Linear LED driver - avoiding a hot resistor?

[Artlav]

Hello.

I want to power an 80W LED, which needs 2A at about 38-40V, from a 43V power supply.
Sounds fairly simple - LM350, a resistor, and a couple of capacitors - a linear constant current source:


However, there is a little problem here.
The resistor gets pretty hot, having to generate 1.2V of feedback voltage for 2A of current.
Certainly, this can easily be solved by using a large enough resistor, but a big LED is hot enough as it is, without the extra help.

So, is there a cooler way to drive an LED in this scenario?
I.e. a 2-3A linear regulator with lower feedback voltage than LM350?

[Marco]

Certainly, this can easily be solved by using a large enough resistor, but a big LED is hot enough as it is, without the extra help.

What the resistor doesn't burn the regulator will ...

[Artlav]

What the resistor doesn't burn the regulator will ...

There are plenty of regulator heatsinks, there are no resistor heatsinks.
Even if it does not improve the efficiency, it should at least make heat management simpler.

[SeanB]

No room for a 5W resistor? Place it next to the heatsink and use a small amount of thermal epoxy ( or a silpad) and it will run cooler. Alternatively clamp it to the heatsink, you do get clamps for that, or you can make them easily enough with a small length of aluminium strip and either long screws or a bending brace.

[T3sl4co1l]

What the resistor doesn't burn the regulator will ...

There are plenty of regulator heatsinks, there are no resistor heatsinks.

No resistor heatsinks?  That's news to me http://www.digikey.com/product-detail/en/PWR220T-20-R750F/PWR220T-20-R750F-ND/2192696

(At a glance, Digikey doesn't have a 0.68 ohm in stock, unfortunately.  Lots of kinds to pick from though.)

Otherwise, if you're having trouble getting the heat out at all, your only other choice is a switching regulator.

Tim

[mariush]

Or, you know, just get 7w cement covered resistors and put some thermal paste between it and the heatsink and use two screws and some metal bar to keep the resistor pressed against the heatsink.

http://www.digikey.com/product-detail/en/2-1623789-4/A103699-ND/2055566
http://www.digikey.com/product-detail/en/PSP700KB-0R68/YAG1122-ND/4135291

or put two 1.3-1.5 ohm in parallel :

http://www.digikey.com/product-detail/en/SQP10AJB-1R3/1.3W-10-ND/18714

[dannyf]

is there a cooler way to drive an LED in this scenario?

Sure.

1) use a smps;
2) use a linear regulator with low Vref.
3) use a linear led driver with low Vref.
4) use some diodes/bjt + mosfet / bjt as regulator.
...

[dannyf]

A typical bjt's b-e junction, for example, as a voltage drop of 0.7v -> that will cut your dissipation significantly.

Having said that, using a linear regulator here is really inefficient.

[Artlav]

2) use a linear regulator with low Vref.
3) use a linear led driver with low Vref.

And that is the essence of my question - what linear regulator/special driver have a lover Vref?
You can't exactly search them on digikey by this parameter.

Having said that, using a linear regulator here is really inefficient.

Hm?
44V, 2A input = 88W; 39V, 2A output = 78W.
78/88=88% efficiency.

It's not exactly simple to get better level of efficiency out of an SMPS, especially in a non-standard range like that.
Add to that a 44V->12V converter to power the SMPS itself, and you total with a whole lot of complexity for a very tiny gain.

No room for a 5W resistor?

Essentially...
The rest of the driver is the size of two 5W resistors, and the place for the resistor is at a somewhat wrong angle to to stick one to the heatsink.
So, i'm looking for simpler ways.

[timelessbeing]

http://www.ebay.com/itm/151172379115

[Artlav]

http://www.ebay.com/itm/151172379115

Hm.
Input Voltage: 100V-240V,  50HZ/60HZ
Output Voltage: 30V-36V
Fits neither input, nor output requirements.

