Talk:Luminous efficacy

Example: Electroluminescent devices

Another example I would find interesting is electroluminescent devices. Night light panels often draw on the order of 0.05W or less but that appears to be due to very low output rather than good efficiency. The web browsing I have done so far suggests 3lm/W or less for panels; I have been unable to find a source for EL "wire".

—The preceding unsigned comment was added by 63.193.211.225 (talk) 16:05, 15 December 2006 (UTC).

Spelling

Is 'Luminous efficacy' correctly spelled, or is 'efficacy' some word with special meaning? Electron9 09:13, 11 January 2007 (UTC)

See definitions. Dicklyon 20:57, 11 January 2007 (UTC)

Merge Lighting efficiency?

Since nobody supported this idea, I'm removing the merge tag. Dicklyon 04:10, 10 July 2006 (UTC)

The merge seems like an obvious good idea to me. On the whole, Luminous efficacy is a better article (gives sources, has more details). Kingdon 20:39, 14 September 2006 (UTC)

As a reminder, the old (June 06) proposal was to merge Lighting efficiency into here; I just noticed that that the other article still has its mergeto tag. It seems like a very distinct topic to me, focusing on practical light sources as opposed to the concept. But they could be integrated if someone was motivated to work on that. Dicklyon 22:14, 14 September 2006 (UTC)

Oppose. The more I look at it the more separate the concepts seem. We should explain that. The efficacies and efficiencies in this article are purely photometry/radiometry relationships on spectra. The lighting efficiency of the other article is with respect to power consumed, not power radiated, and includes effects of energy loss via conduction and convection, if I understand correctly. Dicklyon 22:33, 14 September 2006 (UTC)

This article covers that too. The section on Overall luminous efficacy explicitly deals with efficacy and efficiency of light emission as a function of input electrical power, i.e. including the effect of losses due to conduction, convection, heat, etc.--Srleffler 00:06, 15 September 2006 (UTC)

The overlap is quite strong, and each has tabulated information that the other is lacking. The main difference between the two is mere emphasis. This article focuses on the technical definition of luminous efficacy/efficiency, and compares the standard and "overall" types in one article. The other article focuses more on the practical issue of how efficient a given lighting technology is. Both foci seem useful--one for those interested in photometry, and one for those who want to know more about light bulbs. If anything, I think we should probably split and merge the other way: move the overall efficacy/efficiency information into the other article, keeping the practical layman's terms approach in the introduction and providing a more technical explanation with links between the two articles. This article could be limited to the "true" luminous efficacy/efficiency, with some explanation of the distinction between this and the overall value and links to the other article.--Srleffler 00:20, 15 September 2006 (UTC)

That kind of split makes sense to me. The key thing would be having the two articles link to each other cleanly, and cutting down on the material which is duplicated. Kingdon 14:20, 15 September 2006 (UTC)

I don't see the need for two articles here. This article needs at lease some examples to make sense and there is is no reason to have the information in two places. I support the original merge.--agr 14:07, 21 September 2006 (UTC)

There doesn't seem to be a consensus one way or another here, so I've tried merging the pages, and we'll see if the result is acceptable. This article is not really much changed, and I think the changes are largely ones we would want to keep even in the absence of a merge. The major change is that lighting efficiency now redirects here. I think this is best, since "lighting efficiency" as defined there was really just the overall luminous efficacy, as defined here. It seems better to have one article that explains it all in context, rather than two. I'm certainly open to splitting off the "Lighting efficiency" section of the merged article, and moving that back to "lighting efficiency" with appropriate links, if that seems better to most people.--Srleffler 06:02, 25 September 2006 (UTC)

Xenon Lamp efficiency?

Am yet to see any xenon lamp rated at 150lm/W. Osram XBO line for example has a maximum of 500000lm per 10000W, that's 50lm/W. I would like to see a model rated above 60lm/W in any catalog. Most such lamps are below that, at about 40lm/W. Unless I see any new data I'll correct the article in the next few days. --Rnbc 02:38, 22 February 2007 (UTC)

Read the reference.[1] The lamp with 150 lm/W is clearly a specialty item, with 20 kW total power consumption. The largest XBO bulb is only 12 kW. The efficiencies given in the reference agree with the XBO datasheet for the 1 kW and 5 kW models. --Srleffler 03:47, 22 February 2007 (UTC)

Where can I buy that 3000000lm/20kW widget? I would like to see the catalog and technical specifications... Perhaps it's a model with the same chemistry as car's xenon headlamps, witch are not exactly "xenon", but only use xenon for a quick start effect. And they don't have the nice and continuous xenon spectra also. --Rnbc 01:51, 24 February 2007 (UTC)

No idea. I hunted online but couldn't find anything. IMAX bulbs consume 15 kW, but they are no more efficient than the 12 kW XBO bulbs.--Srleffler 02:00, 24 February 2007 (UTC)

Well, just consider this: the solar spectra itself has an efficiency of about 100lm/W (a little less), and that's as high as you can get with a blackbody-like spectra. Xenon spectra exibits some UV lines, and some very strong lines in the IR, being otherwise rather flat in the visible, even more flat than a blackbody, so it has more blue and red, lowering efficiency greatly from an ideal blackbody due to all these factors combined. I find it really amazing they can get as high as 50lm/W, let alone 150lm/W: I still think that number is bogus.

