Color Temperature Swings

[kenneth_smith7]

<p>Shot a D7000 in a florescent, tungsten mixed environment last night. ( 90% florescent. I doubt the tungsten affected anything )<br>

I shot Raw @ ISO 800 & 1600., with a 35mm f/1.8 @ F/2.8 consistently, and using Auto White Balance, just to see what it would do.<br>

And the surprise was this. A person standing in the same spot receiving three shoots taken a few seconds apart produced first a good balance, almost perfect, but the second shot would always be far more yellowed. This happened repeatedly. Looking at 150 shots in View NX2 it looked as though I had intended to have variances when I shot two or three. Single shots were more random and it would be harder to say whether the light temperature varied enough to throw this off. But three shots of the same area? Why would they vary?</p>

 

[waltflanagan]

<p>Fluorescent lights flicker 120 times a second.</p>

[kenneth_smith7]

<p>So because I'm shooting at 1/125 a second, I'm snatching out different wavelengths. Interesting. That means I'd get a more consistent blend at 1/60th I assume.</p>

[elliot1]

<p>what shutter speed? </p>

[kenneth_smith7]

<p>I was in Aperture Priority and got usually around 1/320 to even as high as 1/800.<br>

I wanted to stay wide open for select focus to blur all the surrounding confusion. The examples show a given shot with 1/500, where the first one is fine, and the identical shot a second later at the same speed is yellow. If AWB is going to grab the first, why not the second? Even it's a different wavelength?</p>

[craig_shearman1]

<p>Actually I think they flicker at 60 times per second, since U.S. electricity is at 60hz (50 in Europe and some other places). But you do need a longer shutter speed to capture the full range of the fluorescent cycle. Definitely not higher than 1/60 and maybe 1/30. Not sure what happens with the fluorescents that are designed for photography.</p>

[kari_oinonen]

<p>Fluorescent can be really bad. I have a feeling and experience that the light colour may change during the flicker cycle.<br>

The other thing to note is the fact that the light colour distribution may not be even, but it may lack some colour components.<br>

A longer shutter speed may help to even the variability (intensity and colour) in the light.<br>

Maybe the AWB could somewhat compensate for single shots you made, but shots in a row had more the same white balance value setting, while the coulour of the lamps varied.</p>

[ilkka_nissila]

<p>You cannot trust general household fluorescent lights or energy-saving compact fluorescents to give any particular color balance from shot to shot - it's a crap shoot basically. I either turn fluorescents off completely to avoid this problem, when I'm photographing, or I simply convert everything to black and white. In my opinion the EU did a huge disfavour to photography when they decided that tungsten lamps need to be replaced by compact fluorescents which are just terrible IMO. Fortunately halogen lamps are at least for now allowed, and they work extremely nicely for both capturing original photography and illuminating printed images on walls (halogen spots), as well as giving a nice clean lighting to the home. The only issue is that they seem to be short lived and they run fairly hot.</p>

<p>LEDs are the rage now but I think they produce quite bad skin tones and I've decided to avoid them completely. Of course, at some stadiums they are used and they're comparatively good relative to alternatives that have been used in the past but I don't much like the skin tones - they're acceptable but not as good as flash or halogen IMO. LEDs are much better than fluorescents though in my experience, but they're in my avoid where possible category in any case.</p>

[luke_kaven]

<p>Auto White Balance is integrated with the color matrix meter and the Scene Recognition System. It makes its decisions partly based upon its own classifier before the shutter is tripped. As such, the information is a few milliseconds old by the time the picture is taken. So it is easily fooled by the spectral shifts within a single AC power cycle.</p>

[kenneth_smith7]

<p>Guess it is a crap shoot then, and given that I could not have chosen the light or allowed flash, it's a bigger color correction job than I anticipated. Oh well. Maybe this would look GREAT in Black and White. Thanks for all the replies. </p>

[rodeo_joe1]

<p>"Actually I think they flicker at 60 times per second, since U.S. electricity is at 60hz" - Nope. Lights flicker at <em>twice</em> the mains frequency, because the AC current swings both positive and negative during each 1/60th second cycle. You can't have negative light, so the tubes or bulbs illuminate once every half-cycle of current, and that equals 120 times per second for a 60Hz mains frequency, or 100 times per second where the mains is 50Hz.</p>

