Project Cree - The LED bench lighting project.

[Kibi]

Right, I have found some time to get this project going. Some of you may be aware of the design / review phase that I went through here: https://www.eevblog.com/forum/index.php?topic=4685.0

Some people have expressed interest in seeing the progress of the project, so here we go. Some of the pictures are appalling because I couldn't be bothered with my SLR tonight, I just used a P&S camera, so I apologise.

So my order comprising of the rest of the components arrived very promptly the other day from Mouser and I could make a start.
Although this project is supposed to functional, it is also a prototype, so I have built the current source on veroboard. It is quite well populated though, so I am pleased with it.
The board is not quite complete, it needs a few more wires, the MOSFETS and decoupling capacitors.
Due to the very high forward voltage (as far as LED's go) of these particular LED's, it was not feasible to wire them all in series and drive the the string that way. The voltage drop across the string would then be over 400V - ja, not so cool. So I have opted to drive each LED via it own "channel". I have also included "soft start" to minimise the effects of inrush currents and it's kind of cool too The circuit takes about 1-2 seconds to supply the full 270mA to the lamps.

So here is the board.





Next, we move on to the case. It has a standard IEC inlet, circuit breaker and 120VA transformer. The two 10 way JAE connectors go out to each of the bars of five LED's.







Now, the heatsinks did cause some issues. The main issue being the price. Some of the heatsinks designed for this type of LED were almost twice the price of the LED's themselves! Sensibly priced heatsinks were either obsolete or out of stock. I settled on these Wakefield heastsinks which are designed for star LED's, so these don't fit quite as easily, but the price was more agreeable. They are rated at 1.55C/W.
I have not quite decided how to mount the LED to the heatsink yet. I am either going to bolt them on with long bolts and nuts or glue them on with Arctic Silver thermal adhesive. That comes at a later stage, so I have some time to think.
At the moment it's just stuck on with sticky backed thermal pad.



I have operated the lamp at 270mA for an hour on this heatsink and it gets up to about 50 degrees C. 270mA (11 Watts) will output ~900 Lumens on this particular LED - which is very bright, it's overloaded the camera lol.



Although it looks like the LED lamp is the only light source switched on, it's not. It's just the brightness of the LED that forces the camera to adjust, so everything else looks dark.

[Kibi]

OK, so I've got a bit more done on this today. So far I have only completed one light rail with 5 LED's on it. I'll have to see if I have some time tomorrow to assemble the second rail.


I completed the board last week some time.


I then crammed the lot into my project box. It just about fits.



I then assembled the wiring loom. I had difficulty finding multicore cable in short lengths. Suppliers that did stock it only sold it in 100m or whatever lengths. I just got some braid and fed 10 strands of instrument wire down it. I have loads of instrument wire like I'm sure most people do.



Then loads of drilling and stuff to mount the LED and heatsink to the rail. The rail is handy because it can contain the wires and allows some space above the heatsinks for air convection. There are two bolts that run through the LED, the heatsink and the rail, clamping the whole lot together. There are star washers too to prevent the bolts coming loose during the heat cycling. It was quite tricky to solder the wires onto the LED because soldering is not possible when the LED is clamped to the heatsink. I had to kind of loosen the bolts off, wedge a screwdriver between the heatsink and LED so that I could solder the wires onto the LED contacts.



And then a picture of the illuminated bench witch just one rail.





Now for some figures.

Currently, with only one light rail, the power supply is pissing away about 2.1W. Although not too bad, this loss may reduce when the second rail goes on and the transformer voltage drops lower, closer towards the LED's voltage drop.

With the three 18W fluorescent tubes (54W) the light meter was reading 1100 lux on the surface of the bench (light meter pointing up towards the light).
With five LED's running at 11W (55W) the light meter was reading 4200 lux on the surface of the bench.
After about an hour I measured each heatsink at between 49 to 49.5 degrees with respect to each other. Not too warm I don't think, the dies are measuring about 59 degrees. Having said that, I am not too sure how accurate an IR thermometer is at reading temperature of an illuminated die. Oh well.
Now, the light rail is closer to the surface of the bench than it will be when it's properly mounted, but the increase in light for (currently) the same input power is very impressive.
I will hopefully get the second rail completed tomorrow and I'll post my experiences.

