Cree XLamp CXA2011 (updated - performance @ 500nA)

[Kibi]

The lighting arrangement for my bench currently consists of a double 18W 60cm fluorescent lamp under the top shelf and a single 18W 60cm fluorescent lamp below the bottom shelf. Very old skool lighting.
I have decided to start a project to drag my bench lighting kicking and screaming into the 21st century.
LED was the natural choice and after a lot of research I settled on the Cree XLamp CXA2011 which was released earlier this year. It is an 11W LED array module. It's basically got loads of small LED's on the same die.



The plan is to have 10 of these LED's lighting my bench. Five will be mounted to a long (60-80cm) heatsink and then replace the fluorescent lamps with the heatsink setup.

Of course I had to have a little play with it to see how well it performs.
I set my camera up on a tripod so that I could get the same shot of my bench with fluorescent lighting and LED lighting. I also set my camera up manually so that I could have the same exposure settings for both shots.
The aperture was fixed at f4.0, shutter speed 1/20th of a second, ISO 100 and white balance set at 4000K for both shots.

Fluorescent lighting:



LED lighting:



The LED shot is 1 stop underexposed, so basically there is half the amount of light falling on my bench from the LED lamp compared to the three fluorescent tubes. This is just a basic assumption, the rear of the bench is in darkness because of the shadow caused by the lower shelf. It is still very impressive since this is just 1 LED, so by the time I have all 10 up and running, I should have some very descent light on my bench.
I am driving this LED at 250mA, just below the 270mA that the data sheet suggests. At 250mA the lamp is kicking out 10.27W. With it's temporary heatsink it's operating at about 45 degrees C.

Here is the lamp "mounted" more or less where it will end up.



Now I just have to design a driver for the LED's. It would be easier to wire them all up in series and have one driver for the lot, but the voltage across the series string would be over 400V - not very good. I am thinking of having one driver for each LED but running from a single power supply, unless anyone else has got a better idea.

[IanB]

Now I just have to design a driver for the LED's. It would be easier to wire them all up in series and have one driver for the lot, but the voltage across the series string would be over 400V - not very good. I am thinking of having one driver for each LED but running from a single power supply, unless anyone else has got a better idea.

I think it is usually more convenient to source a commercial driver ready-made for LED lighting installations rather than designing and building your own (it will come nicely packaged with terminals for wiring and screw holes for mounting), but your choice of course.

I also think you are right not to create a constant current supply with 400 V behind it. The danger is that the current will try to arc across any failure in the system, and that could be a safety hazard.

[Psi]

10 of them is probably a bit more than you need (110W of leds).
I found that 30-40W of LED to equal a 100W lamp which is fine for a small room.

That said though, i have around 100W of leds in my room too and it's nice to have the option of switching them all on and having lots of light when you need it.  I have ~40W of pure white and ~60W of RGB.
http://psi.abcom.co.nz/leds1s.jpg
http://psi.abcom.co.nz/leds2s.jpg



I couldn't be bothered making a proper led powersupply at the time so i just put them in strings up to 13V and used large resistors mounted to the LED heatsinks to limit the current from a high power 15V supply i had available.

Obviously the more you can get in the string the better but yeah, 400 is a bit dangerous.
If i was you i'd probably do multiple strings up to ~45V which isn't to bad safety wise.

[IanB]

If i was you i'd probably do multiple strings up to ~45V which isn't to bad safety wise.

Amusingly, a single one of these devices appears to need ~40 V to drive it (10 W @ 250 mA = 40 V) 

[Psi]

If i was you i'd probably do multiple strings up to ~45V which isn't to bad safety wise.

Amusingly, a single one of these devices appears to need ~40 V to drive it (10 W @ 250 mA = 40 V) 

Oh, so it does, hahah

Well i guess that's probably why they're 40V.
Any more than that and it starts to give shocks

40V / 0.250mA led from 45V with resistor is only going to burn away 1.25W, which isn't that much.
3-5W resistor would be fine.