I changed the wording to de-emphasize the high value and give a more typical range. I agree with you that the 150 lm/W number is suspicious, but this is original research. The 150 lm/W number is supported by a reference. Before we change it, we should try to find a reference that explicitly gives the range of efficiencies that can be obtained with Xenon (or quotes a value for the highest obtained, etc.). The XBO catalog doesn't do it, since that only tells us what their range is; they don't assert that there aren't more efficient Xenon lamps in existence.--Srleffler 04:12, 25 February 2007 (UTC)

Ok, but let me tell you that the reference mentioning 150lm/W contains no part numbers, no model references, nothing: It's just a number. I wouldn't consider that "reliable". Also, I've looked at a few catalogs (from philips, osram, ushio, hanovia) and found nothing above 50lm/W. I found a few references for 60lm/W when in the context os Xenon-long-arc lamps, but nothing solid with part numbers and model specificaations.

Yes, it's probably just a typo—they probably meant 50 lm/W.--Srleffler 02:52, 26 February 2007 (UTC)

Tungsten lamp curve

This curve does not accurately reflect the emission of a tungsten lamp for reasons explained in the talk page for Blackbody radiation.[2] Furthermore, it contradicts the text of the article, viz., "...most of its emission is in the infrared." This graph should be removed and replaced by a more accurate one. Drphysics 18:54, 15 March 2007 (UTC)

Link to discussion--Srleffler 20:31, 15 March 2007 (UTC)

add clarifying details

This is a very important, difficult, and confusing subject. Many lighting-related articles are involved, and share tables as templates. The "Lighting efficiency" table needs more columns, so that those of us trying to make sense of this have some hope. The various sources listed should be characterized with deg-K range etc. The text says that "True luminous efficacy is a property of the radiation emitted by a source" and then leaves us hanging. Please add a column to the table that gives this actual "true luminous efficacy" for each source. Please add the total W radiated by each source, and how much (% or W) of that is in the visible spectrum.

Electric power goes into an incandescent bulb. Some of the energy is radiated. Some of that is in the visible spectrum. Some of that the human eye sees well, some barely at all. That is three levels of nested loss. A fairly simple and important concept. But not clear in the WP lighting articles, and the relevant numbers are not given, only some sorts of "final" numbers that confuse because it is not clear what/which factors they are including how.

The table should probably also list a few monochromatic color sources to help with understanding the color-weighting aspects.

Without an extremely clear exposition of these matters, we cannot understand the "efficiency" aspects of incandescent lighting, nor understand comparisons with other types of lighting.

Also for the fluorescent tubes, we need a note clarifying whether these numbers are for the tubes themselves, or for the whole tube and ballast system as a whole. (And I assume these numbers are not for typical lighting devices installed in typical luminares, which should also be at least mentioned in a footnote, mentioning typical additional losses - these "overall" numbers are not truly "overall" from an applications perspective.) -69.87.201.16 11:01, 25 May 2007 (UTC)

Add theoretical maximum efficacy for white light?

I notice the theoretical maximum efficiency for lumens per watt is 683. However, this is only for a single monochromatic wavelength of green. Would it not be feasible to add another entry to give the theoretical maximum of white light to around 227.67 (which is basically 683/3)? --Skytopia 06:36, 4 October 2007 (UTC)

Where do you get that number from? It's not clear to me that there even is a theoretical maximum of white light, since it would depend on what your definition of "white" is. In any event, a citation supporting the new information would be required.--Srleffler 17:26, 4 October 2007 (UTC)

Note also that, as mentioned in the article, an "ideal" white light source with a perfectly flat spectrum has a luminous efficacy of 242.5 lm/W, which is higher than your supposed maximum.--Srleffler 13:19, 7 October 2007 (UTC)

Is there any actual WP:RS for this number, or for this concept of an ideal white light source? Does it mean equal energy per wavelength from 400 to 700 nm? Or what? I can find no source that says. Also, what about the maximum luminous efficiency of an ideal blackbody? There must be some temperature at which efficiency is maximized, and the efficiency at that point is well defined and probably well known. Do you know? Looks like maybe near 14.5% at 7000K, but do we have an actual source and actual numbers? Here is a book that says it peaks at 6500K, but doesn't give the efficiency number. Is it possible that the "ideal white light" being discussed is in fact a blockbody, not a flat spectrum? Dicklyon (talk) 01:27, 6 January 2008 (UTC)

I suspect it's equal power per unit frequency in that range, actually. In audio and electronics "white" noise is noise with equal power per unit frequency. Note that the article already gives 95 lm/W at 6300°C as the maximum efficacy for a blackbody, and it cites a source.--Srleffler (talk) 18:23, 6 January 2008 (UTC)

Ah, I see it now. It says 6300 C or 6600 K, which agrees with my calculation. Why does it say "[sic]" in the quote? I don't think you're right on the equal energy per frequency; I've never seen such a thing in talk of optical spectra, but equal per wavelength is common and gives very nearly the answer quoted; still need a source and a better number for that one, though. Dicklyon (talk) 20:53, 6 January 2008 (UTC)

I added the [sic] because he uses "efficiency" to describe what should, in the context of this article, be "efficacy".--Srleffler (talk) 01:28, 7 January 2008 (UTC)

Seems like it implies he said something wrong, when it's really just an alternative usage of a correct term. OK if I take it out? Dicklyon (talk) 01:49, 7 January 2008 (UTC)

No. Given the definitions in this article, his usage is wrong. It would potentially be confusing to have "efficiency" used to mean lm/W without marking it. An alternative might be to do an editorial correction in the quote: "...and the theoretical luminous [efficacy] is 95 lumens per watt."--Srleffler (talk) 05:04, 7 January 2008 (UTC)

I understand it's wrong [sic] per the definitions in the article. But that's not his fault; what he said was correct and consistent, and the "sic" makes it look like we think he said something wrong. It's pretty common to use it that way. Dicklyon (talk) 05:10, 7 January 2008 (UTC)

