Informal test of best replacement for 60 W incandescent bulbs

<p>Several weeks ago many of you provided very helpful information after I posted a question about the best replacement for 60 W incandescent bulbs for ordinary household use where available-light photography is a significant consideration:<br /> http://www.photo.net/casual-conversations-forum/00cHcE</p>

<p>I finally got around to running an informal test, and as promised, here's the report, for whatever it's worth. The tested bulbs were:<br /> (1) G.E. Reveal 60 W incandescent bulbs that had been in use a while (yes, I know--but I don't have new ones to test!), rated at 630 lumens<br /> (2) G.E. Reveal 53 W halogen bulbs, new, rated at 790 lumens (this is the nominal replacement for a 75 W incandescent, but the nominal replacement for the 60 W incandescent only puts out 565 lumens, and I am not accepting <em>less</em> light)<br /> (3) Cree Daylight 9 W LED bulbs, new, rated at 800 lumens (this is the version rated 5000 K; they also make a 2700 K version)</p>

<p>The short answer, suggested IIRC by Sarah Fox, was that I'll be going with the halogen bulbs for now, at least for any application where the bulb is not behind a frosted globe (which is most of my house). The LED bulbs produced what I would describe as glare; they tended to make you want to avert your eyes. For this reason they were borderline-intolerable where I tested them (and I wouldn't think the 2700 K versions would be appreciably better in this regard). As expected by the output ratings, both the halogen and the LED put out visibly more light than the incandescent; the LED seemed to have an advantage greater than its slightly higher rating suggests. As expected, the LED produced very white light, although the halogen was visibly whiter than the incandescent. Unfortunately, I forgot to use the actual light meter (I have a Sekonic L308) until after I was well into the test, so no meter data. Also, I have no convenient way to test <em>overall</em> color accuracy or continuity of output spectrum.</p>

<p>I did however run a picture test. I let each set of bulbs run for more than five minutes so hopefully they'd 'warm up' or whatever. I put a gray card directly under the fixture (three bulbs in clear glass, open-end settings on a ceiling fan) and shot it with a camera on a tripod with manual exposure (ISO 100, 1.6 s, f/8), manual focus, and auto white balance (exact same for all test shots). I then processed the raw files using both daylight white balance and auto white balance. I cropped the middle ninth (middle third of the width and middle third of the height) out of each file and cubic-scaled it down to a single pixel to check the RGB and HSV values. So this gave me a reasonably objective way to test overall output and color balance, both relative to a fixed standard (what the raw converter calls daylight) and to the camera's auto-white balance performance (for each part of the test, I took three shots in fairly quick succession and used the third, because I find auto-white balance tends to calibrate in after a shot or two).</p>

<p>Here are the photo test results.</p><div></div>

<p>Interesting test, but I'm surprised you didn't try a daylight fluorescent. I would think something above 5000K and 92+ CRI should look pretty good. </p>

<blockquote>

<p>Several weeks ago many of you provided very helpful information after I posted a question about the best replacement for 60 W incandescent bulbs for ordinary household use where available-light photography is a significant consideration:<br>

...<br>

The short answer, suggested IIRC by Sarah Fox, was that I'll be going with the halogen bulbs for now...</p>

</blockquote>

<p>Hi, I think you made the right choice, but I'm not a fan of your testing method. I appreciate that you DID get a sense of how much light of each color is delivered by each bulb by comparing under the fixed "daylight" camera setting. But that is not the real gist of what makes a lamp good for photography. (By the way if, instead of auto WB, you had done an in-camera "custom white balance," then shot jpegs, you would have gotten equal R=G=B values for each shot.)</p>

<p>A better test would probably be to custom white balance for each of the lamps, then shoot a variety of colored objects under each one. It's too complicated to explain all of the details, but if you can't see any difference between any of the lamps, then it probably doesn't matter much to you. If you see any differences, then you have to decide which one is right. Almost by definition, the halogen lamp will be, but you can use whatever you prefer.</p>

<p>If you were shooting color film, then it's pretty important to get close to the aim color balance for the film, which nowdays is almost always daylight. With the halogen lamp, you'd have to filter the camera heavily, and lose a handful of stops of light. So in the case of film, it might be better to just use the LED and see if you can live with the results.</p>

<p>With the LED, I'd expect to see color deficiencies in the zone between blue and green, because they all have a dip in their spectral output near there.</p>

<p>Hi Dave,</p>

<p>I'm flattered that you took stock in my advice, but my advice was actually to use different bulbs for different purposes. I'm all about minimizing carbon footprint. I only pull out the incandescents and halogens for photography, showing artwork, and lighting closets and bathrooms (brief lighting, cycled a lot). But thanks for sharing your results and impressions with us. I have heard others also remark that the LEDs are quite glary. I think it depends on whether you look straight at them. I think that glare can't be good for your eyes!</p>

<p>As Bill C said, a more interesting test would have been shots of a colorful assortment of items under various lighting conditions, each with a custom WB (to be fair). I'd compare daylight to old-style incandescent to halogen to CFL and LED. I think what you'd find is that the incandescent and halogen look a lot like daylight, but the CFL and LED, while appearing superficially similar, would have a lot of color inaccuracies -- this item that looks too orange, that item that has a strange, green cast, etc. Color inaccuracies (metamerisms) would be the primary issue.</p>

<p>BTW, and easier way to get an RGB value off of a surface is using the color selection tool in Photoshop. You can change the size of the eyedropper, so that you're averaging over a very large number of pixels.</p>

<p>Another odd issue with LEDs is that the light can be a different color temp in different directions. This is because the blue light used to excite the phosphors tends to escape more perpendicular to the surface of the phosphors, and not so much at grazing angles, while the yellowish components of the phosphorescence is emitted equally at all angles (not taking into account the diminished apparent size of the element from the sides). This means that the center spot is more blue, surrounded by a ring of light that's more yellow. This problem is diminished with diffusion (e.g. through frosted glass).</p>

<p>Since you made all shots in raw and were shooting a (presumably neutral) gray card, I wonder why you did not also include an example of WB performed with "eyedropper" in ACR or LR. Had you done so, my guess is that all samples would appear neutral gray (R=G=B). Actually that would not have given you much information, thus the need for various colored samples as Sarah and Bill have suggested. Ideally a Color Checker Passport would be the preferred target. Then after white balancing you would look for different rendering of the various color patches.<br>

@Sarah-<br />I was not aware that LED's employed a phosphor. Can you elaborate?</p>

<p>Thanks all for the replies. For a variety of reasons, I've concluded that CFL's are a nuisance and for my purposes a fail on the cost/benefit level. There may be some CFL's that would be pretty satisfactory, but not ones available at my local grocery story / drug store / Lowe's / Home Depot.</p>

<p>I used AWB instead of custom calibrated WB because the goal was in part to simulate what I'd get in typical use. Snapshots around the house generally preclude effective use of setting a custom WB setting in the camera. (In this day and age, to me color film and snapshots are almost mutually exclusive. A decade ago Fuji's color negative films with the fourth color layer were a definite improvement over most other films, but in those days, color correcting filters just robbed too much light for normal use.)</p>

<p>The use of colored objects (I don't have a color test chart) would have been a very informative addition, but time precluded it. Also, given that I'd already ruled out CFL's, and the LED's were unacceptable from a glare / harshness standpoint for most of our fixtures, their relative color rendering aspects ended up being not too relevant to my decision.</p>

<p>I don't use PS, but the color picker in GIMP served easily enough to let me check. However, I think it always reports a single pixel, hence the averaging provided by cubic scaling down to a single pixel.</p>

<p>By the way, I think the ideal color for indoor lighting is probably around 4000 K. Obviously the 5000 K LED's were pretty close to full sun, but that color looks a bit odd. My leading theory is that the total amount of light is much lower than what you'd get outside in full sun, and your brain expects less blue / more orange light at the levels typical in indoor artificial lighting. Clearly 2700 K is too yellow-orange, so I'm thinking 4000 K would be about 'right' for most uses.</p>

 

<p><em>Since you made all shots in raw and were shooting a (presumably neutral) gray card, I wonder why you did not also include an example of WB performed with "eyedropper" in ACR or LR.</em></p>

<p>A fair question. Very, very few of my home snapshots include anything that could function as a gray card. The point was to test 'actual' light (relative to 'daylight') and also to test AWB performance relative to the different light sources. I have occasionally used the gray-point method of color balancing where the picture happened to include something suitable (a concrete sidewalk can work pretty well), but that's a very small fraction of my pictures.</p>

 

<p>Mike, a "white" led is built upon a royal blue LED. Phosphors are coated onto the surface of the LED. When the blue light strikes the phosphors, those emit other wavelengths, mostly in the yellow part of the spectrum. In addition, some blue light penetrates the phosphor layer and contributes to the "white" spectrum.</p>

<p>A printed color checking card might not have been as informative as real objects of different colors. Each patch on a printed card is printed from the same small selection of inks, which are subject to the same metameric effects. But real world objects get their colors in all different ways, so the possibilities for metameric "errors" would be much greater. </p>

<blockquote>

<p>[Mike B] Ideally a Color Checker Passport would be the preferred target.</p>

</blockquote>

<p>To be clear, the Macbeth ColorChecker and presumably subsequent versions such as the Passport are not simple printed targets such as Sarah is talking about.</p>

<p>The ColorChecker used carefully selected colorants, in an attempt to mimic, as close as possible, a number of real-world colors. A 1976 technical paper by CS McCamy et al describes the original objectives as well as an evaluation of how close they came. I've put a temporary excerpt showing the spectral reflectance of skin vs their result here: <a href="/photo/17667978">http://www.photo.net/photo/17667978</a></p>

<p>So Dave, what did the images I posted of the Color Checker chart and Skittles bag color rendering test shots using the Philips 3000K LED posted in your prior thread convinced you to go with halogen?</p>

<p>You're right about the LED being overly harsh in brightness but I have it behind an audio speaker bouncing against corner walls that provides a nice warm (non-yellowish) glow throughout my living room. It's definitely better looking than regular incandescent living room bulbs.</p>

<p>Skin tones came out pretty good using a custom WB, a lot better than what I got with CFL's.</p>

<p>Glad I could help you with your decision on which bulb to get. Good luck with it.</p>