<p>Anyway, I think we're all agreed that mains flicker is the issue here, and that you need to capture one full cycle of mains to avoid colour shifts or inconsistent exposures. Anything less than one full cycle and you can't be sure which part(s) of it you'll capture. So, where the mains is 60Hz the only 'safe' shutter speed is 1/60th or longer; and in 50Hz regions it's the non-standard speed of 1/50th second. For the same reason you have a choice of 25/50 or 30/60 FPS video capture. Some cameras also let you choose the EVF or Live View refresh rate to match the local mains frequency as well, so that your viewfinder doesn't 'strobe' with the lights.</p>

<p>The above light flicker can present a safety issue in workshops where there's fast moving machinery, and also presents a televising problem in professional sports stadia. In those situations you'll likely find that the lights are driven from a multi-phase source such that the flicker is much reduced or eliminated.</p>

[mike_halliwell]

<p>I guess the question is, is this the result of AWB being fooled by the actual lamps changing colour at a frequency related to the flicker of mains power or is it a true record of the colour changes.</p>

<p>As you shot in RAW too, if you do a conversion of say 10 sequential images with a new WB, say Fluorescent, ie override any 'WB as taken issues', do they display as much colour variation as the AWB set?</p>

[mike_halliwell]

<blockquote>

<p>The above light flicker can present a safety issue in workshops where there's fast moving machinery</p>

 

</blockquote>

<p> In my old workshop, they had mains powered fluorescents that had pairs of tubes together, but were deliberately of a mirrored phase, ie out of phase, so when one was peaking the other wasn't etc etc. it stopped fast-spinning lathe chucks appearing to be stationary.</p>

[kenneth_smith7]

<p>I tried a fluorescent conversion and the variance degree is the same, and worse actually as the correct ones are too pink, while the off ones are still yellowish. I'll tell you the absolute worst thing about this take is not even that, it's the fact that the top half of a shot is cooler than the bottom. I'll get a face looking good, but the shirt will still be 40CC too yellow and uneven to boot. Bounce from the tungsten off the floor? I don't know. It's all a bit much. They should have given this to Joe McNally. Kidding. Small stuff here.</p>

[mike_halliwell]

<p>I guess as the shutter moves up/down and the colour changes whilst the blades are moving, colours will appear different at the 'start' of the exposure to the 'finish', ie top to bottom.</p>

<p>Any portrait orientation shots with a side to side variance? Now that would be a real bu**er to fix in post!</p>

[ellis_vener_photography]

It's not different wave lengths of light you are seeing being captured, it's different points in the voltage phasing. At at the

peak voltage you are getting the neutral balance but when it appears more of a muddy yellow the voltage is dropping

below that. Fluorescents with excellent ballast do no have this issue.

[nsfbr]

<blockquote>

<p>In my old workshop, they had mains powered fluorescents that had pairs of tubes together, but were deliberately of a mirrored phase, ie out of phase, so when one was peaking the other wasn't etc etc. it stopped fast-spinning lathe chucks appearing to be stationary.</p>

</blockquote>

<p>Actually, they weren't on different phases, or at least didn't need to be. They would have just been put on a lead-lag ballast. </p>

<p>And modern CFLs don't have this problem because they use high frequency ballasts, up around 5kHz. </p>

[Tim_Lookingbill]

<blockquote>

<p>I tried a fluorescent conversion and the variance degree is the same, and worse actually as the correct ones are too pink, while the off ones are still yellowish. I'll tell you the absolute worst thing about this take is not even that, it's the fact that the top half of a shot is cooler than the bottom. I'll get a face looking good, but the shirt will still be 40CC too yellow and uneven to boot. Bounce from the tungsten off the floor? I don't know. It's all a bit much. </p>

</blockquote>

<p>I gave it go with a mix of my bathroom sink (4) vanity globe 2800K CFL track lighting against a 100 watt GE Soft White 2800K bulb installed in a utility lamp and didn't get top half cooler than the bottom issues. You didn't mention the color temp of the fluorescents you were shooting under, so my results (shown below) may not be what you get, but I did take several shots at 1/80th, f/6.3, ISO 800 spaced randomly minutes apart and never got two tone results.</p>

<p>I did start out setting a custom white balance sampling off a WhiBal card positioned where both lights were equally lighting the gray card in order to get an equal mix shooting Raw in order to have something decent to start out with in editing which did require additional white balance adjust in ACR as well as some heavy HSL tweaks. I found going for a warm white balance instead of R=G=B sampled off the WhiBal card sort of perceptually smoothed out color hue differences. </p><div></div>

[CvhKaar]

<p>If you still have those differences at longer ( 1/30th plus) exposure times, you might also want to check something else ..<br /><br />(Not sure about the D7000) edit : On the D7000 setting E5 , but several Nikon DSLR's (On the D7000 setting E5) can also do "White Balance Bracketting" , so if that is swithced on, depending on the cycle settings, you would get sets of 2 or 3 subsequent different white balances ....</p>