I just calculated the blackbody max. Based on 1 nm data and simple sums, using 1931 y(lambda) and a matlab blackbody code snippet that I found online, the max luminous efficiency is 13.97% (95.4 lm/W), attained for temps from 6500 to 6700 K (that is, it varies less than 0.01% across that temp range, with a peak near 6630). And I get 243.13 for the mean y(lambda) from 400 to 700 nm, so you're right that the 242.5 number probably does come from a flat spectrum over that range. It would sure be nice to have a source for these things, and maybe better numerics for more definitive numbers. Dicklyon (talk) 03:40, 6 January 2008 (UTC)

Here is a source that says 93 is max, at 6300 K. And here is one that shows the peak at 6600K but doesn't give the exact value, just a plot. And this one says about 6800 K. Dicklyon (talk) 04:08, 6 January 2008 (UTC)

You're right, 242.5 is mentioned higher up the article. I hadn't seen that previously. Do you think it would be a good idea to list it in the lower table as well? I know it's a slight duplicate of information, but it's nice to compare the theoretical maximum against the other white-ish lights in that table. Also, the information 683 lm/W is duplicated as well.

Also in regards to the point that a more efficient white light source can be made more efficient with say... 3 wavelengths, instead of a multitude of wavelengths, I'm not sure that's the case. If it's true, then they've got to be very close in efficiency anyway (do you know the 3 wavelength stat by any chance?)

Finally, I thought the link to Luminosity_function link was appropriate along with the other three. In this kind of article, confusion can stem from the finer points of what lumens actually are, so it's good to include it in there.--Skytopia (talk) 18:06, 29 March 2008 (UTC)

I took it out not just because it's a duplicate, but also because it is misleading. It's not really correct to talk about a "theoretical maximum" efficacy for white light, since "white" is not a well-defined thing. The value we are tossing around (242.5 lm/W) is the luminous efficacy of a perfectly flat spectrum through the visible, with no light outside the visible spectrum. This is "ideal" in the mathematical sense: an idealized or optimally simple form of "white" light. It's not necessarily the best white light source possible. One could make a lamp that produces a nice smooth continuous spectrum that is humped in the middle; something like the luminosity function, but perhaps with not as big a ratio of green to red or blue. Such a source could have an efficacy greater than 242.5 lm/W, and would appear nice and white to an observer.

I removed the extra link to luminosity function, since it already appears in the article. The "see also" section is for linking to relevant pages that are not already linked.--Srleffler (talk) 18:21, 29 March 2008 (UTC)

I think the main reason I would still like the (admittedly somewhat engineered) value for maximum white light efficiency in the main table as well is so that one can see how much more technology can progress before 'white light' efficiency maxes out. You're right that a slight hump towards green will still be considered by the average observer to be more or less 'whitish', but let's assume that 'pure white' (whatever that is) is actually a goal. For that goal, I think the definition of a flat line throughout the visible spectrum fulfils that quite well.

If a light is tinted slightly red, green or blue, it can be difficult for people to see that. But that doesn't make it any less tinted; it's just that the eye has temporarily 'adapted' to that new level as white. However, the illusion is lost more and more as you keep adding more saturation to the light, which is why the best value for perceptive white light would be in the 'middle' of the illusion area (around 6000-6500K). Therefore, this makes perceptive white somewhat an absolute value, as well as just a mere relative one. --Skytopia (talk) 12:53, 16 April 2008 (UTC)

There is nothing special about the flat spectrum. It is neither the maximum in efficiency nor a good definition of "white". It is completely unnatural. Remember that our eyes are designed to see primarily with a ~6500 K blackbody illumination. That strongly-humped spectrum is probably the best definition of what "white" is. A blackbody spectrum truncated to cover only the visible range would probably have a higher luminous efficacy than the flat spectrum, too. The flat spectrum is by no means an indicator of the maximum luminous efficacy that can be achieved by a "white" source.

I think you seriously underestimate the human brain's ability to adapt to different ambient lighting. Our vision system is quite adept at "retuning" to perceive ambient illumination as white, even when the same light would be perceived as quite strongly coloured when compared to more typical illumination. This is not a trick or an illusion—it is an essential feature of how the human vision system works. White is relative.--Srleffler (talk) 04:39, 17 April 2008 (UTC)

I agree with Srleffler, that white is not any particular color. However, once you pick a chromaticity for white, say D50 or D65, or whatever you want, then you have a well-defined optimization problem to find the maximum lumens per watt. It would be interesting to work out. I wonder if that's been done. If not, such original research has no place in the article. Dicklyon (talk) 05:07, 17 April 2008 (UTC)

I study and create optical illusions, so I am well aware of how colours (including white) can fool the visual system. However, if the sensation of '100% white' was relative only, then why is it easy for anyone to tell the difference between illumination of a ~6500K light bulb compared to that of the more orangey 2700K illumination provided by incandescent bulbs? I can be in a room with either illumination for hours, and while my eyes can adapt to each, it only takes a moment's thought to tell which illumination is closer to actual white (whatever that is). Therefore, closer values to the 6000-6500K ballpark would also be distinguishable from 6500K to lesser degrees, even after acclimatization. --Skytopia (talk) 08:07, 17 April 2008 (UTC)

Should the "ideal white light" entry on the first table be changed to something like "flat spectrum 400nm-700nm" or "constant power 400nm-700nm" then? I found the entry confusing without a definition of what "ideal white" is.Totsugeki (talk) 21:01, 16 October 2008 (UTC)

