical experience, no lab, most of my theoretical background needs to be reaquired, and I am not willing to spend boatload of money paying someone to design and build this.

And here comes my question. Is this even feasible to attempt? As I am in the beginning of all this it all seems like a bunch of math and a reading on the internet. Both I can do. However, what I don't know is what bugs me. Do you feel, from what you've read this far that this is something that can be accomplished? It's two PCB's one of which is a SMPS with as of yet unknown feature... It does seem a lot.

Thanks!

[Ice-Tea]

Just a thought: pick 12V strips. Safe voltage. Hack a ATX PSU and pronto: dirt cheap, highly fairly efficient supply.

[botcrusher]

While i love ATX supplies, remember that a 400W supply can't output 400W on the 12v rail. It could be more like 200W, because therr are other lines it has to drive.

[Rerouter]

actually most 400W ATX supplies built in the past decade would have about 300W on the 12V rail, and far less on the rest, most computers have been relying less and less on the other rails due to the efficiency benefits of starting with a higher voltage.

[Kjelt]

First of all what kind of fixture are you building?
155lm/W is silly that is not realistic unless it is cold led light (6000K or higher) is that what you want?
So say that it is true, then the 50000 lumens is about 350Ws of leds and generating enormous heat, you need watercooling or a giant heatsink. Are the leds and strips mounted on aluminium carrier?

Then the strips, you say they are 72V, 0,4A each what is their configuration how many leds in series (S) and how many strings in parallel ?
Does the fixture needs to be dimmable or only full on/off?
Building a driver yourself is probably outside of your ability, it takes a lot of time and experimenting to get things right, esp. with such high power outputs.
There are commercial drivers available that can do this and if it is only on/off they are pretty cheap.
For instance this one if programmed to output 0,4A with an output voltage of max 300V can drive  4 strips in series per led driver. But it all depends what you want and are able to do.
http://oemwebshop.philips.com/xitanium-150w-02-07a-300v-ixt-230v.html

[RGB255_0_0]

12V is easier. An EVGA Supernova G2 750W can output over 60A on its 12V rail and over 90% efficient (150W and over). All for roughly £100.

[Kjelt]

12V is easier. An EVGA Supernova G2 750W can output over 60A on its 12V rail and over 90% efficient (150W and over). All for roughly £100.

You need a current controlled supply , if you use a fixed voltage output supply you loose your efficiency in the resistor(s) that is needed in front of the ledchain.
If you want large lumens like TS you have to go to HV in order to keep the currents down, keep the costs down and make it feasible.
If you look at led implementations of street lighting and stadion lighting the highest efficiency is to get as close as possible to the rectified mains.

[beelzebozo]

First of all what kind of fixture are you building?
155lm/W is silly that is not realistic unless it is cold led light (6000K or higher) is that what you want?
So say that it is true, then the 50000 lumens is about 350Ws of leds and generating enormous heat, you need watercooling or a giant heatsink. Are the leds and strips mounted on aluminium carrier?

Then the strips, you say they are 72V, 0,4A each what is their configuration how many leds in series (S) and how many strings in parallel ?
Does the fixture needs to be dimmable or only full on/off?
Building a driver yourself is probably outside of your ability, it takes a lot of time and experimenting to get things right, esp. with such high power outputs.
There are commercial drivers available that can do this and if it is only on/off they are pretty cheap.
For instance this one if programmed to output 0,4A with an output voltage of max 300V can drive  4 strips in series per led driver. But it all depends what you want and are able to do.
http://oemwebshop.philips.com/xitanium-150w-02-07a-300v-ixt-230v.html

I'm on the road right now and writing from phone so I'll be brief.

The strips are 5700k and I'm trusting the manufacturer for the lm/w rating. Actually the rating is what led me on this path... I basically selected the best lm/w item on their portfolio and took it from there. As for the driver I'm hoping I could get  TI's WEBENCH to do the heavy lifting... Still the supply though.

I need to run. Be back in a couple of days...

[beelzebozo]

Do you think it is feasible to buy a boost converter, such as this, modify it to accommodate the Dynamic Headroom Control feature of the LM3464, and power it with some computer PSU with >30amps on the 12V rail (like this)? That way I build the LED driver side, power it up, have a lamp ready and work on the SMPS design in parallel (I've been looking at TI's UUC29950 for the past couple of days).

What do you think?

[Kjelt]

What do you think?

That you are building a constant voltage source while a led lamp needs a constant current source.
I would buy an off the shelf Meanwell or other brand Led driver suitable for your strips and enjoy the light.

[beelzebozo]

What do you think?

That you are building a constant voltage source while a led lamp needs a constant current source.

I'm not sure what you mean by this. I still plan on using cascaded LM3464s to provide a constant current to those strips. The post above was more about the Dynamic Headroom Control capable power supply for the LM3464.

[Kjelt]

Ah ok, then i do not understand why you not follow the datasheet. The only reason to make it more difficult by starting with 12V is if you want to power it off carbattery or solarpanel. If mains is you're powersource than follow the datasheet use a mains regulated smps with enough headroom (85V) preferably with an control input.

[beelzebozo]

Ah ok, then i do not understand why you not follow the datasheet. The only reason to make it more difficult by starting with 12V is if you want to power it off carbattery or solarpanel. If mains is you're powersource than follow the datasheet use a mains regulated smps with enough headroom (85V) preferably with an control input.

Basically because I have a computer PSU lying around and I thought I could use it + a cheap boost converter to have a power supply for the LM3464 board until I get my AC/DC SMPS game together... That way I could focus on the LM3464 stage and have the power supply move to the future..

[CJay]

Ah ok, then i do not understand why you not follow the datasheet. The only reason to make it more difficult by starting with 12V is if you want to power it off carbattery or solarpanel. If mains is you're powersource than follow the datasheet use a mains regulated smps with enough headroom (85V) preferably with an control input.

Basically because I have a computer PSU lying around and I thought I could use it + a cheap boost converter to have a power supply for the LM3464 board until I get my AC/DC SMPS game together... That way I could focus on the LM3464 stage and have the power supply move to the future..

Occasionally it's just better to (metaphorically) throw out the bits you have because they will  cause so many problems later on that you might as well have not bothered.

[tron9000]

you want proof its feasible? Here you go: