LEDs as lamp replacements

snipped-for-privacy@manx.misty.com (Don Klipstein) wrote in news: snipped-for-privacy@manx.misty.com:

Ok, but try thermally coupling a CFL, or any part of one, to a heatsink. LED's and their drivers are much more easily adapted to use existing structures to carry the heat away. That's why they can occuppy smaller volumes. This is actually done, I mentioned the Clifton Suspension Bridge in another post, that's covered from end to end in them, they're tiny, and extremely bright, and they use a small metal cowling to carry heat from the emitters. They're not 20 watts, more like 10, but the total size of the lamp is similar to a low-volt halogen, far smaller than an 11 watt CFL.

snipped-for-privacy@manx.misty.com (Don Klipstein) wrote in news: snipped-for-privacy@manx.misty.com:

I think so too. And I hope so. Phosphors don't allow colour mixing, and that's one of the biggest strengths of LED's.

snipped-for-privacy@manx.misty.com (Don Klipstein) wrote in news: snipped-for-privacy@manx.misty.com:

Ok. I thought more laser diodes were but never mind.. Aren't most class 3B visible red diodes around 20% efficient or more though? That still leaves a lot of headroom. Tungsten is often said to be 1% to 2% efficient at making visible light. So a 100W incandescent 17 l/W at 1% to 2% places the Cree XR-E's 50+ l/W at 3 times that, up to 6%. These are loose figures but they suggest that if LED's reach efficiencies like DVD writer diodes, maybe 3 to

Wikipedia says 2.6% for the ubiquitous 100W tungsten filament bulb and 3.5% for quartz halogen.

Graham

some have yellow fluorescent dye and a blue LED die, it's sort of the same, but I think the fluorescent lamp has a better spectrum.

Bye. Jasen

Eeyore wrote in news: snipped-for-privacy@hotmail.com:

For a 110V type at 1700 lumens, perhaps. 240V types only put 1200 lumens.

More like 6-7%. Each watt of tungsten radiation in the 400-700 nm range is around 250 lumens.

Figure around 250-300 lumens per watt of "white LED light". Looks like those achieve about 20%.

- Don Klipstein ( snipped-for-privacy@misty.com)

snipped-for-privacy@manx.misty.com (Don Klipstein) wrote in news: snipped-for-privacy@manx.misty.com:

So why do Cree, and Wikipedia, and probably many others, cite only 17 l/W? Every time I've seen the efficiency expressed as a percentage it's been '1 to 2%'. Wikipedia state 2.6% which is a tad higher than I've ever been told before, and even that's nowhere near 6-7%! It's not going to be easy to learn if everywhere I turn there are figures differing by factors of three or more. What makes all the others I've seen wrong?

I have a homebrew BASIC program with the blackbody function and the photopic function.

A USA-usual "Big-3" brand 100W 120V "standard frost" or clear incandescent rated 750 hours average life and with a coiled-coil filament is rated to produce 1710-1750 lumens, traditionally 1710. (The "Soft White" version achieves 40 lumens less.) The color temperature of that one is 2865 K.

My homebrew program says 16.7 lumens per watt (pretty close) and that

The discrepancy is caused by tungsten having emissivity varying with wavelength - generally inversely. Infrared radiation is suppressed enough to get 17.1 lumens/watt instead of 16.7 despite the lamp having some heat conduction loss. (For that matter, color temperature does not exactly match filament temperature - filament temperature is slightly lower.)

So an ideal blackbody at 2865 K receiving 100 watts and radiating 100% of this produces 6.63 watts of visible light and 1670 lumens. The ratio of lumens to watts of visible output is 252, not 683. 683 lumens in a watt of visible light is only true for yellow-green light of wavelength around 555-556 nanometers, where this figure is maximized. Those saying that incandescents are only around 2% efficient are assuming that a watt of any kind of visible light has 683 lumens.

Assuming my 252 lumens per watt figure for the visible portion of 2865K blackbody radiation is true for a 1710 lumen 100 watt lightbulb, that means a 1710 lumen 100 watt lightbulb is about 6.8% efficient at converting electrical power to visible light (400-700 nm). The truth won't be far from this.

- Don Klipstein ( snipped-for-privacy@misty.com)

It's really very orange in the evening, at least around here. The 5000K fluorescent in my kitchen looks downright blue compared to evening sun.

Incandescent lamps that most people are accustomed to are very orange, around 2700K.

snipped-for-privacy@manx.misty.com (Don Klipstein) wrote in news: snipped-for-privacy@manx.misty.com:

This makes sense, in a way, though the actual assumption is surely a misinterpretation. In the context of lasers it makes sense now, because those are usually monochromatic (or take pumping on narrow bands of lines), and the maximum efficiency of any 'line' drawn from that lamp will be around 2% at best. Discussions of efficiency for narrow bands or lines in lasers or LED's or phosphor or sodium sources dominate a lot of reference material, so that's probably why this figure arises so often.

Even so, it's harder to see how that hasn't been corrected in something like Wikipedia by now. I guess a lot of people don't think of light below

A lot of the heat energy is carried to the bulb by convection and emitted as IR, so the temperature will be lower than than if the filament was heated in vaccuum. It's not an ideal blackbody radiator. That could make a likely average fall well below 6%, especially if you consider that the world has a lot of 240V lamps too. The steepness of that curve alone is enough to make large changes in output of visible lumens with small changes in voltage.

In short, I guess that the figure of 2.6% and others similar might not have been gained by calculation at all, but by measurement. I don't know what the conditions for that were though, so I can't comment on them.

I have been happier with the color rendering of CFLs than that of most white LEDs. I have found most white LEDs to make reds and greens appear duller. I have also seen the color-dulling effect of most fluorescents with color rendering index outside the range of 82-86 (though high is better than low). The key here appears to be ratio of yellow content to red and green content.

- Don Klipstein ( snipped-for-privacy@misty.com)

Wikipedia articles are written by anyone who wants to write them, and with a few exceptions can be edited by anyone who wants to go in and edit them - anonymously even, without even creating an account and signing in.

Tungsten deviates from blackbody largely by some supression of infrared, enough to slightly outweigh heat conduction and convection by the fill gas in some incandescents.

I think 240V 1000 hour 100 watt would be more like about 5.5%.

Yes, quite true. But at full voltage most 120V incandescents 60 watts or more are about 4.5-8% efficient at converting electricity to radiation in the 400-700 nm range.

- Don Klipstein ( snipped-for-privacy@misty.com)

snipped-for-privacy@manx.misty.com (Don Klipstein) wrote in news: snipped-for-privacy@manx.misty.com:

Of course. It's also more likely that such an item would be corrected than vandalised. If we were wanting a verdict on the mental health of Micheal Jackson the laso place I'd expect to find a sensible judgement would be Wikipedia but for technical things that don't have any salacious or celebrity entertainment value it's fairly good, or at least likely to improve over time.

It's not the only source I rely on either. It's actually slow to load and not the first I turn to, most times.

snipped-for-privacy@manx.misty.com (Don Klipstein) wrote in news: snipped-for-privacy@manx.misty.com:

Is this by calculation or measurement? I was hoping you or someone else would comment on that from experience with it. Those lower figures are so common that they must have come from somewhere, and not all from looking only at the strongest wavelength or omitting something in calculation. I've never seen claims of 6 to 7% for a 100W lightbulb before, and I'm sure I would if measurements routinely reported it.

snipped-for-privacy@manx.misty.com (Don Klipstein) wrote in news: snipped-for-privacy@manx.misty.com:

There is a sharp dip in green in the LED's. I agree it's not nice, nor is the muted red. It's still a continuum though, and if improved will be extremely easy to live with. Right now it's almost discontinous because the dip is so great, but the discontinuity in CFL's seems very un-natural to me, I notice it especially if I'm carrying something brightly coloured between rooms. There's an almost disturbing 'filtration' effect on the various colours. Some of my CFL's are way too heavy in red too. It's not a pleasant colour, it's sickly, it makes normally innocous marks like pine resin under pain on wood look like inflamation on diseased skin.

I don't think the key is any specific colour absence or presnce, but the presence of any sharp absence. CFL discontinuities drop out like digital bits, but those of LED's are like analog curves, and inherently easier to compensate for.

snipped-for-privacy@manx.misty.com (Don Klipstein) wrote in news: snipped-for-privacy@manx.misty.com:

Watts of emitted light? I just saw a later post of yours that mentioned "lumens per visible radiated watt". I think that's why we're discussing such different values. I'm talking about input watts. I thought we all were, at least Eeyore certainly was, as that's ultimately watt (haha) is consumed no matter watt is emitted.

Cree themselves don't claim anything like 250-300 l/W for input watts, at least not yet, though that might not be long awaiting.

So how does a 100W incandescent look in that context?

They do lots of photometric measurements, while radiometric figures for incandescents appear to me to be rather rare. More common than actual measurements I see comments in the direction of "close enough to blackbody".

Osram says 5% here:

- Don Klipstein ( snipped-for-privacy@misty.com)

I am saying that a watt of white light is about 250 lumens, not the 683 some use as the lumen/watt figure for a 100% efficient light source. A

Most of those generating figures of incandescents being 1-2% efficient are assuming that they would achieve 683 lumens/watt if they were 100% efficient.

- Don Klipstein ( snipped-for-privacy@misty.com)