I agree, but we need a source that is clearer about how this "ideal" light is defined. The source cited in the article doesn't say. Looking around for this number on the net, I find lots of people who have obtained it from Wikipedia, many of whom are confused about what it means (e.g. assuming that it is a theoretical maximum for "white" light, which it is not.) I'm going to delete it from the article until a better source can be found.--Srleffler (talk) 23:03, 16 October 2008 (UTC)

Type	Luminous efficacy (lm/W)	Luminous efficiency
ideal white light source	242.5	35.5%

Luminous efficacy of display screens

I would find it interesting to see figures on luminous efficacy and efficiency of different display technologies. As I understand some of the screens produce light pretty effectively. --Khokkanen 11:14, 7 October 2007 (UTC)

Comparison of tungsten lamps needed

100Watt tungsten lamps are more efficient than 60Watt lamps. Ordinary "pear shaped" bulbs are more efficient than "candle" bulbs. Quartz is better than glass. 110V lamps are better than 240Volt. Presumably, 12Volt-halogen are better than mains powered halogen. Is this simply to do with the filament temperature, and that a higher-resistance filament must be thinner, therefore must run cooler lest it burn out rapidly? —Preceding unsigned comment added by 81.187.40.226 (talk) 05:31, 14 November 2007 (UTC)

Some extra data points (information from the packaging of the light bulbs), all referring to 220/240V bulbs:

• A 60 Watt "16x life" (Greenstock) bulb emits 320 lumens, whereas a regular 60 Watt bulb emits 700 lumens, and a HalogenA bulb emits 840 lumens.
• A 100 Watt GLS "triple-life plus" bulb emits 1120 lumens, whereas a regular "single life" (1000 hour) bulb emits 1330 lumens.
• A 25 Watt Greenstock 8x candle bulb emits 160 lumens, whereas a 28 Watt Osram "Halogen Energy Saver" candle bulb emits 340 lumens. —Preceding unsigned comment added by 87.194.171.29 (talk) 00:26, 6 January 2008 (UTC)

Reference no longer exists!

http://www.ts-audio.biz/tsshop/WGS/411/PRD/LFH0324408/Osram_6406330_500mA_52V_E10_BLK1_MINIWATT-Halogen-Gluehlampe_f.Taschenl..htm points to a page which no longer exists. This is currently reference #12 (#11, in German, is dead, too). That is all.65.183.135.231 (talk) 16:47, 2 June 2008 (UTC)

Thanks. I have marked the links as dead.--Srleffler (talk) 17:32, 2 June 2008 (UTC)

some numbers for non-white LED's

non-white LED's work in a different way from white ones (they don't have the coating on top of the LED), and all the figures for efficacy are available on the net. It would be good if someone could source some numbers and add them to this article. I may if I have some spare time. 192.102.214.6 (talk) 10:25, 19 June 2008 (UTC)

Sounds like a good idea, but please limit it to just a few entries—perhaps red, green, and blue.--Srleffler (talk) 11:29, 19 June 2008 (UTC)

I'd love to know the "wall-plug efficiency" of monochromatic light sources such as LEDs and lasers (especially green ones at ~555nm).Totsugeki (talk) 20:11, 16 October 2008 (UTC)

Me too, but I think we should just limit the article to a couple types of LED that are intended for use for illumination or indicator lights.--Srleffler (talk) 22:36, 16 October 2008 (UTC)

Values in examples table

Regarding the table listed in the Examples section

I disagree with a lot of the information in the table above. I'm not deleting it since someone put a lot of time intor creating it and some of the info may be useful to others. Efficacy would be calculated based on lamp lumens (either initial or average) / wattage of ballast when paired with the lamp * ballast factor. Example for a 32W T8 fluorescent lamp with a 2-lamp electronic instant start ballast. 2714 lumens (mean) per lamp, 2 lamps, 59W for ballast, 0.90 ballast factor. 2714 lum * 2 lamps / 59W * .9 BF = 82.8 lumens/Watt. Initial lumen efficacy would be about 90 lumens/Watt. While the table above claims 60 lumens/Watt) — Preceding unsigned comment added by 204.177.188.65 (talk • contribs)

This may need to be looked into. I assume the "ballast factor" you're referring to is a power factor. The efficacy is the flux in lumens divided by the actual power consumed, in watts. It's possible some of our sources have reported volt-ampere values incorrectly as "watts" (failing to allow for the power factor of the ballast), but the error is in using watts for something that is not actual power. Do ballast manufacturers report volt-amp ratings for ballasts in "watts"?--Srleffler (talk) 18:17, 2 July 2008 (UTC)

I notice that, using your example above, if a 32 W lamp emits 2714 lm, the efficacy is 2714 lm / 32 W = 85 lm/W, which is not far from the value you give. This may neglect losses in the ballast however.--Srleffler (talk) 18:24, 2 July 2008 (UTC)

I removed the following entries from the table, as uncited:

Category	Type	Overall luminous efficacy (lm/W)	Overall luminous efficiency
	34 W fluorescent tube (T12)	50	7%
	32 W fluorescent tube (T8)	60	9%
	36 W fluorescent tube (T8)	up to 93	up to 14%
	28 W fluorescent tube (T5)	104	15.2%

Looking around, I found a US government publication that gives some recommended efficiency specs for lighting. From their examples, it appears that not only is there some electrical loss in the ballast, but the actual light output of the lamps as driven is lower than their rated values. Crunching their numbers gives "real" values of 58 lm/W for "base" 34 W T12 lighting with magnetic ballasts, 81 lm/W for a "recommended" 32 W T8 system with electronic ballasts, and 92 lm/W for a "best available" 32 W T8 system. I updated the article with this info. I left it generic regarding the length and rated power of the tubes, just giving a range for T8 tubes with electronic ballasts.--Srleffler (talk) 04:59, 4 July 2008 (UTC)

