Confused. Even if you are using 20mA emitters, thermal coefficients do apply and I don't see why this matters. You still have to remove the same 24-watts of heat and I don't see how using a horde of smaller emitters changes the equation. Energy in = energy out. Although, this does explain why the cheaper versions of these lights have poor lifespans.
Yes, you're quite correct about that. The thing is given the surface area of a T8 fluorescent tube replacement, you can keep temperatures reasonable with roughly 3-4 watts of heat per linear foot ( this assumes the tube is in open air of course-an enclosed fixture would require lowering the wattage ). The two-foot flouro tube retrofit I have draws about 9 watts from the AC mains and is speced for 700 lumens ( and yes, it does seem to meet this spec although I didn't lumens test it ). So figure about 2 watts of those 9 going in come out as light, the remaining 7 as heat. The outer part of the tube is roughly body temperature. Inside is a little warmer, perhaps 40°C. Given the junction to ambient impedance of typical small LEDs ( on the order of 200 to 300 °C/W ), I'd expect the junction temperatures to be about 60°-65°C, which is plenty low enough for long life.
Again, confused. Right now SMDs in the ~100mA to one watt class seem to be ruling linear lighting, correct? I see them everywhere from lights like these to grocery store freezers. Again, where does 20mA lighting come in, what problem does it solve over ~100mA SMD, and why would using 5-10x as many of them be cheaper?
Correct, using 100 mA SMDs over 20 mA gives you about 1/5 the number of parts, so it would be cheaper. 100 mA LEDs represent a good intermediate point between the more difficult assembly using power LEDs, and the much greater number of LEDs using 20 mA types. That being said, as Barbarin mentioned in some instances you might still want to go with the 20 mA type for better light distribution and heat spreading. I've been modding nightlights with 20 mA SMD LEDs. I don't really have any 100 mA types anyhow, and I find the 20 mA ones give more even light distribution and less complex drive requirements. It turns out in order to reach the desired output you often need to run enough emitters so that the total Vf is an appreciable fraction of the line voltage. In some cases you might even get by with a simple filtered full-wave bridge. Worst case a capacitor-fed ( i.e. current-limited ) full-wave bridge is all you need.
In theory, but can you run any type of LED this way, correct? Again, 24-watts of 20mA emitters consumes the same power as 24-watts of Bridgelux or Crees. The voltage drop is determined by the series value, and obviously the less you have to muck with native voltage the less penalty there is along with making your driver cheaper. Still, we're talking about lumen values in the thousands at 24-volts and not 180 lumen retrofit bulbs. Plus, isn't a regulated driver mandatory for safety, or are we talking strands of xmas lights here?
Yes, you can use a capacitor-fed full-wave bridge to drive even power LEDs. The kicker is that the current-limiting cap has to be on the order of 10 uF or so, if not larger, and rated at 250 V or better ( for 120 VAC line use ). For 240 VAC the requirements are even more onerous. Such caps cost quite a bit more that the 0.47 uF or 1 uF ones needed to drive 20 mA LEDs. Not an issue for a one-off DIY project, but a problem if you're making hundreds. Given the cost and size of the needed cap, in most cases you're better off going with regulated switching supplies for driving power LEDs off line voltage. And no, a regulated driver isn't mandatory for safety, but honestly I wouldn't make a power LED setup for anyone, except maybe myself, without one. You can fuse and protect a capacitor-fed full-wave bridge, of course, but this type of circuit really mostly makes sense for driving 20 mA, and perhaps 100 mA LEDs. Just because it
can be made to work with power LEDs doesn't mean it's the most optimal solution.
Spreading the light though has a substantial benefit. I'm trying to figure out how to do this efficienctly with larger emitters.
And if you ever do it might well be patentable. I'm not kidding. One of the big problems with power LEDs is the fact that they're an obnoxiously bright point source. This is one reason alternate solutions ( 20 mA and 100 mA ) are still being used ). Even in cases where the heatsinking requirements aren't onerous power LEDs might be a nonstarter on account of not easily being diffused. Yes, you can go the indirect lighting route, aiming the light bars at the ceiling, but you squander a large portion of your efficiency in the process. Some solution of diffusion so you can use the light directly would be quite welcome. I have seen light guides where a light source is put on the end of a tube, and then the tube lights relatively uniformly, but IIRC the efficiency of those is horrible. I was actually trying a homemade version of this a few years ago when I was trying to light HO scale coaches because all I had at the time were narrow-beam 5 mm LEDs. Of course, that project got shelved once I found the wide-angle Osram SMD LEDs on eBay.
Would you care to post pictures of the fixture itself? 
