Some energy is, of course, converted to light. The percentage of it, however, is quite minimal. White LEDs are still not very efficient at turning electricity into lumens - they’re far more efficient at turning electricity into heat. It’s not 30 watts of heat… probably more like 26 or 27… For thermal analysis, however, I usually just assume it’s all heat - gives me a little margin or error on the favorable side anyway.
PPtk
Hm... Am I correct in assuming that LEDs are most efficient? Also are other colors more efficient. Thanks.
Some colors are more efficient than others. Royal blue, for instance, is quite a bit more efficient than white. I can’t speak intelligently about other colors, however, I just don’t know…
Thanks, I was hoping they were more efficient than that!
Still more efficient than any other lightsource iirc..
Underneath the skin of a white LED is a blue led. They use a phosphor to convert the blue light into white… and, like everything in the universe: a) You can’t get something for nothing. b) You can’t even break even. c) You will die trying. (The three laws of thermodynamics)
Correct. Hence the reason I speciffically used Royal Blue as the example. The laws of physics demand that the blue must be more efficient than the white since a conversion(energy loss) has to take place in order to make white out of the blue.
So would a blue CFL be more efficient or not since they normally put out white light?
Maybe - i’m not sure. But if it were more (or less) efficient, it wouldn’t be for the same reasons. CCFLs naturally create a white’ish light - there is no phosphor conversion taking place. I strongly suspect that if there is a gas that creates natural blue light it would be very similar in efficiency to a white ccfl - but that is no more than a guess. Either way, what I can say for sure is that no matter the answer, the reason for the difference (if any) would be a completely different one than in the white LED situation.
PPtk
Actually fluorescents do have a phosphor conversion. That tasty mercury filling makes ultraviolet light that excites the phosphor. Leakage of UV light from CFL tubes can be a problem. It can cause degradation/fading of pigments, paints, and dyes. CFL tubes that emit minimal UV light have extra filtering built in.
And there are ways of upconverting long wavelengths of light into shorter wavelengths. Most cheap green lasers work on this principle. The output of an 800 nm laser diode is down converted 2:1 to form 1600 nm light in a non-linear crystal. That 1600nm light is then upconverted 3:1 into 533 nm green light using a different non-linear crystal. The overall efficiency is very poor and the required alignment and temperature stability of the crystals is very critical.
You learn something new every day. I’m quite familiar with laser conversion - I work with lasers a lot… CCFL, however, I’m no expert on at all. I didn’t realize that UV was the native output. Cool.
Here is a housing I have drawn up in Solidworks. It weighs 90g and can easily dissipate the 30W of heat when riding at 12mph.
Rendering takes ages so there is actually one more shallow fin which you can see in the following pics. Simple little adjustments meant the stabilised temperature fell from a hot 114C to 65C. A chamfer here, extra depth groove there makes all the difference. I found the 'firewall' as I call it, the depth of solid metal behind the LEDs, to be at the mose efficient at no more that 4.5mm. Any more and the weight went up with no noticeable improvement in cooling.
The test was done with an initial solid temp of 60C to speed up the calculations. Air speed at 12mph, my normal riding speed when needing 3000 lumens! Air temp at 20C, a little hopeful here in England. Three LEDs on a copper disc simulating roughly Pilots module, each at 10W.
More pics to follow...
Here is the side view. Note the extra shallow groove a third of the way in from the left. It's not shown on my first render. The hole on the the thicker portion is for mounting. There is one the other side too as it will be mounted with a cradle held on by the bolts on my stem/face plate.
The scalloped portions seem to help with the turbulence, breaking up the boundary layer to draw the heat away. Without them, the air seemed to flow over the fins.
I know it would be much more efficient to have the fins running into the air flow but there is no way I'm going to be able to easily do that on my manual lathe! The model said it would stabilise at 34C though.
There is a hole at the back, either for a plug of a cable gland. I'll have the switch on a fly lead.
Here is the graph. I didn't run a transient (timed) test at my pc would have thrown its toys out of the cot, but you can still see the rise of the LEDs in red against the max temp of the housing in green. Once the templevelled out, the test stopped. Like I said earlier, I started at 60C to speed up the calculations. With a fine mesh setting and narrow channel refinement it could take upto 2 days to complete! I placed a local mesh at the front as I noticed that portion played the biggest part in getting rid of the heat. The rear section had a rough mesh to speed up things.
Its just plain awesome what today's software can do in gifted hands..
PilotPTK - I know I am late to this thread, but that is a very nice design. Great job with the thermal vias and the large copper areas for thermal transfer.
I really liked the small oval metal “posts” that solder to the board, yet by being hollow make it easy to attach a test probe - where can I find those? Who makes them?
Same here. I try/look for it first on Ebay :bigsmile:
Will
Thanks Will - A lot of time and thought went into the design, and I’m pretty happy with it. The software/firmware is just as neat as the hardware though.
The test points are awesome. I use them for practically everything. Small enough to not be a real-estate burden, big enough to probe and super easy to attach test-leads to. They’re keystone part number 5015
http://www.keyelco.com/products/specs/spec130.asp
Mouser usually has them in stock, and I’m sure other places as well.
I order them by the thousand, because I really do put them on everything. One board I designed that’s full of analog and needs to be calibrated has over 800 of them on it. One Board. Ugly. Takes a good tech about 2 days to calibrate it. Couldn’t imagine doing it without good test-points.
PPtk
Awesome - thank you very much. The links you provided also have the SMT pad, so I will try to incorporate the pads in future boards :bigsmile:
Will
The keystone suggested pad is .135 x .070. I use a smaller pad, however, as I had trouble with the test points floating around too much.
My pad is:
Copper: .120 x .055
Mask: .125 x .060
Paste: .105 x .040
PPtk