No definition of Lumen or lm is provided

It would help if this article included a definition of Lumen and that lm is its abbreviation. A link to the detailed article about them would be good, too. http://en.wikipedia.org/wiki/Lumen_%28unit%29 —Preceding unsigned comment added by Seldenball (talk • contribs) 12:01, 6 August 2008 (UTC)

Lumen is linked the first time the term is used. Same with Luminous flux. The "Explanation" section explains the concepts involved, and defines lm/W as an abbreviation for lumens per watt.--Srleffler (talk) 04:48, 7 August 2008 (UTC)

Image from German Wikipedia

The German page has this image which I think would make a useful addition to the article. It shows the luminous efficiency of a blackbody radiator as a function of its temperature (I think). Unfortunately the captions on the image are in German and there is no source data for the image so that I could recreate it. I replaced the captions with English equivalents, but I won't add it to the article in case I misunderstood the German version.

Totsugeki (talk) 20:40, 16 October 2008 (UTC)

This is great. The axes need units, however: % for the y-axis and kelvins (K) for the x-axis. --Srleffler (talk) 22:58, 16 October 2008 (UTC)

Updated it. Totsugeki (talk) 00:32, 17 October 2008 (UTC)

This is a great addition...but are we sure the units are % on the vertical, and is that defined the same as in the article, that is relative to a green line? —Preceding unsigned comment added by Ccrrccrr (talk • contribs) 07:30, 14 November 2008 (UTC)

Definitions: overall and per radiant watt

There are two definitions of luminous efficacy that are commonly applied: lumens per radiant watt, and lumens per input power to the light source (overall luminous efficacy). This article treats the first as being the primary, correct usage of the term, and denigrates the other as being confusing and implies that it might be incorrect to call it "luminous efficacy".

I think that this is a biased point of view. The term "luminous efficacy" without any qualifiers is widely used to mean what the article calls "overall luminous efficacy". Regardless of what we as editors think should be the correct usage, we don't get to decide what is correct.

I agree with the article that the conflict between the two usages is a problem. However, I have more often seen people confused in the other direction--for example, people from the lighting field reading a lm/W value on an LED datasheet and thinking it means overall luminous efficacy when in fact it is being used for lm(radiant watt), which we might call spectral luminous efficacy. Making up our own rule about what is correct and what is not will not help the situation, and it's against WP policy. Unless there are authoritative sources that back up one as being correct and the other being incorrect (and there aren't authoritative sources with the opposite POV), we need to represent the real state of the world here, not how we wish it to be.

I propose that the article should be re-written to mention both definitions in the lead, to put them on equal footing, and to warn against the potential confusion. I think that it is easy to find authoritative sources that use the terms both ways.Ccrrccrr (talk) 20:45, 14 December 2008 (UTC)

ps: two more minor notes. First, "overall" isn't necessarily the right term. In ligthing, people distinguish between lamp efficacy (lamp lumen output/lamp electric power input) and system efficacy (fixture lumen output/ballast electrical input); "overall" might imply the latter. Also, I think that using "lighting efficiency" for any kind of lumens/watt is a worse choice of words--efficiency should be unitless. For example, radiant watts per input watt, or lumens out of of fixture per lumens out of a lamp for fixture efficiency.Ccrrccrr (talk) 20:45, 14 December 2008 (UTC)

I'm fine with the rewrite you propose. Your post reminded me of the LED issue, of which I had become aware only recently. Someone needs to review the LED entries in the "overall efficacy" table, and review whether the citations support inclusion of the cited numbers in the "overall efficacy" table. I suspect that the efficacies for the "prototype" LEDs listed are actually lumens per radiant watt rather than "overall" efficacy, as defined here.--Srleffler (talk) 04:30, 16 December 2008 (UTC)

I'm OK with the proposed change, provided that reliable sources are cited in support of both definitions. Alternatively, if had reliable sources with sufficient weight showing that one use is preferred, we could go with that. I agree it's basically not up to us, but up to sources. Dicklyon (talk) 05:11, 16 December 2008 (UTC)

I found a good source that even provides that official international standard terminology, so I edited the whole article to explain and use that terminology. Additional editing might be called for in light of the information in that article (no pun intended). If nothing else, it has a great table of some values of luminous efficacy of radiation that we could use to expand our table of those.Ccrrccrr (talk) 16:09, 26 December 2008 (UTC)

Srleffler, thanks for your copy edits. They are quite helpful. Two minior changes that I have doubts about: First, shouldn't we include the information about the international standard terminology? Maybe that goes in the section with that table of units? But I thought it was a nice fit with the discussion of possible confusion. I don't like the way you rephrased that discussion. Second, the heading on the table for luminous efficacy of a source: I know it sounds a little funny there, but I thought it was best to stick with standard terminology, given that the potential for confusion is strongly in evidence.Ccrrccrr (talk) 18:21, 26 December 2008 (UTC)

CFL efficacy and efficiency

I have just added a paragraph to the CFL article based on some of the data in this article. The bit that worries me is this use of that figure of 683 lm/W, the 'efficacy of an ideal light source'. You guys seem to reference that to the Luminosity function article, which in turn cites it to the Commission Internationale de l'Éclairage... At that point, what I remember of A-Level and University physics is getting spread too thin... Without being too picky, we only need a bottom line or a headline figure and this article is referenced as the main source of details, is what I wrote for CFLs acceptably accurate for a lay readership? (This link may be useful if what I wrote is already hacked away) --Nigelj (talk) 16:40, 24 December 2008 (UTC)

