Step 6: Test

With the driver mounted, you are ready to test. You can do this at your bench or installed in the fixture or choice. If you see a lot of flicker then your fixture may not be low voltage DC but rather low voltage AC. You need to modify the circuit to handle the AC by using a full wave rectifier and some capacitors, not too bad but not ideal. Enjoy, and be safe!
Do you need a heatsink that large? I thinking about using these for cabinet lighting and something that large would be hard to hide.
A good heatsink is really important to maintain as low a junction temperature as possible. Airflow helps a lot but in a passive cooled heatsink size and surface area are the only ways to improve cooling. You could find a larger area but shorter height heatsink that might be easier to hide. Let us know if you find anything that works.<br/><br/>I have been working on an undercabinet design for a while trying to come up with a good solution. The best that I have come up with is to use an aluminum extrusion (u-shaped) as a mounting, housing and heatsink and to use lower power LED strips rather than discrete high power emitters like the Rebel. Such as <a rel="nofollow" href="http://ledsupply.com/elarastrip.php">the elara strip</a>. It is more expensive but has a built in driver ready to take low voltage DC and can be chained. Not a completely satisfying solution, but the best I have found so far. You only get about 150 lumens per foot in warm white so more of an accent rather than task light.<br/>
Where do you get your warm white LED strips? I'm confused by the enormous price range and am never sure if the high-priced strips that claim to be better, are really only better at marketing.
Oops. Just realized you linked to the LED strip.
what about using a old cpu heatsink
Sure, if you can find one with the right form factor to fit. The circular Zalman heatsinks might be a good but pricey choice. The traditional extruded, square with lots of fins, heatsinks are not suited for integration in residential lighting fixtures. For making a bright light that you hook up on your benchtop to see it light up they work fine, but replacing an incandescent, halogen, or CFL bulb is another story.
Those heatsinks might not work as a typical bulb replacement because of the shape, but how much surface area do the leds need at this current? I went to the data sheets, but all they said was you should use something and there will be more info here later.
It will be the heatsink rating that will tell you the performance of the heatsink and its suitability for use in the LED lamp. The number we are interested in is the passive degree C per Watt rating (from here on out I will abbreviate this as C/W), the rise in temperature for a given input power in still air. In general, the bigger the heatsink and the more fins, the lower the C per Watt rating and the better the heatsink. If you add a fan, then you will need CFM airflow numbers and commensurate heatsink numbers at that flow.<br/><br/>For example, the heatsink I used is rated at 5 C/W. The LED thermal pad to thermal backplane on the star is rated at 6.3 C/W. The LED junction to thermal pad on the LED is rated at 10 C/W The star to thermal grease to heatsink transition can be guestimated to have a thermal resistance of .05 degree C square inch per Watt and if we assume 1 square inch contact area between the star and the heatsink then 0.05 C/W shows that this is pretty negligible and will be ignored in the rest of the example. So we have a total thermal resistance of 6.3+5=11.3 C/W from the LED thermal pad to the ambient and another 10 C/W from the thermal pad to each LED junction. The manufacturer lists the max LED junction temp as 150C, but you will want to stay well away from this level in order to maintain high light output and long life. The light output will fall as the junction temp increases; see the datasheet for the Rebel for more info. For reference, the light output drops to 90% at LED thermal pad temps of 60C. In this case if we assume that the heatsink is at the star temp and ambient is 25C, then we can assume that there is a 19 degree drop across the star ((60-25)*6.3/(5+6.3))=19 and the heatsink is at about 45C. This gives a rough guess at power flux of 4W since the heatsink is rated at 5C/W and the temp of the sink is 20C above ambient. This is just a back of the envelope guess, as the sink will not be at uniform temp as well as other factors we will gloss over. At this power level, each LED will be outputting 4/3=1.33W. Based on the LED junction to LED thermal pad data this means the LED junctions will be 13C above the thermal pad (10C/W) and at 73C which is well below the max rating.<br/><br/>If we start with a 9W input power, then the heatsink will be 45C above ambient, at 70C (hot!) and the LED thermal pad at 107.5C and each LED running 3W will have a LED junction temp of 137.5C which is getting close to max. The light output will be down to ~80% as well.<br/><br/>Of course this is just a rough estimate but you can see the problems inherent in designing a good LED bulb and why the commercial fixtures have such big heatsinks. Bigger finned heatsinks (e.g. 4 by 4 inches and 2 inches tall) have natural convection ratings that approach 1C/W. If you add a fan, then everything changes and the heatsink rating drops by an order of magnitude.<br/><br/>Good luck, I hope this info helps.<br/>
You might actually be able to use one of those mini 12v blowers instead of a heatsink, or along with a TO-220 chip heatsink stuck to the back of the star.
