Surprisingly difficult drive problem

[dbvanhorn]

I need to drive a single Osram Golden dragon IR emitter, with controllable intensity, from "off" to full power.

I need it to run from 5V - 10V input.

I need the intensity control to be relatively simple, like a variable resistance to ground.

Now the kicker: I need it to be efficient.

You wouldn't think that was so hard, would you? I've searched for hours, and not found anything much over 70% efficient. All the drivers seem to be optimized for white leds (higher Vf), lower currents, and strings of many leds.



The National Semi web seminar on LED drivers has all the same issues, nothing that works for me.

Could anyone point me at a possible solution?

 

[mds82]

http://ledsupply.com/03023-d-e-700p.php

a 700 ma driver, with a external pot for dimming - and they are efficent.

 

[dbvanhorn]

And they are enormous, and about half the output I need.

I've used them in prototypes on other projects, but I need a driver IC like the ZXLD1360 from Zetex, only much higher efficiency. The 1360 is what I'm using now. It's nicely small, but efficiency is low.


Page 9 of the data sheet shows efficiency, but they are assuming white LEDs. As the Vf gets lower the efficiency gets markedly worse. The Vf on the SFH4230 is spec'd at 2.4V max, but is lower in practice, about 1.8V or lower as the device heats.. This is a 1A diode.

I'd put two dragons in series, but I don't have room.

http://catalog.osram-os.com/catalogue/catalogue.do?favOid=00000000000207c000ed0023&act=showBookmark

 

[Foxx510]

PT4105? I don't know a lot about it, and it might be a stupid suggestion, but it can't hurt hey! They are pretty efficient at 700ma, on paper at least, but might be far worse at the higher currents you are after.

 

[SemiMan]

Find yourself a synchronous buck convertor designed for lower voltage. It will likely have a 0.6 - 0.8 feedback voltage. You can use an op-amp to lower the feedback voltage down into the 25mV range or so. That should get 90% efficiency pretty easily.

Semiman

 

[HarryN]

As you note, it is not so easy to make a highly efficient LED driver, and more complex when you go to low Vf. The good news, is that you only need buck, not buck / boost. One of the problems right off the bat, is that protecting the driver from accidental open circuits takes diodes, which have almost as much drop as the LED in the IR range.

It gets even more exciting as you start varying through the power range to dim, as there are nearly always spots with obvious strobing.

It might be nearly as efficient to use a single chip 1.8 V constant voltage system and a small variable resistor as to design a constant current unit for that range. Designing a constant current setup to hit 80% or higher is a serious engineering effort - almost anal retentive - not one likely to be achieved without very specific parts selection and testing.

Talk to George at TaskLED.com - he has hardware, software, and years of effort / pain on how to make highly efficient setups. If you are looking for a commercial design, he might license - you can always ask.

 

[SemiMan]

Really with most high efficiency synchronous buck convertors, the design effort is minimal. All that is needed is an op-amp such that you can use a small feedback current sense resistor to amplify that small voltage up to the feedback voltage for the chip. Something like a National LM3100 I could hit 80%+ with little work. A little more thought in the selection and should be able to hit higher. It would be beneficial to bound your input voltage. Do you really need 10V input?

Semiman

 

[snarfer]

For a monolithic solution, Linear LTC1879 plus opamp. Use a resistor divider across the opamp to vary the amplification of the current sense feedback circuit and implement dimming control. 90% efficiency and extremely small size is easily possible with some attention to the inductor. If you don't want the extra expense of the opamp you can just use a matched dual pnp transistor array with a few resistors.

Synchronous with external MOSFETs there are too many possibilities to list. I happened to look on Linear's site so, perhaps LTC3835, good for input voltage up to 36v. I'm sure National has some too. Efficiency can be even better.

You will definitely not find anything of much use specifically intended as an LED driver. So that means you need to build a current feedback circuit and use a voltage mode buck controller.

 

[dbvanhorn]

As you note, it is not so easy to make a highly efficient LED driver, and more complex when you go to low Vf. The good news, is that you only need buck, not buck / boost. One of the problems right off the bat, is that protecting the driver from accidental open circuits takes diodes, which have almost as much drop as the LED in the IR range.

I don't really care if a bad LED destroys a driver. The failure is the same to me, one dead channel. I could use boost, if it works out, returning the LED to VIN instead of ground.

It gets even more exciting as you start varying through the power range to dim, as there are nearly always spots with obvious strobing.

? The Zetex ZXLD1360 I'm using now isn't having that problem. It's just woefully inefficient. It's a hysteretic converter. Even using the video camera, we aren't seeing strobing.

It might be nearly as efficient to use a single chip 1.8 V constant voltage system and a small variable resistor as to design a constant current unit for that range. Designing a constant current setup to hit 80% or higher is a serious engineering effort - almost anal retentive - not one likely to be achieved without very specific parts selection and testing.

I'm ok with that. The system has over 40 power supplies now, with nine switchers for the cameras, and computing platform.

The space I have to work with is very constrained, and very oddly shaped, so I'm pretty much stuck with a "single chip" implementation. I used the best inductor I could find, with minimal loss and resistance, and I'm getting what the data sheets tell me to expect, but it's still pretty dissapointing.

Talk to George at TaskLED.com - he has hardware, software, and years of effort / pain on how to make highly efficient setups. If you are looking for a commercial design, he might license - you can always ask.

I'd have to do the design work, multiple issues with outsourcing this.

I'm just surprised at how little real variety there is in the total zoo of "LED drivers", especially at the higher current end. Everyone wants strings of white leds, nobody seems to look at single leds, much less single IR leds.

 

[dbvanhorn]

Really with most high efficiency synchronous buck convertors, the design effort is minimal. All that is needed is an op-amp such that you can use a small feedback current sense resistor to amplify that small voltage up to the feedback voltage for the chip. Something like a National LM3100 I could hit 80%+ with little work. A little more thought in the selection and should be able to hit higher. It would be beneficial to bound your input voltage. Do you really need 10V input?

Semiman

I may have to go that route. I'm using the LTC3602 in a bunch of slots, but I don't know if I can fit that solution in the space I have to work with.

Yes, I need 10V. The source is three 123A cells in series, and I have to have a little safety room at the top. The battery voltage under load drops quickly to about 2.5V, but I have to withstand three fresh cells, hot day, etc..

 

[dbvanhorn]

I was looking at the LM3401 this morning.
Nice low Vref (0.2V) and high efficiency on the cover page, but when you get into the data sheet, with a 1A led, they show about 80-85% over my input voltage operating range.

Not horrible.. but not impressive.

Price of components isn't much of an issue, efficiency is.

 

[alexglvr]

hello dbvanhorn,

i need help please to make this SFH4230 work...

Are you still there?

sincerely,

Alex