[Kibi]

I think this will be a good cheap design to drive the LED. It has a very low dropout so if the power supply voltage was just 2v above the voltage drop across the LED, this circuit would waste 0.5W at 250mA. This equates to a 95% efficiency of the current limiter circuit. The MOSFET won't even need a heatsink

I have tested it and it works very well. I did sub out the 2N5088 with a PN2484 because the 2N5088 is only good up to 30v. The MOSFET doesn't even get warm without a heatsink.

[Zero999]

10 of them is probably a bit more than you need (110W of leds).
I found that 30-40W of LED to equal a 100W lamp which is fine for a small room.

Is that compared to a 100W standard incandescent, halogen or fluorescent?

[DaveW]

That looks familiar, I've been working with that LED as well for under bench lighting; they'll send out free ones if you go to this link

http://www.cree.com/products/xlamp_cxa2011.asp

I've been playing around with this quickly-nb. this is not currently isolated, so not currently safe to build for everyday use...

[Psi]

10 of them is probably a bit more than you need (110W of leds).
I found that 30-40W of LED to equal a 100W lamp which is fine for a small room.

Is that compared to a 100W standard incandescent, halogen or fluorescent?

Incandescent.

A 100W incandescent lamp is 1600-1800 lumens
With the CREE leds i bought i could get 1700 lumens from 34W.  (50lm/w)

I believe CREE now have even better leds, a quick google seems claims they have some which do 231 lumens per watt. If you had some of those then you'd only need 7.1W of leds to equal 100W incandescent.

Looking at the datasheet for the CREE XLamp CXA2011 they range from 680 lumen to 1040 lumen depending on the color temp you buy. So you'd only need 2-3 of those to get equal to 100W incandescent. (assuming their figures reflect real world performance)

[Kibi]

The ones I got are 4000K colour temperature and 900 lumen. I figured that higher colour temperatures would be a bit"cold".

I might also modify the driver circuit to employ "soft start". I don't know, it'll probably be more of a gimmick that anything useful.
I have just got to wait for the heatsinks to arrive which are on back order for a few weeks before I cam complete this project.

[Zero999]

A 100W incandescent lamp is 1600-1800 lumens
With the CREE leds i bought i could get 1700 lumens from 34W.  (50lm/w)

I believe CREE now have even better leds, a quick google seems claims they have some which do 231 lumens per watt. If you had some of those then you'd only need 7.1W of leds to equal 100W incandescent.

Looking at the datasheet for the CREE XLamp CXA2011 they range from 680 lumen to 1040 lumen depending on the color temp you buy. So you'd only need 2-3 of those to get equal to 100W incandescent. (assuming their figures reflect real world performance)

I'm often cynical of figures like that. I would be interested to see what the real world performance is like. My gut feeling is you need 18W to 25W of LED power to get the same amount of light as a 100W incandescent, all other things being equal.

[alm]

Why does everyone keep comparing the efficiency to incandescent? Everyone knows incandescent lamps suck in that regard. Car manufactures don't claim their cars to be ten times faster than a horse-drawn carriage, do they? Comparing to fluorescent would be more informative.

[Zero999]

Why does everyone keep comparing the efficiency to incandescent? Everyone knows incandescent lamps suck in that regard. Car manufactures don't claim their cars to be ten times faster than a horse-drawn carriage, do they? Comparing to fluorescent would be more informative.

The power output of car engines is often measured in horsepower.

I think it's more practical to compare with an incandescent than a fluorescent.

The luminous efficiency of a fluorescent is dependant on more factors than an incandescent the: ballast, phosphor, colour temperature and tube dimensions.

Incandescents have been used for domestic lighting for years and people are familiar with the amount of light they get from a 100W bulb so it makes sense to use it as benchmark for domestic lighting.

[IanB]

Incandescent bulbs are also the gold standard when it comes to quality of light. They really are the light source other sources of illumination should be compared to, in spite of idiotic politicians banning them all over the place (yes, I'm looking at you, Europe!)

[Mechatrommer]

someone should make cost/time analysis too. i got few CFLs that burnt out after few weeks/months of use.

[ciccio]

someone should make cost/time analysis too. i got few CFLs that burnt out after few weeks/months of use.