Seems generally OK. I tweaked the wording a bit. One sometimes sees a different "efficiency" figure used, which is the fraction of the energy that is emitted as photons within the visible spectrum. Your wording suggested that this was the definition being used. The efficiency numbers, defined that way, are higher. The definition used here takes into account the fact that the eye is not equally sensitive to all wavelengths of visible light. Red light is less useful for illumination than green light, because the eye is much less sensitive to red. A source with more green will illuminate a room more efficiently than a source with less green and more red and blue. The 683 lm/W is the physical upper limit. No source can produce a higher efficacy than that. The number corresponds to the efficacy you get when all of the light is emitted at exactly the wavelength to which the eye is most sensitive, and no energy is lost or wasted.--Srleffler (talk) 04:15, 25 December 2008 (UTC)

Thanks for that, Srleffler. You're right, I think what the average, technically literate, but non-specialist, reader would like to see - both here in the tables and in places like the CFL article - are some figures for "the fraction of the energy that is emitted as photons within the visible spectrum". Should we use a lower constant - there are a few quoted in the 250 lm/W range in places?

I think it is not actually possible to get to the holy grail figure: to say, "this lamp converts x% of incoming electrical energy into photons of visible light" regardless of whether they are red, green, blue etc. A lot of energy may be put out as almost-invisible near-UV by some lamps. So, we would need a curve representing the average human eye's response to each part of the spectrum (i.e. not necessarily a black-body curve). But then, CFL and LED lamp manufacturers at best give us an overall efficacy figure with no spectral curve to work from. So we can't proceed that way.

If we can pick a constant figure, either 683 or something lower, I think it should be used consistently in several articles including for the tables in this one. As it says above, other sources on the web are now starting to pick up such headline 'facts' from Wikipedia now, so it really is worth putting some effort in to get to the bottom of this, along with the right descriptive words to make it clear what we are saying and what we're not in every case. --Nigelj (talk) 10:01, 26 December 2008 (UTC)

We can't just pick a number out of the air. It has to be correct, and supported by some source. In fact, there is no such number. There is no way to convert an efficacy in lumens per watt into a "percentage of energy emitted in the visible spectrum". You have to know the actual emission spectrum of the light source to get the latter. I would be fine with quoting figures for efficiency defined the latter way, but the figure for each type of lamp has to be supported by an outside source.--Srleffler (talk) 14:10, 26 December 2008 (UTC)

Srleffler is quite correct. We don't get to pick a number and use it consistently. That's original research. So I did the necessary unoriginal research, and put in correct numbers with an authoritative reference. Which also turns out the to be an excellent ref. for this article... Hopefully what I wrote is understandable--if not please edit and/or complain here.Ccrrccrr (talk) 15:21, 26 December 2008 (UTC)

There's still a problem. The two factors Ohno talks about are the LER, and the conversion efficiency from electricity to EM radiation. Not all EM radiation is in the visible spectrum. A figure for efficiency of conversion from electricity to visible photons would still be useful, but you have not provided that. I don't think the conversion efficiency to all EM radiation is sufficiently useful to mention in the CFL article. I have rephrased to give efficiencies as defined here (% of 683 lm/W).--Srleffler (talk) 17:25, 26 December 2008 (UTC)

Thanks for the prompt attention to this. Actually, the numbers I provided are based on considering only the visible spectrum, not total radiation (see the "illuminant A" entry in the table in the reference--it says 400 nm to 700 nm only, and the 248 lm/W is consistent with that). So changing the description of the numbers without changing the numbers themselves doesn't work. So I changed some of the description back but left your other editing in place (thanks for your attention to the details).

I'm not really comfortable with including "overall efficiency" numbers there without a discussion of what they mean. Problems: 1) They are OR: both the numbers and the whole concept--we'd need a source for the concept. 2) They are relative to ideal monochromatic green light. I don't think that's a relevant comparison for considering residential illumination; and I don't think it's fair to reader to shift the comparison to green light without any mention of that. I didn't delete that yet, but might unless it gets more support here.Ccrrccrr (talk) 18:09, 26 December 2008 (UTC)

Perhaps we can remove the overall efficiency numbers or do something different, but your last edit there introduced what seem to me to be factually incorrect definitions. I believe you are misinterpreting the reference. In splitting up the luminous efficacy into radiant efficiency and luminous efficacy of radiation, Ohno makes the radiant efficiency the conversion efficiency to all EM radiation. It cannot be otherwise, because the luminous efficacy of radiation is defined by integrating over the whole EM spectrum. This is also the standard definition of radiant efficiency. I don't see anything in the paper that contradicts this interpretation or implies that he is using nonstandard definitions. The distinction is important because most light sources emit some radiation outside the visible spectrum. Incandescent lights emit lots of infrared, and possibly some ultraviolet. CFL's emit some UV. The radiant efficiency is the efficiency of conversion of electrical energy into radiation. The LER then determines how efficiently that radiation is detected by our eyes, taking into account both the variation in sensitivity of the eyes over the visible spectrum, and the falloff of sensitivity outside the visible range. Note that the variation in sensitivity over the visible range is not distinct from the falloff outside; there is no sharp cutoff.--Srleffler (talk) 21:10, 26 December 2008 (UTC)

I absolutely agree with all of your statements about the reasons this is messy. I think that is why Ohno chose to truncate. If you read p. 95, you will surely agree that the LERs in there are for truncated spectra. Although Ohno does not include limits on the integration, I agree with your interpretation that from an idealized physicist point of view, they should be 0 to infinity. However, for reasons explained below (next section), I don't think there's any way to take that 100% seriously and get to two separate numbers for radiant efficiency and luminous efficacy of radiation. I think that what people are interested in in practice generally corresponds to integrating from 400 to 700, as Ohno actually did. If you have a proposal for a way to do it with integrating from zero to infinity (or at least 20 um to 300 nm), that could be more elegant, but I don't think it's of any engineering importance; nor do I think it's easy to do.Ccrrccrr (talk) 23:42, 26 December 2008 (UTC)

