17
\$\begingroup\$

I just bought a Nikon SB-910 speedlight.

The filter that comes with it to use under fluorescent lighting it green, why is this? because certainly to my eye there is not a hint of green under normal fluorescent lighting.

\$\endgroup\$

4 Answers 4

23
\$\begingroup\$

Balancing for fluorescent lights is harder than say tungsten. The reason for this is that tungsten bulbs produce the same sort of spectrum (set of intensities at different wavelengths) as a daylight balanced flash, just shifted.

A fluorescent light doesn't have the same bell curve shaped spectrum, it produces a set of spikes at very particular frequencies. In particular there aren't many spikes in the red part of the spectrum. The reason you don't see a green tint is probably that the brain is filling in the missing information for you.

Using a green gel on the flash allows you apply a magenta colour tint to your image (to cancel out the green) which helps restore some red that will be missing from the fluorescent lit skintones.

You can never fully replace the missing frequencies with fluorescent lights, and some are much worse than others, for example sodium lights produce very few frequencies, no matter how you try and adjust the colours, there's no information there to recover.

Here's an example of an incandescent light source (a fire!):

Now because this source produces a similar spread of frequencies as the sun, albeit shifted toward orange, we can correct this to a achieve a daylight white balance:

Now let's take a shot under the worst kind of fluorescents:

It looks orange like the first shot. However, even if we shift the image by the same amount we don't get any colours, they simply aren't there in the first place:

So while the filter may help you take out the slight green tint that results from missing certain red frequencies, it wont replace certain colours that are lost.

Fluorescent lights are good for the environment, but terrible for photography. There is hope however, newer designs are improving the width of the spectrum, as defined by their CRI (colour rendering intent) number.

\$\endgroup\$
4
  • 1
    \$\begingroup\$ Answer deja vu! :) Do you think these questions should be merged? \$\endgroup\$
    – mattdm
    Jun 21, 2012 at 11:40
  • \$\begingroup\$ @mattdm It's a different question I just thought it's worth reiterating that you can never properly colour balance fluorescents, unlike incandescent lightsources. \$\endgroup\$
    – Matt Grum
    Jun 21, 2012 at 13:43
  • 2
    \$\begingroup\$ Nice job of colour correcting the firelight scene :-) ! \$\endgroup\$ Jun 21, 2012 at 19:31
  • \$\begingroup\$ "Worst case of flourescents"? No, bad case of sodium vapor lamp, which isn't a flourescent. \$\endgroup\$ Nov 20, 2019 at 21:25
8
\$\begingroup\$

I think in this case, a diagram or two is probably the easiest way to get the point across.

A typical "cool white" fluorescent bulb produces an output spectrum something like this:

enter image description here

In this diagram, blue is to the left, green in the middle, and red to the right. As other answers have already pointed out, your eye/brain can/will adjust so you usually see the dominant light as "white", almost regardless of its actual color (unless the spectrum is really narrow, like sodium vapor or mercury vapor bulbs produce).

Anyway, the problem you run into with using flash under fluorescent light is that you end up with some parts of the picture lit by the flash, with one color balance, and other parts of the picture lit by the fluorescent bulbs, with an entirely different color balance. If you adjust the balance for the "flash" part of the picture, the part that was lit by the fluorescent bulbs will look sickly green. If you adjust for the fluorescent-lit part of the picture, the part lit by flash will look quite purple.

To avoid that, the green filter on the flash shapes its output to be at vaguely similar to the fluorescent light's. With a typical filter you won't get the spikes of a fluorescent output, but might get something more like this (the black line overlaid on top of the fluorescent spectrum).

enter image description here

You won't get (and generally don't want) a perfect match, but this will at least let you get something approaching an even balance across the entire picture.

I should probably add that I've probably exaggerated the height of the "peak" in the middle of the output produced by the filter. You're not really trying to match the height of the spikes, but the approximate energy output in that general region. Tall, narrow spikes mean high brightness at a specific wavelength, but not all that much overall energy.

\$\endgroup\$
6
\$\begingroup\$

The key words in your question are "to my eye". The human vision system is fiendishly good at adjusting white balance. Fluorescents are anything but white, but your eyes perceive them as white nonetheless.

If you look at some nighttime photos of a cityscape or a building exterior, photos with a lot of different light-sources visible, you will see that most artificial lights have a dramatic colour cast. Incandescents are orange, fluorescents are - yes - greenish.

\$\endgroup\$
4
\$\begingroup\$

This is really a technical comment to accompany Matt's answer, not intended as a complete answer - others do that well enough already.

Matt did a nice job of colour correcting the firelight scene :-) !
A look at the luminance and RGB information of the original show a very sorry looking light mix which he has redeemed very nicely.

The orange lit street and tree version is actually lit by a Sodium vapour lamp which excites Sodium vapor to emit light at essentially a single spectral line (for practical purposes). The result is worse than fluorescent - the principle is the same - with fluorescent you have a limited number of spectral lines available and colours in between spectral peaks are not matched by light in the source.

Fluorescents try to produce light which the eye-brain sees as white with a limited mix of colours produced by reradiating visible light from phosphors excited by the tube's native UV. BUT with sodium vapour lights the overall aims are high efficiency and good visibility with colour rendition being a non-issue, so they have an almost monochromatic source. With fluorescent there is "not really enough spectral content to properly restore the tru balance". With gas discharge based light, there is about zero spectral content away from the main emission line. Which is what Matt said :-).

\$\endgroup\$
7
  • \$\begingroup\$ Excellent! Now I see who it is who has been revenge downvoting my electrical engineering posts :-). \$\endgroup\$ Jun 21, 2012 at 20:04
  • \$\begingroup\$ No idea why you were downvoted, I must admit I tend to lump sodium lights in with fluorescents even though the process by which they produce light is quite different. In any case they are both quite different to incandescent light sources. \$\endgroup\$
    – Matt Grum
    Jun 21, 2012 at 23:06
  • 1
    \$\begingroup\$ @Russell: I don't know how you can see who it was without additional information not available to bystanders. But care to share who it is and how you know? I've been wondering who has been doing this too (No it's not me, really.). I get a few for no apparent reason, but you seem to get more, I suspect in part because you get so nicely upset about it. \$\endgroup\$ Jun 22, 2012 at 0:03
  • \$\begingroup\$ @OlinLathrop - no detailed elucidation in public :-). There is a degree of Gurney Halleck action in there but the conclusion is going to be about as good as his are. Needs a person who is active in two SE areas and dissentious in both and actively so at present. Do your own Mentating :-). \$\endgroup\$ Jun 22, 2012 at 1:48
  • \$\begingroup\$ You'll note, the downvote button says "this answer is not useful" - it does not read "this answer is technically incorrect". Your assumption that you've added into your answer leads people to believe they should only vote when an answer is technically incorrect and thats not true. And it will be removed. \$\endgroup\$
    – rfusca
    Jun 22, 2012 at 4:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.