You can certainly use a fan, it just increases the size, complexity, and noise. If your fan fails without you noticing, the LED will soon follow.
Size? You can find very quiet, tiny fans for pretty cheap.
Your best bet is a fan from a laptop. They are tiny, and generally have integrated heatsinks.
Sounds ideal if you can fit all the parts in. Prolly not difficult if you dremel away some fins in the middle.
well i was thinking more along the lines of OLD intel pentium oem sinks, or something found on a amdk6/2 or penitum 3, where the footprint is ~ 2 inches square and ~an inch tall<br/><br/>something like <a rel="nofollow" href="http://content.etilize.com/Large/11395738.jpg">http://content.etilize.com/Large/11395738.jpg</a><br/><br/>not a huge modern one<br/>
That should be even easier.
Anything that fits where you need it will be OK as long as the thermal performance is good enough. See my answer to e_lectro above for more discussion on heatsink rating.
If anything, you need a heatsink larger IMHO. The cooler they are, the better LED's run (and 50,000 hours is only when kept cool- overheated, you'll be lucky to get 1000 hours). Passive heatsinking (no fan) will require a very large heatsink. If you can find a small, quiet fan, then you could use an active heatsink- the fan rigged to the smallest CPU heatsink would be sufficient to keep the LED's running nice and cool. The cool thing about LED's is what they can if integrated properly- i.e, water cooled with the water that's entering the water heater. Now that'd be pretty swish : )
I agree, I need a bigger heatsink for the 3-up rebel star. My initial reason for purchasing these heatsinks was to use a single Cree (or Rebel) on each sink and to build a linear array fixture for over a sink. That project is still underway. Stay tuned. For the above halogen replacement with 3-up star, the goal is to run it and see how long it lasts. This fixture serves as our "night light" in the house and is on more than any other in the house so I hope to get data from the experiment more quickly while reducing our energy usage by replacing the 50W bulb with LEDs. Water cooled? I can see it now, special 14/2 ROMEX wiring with integrated PEX 1/8" water tubing for cooling. Just don't kink or nick it. Nightmare on my street. Fans are a tough thing to integrate into an existing fixture, and no one wants to hear their lights. In a fixture designed from the ground up to use LEDs it would be much easier to build in fans, but if building from a clean slate then I would design a fixture that leveraged the distributed light capabilities of small LEDs rather than the single point source of bulbs. For example, spreading a large number of small LEDs over a long strip or plate rather than a linear array of 3 bulbs. You could build the structure to serve as a heatsink and cool it passively while using a large number of LEDs spread out to reduce local heat load.
We have bought a lot of <a href="http://www.hero-ledstore.com">G4 Led Bi-pin halogen bulbs</a> from the hero led store. They have a wide range of halogen led replacements bulbs. <br> <br>
I'd like to do this for a light that runs GY8.6 75W 110-130V hallogen bulbs. Do you think it's possible or practical? What would I need to do in addition to this protocol?