Measuring from 300 nm to 20 µm and perhaps much beyond is a pain, but it isn't impossible. Such measurements are done through calorimetry: the light is absorbed by a surface with uniform (ideally high) absorption over a broad spectral range, and you measure the temperature change in the absorber. One doesn't need to do this, of course, to get the luminous efficiency of a source, which is the figure most commonly quoted. You only need to do it to get the denominator for a LER measurement (since the numerator is not affected by radiation much outside the visible range). In measuring LER, one would certainly not truncate the spectrum to the visible range. This would be terribly inaccurate, and it would be misleading the call the result "luminous efficacy". Doing so would not comply with international standards. It's not clear how Ohno's LER values for incandescent and fluorescent sources are obtained. He was working with tabulated PSD data for some example sources. He may well have simply integrated over the available data to get an "LER". Without information on what range his data happened to cover, though, it seems inappropriate to presume that it only covered the visible range. It wouldn't be unusual to have data extending into the IR and UV for these sources, especially if one plans to calculate LER.

I've taken another look at pages 94–5 in Ohno's paper. He is not truncating a spectrum for purposes of analysis, as you propose. Rather, he is modeling a source (an idealized LED), where one can in principle engineer the output spectrum to be anything one desires. Obviously, designing the device such that its spectrum includes no emission outside the visible band is desirable, since this increases the efficiency. Since the idealized design doesn't include any power outside the visible band, there is no need to consider radiation outside the visible band in the analysis.--Srleffler (talk) 07:14, 27 December 2008 (UTC)

I agree about what Ohno says and does, and I agree about how one could make the measurement. So rather than continuing to have expend effort explaining on the talk page explaining to me stuff I already know, let's figure out how to work together to make the two pages, CFL and this one, explain this stuff to people who don't yet know about it. The issue we want to get at here is what can we say in the CFL article that is helpful to people understanding the efficiency of CFLs and comparing them to incandescents. The fact that one could measure true LER doesn't get us very far if we don't have that data. So here are our choices: 1) include the information there now, with explanations somewhere about LER with integration over limited bands... 2)Provide luminous efficacy of a source, and don't discuss efficiency at all, 3) Find data on broadband LER or efficiency to provide the data in the form that you prefer. Or perhaps there are some other options or compromises between those? Perhaps only discuss CFLs where the "optical" and "thermal" bands are well separated, such that an "optical" LER is more clear cut, and doesn't involve an arbitrary cutoff (e.g. 700 nm) upon which the results are highly sensitive? What do you think we should do?

In terms of making the CFL article useful to people for understanding the technology, it would be great to have a breakdown of 100 units of energy, going in the ballast, 90 units going into the lamp, 40 coming out of the arc as UV photons, and then 20 coming out of the phosphor as visible photons. Those are made-up numbers so we can't but that in, but to me, that means that the engineered system is 20% efficient--20% of the energy going in comes out as the final product that we want. Then the next issue is the LER of the spectrum emitted by the phosphor. Inserting thermal radiation back into the equation at that point doesn't really help understanding the CFL technology is a systematic way.Ccrrccrr (talk) 15:17, 27 December 2008 (UTC)

Another option would be to drop LER and talk only about the overall luminous efficacy (l.e. of the source), since that is the quantity for which lots of data is available and the interpretation is clear. LER data are much more limited, and we have this question of interpretation, regarding how the values were obtained. I think we should remove the radiant efficiency values from the article, since their interpretation is unclear. The current explanation (unless it's changed since yesterday) is misleading, since it implies that this is the efficiency of production of visible radiation. It would be good to separate ballast loss from lamp loss and have data for both. I agree that if we use the luminous efficiency we should explain it.--Srleffler (talk) 17:51, 27 December 2008 (UTC)

I don't think the "overall efficiency" used here is OR. I'm pretty sure we can find sources that use this definition of efficiency. I don't know of another way to obtain an efficiency figure from the luminous efficacy of a source. Your objection to referencing the efficiency to monochromatic green light is strange. The efficiency is referenced to the maximum efficiency possible. If you want to know how well you can see by any given source (independent of colour rendering), what you want to know is how many lumens of light that source produces. The luminous efficacy is the efficacy with which the source produces lumens of light. No source can be more efficaceous than an ideal monochromatic 555 nm source. You might not want to illuminate your house with it, but that isn't because it's not efficient, it's because we also care about colour rendering. As Ohno notes, there is a tradeoff between efficiency and colour rendering.--Srleffler (talk) 21:10, 26 December 2008 (UTC)

Great, it would be useful to have those sources referenced. Again, I agree with all of your facts. I would be OK with including it with an explanation of what it is, and a reference documenting that it is not OR.Ccrrccrr (talk) 23:42, 26 December 2008 (UTC)

I'll look into it when I have a chance.--Srleffler (talk) 07:14, 27 December 2008 (UTC)

Stuff to do

Notes for myself or for anyone else who cares to work in this on things to improve:

• Explain the possibility of quoting LER with total radiation in the denominator, or only based on the visible range. This is the root of confusion that just arose in the CFL article, so it probably needs discussion here. e.g., 248 lm/W for "illuminant A" in the visible range, vs. about 17 lm/W for a 3000 K blackbody including all radiation.
• Explain the different between lamp and system efficacies (in particular, including the ballast losses or only the lamp).