More to the point, an assembled LED floodlight - driver, case, LED and all - cost much less than it's components, yet i didn't buy one.
That's a hint about my purposes, ain't it?

[dannyf]

what linear regulator/special driver have a lover Vref?

It was offered right there in the post.

You can't exactly search them on digikey by this parameter.

Sure. But you can search for Vout, and think about the relationship between Vout and Vref.

[Artlav]

Sure. But you can search for Vout, and think about the relationship between Vout and Vref.

Ok, i'm officially dumb.

However, it's no big loss, since i still can't quite find anything that would fit.
There are ones with low dropout, but their feedback voltage is still 1.2V.
And there is LP38853, which kind-of, sort-of looks like it might work if the LED won't get hot enough to go too far down.

All in all , the path of least resistance appears to include a bigger resistor stuck to the heatsink.

Thanks all.

It was offered right there in the post.

Discrete transistor regulator of some sort?
Can't quite see what you mean.

[muvideo]

I would also sense the LED's heatsink case and provide a means to throttle back the current if needed.

In your circuit if Q2 is coupled to heatsink, the current will decrease of about 0.3 to 0.4% per °C
But a decent protection need to be more sharp...

[NiHaoMike]

Use several resistors instead of one. Or use thermal epoxy to attach a heatsink.

[dannyf]

In your circuit if Q2 is coupled to heatsink, the current will decrease of about 0.3 to 0.4% per °C

Depending on what you are protecting against. Q2 sets the current for the whole thing. The whole set up is immune to Vbe decline in Q1 but does not protect against Vfwd decline in the diode (but that's irrelevant here).

So if anything, you should NOT thermally couple Q2 to Q1 - if you do that, the current will actually decline (2mv / 700mv per C) when Q1/Q2 heat up.

[muvideo]

In your circuit if Q2 is coupled to heatsink, the current will decrease of about 0.3 to 0.4% per °C
But a decent protection need to be more sharp...

Thats not my circuit. It's a very basic/CRUDE current limiting circuit thats been around probably since before I was born. That was just the first pic google led me too for an example for the OP.

Sorry, I should have said "the circuit you posted"
as for the right approach for the op problem, I said nothing about

[muvideo]

Sorry, I should have said "the circuit you posted"
as for the right approach for the op problem, I said nothing about

No prob. . I didn't mean to come off as hostile.

Ok, so now I can say something about:
I like the minimalistic approach      

Anyway, more seriously, a 40V led array will have widely varying forward voltage,
the voltage will depend on the particular unit and on the temperature, temperature
alone will probably change forward voltage around 1-2V from room temp to full power.
So I agree with you that the current regulation with such low headroom, better be
smarter than a pair of transistors.

[Artlav]

The LED is a nice Cree CXA3050 series one, about $30 equivalent each, i got two of them more or less randomly, of different colour temperature.
I wanted to see what modern high-power LEDs look like, how hard is it to keep them cool, how to drive them, etc - get familiar with the technology.
If it was to survive all that, and be viable, then it would end up as a light in a room (outshining everything already there ).

There aren't exactly ready-made drivers to fit it's specs around here, i'll have to order one online, and i hate doing that.
Therefore, i hook it up to a tweaked-down 45V power supply (which are found in abundance everywhere), with a simple current regulator.

The reason for simplicity is - there is pretty much nothing that can go wrong with that LM350 circuit, unless i screw it up (or rather not screw it tightly enough to the radiator).
My trust in things i make is proportional to their simplicity - a constant current buck converter is not that hard to make, but i would trust it much less than a simple, stupid linear one.
I've already verified that the cooling is sufficient with a huge safety margin, so no need for temperature monitoring.

So, that's all there is to it.

[dannyf]

There aren't exactly ready-made drivers to fit it's specs around here

SMPS can be easily repurposed to drive those leds, efficiently as well, in a fashion highly similar to what you are doing with a linear regulator.