Yes it is possible. The main difference will be the driver circuit. You will need a driver that has a step-down transformer. DealExtreme sells an inexpensive module, linked below, that can drive ~9W of LEDs. However, if you are looking to get the same light output from your LED replacement as you got from your 75W halogen you may need to go a different route. With a 75W halogen putting out around 1200 lumens of warm light, you will need about 15W of LED output at 80 Lumens per Watt. This is more than the driver circuit linked below can handle, more heat than the heat sink shown in the above Instructable can handle, and more light output than a single 3-up Rebel star or Cree MC-E can output. <br/><br/><a rel="nofollow" href="http://www.dealextreme.com/details.dx/sku.13556">http://www.dealextreme.com/details.dx/sku.13556</a><br/><br/>So, to summarize, you can use a different driver (linked above) modified to increase the set current (change out resistor) and make your own pins for insertion into the fixture and you will then be able to make a LED &quot;bulb&quot; like the one in the Instructable that runs on 120VAC and outputs ~300 warm lumens. If you need more light, you can try a Cree MC-E for another 100 lumens or so, you can switch to cool white for another 300 lumens, or you will need to come up with a better heatsink, a beefier driver, and more LEDs or higher power LEDs (see link below).<br/><br/><a rel="nofollow" href="http://www.ledssuperbright.com/25w-led-c-22/25w-high-power-led-1400-lumens-p-108">http://www.ledssuperbright.com/25w-led-c-22/25w-high-power-led-1400-lumens-p-108</a><br/>
Something seems not to add up here: A typical halogen is about 50 watts & its efficiency is not hugely different to LED's as (I understand things), as for example a compact fluorescent is, being 3 to 4 times as efficient as a halogen. So how on earth can an array of three LED's, total 10 watts, be anywhere near the light output of the replaced halogen ?
The thing that you are missing is that an LED can be hugely different than a halogen in terms of efficiency. Cool white LEDs are available that are rated 100 lumens/watt which is greater than the typical CFL at 50-75 lumens/watt and a typical halogen at 15-20 lumens/watt. Warm white LEDs are not as efficient and peak at around 80 lumens/watt on the market today. Certainly there are less efficient LEDs, and the bulbs out there on the market that use a huge collection of small through-hole lead LEDs do tend to give LEDs a bad, anemic rap. However, the latest high output high efficiency LEDs (Philips Rebel, Cree XR-E, etc) can put out an amazing amount of light. So that is how a 10W LED array putting out 1000 lumens is near the output of a 50W halogen putting out 1000 lumens.<br/><br/><a rel="nofollow" href="http://en.wikipedia.org/wiki/Luminous_efficacy">http://en.wikipedia.org/wiki/Luminous_efficacy</a><br/>
y not just use this on a bike with a bike generator????????
There is a really cool looking heatsink available that is for LED stars to be mounted to. The picture is only different from the ones I have purchased as they came Black Anodized(Way cool, has a Star Trek-Klingon kind of thing to it). This one is 12.7mm Height, 70.1mm Diameter and a bit under three bucks. Also available in 35.4mm, 50.8mm and 76.2mm H., all 70.1mm Dia. Largest is above nine bucks. I'm about to use them for flashlights but am having a small artistic brain freeze design wise as I want them to be exposed because of their unique aesthetic appeal. But as you can see there are a number of mounting options and multiple emitters is what I am trying to make happen so bolting a number of them together won't be a problem. Lux Overkill is what I have appropriately named this project and it does have a certain ring and truth to it don't 'cha know. I am not sure is it is allowed to post the retail outlet site that I got them from in this forum but it's all in good fun and building stuff so here goes. Hope this helps out. Newark.
Thanks for the heads up. I have used the Lamina radial heatsinks in the past (also available from Newark) but they are really expensive. Their ratings are also not much better than the Wakefield parts you suggest, which are new to me. They look like a great fit, I will order some ASAP and give them a try. Their thermal performance is a little better than the sink I used above, which is to be expected due to the larger size.
Your Welcome. These are really interesting in that they can be stacked so if one isn't quite getting the job done no need to even pull the star off. I have 6 of the 12.7mm and 4 of the 35.4mm purchased five or six weeks apart and the manufacture is perfect. Running wire is a cinch and easy to hide. With a bit of felt on one side make table coasters that gets noticed like you wouldn't believe and the best serving trivet ever. I put three 12.7mm into a trivet as a joke but after adding rubber chassis box feet rather than felt turning it machine washable... Never thought I'd see the day a heatsink would get me excited. Go figure. >:)
Any idea if the bi-pin drivers can be found (or modified) to run off 6v?