[rodeo_joe1]

<p>Tim, a 1/80th second shutter speed is too long to see much of a colour shift across the frame. The shutter speed needs to be quite a bit shorter than one-half mains cycle to get any strong colour shift. It also needs to be above the X-synch speed of the camera. The OP reports he was shooting at upwards of 1/320th second I believe. Not only that, but as Joel said, energy-saving compact fluorescent tubes (CFLs) have an inbuilt voltage converter that runs at a much higher frequency that that of the mains. And tungsten bulbs have a thermal lag that makes their flicker much less troublesome than that of fluorescents. So I'm afraid your experiment doesn't count for very much - sorry!</p>

<p>If you can find some old 6' cool-white fluorescent tubes and shoot at 1/250th and above, then I think you'll see the effect quite readily.</p>

[Tim_Lookingbill]

<p>Thanks for summarizing and making it more clear for me what Kenneth was attempting, Rodeo Joe.</p>

<p>You're right the description of Kenneth's results I've never encountered because I've never shot at ISO 800-1600 under any kind of fluorescent lights that allowed 1/320th shutter speed. Since Kenneth doesn't exactly describe the location and type of fluorescents, I was under the impression he was shooting at someone's home.</p>

<p>Trying to keep in mind the various cycling of the electronics of these type of lights in order to anticipate the various visual anomalies they create is way too much to keep in my head.</p>

<p>At least I now know to never think I can be knowledgable enough to add to these types of discussions on lighting mainly because I'm never going to know or remember all the technical facts of the situation discussed nor will I want to shoot under such conditions where I have to remember the technical limitations in order to prevent or mitigate against the botched results the OP describes.</p>

<p>But then I don't think in all the different lighting situations I've shot under using available light which makes up about 90% of my 3000 or so Raw's I've shot so far as a hobbyist that I've had to say..."I can't shoot under these lights because XXXX lights will give me terrible results I don't want to go to the trouble of fixing in post".</p>

<p>I just wait for a better time and better place to shoot under available lights for fine art type images I aim for. This is why I've turned down several requests from friends and relatives to shoot an important event in their life taking place outside and in their home. Too many unknowns. In fact my most recent high school reunion that took place at a classmate's home I was turning down requests from those with P&S asking me to take pictures with their camera and I declined because I knew I couldn't get any better results, but knew I'ld get the blame for any screwed up images because of my unfamiliarity not only with their camera but the constantly changing mixed lighting.</p>

<p>It would be interesting to see a sample of Kenneth's results he describes here.</p>

[kenneth_smith7]

<p>I checked. No, I didn't have any bracketing set.<br>

As for the location. Record store. Probably a mixed bag of various aged tubes bouncing off the covers of Frank Zappa and the Mothers of Invention ( aha !!!) <br>

It's not enough to fear shooting anywhere. We've got astounding editing abilities. This is all fixable, albeit a bit time consuming.<br>

I'd send an example, but I'm computered out. Trying to figure out today why an online back up of 1200 RAW files that took four days is not showing up anywhere. <br>

Technology... head... exploding...</p>

[dwight200]

<p>I took my camera, set the shutter to 1/1000 second, and took several shots of a fluorescent bulb in an orientation such that the shutter was travelling along the length of the tube. That meant the shutter was taking small samples of the tube at different times. I stacked several of them into an image which you can see below. You can see the color shifts of the lamp with time. The shutter is travelling vertically in the image.<br>

<img src="http://farm6.staticflickr.com/5041/5250301564_9444e997e4_b.jpg" alt="" width="1024" height="449" /></p>

[kenneth_smith7]

<p>Wow, that looks familiar. What a telling test. Thanks Dwight.</p>

[Tim_Lookingbill]

<p>Yikes! So that's what it looks like. Thanks for going through the trouble and including shooting specs, Dwight. I sometimes get a light and subtle pinkish bar in one half the frame shooting under my two 18" T8 5000K fluorescent tubes I use for viewing prints.</p>

<p>I tried to make my bathroom vanity track lighting globe CFL's give inconsistent color variances shooting the same scene I posted above at shutter speeds of 1/320th, 1/160th, 1/60th and 1/30th, three exposures a piece one after another for each speed and all the shots had the same WB with no other artifacts using custom WB with WhiBal card. Never considered my GE vanity globes were that reliably consistent. I might just start using them for quick online show and tell. My daylight balanced tubes are sometimes hit or miss with the pink area if I don't adjust shutter speed just right.</p>