Ccrrccrr (talk) 18:27, 26 December 2008 (UTC)

Your first point above makes no sense. You can't simply cut the spectrum off at the edges of the "visible range". There is not really any sharp edge. The eye's response falls off gradually with wavelength, following the luminosity function. We obtain efficacies and efficiencies by integrating the spectrum the source emits (all of it, visible or not) times the luminosity function, over wavelength. Someone who cuts off radiation outside the visible range before calculating LER is simply making an error.--Srleffler (talk) 21:15, 26 December 2008 (UTC)

If that is the case, then my reference, which is a proper peer reviewed reference "makes no sense" and "is simply making an error". It includes an LER number for illuminant A (see standard illuminant) truncated to 400 to 700 nm. You are right that truncation to 400 to 700 nm is arbitrary because the eye doesn't in fact cut off abruptly--there's a very long but very small tail. However, you'll have a hard time finding data on how much thermal radiation a CFL emits in the 7-14 micron range. Without that, there would be no way to find a value for the "proper" taking a purist physicist approach that you must include the whole spectrum. (I'm not going to insist that the EMI radiated at 10 kHz to several MHz be included, but the thermal radiation could easily larger than the radiation in the 400-700 nm band.) Likewise, although it's possible to calculate the LER of a blackbody at 2850 K, it would be very hard to find it for an actual incandescent lamp--it's affected by the IR transmission curve of the glass, and the thermal radiation from the glass itself, which is warmed both by convection from the fill gas and by radiation at wavelengths it absorbs....

For an idealized theoretical construct, it makes sense to include all radiation in LER. For a source like an LED, it can make sense to include all the radiation in the peak of the LED emission, whether all within 400 nm to 700 nm or not, and to neglect the thermal radiation. For a fluorescent, one would neglect thermal radiation; for the 365 nm and shorter wavelengths, I could see arguing that either way depending on context, although I think my reference neglects them. But I don't know how to do anything sensible with an incandescent lamp other than picking an arbitrary cutoff for what's visible if you want to talk about the efficiency rather than the efficacy. (Unless you just want to report efficacy in other units.)Ccrrccrr (talk) 23:24, 26 December 2008 (UTC)

Remember, you don't need a spectrally-resolved measurement over the whole spectrum. All you need is the total power emitted as EM radiation, and a spectrally-resolved measurement over the band where the luminosity function is non-zero. You can get the total radiated power over a pretty large band just by absorbing the emission and measuring temperature change. Things like absorption and thermal emission from the glass envelope of a bulb are not a problem either. They are just part of the performance of the bulb itself. There does have to be an arbitrary cutoff (you're not likely going to include RF emission, for example), but the cutoff one would use when doing this right would extend far beyond the visible band.

Ohno's truncated illuminant A is an interesting case. It would be wrong to say that the value given is an LER for the illuminant A spectrum. LER has a particular definition, and this does not comply. It's fine, however, to propose an idealized source that has the illuminant A spectrum in the visible band and no emission outside, and to calculate the LER of such a source. It makes sense for Ohno to do this, because he is modeling LED sources where the emission spectrum can be engineered. So, it's not an "LER" of an illuminant A source with the spectrum truncated in the calculation of LER. Rather, it's the standard LER of a source that is like the illuminant A source in the visible, but which has no emission outside. --Srleffler (talk) 07:37, 27 December 2008 (UTC)

You are right that Ohno calculates LER of a truncated spectrum, rather than calculating a "limited bandwidth LER" of illuminant A. Thus, it is not an example of the "limited bandwidth LER" that I am proposing to discuss here. And you are right that he's not specific about the band over which he models LED spectra, although if you look at the plots, they all pretty much go to zero by the edges of the 400 to 700 range plotted, so whether or not there is some tiny tail of energy modeled at 390 nm for a blue LED makes no significant difference to value of LER calculated. What would, however, make a large difference would be whether or not the far IR (~7 to 14 micron) was included. For an LED with 30% efficiency at producing radiation in or near the visible range, 70% of the input power is "lost as heat". Most of the leaves the LED chip via conduction; then it goes to a heat sink that delivers it to the ambient via some mixture of radiation and convection. It might be half and half convection and radiation, but even if it were only 10% radiation, that would make a 20% difference in the LER value, so it can't be neglected. As you explain above, it is possible to make appropriate measurements to find out the "true" LER, but I don't think anybody would do that for an LED, because 1) It doesn't help advance the engineering--from an engineering point of view, that IR is not part of the radiation you were producing with the engineered light production system--it's part of the waste heat that you are trying to get rid of, and 2) The results would depend on the particular heatsink configuration used. A manufacturer reporting an LER number on a datasheet would need all kinds of detail about the setup; Ohno would need models of heatsinks in the paper, and would need to specify what I'd call the efficiency of the LED, to determine how much exits as heat that might become thermal radiation. It's true that formally, the proper definition of LER includes all radiation, but maintaining that stance gets silly when the optical radiation is so clearly separated from the thermal radiation. And it's not just Ohno that quotes LER for LEDs based only on the optical radiation. As a random example, here's an LED datasheet. On p. 6, the 500 and 155 lm/W use the integral of the optical range data, shown on p. 7, and would be substantially smaller if they included thermal radiation as well. Ccrrccrr (talk) 14:43, 27 December 2008 (UTC)

What you've written makes sense. I think this is why LED manufacturers use LER and nobody else does. Since the emission spectrum of LEDs can be engineered to not extend much beyond the visible band, they are amenable to treatment using this truncated LER that considers only that direct emission. --Srleffler (talk) 18:01, 27 December 2008 (UTC)

new record for LED efficiency

As referred to in LED, Cree Achieves 161 Lumens per Watt from a High-Power LED. We should probably be consistent and edit Luminous efficacy accordingly.