[SgtRock]

Dear Kibi:

--Congratulations on a fantastic series of posts. You are truly doing engineering, which is often what you can make with what is available at a reasonable price.

--When you are all done designing, please consider posting a schematic.

--To get the absolute best thermal transfer, I would recommend you use diamond thermal paste (see link below) and bolt the dies down tightly (as recommended by Cree).

http://www.heatsinkfactory.com/ic-diamond-7-carat-thermal-compound-15-gram-p-16605.html

"A low voter turnout is an indication of fewer people going to the polls." Al Gore


Best Regards
Clear Ether

[Kibi]

Dear Kibi:

--Congratulations on a fantastic series of posts. You are truly doing engineering, which is often what you can make with what is available at a reasonable price.

--When you are all done designing, please consider posting a schematic.

--To get the absolute best thermal transfer, I would recommend you use diamond thermal paste (see link below) and bolt the dies down tightly (as recommended by Cree).

http://www.heatsinkfactory.com/ic-diamond-7-carat-thermal-compound-15-gram-p-16605.html

"A low voter turnout is an indication of fewer people going to the polls." Al Gore


Best Regards
Clear Ether

Thanks

I'll post my final schematic when I'm done.

That thermal compound looks good. It's better than what I am using now which is rated at 2.9W/m-K. However, as it stands, with the die running at 60 degrees, it's still well below Cree's recommendations, so I don't really need to worry.

[Psi]

yeah, my heatsinks are around 50deg too. And it's my room lighting, so its used for many hours a day over the past 4 years or so.
Not a single led has failed.

The Cree LEDS i'm using are old XLamp 7090 XR-E (pure white)  Driven at 1A

[Hypernova]

If you got a toaster oven have you considered roughing out the contact surface and reflow the module to the heatsink? This would give you the best possible transfer and you would be able to use smaller heatsinks.

[Kibi]

If you got a toaster oven have you considered roughing out the contact surface and reflow the module to the heatsink? This would give you the best possible transfer and you would be able to use smaller heatsinks.

That would be cool, but the heatsinks are aluminium and so are the bases of these LED's, so solder is not going to work on them.

[Psi]

Thermal epoxy works quite well, but its pretty hard to get them off again.

There is also 'Thermal Glue' which is much easier to pry apart later (compared to thermal epoxy)

And then there is double sided sticky thermal tape. I'm not sure how well that stuff works though.

[Kibi]

Thermal epoxy works quite well, but its pretty hard to get them off again.

There is also 'Thermal Glue' which is much easier to pry apart later (compared to thermal epoxy)

And then there is double sided sticky thermal tape. I'm not sure how well that stuff works though.

I have some of the thermal sticky tape. It's not too bad for what it is, but it did not work so well with these LED's.
What I have done is working well, but I'll keep an eye on it. If it starts causing problems then I'll think about the adhesive options. That "Thermal Glue" looks like interesting stuff.

[Kibi]

So, I did complete the second bar of LED's over the last weekend, but that was all.
Work has been pretty hectic this week with 18 hour days combined with heavy drinking at go-live parties. Oh well.

I have connected it up and now there is a problem with the power supply. The voltage drop out of it is too high. The transformer is still operating within spec. 33vAC under the full load of the circuit. The ripple voltage across the filer capacitor is 0.56V. However, I've misplaced over 2V somewhere. Like a dumbass, I completely neglected the parameters of the bridge rectifier. It's an 8A device which I had in stock. The is fine because this power supply operates at under 3A, but without a heatsink the rectifier is currently operating at 90°C! I decided to consult it's data sheet and it confirms my suspicions. The data sheet says that it will reach 100°C at 3.2A without a heatsink and will drop 1V per diode at 4A @ 25°C.
So basically I need to find an extra 0.6V from somewhere. I think a better rated bridge rectifier that will operate at a lower temperature and drop 0.6-0.8V per diode. Then I will reach a high enough voltage (just oboe the voltage drop of the LED's @ 270mA) for the current limiter circuit (which is doing the square root of bugger all at this time) to operate. The LED's are currently only operating at 210mA per device. I need them to operate at 270mA to gain another 1000 lumens overall and have them operating at their optimum lumens / watt.
I'm thinking one of those big square 50A rectifiers will sort me right out or go down the schottky route.

Oh well, we're getting there. Electronics Engineering - who'd do it.

[Zero999]

Why not just replace it with a Schottky rectifier? The voltage drop will be around 0.6V at full load.

[Kibi]

Mmm,I like your thinking. Tthe wheels are turning. Just have to think about heat wastage.
The idea of this project is efficiency. I am leaning towards the schottky route.

[Psi]

yeah, try some schottkys, there lot better than normal silicon diodes. You'll probably get an extra 0.3V on each side of the rectifier.

You can even put a few schottkys in parallel if you need just a tiny bit less voltage drop.
The more you share the current between them the less current flows through each and the lower their voltage drop is.
You can't guaranty the current will split evenly between them all so each one should really be rated to handle all the current. Also the effect doesn't really improve much past 5-10 diodes, (law of diminishing returns).  It's only really useful to gain another 0.1 -> 0.2 volts

I used 10x 1N5817 (20V 1A) in parallel for a project a while back and it worked really well. I got another 0.150v  @ 1A which was all i needed.
I wouldn't recommended it for commercial products though, diodes in parallel isn't really good practice.

[Kibi]

Thanks very much for the input people.

I am going to use shottky diodes. I've ordered some and they should turn up tomorrow.
The next idea that I cam up with in the shower (as you do) was to use just two diodes for full wave rectification utilising the twin secondaries on the transformer. That way, I don't have the current flowing through two diodes at a time as is what happens in a bridge rectifier.
I'll also wire up two schottky's in parallel.

[Alex]

There is aluminium solder but I would not risk heating the LED at those temperatures.

How about a structural I-beam for heatsinking and mounting? You can place the LEDs a bit higher this way not to obstruct access to equipment. One side of the I-shape can accept the edge of the shelf. The other side you can drill holes for extra heat convention but I doubt you will need it with such a large surface area exposed to an air stream already.

About diodes, you can also consider active ones. These utilise a low ON-resistance mosfet and some control circuit. You can make your own, just remember those body diodes...Schottky, as proposed, should make your constant current circuit the limiting factor in terms of efficiency in the entire design.

Nice project. You could probably buy two more LEDs for the cost of those connectors!

[Kibi]

There is aluminium solder but I would not risk heating the LED at those temperatures.

No, I wouldn't want to risk that either

How about a structural I-beam for heatsinking and mounting? You can place the LEDs a bit higher this way not to obstruct access to equipment. One side of the I-shape can accept the edge of the shelf. The other side you can drill holes for extra heat convention but I doubt you will need it with such a large surface area exposed to an air stream already.

I had given this some thought, but a metre long strip of aluminium that I thought "might" do the job, was the same price as five heatsinks that I "knew" have the right thermal charactaristics.
I have since found some cheaper heatsinks from RS of all places.

About diodes, you can also consider active ones. These utilise a low ON-resistance mosfet and some control circuit. You can make your own, just remember those body diodes...Schottky, as proposed, should make your constant current circuit the limiting factor in terms of efficiency in the entire design.

Sure, the excess energy is going to go somwhere be it in the rectifier as it currently is or in the current limiting circuit. I'd rather it went in the current limiting circuit.

Nice project. You could probably buy two more LEDs for the cost of those connectors!

Yes, those connectors. I saw more expensive connectors. I saw cheaper connectors. Umm'ed and aaah'd a bit and settled with these even though they were still a bit pricey.
The IP68 connectors that I will have to use if I go for an outdoor lighting project seem expensive....... and then you have to buy the pins for them which cost even more that the connectors!

[ashley.hughes]

Awesome looking lighting jealous as.... Sitting in my dark workspace in my room (Tv blocks out most of the light.... and sun) 

[Kibi]

Right, so it's finished!