Me too.. The just don't last the time they declare. Some lose intensity to the point of requiring replacement.
Maybe they will last long if always on, but so where is the power saving?
I'm also worried about the pollution if not correctly recycled and about the cost of recycling.
This  site has a good analysis of the subject: http://sound.westhost.com/lamps/index.html

Regarding LEDs, I think that they can be a solution if there is no need for a power supply: ideally they should be series connected in great numbers, and driven directly with rectified mains (300 V dc?).
If driven at low voltage, then a power supply is involved, but  the new EU regulations require a SMPS, that will last as long as a CFL, with the same recycling problems.
The toroidal transformers that were used for powering 24 V  filament lamps where practically eternal. Now they have been replaced by "electronic transformers"  that often fail, sometimes with flames (and the flames propagate to the ceiling).

[jahonen]

Even the energy saving point is not always true, when the whole picture is taken into account.

Here where I live, we mostly need lights when it is also very cold outside (something like -20...-10°C is quite normal at mid-winter, with lowest temperatures around -30°). So heat produced by the lightning will be used for warming the house and not wasted (really basic physics!). Even more so if one has direct electrical heating, which is often only practical option in sparsely populated areas, but even water-based systems benefit from the heating effect. And at summer, we have almost zero need for lights, as the nights are quite bright, thus it doesn't matter if the lamp is 0% or 100% efficient. At southern Europe it makes sense, but not here...

I think it has been shown that whenever politicians interfere with the engineering, the result is a miserable failure. The problem with new lamps designed to replace the faithful incandescent lamp is that most of them have electronics built in, which have quite low tolerance to heat compared to incandescent lamps. This is magnified by the fact that most lighting armature is not designed to remove this heat very well, as there is no need to do so when incandescent lamps are used, and that only possible integrated location for electronics is just in the hottest place, between socket and led, cfl or whatever. One can calculate how long a 5000 hour @ 105 °C specified electrolytic capacitor will last at something like 100°C ambient temperature, which is quite plausible in closed lightning armature. Add the enormous cost pressure, thus tendency to use cheaper than cheapest possible components. So much for promised tens of thousands of hours.

It would be best if the power supply would be located far away from the lamp itself (hottest point), and that the power supply would be easily replaceable regardless of the state of the lamp itself. But then as the industry likes the idea of "planned obsolescence", who will dare to manufacture a lamp (or lightning system) which would really last those promised tens of thousands hours?

Regards,
Janne

[Kibi]

Just a quick one, may be interesting to some people.
I was messing about and I wondered what the lowest current I could pass through this LED and still have it emit light.
At ~500 nano amps the eye can still see light coming out of the LED in the dark. The voltage drop across the LED is 29.44v
Camera's are not so sensitive, but I set my camera up with the lens wide open, ISO 400 and 30 second shutter in a fairly dark room. I shone a torch on the multimeter in the last 5 seconds of the exposure so that the reading could be captured by the camera (old meter - no backlight).



I'll post another thread in the projects section on the progress of my lighting project in the near future.

[SgtRock]

Dear Kibi:

--This is a fascinating post and a fascinating topic. I very much want to do something very similar to what you are planning to do. Being very ignorant on this topic I have a number of questions. I will post them one or two at a time in this thread and see if people who know more than me can help. I am sure that there are a lot of others out there in EEVBLOG land that have the same questions. See below for link to CXA2011 data sheet:

http://www.cree.com/products/pdf/XLampCXA2011.pdf

1)Assuming (in round numbers) that this LED package is operating at 10 Watts. How much of that is light, and how much is heat?

2)How many lumens are being produced?


"Three weeks in the lab will save you a day in the library every time" Stanley Williams, HP Labs

Best Regards
Clear Ether

[Kibi]

Dear Kibi:

--This is a fascinating post and a fascinating topic. I very much want to do something very similar to what you are planning to do. Being very ignorant on this topic I have a number of questions. I will post them one or two at a time in this thread and see if people who know more than me can help. I am sure that there are a lot of others out there in EEVBLOG land that have the same questions. See below for link to CXA2011 data sheet:

http://www.cree.com/products/pdf/XLampCXA2011.pdf

1)Assuming (in round numbers) that this LED package is operating at 10 Watts. How much of that is light, and how much is heat?

