Folks, I ran into a bit of a problem Saturday - nothing mindboggling, but one that I'd like to know the answer to for future reference.

Simple situation: I found out about a badminton practice on way-late last-minute notice, (as in, an hour before it was to start.) Important, but jammed between other assignments Saturday.

I knew the gym they were practising in was lit by ugly yellow flourescents. Yes, yellow - with some windows adding a touch of blue to the place. Now, even though I knew the pics were running black and white from raw images, I tried to balance my colours (manual white balance on a nearby white object) which gave me a temperature of 2500K as shot. No problem with skin tones OOC there, but I was shooting action, and I needed some more light.

Hence, I threw on a Roscosun "Straw" gel to the 580EX flash, set to high-speed synch so I could go beyond 1/250 sec. I figured a yellow gel would do the trick for colour matching .... Bzzzt. Not really. It kinda gave a green skin cast. Again, I took risks I knew I could get away with (I knew I'd run the pics b/w so I don't care), but still ... the gel didn't take off enough blue.

The question here is: what should I have gelled the flash with to bring the output to the same yellow as this gym? Double up a 1/2 CTO as well as a yellow?

Input welcomed. This was well beyond the normal "add green for fluorescent, cto for incandescent" bag of tricks.

When you use high shutter speeds - shorter than 1/125th or so - with fluorescent lights ( are you sure these were fluorescent and not Sodium Halide ) the color temperature varies/cycles with the 60 cps AC current, and the color temp in the image will vary from frame to frame with too short a shutter speed.

If you decide to go to fill flash, keep your shutter open long enough that you avoid this color temp cycling.

The other alternative is to shoot a jpg of a neutral non specular white sheet, and create a custom white balance, but you still need a full cycle of the light from the fluorescent light so that the color temo does not vary from frame to frame.

Gym floors are yellow and reflect a lot of yellow light, so I am surprised you had that much blue left.


You might post the image here so folks can actually see the colors you are describing with text.

Originally Posted by pathfinder When you use high shutter speeds - shorter than 1/125th or so - with fluorescent lights ( are you sure these were fluorescent and not Sodium Halide )

Might be better if I go in the gym when I have more time and capture some establishing shots ... I'm not 100 per cent sure they're fluorescent, tho ... they are awfully yellow, but don't look like sodium vapour to me. And I seem to recall one of 'em was a newer bulb with a standard blue/green cast to it.

Let's just say this place is a white-balancing nightmare. I usually shoot ambient with a manual white balance, and that's it ...

Shoot a series of shutter speeds, from 1/30 to 1/500 of a neutral wall and see if the color temp is stable, or variable too

Sounds like your choice of B&W is a good one

In the USA or any other place with 60Hz electric service, lamps flash 120 times a second exactly. If the color of the light shifts during each electrical AC cycle (flashes) and you need consistent shots, you must therefor use a shutter time that is an exact multiple of 1/120, like 1/60, which captures exactly one or two complete wave-periods. If you would shoot faster than 1/120, like 1/240, you only capture during a part of an electical AC cycle and different shots will be from different parts of the cycle, leading to different color casts.

This all is asuming that the lamps indeed shift olor during cycles.

Here's a quickie - little post-processing from raw ... It's a reject, but I include it so you can see the light coming in from the windows beyond.

I think this gives reasonable sense of the colour cast in the gym, and the goofed gel on the flash making badminton guy a little too Spocklike.

You'll note that the lines on the floor are mostly white. Mostly.

As a start, try using the white strips in the floor for a custom white balance in Photoshop or ACR. If this was shot in RAW, do this color balance in ACR.

You might even try balancing off his black shirt, or the silver stripe in his shorts. Those all seem like good potential neutrals to color balance off.

When I read the pixels in the white line on my computer screen, calibrated with an i1Display, it reads R 243, G 252, B 247 which to me is too green, not too yellow.

The silver stripe in his left leg reads R 231, G 240, B 223: again this is too green. The numbers vary a bit depending on which pixels I read, but the general overall tone is as described.

The wall behind him is warmer in the red channel, so this image has different color temps in different places, but the important places are his face, skin, shirt, and the white lines in the floor which we are pretty certain as to appearance.


When I open your image in Photoshop I get informed that this jpg has no color profile. How can we discuss color balance of an image, if it has not been assigned an appropriate color profile like sRGB for the web, or aRGB for printing??

Originally Posted by pathfinder As a start, try using the white strips in the floor for a custom white balance in Photoshop or ACR. If this was shot in RAW, do this color balance in ACR.

I think you've mis-read my query. It's not about how I could post-process this spocklike abomination, I want to know what gels I should have used to obtain a good balance in the first place. (As noted, this is OOC, other than pushing through raw conversion to create a .jog.)

When I open your image in Photoshop I get informed that this jpg has no color profile. How can we discuss color balance of an image, if it has not been assigned an appropriate color profile like sRGB for the web, or aRGB for printing??

