16.7 Watt LED Lamps
I'm currently building a couple of high-power LED work lights, using some Cree modules capable of around 2000 lumens of flux. I opted for 5000 K colour temperature because these lamps are for working under, not for mood lighting. I also opted for two parts with identical Vf of 37 V and max VI of 450 mA. Where the parts differ is their CRI ad luminous flux.
I opted for one high CRI (90), which has a reduced flux of 1590 lm, and one lower CRI (70) with an increased flux of 2260 lm The parts also have a sensible efficacy of about 140 lm/W for the low-CRI part and ~100 lm/W for the higher CRI.
Heatsinking is a must with LEDs of this power. Some LEDs today will reach as high as 50% efficiency, so when the power of the LED climbs beyond any small amount, some heatsinking is required to move this excess energy away from the device. I am not a proponent of sticking fans on something just because it produces heat energy at all, and I prefer to run heatsinking without fans where it is allowable. They can create undue noise, and can quickly become a maintenance item where dust filtering is not used. Additionally, any moving part can potentially decide to stop moving, so there is the issue of maintenance with respect to bearings or bushings, too.
Knowing this, I investigated repurposing CPU heatsinks for my LED lamps. When coupled with their matching cooling fan, they are often designed to dissipate sometimes up to 80 or 90 W of heat energy for "stock" coolers. Without the fan, some sources on the web recommend a derating their cooling capacity by 90%.
The heat waste that needs sinking from the parts is reasonable. Cree's fine application note on thermal management of their XLAMP parts provides a rough figure of 60% of the input electrical power being wasted as heat, but they recommend a conservative figure of 75%. Much better to slightly over-engineer the thermals than to under-engineer.
The total power of the LEDs is approximately 16.7 W, and using Cree's recommended factor of 0.75 gives a total thermal power of about 12.5 W.
I am unable to run thermal simulations to determine the genuine derating factor when running my chosen heatsinks without fans, nor do I have information on their rated power, since I don't have the equipment they were removed from. It shall have to be determined experimentally. If the chosen heatsinks prove unable to dissipate the required 12.5 W passively, then I am prepared to experiment with running the fans at partial power so that they do not produce perceptible noise.
These sorts of LED cobs are popular with other electronics hobbyists, so there is a fair amount of information on the web relating to driving them. Many people buy dirt-cheap 100 W (claimed) modules from eBay sellers, and there have been mixed reviews on them. To drive them, people generally also buy dirt-cheap DC-DC constant current regulators (from eBay again), which are generally based around TI chips like the LM2595 and its cousins. These types of boards are of severely inferior quality and they use fake chips which cannot handle short-circuit conditions, amongst other problems.
For this reason, I didn't want to touch the boards with a barge pole, and set about making my own driver. For now, I will drop the schematic and a link to the Kicad (eeschema, pcbnew) files. I intend to revisit here once I have built the boards, let some magic smoke out, and revised the design to tell the story from design to high-brightness desk lights.