2)How many lumens are being produced?


"Three weeks in the lab will save you a day in the library every time" Stanley Williams, HP Labs

Best Regards
Clear Ether

This document highlights the efficiency and efficacy of this particular LED at various drive currents.

http://www.mouser.com/pdfdocs/CXA20_Customer_Presentation.pdf

In my design, I am driving this device at ~270mA which according to the data sheet will output ~1000 lumens per device. I am using the 4000K module.
Warmer light modules output less lumens.

[SgtRock]

Dear Kibi:

--Thanks for the help. The mouser document is very helpful as well. I will restudy the DJ video on LEDs and heat sinks, and try to educate myself further on disposing of waste heat. In your present configuration, assuming a nominal efficiency of 80% I calculate you are having to dispose of about 2.2 Watts of waste heat, if my understanding and math is correct.

--Your work appears to be very clean, and very professionally done. Many other people are going to be making their own lighting systems. I am going to try to find some way of using bar stock (aluminum or copper), or maybe heat pipes, as those huge star shaped heat sinks seem rather cumbersome for under the shelf installation.

--Please keep us apprised as you go along, if you wish, and if you have the time.

"Three weeks in the lab will save you a day in the library every time" Stanley Williams, HP Labs

Best Regards
Clear Ether

[Kibi]

In your present configuration, assuming a nominal efficiency of 80% I calculate you are having to dispose of about 2.2 Watts of waste heat, if my understanding and math is correct.

The power wasted by the current source circuit depends a lot upon the difference between the input voltage and the voltage drop across the LED. This design does have a low dropout which is handy. It means that the input voltage only needs to be a little bit higher that the LED voltage.
In this case, I am using a 30V transformer. 30x1.414=42.42vDC at full load. The voltage is more like 46v at no load. The voltage drop across on of my LED's is 42.4v at 270mA.
So, worst case, 46-42.4=3.6v. This is a 3.6v drop across the current limiting circuit. 3.6x0.27=0.972W. This is for one LED, I have ten. So, 9.72W is lost when driving 10 LED's.
But this is worst case, the voltage of the power supply will drop below 46v when under load, so the dissipation figure will be lower.

As for heatsinks, yes, this was a bit of a problem for me. I would have preferred a flatter profile, but the heat dissipation factor of a flat heatsink is a lot less than it's rated figure when it is not orientated the way it was designed to be. Most of them need to be vertical. I could have got a better rated heatsink to compensate, but then the price went up. The star heatsinks offered a much better price / performance ratio. I could have got lower priced star heatsinks at a similar performance, but I couldn't wait 2 months for them to come in stock.
Although the star heatsinks that I am using are quite a deep profile, they are still shallower than the fluorescent lamp that is currently fitted under my shelf. So I am still winning.

[IanB]

In this case, I am using a 30V transformer. 30x1.414=42.42vDC at full load.

Actually, that's not quite true. The transformer would be 30 V RMS AC at the rated output, which for the same power output would correspond to 30 V DC. The 42.4 V DC is the peak DC voltage under no load conditions. As soon as you attempt to power a load the peak voltage will become lower in a complicated way depending on the size of the smoothing capacitor.

[Zero999]

In this case, I am using a 30V transformer. 30x1.414=42.42vDC at full load.

Actually, that's not quite true. The transformer would be 30 V RMS AC at the rated output, which for the same power output would correspond to 30 V DC. The 42.4 V DC is the peak DC voltage under no load conditions.

The voltage will probably be higher than that under no load conditions, 45V to 50V..

[IanB]

In this case, I am using a 30V transformer. 30x1.414=42.42vDC at full load.

Actually, that's not quite true. The transformer would be 30 V RMS AC at the rated output, which for the same power output would correspond to 30 V DC. The 42.4 V DC is the peak DC voltage under no load conditions.

The voltage will probably be higher than that under no load conditions, 45V to 50V..

I know, but I was just going with the assumption of 30 V RMS