The schottkey diodes only arrived the other day, so I was messing about and hacked the transformer to up the voltage a bit. Got to love toroids, very easy to hack.



So when the diodes did arrive, I soldered them on. I could only find SMT ones at short notice. Much better, they are arranged two in parallel as some people suggested I do and re-configured the wiring from the transformer so that the rectifier is not a bridge configuration anymore, but it's still full wave. That way only one diode (or two in parallel) is in the path of the current as opposed to two in bridge mode. The rectifier only gets slightly warm now.



Now the readings are much better, the LED's are operating at the full 270mA I want them to and the power supply voltage and LED voltages are good.
MOSFETS are running warm too, so I am not losing much energy by means of heat.



So, the lighting project is ready just in time for me to dismantle the whole lot and move house - again. Oh well, at least I'll have cool lights in the new house.

Thanks to those who provided me with valuable advice. 

[logictom]

Nice project and great result
If you have the time would you throw together a schematic and list of parts? Which heatsinks did you use and could you link to the one you mentioned you found from RS?
Trying to gather as much info before I embark on an LED project of my own

[Kibi]

Sure, I'll endevour to get a schematic posted up. It's a bit awkward now that my lab is all getting packed away in crates.

I used these heatsinks:

http://uk.mouser.com/Search/ProductDetail.aspx?R=882-300ABvirtualkey56720000virtualkey567-882-300AB

The page below has a list of heatsinks on RS that are a fair bit cheaper. I think you'll have to drill them yourself to suit though. However, I'd use these next time as the price is better. Scroll down to find the Fischer ones, quite a choice of funky shapes and sizes:

http://uk.mouser.com/Search/ProductDetail.aspx?R=882-300ABvirtualkey56720000virtualkey567-882-300AB

[JimmyM]

If all of your emitters in series would need ~400V, why not just arrange them in series strings that will run just below rectified mains voltage. That way a constant current buck converter can be used. Eliminating the high current bridge rectifier for a lover amperage/higher voltage one and eliminating the transformer altogether.
A simpler, although less efficient method would be to try to match the number of LEDs in a string such that they run at a little more than 270mA when cold (higher voltage). When they heat up, their Vf at 270mA will drop. Use an LM317 in constant current mode. As long as the input-to-output voltage doesn't exceed 37V (actually .270A*Vdrop < Max dissipation) it should work. Right? Or am I completely missing something?

[M. András]

if you look at the datasheet of these led arrays, there is a maximum current rating which is 1 amps at the graph almost at the end of the datasheet you can see the forward voltage versus forward current and on the other graph there is the rel. luminous flux which is 100% at their 270ma/40volts test condition, but its works up to 1 amps max with almost 2.75x the test results at lower ratings which means less led array for the same output level, even there is a thermal vs flux graph if you somehow could keep the junction temp of the array below 30 degree celsius at your 270ma drive current would be approx 10-12% brighter (hard to read that pdf graph). anyway its all of my opinion

[Kibi]

So the heatsinks I was using before, although shallower than a fluorescent lamp unit, protruded too much and got in the way.
I have re-thunk the idea and have now mounted the same LED's on a 25mm x 25mm square tube. It is much shallower and neater.
The tube with 5 LED's on it does get about 15 degrees hotter than the original heatsinks and the LED dies get up to about 65 degrees now, but that's still OK for them.

[M. András]

isnt it too bright a little bit or its just the picture, btw they bin the cxa2011 at 85 degrees. anyway nice work

[terabyte]

So the heatsinks I was using before, although shallower than a fluorescent lamp unit, protruded too much and got in the way.
I have re-thunk the idea and have now mounted the same LED's on a 25mm x 25mm square tube. It is much shallower and neater.
The tube with 5 LED's on it does get about 15 degrees hotter than the original heatsinks and the LED dies get up to about 65 degrees now, but that's still OK for them.

VERY NICE!  I want to do something like this for my bench.  Is that aluminum bar solid or a tube?  If a tube what is the wall thickness?