My normal workflow pushes to print, where a CMYK profile would have been provided. Not that I would ever print Spock, here, in the paper.

Then you needed more magenta in your gel ( to kill the green ), in this frame. Your shuttlecock reads an excess of green also.

Magenta alone, or magenta with the yellow?

(I'm not skilled with gels other than CTO for incandescent, green for fluorescent, and purdy colours for effects.)

You actually use a green gel with flash to make your flash match the ambient fluorescent, and then set your camera for Fluorescent color temp, but it is not always that easy. (Your posted image was too green, and that needs magenta to kill the green. Your image also need a color profile as well.)

A quick google of Balancing Flash with Fluorescent lights returns several very useful links

http://strobist.blogspot.com/2008/04...uorescent.html The Strobist is always a great source for flash help.

http://www.photographytalk.com/photo...orescent-light

http://www.kenrockwell.com/tech/whitebalance.htm Ken Rockwell's suggestion for gyms lit with 120cps fluorescent?? "Good Luck!"

http://www.scantips.com/lights/flashbasics5.html A good discussion of Fluorescent light flicker of color temp

http://jwadkins.blogspot.com/2010/01...uorescent.html


If shooting in RAW, you could always shoot a WhiBal target, or use an Expodisk also. But a green gel will help your flash color temp more closely match the ambient fluorescent, and then you can set your camera to Fluorescent color temp.

Originally Posted by pathfinder You actually use a green gel with flash to make your flash match the ambient fluorescent, and then set your camera for Fluorescent color temp.

But the fluorescents (if that's what they were) were pumping out yellow light ... not the standard icky white-green...

I'm wondering, if as another poster suggested, these are some form of sodium lights ...

Quite possibly, but your image measures green, not really yellow, on my screen when I read the individual pixel data.

OK. Following up on this issue, I'm going to try to hammer a few things out.

We have a gym in town that's lit by the same butt-ugly, really dark yellow lights as the the one in Saint Andrews posted earlier.

To provide perspective, this is an OOC (other than resized) JPG image, with the camera set at Fluorescent temperature, 1,250 ISO, 1/250 @ f2.8.

Note the two lights on the ceiling. The left light is a standard fluorescent, and the parts that aren't blown out measure mostly white - not surprising for a fluorescent balance. The right light is one of those butt-ugly yellow lights I've been talking about.

Its reading is about 20 points off on the blue channel. And you can see how (a) dark and (b) yellow the gym is.

OK, and here's an action shot, manual white balance to a handy-dandy chunk o' paper.
ISO 5,000 (Did I mention it was bloody dark?), 1/250, f2.8.
The white of the stripe on the setter's left shoulder pulls R223, G221, B223; a part of the white on the upper part of the ball registered R247, G246, B242 - so the manual white balance is pretty good, OOC. Were this a portrait, it would likely require a wee bit of fine-tuning on the levels, but it's pretty good for what it is, and the situation it was shot in.

So ... you'll note a wee touch of motion blur on the girl's right hand. Let's assume I wanted to use flash to give me some extra light, say a half stop or so in order to get near 1/400 sec.

Clearly, from the previous experiment, a "straw" flash will turn our middle-school volleyball player into T'pal from the Vulcan Volleyball Academy.

So here's the question that has yet been answered ... if you were to try to use strobes to add some light in this gym, what gel would you put on your flash that would come pretty close to registering white properly, assuming the manual balance had been set to match the piss yellow?

What is the Kelvin temperature of your custom white balance in your camera for this image shot in dark, mixed fluorescent/Sodium lighting?

Your speedlite should be about 5600K unfiltered, and that will have to be filtered to match the custom Kelvin temp your camera is adjusted to.

Will have to get back to you - those pics are on the work 'pooter.

Here is where a custom DNG profile would probably help (especially if you build a dual illuminant profile). If you have a MacBeth color checker (24 patch), you have all the hardware you need. From there, after loading the custom profile for the raw data, a WB on one shot under that scene and you should be in better shape. With mixed lighting, you can’t have it all. IOW, if you make the Fluorescent look good, the daylight will still have a cast (even with a dual illuminant DNG profile although that can help a bit).

Have the work pooter: the yellow came in at 2500K.

Originally Posted by arodney Here is where a custom DNG profile would probably help (especially if you build a dual illuminant profile). If you have a MacBeth color checker (24 patch), you have all the hardware you need.

Once more, I point out, I am not asking how to fix the colour cast in post, or to render a white balance in post. I'm reasonably aware of how to fix colour casts and missed temperatures in post ... at least sufficient for newsprint, anyhow.

What I am asking is what blasted gels to use on a flash so that it will match, somewhat, the above temperature.

A DNG profile is NOT a fix in post. It happens before you even render the raw data. It is used as part of the raw processing engine.

If you want to use fill flash, you simply have to filter it to match the lighting in the room, use a shutter speed such that the ambient light and the flash record. Did this before the digital camera was invented years ago. Under most Fluorescents, something like a 40M on the lens, 40G on the flash (best to test first).