The driver IC on these circuits is specced to work on 8-30V input, so no it will most likely not work. That said, it might work for a single LED with a reasonable Vf (~3V) and a low current. My experience with these buck circuits has been that as input voltage drops below the minimum overhead plus Vf total for the LED string that the drive current will drop but the LEDs will remain lit. The Vf is a function of drive current and typically is lower for less current. Eventually if the input voltage drops below minimum overhead plus Vf at very low current the LEDs go out. For example, a 3-LED Rebel star has a Vf=10.2V at 700mA and Vf=9.5V at 300mA. If the buck driver overhead is 1.5 volts, then input must be 11.7 or great to maintain 700mA. As input (e.g. from a battery) droops below this value the maximum drive current will drop as well into equilibrium. This ignores any protection circuitry that might be present in your battery pack, such as in a Li-Ion pack that cuts out below a minimum charge state. <br/>
I always find the colour of the light from LED's isn't particularly nice on the eyes. maybe its just me?.
hence the need for using the warm white LEDs. Their CRI (color rendering index) is much better than the typical cool bluish white LEDs, 80 vs 70. For comparison, some special ($$$) fluorescent bulbs can produce CRIs in the 95 range, but most are in the 80s as well.
Thanks, cool instructable. What is the color temperature in this setup, I gather its the bluish light , 6500? and if we wanted the warmer light it would throw off 2-3 times the heat? Can this be dimmed?
Your best bet there would be to use a yellow filter over the light. I personally like the white glow of the LED's, but I can see where some people wouldn't like it.
Filters decrease efficiency by blocking light, so I can't recommend that approach. The mix of phosphors used in the warm white LEDs gives a warmer look without the use of filters but with less luminous efficiency. An interesting test would be to measure the difference in effective luminous flux of a warm white LED and compare it to a cool white LED with filter. My guess is that the warm white LED without filter would be more efficient. Otherwise, the LED manufacturers would not mess around with making different LEDs, and would just apply different lenses to the LED die. Hmm, maybe a new idea for LED manufacturing...
This is a warm white LED, 3000K as noted by CameronSS. I would rate the light as similar to the halogen it replaced in terms of color. I have Rebel stars that put out much more light (100 lumens/watt) but these are the 6500K cool LEDs with bluish tint. The warmer LEDs are less efficient, so they will get hotter at the same luminous output as the cool white LEDs. For example, to get 1000 lumens using a 100 lumen/watt cool white LED would take 10W, most of which would end up as heat. Similarly, with a 60 lumen/Watt warm white LED would require 16.7W, most of which would end up as heat. The difference in heat output is basically the ratio of luminous efficiency, so 1.67 times more heat in the above example.<br/><br/>The driver circuit as shown does not support dimming. The drive IC can support dimming but uses a PWM approach that would not be compatible with the typical triac dimmers found in most residential applications. I am working on a triac dimmable driver circuit based on the LM3445 reference design provided by NSC.<br/><br/><a rel="nofollow" href="http://www.national.com/ds/LM/LM3445.pdf">LM3445</a><br/>
According to the website to which he linked in the Materials step, it's an average color temperature of 3000K.
<a rel="nofollow" href="http://www.instructables.com/id/Bright_Luxeon_LED_Bike_Light/">Here's another heat sink idea</a> from Instructables. <br/>
do the provide the same light radius as a CFL? and also these are mainly for celing fixtures beacuse the only light up on 1 side right?...
The light distribution is given by the manufacturer on the spec sheet. Basically the light intensity is highest perpendicular from the plane of the board and falls off to ~5% at + and - 90 degrees. The 50% angle is pretty good, at + and - 60 degrees. As you note, these only light on 1 side, but without optics the three LEDs give a really wide spread of light . Not as much light radius as a spiral CFL, but since the pendant I was using had a shade, any additional light would be underutilized.
What about the possibility of increasing the reflectivity of the inside of the lamp shade (chrome or other shiny paint) and mounting the LED package up-side-down? That might increase you dispersion some, but it also might increase the smoothness of the dispersion, making it more even throughout its dispersal pattern. Maybe.

About This Instructable


154 favorites


More by jmengel: Laser Cut Front End Loader Toy Laser Cut Ukulele Electric Brewery Control Panel on the Cheap
Add instructable to: