I've a white balance problem - any ideas?

Yesterday I was taking pictures of a fencing event with my Ist DL2 - set to jpeg, and using a 50 1.7 lens. Everything was manual, and the settings were the exactly the same for the two pictures below. 1/200s, f/2, fluorescent wb, iso 800. However, they've turned out to have different color casts, despite being of the scene under the same lighting.




I'm wondering if theres anything I could do to avoid this - or is the brain of my dl2 just beginning to struggle?

I know one of the suggestions is likely to be to use Raw, and i often do but on this occassion jpeg was more convenient as some of the pics were distributed immediately.

that's one setting that appears to be too tough for the camera to handle.

I'd try to find a custom white balance setting that works.
or even better a manual white balance off white paper (or grey card).

next up: set camera to manual, take an exposure reading, then pop up the flash.using flash this way will get you a decent color balance on your subject, but lets the areas your flash won't reach go off.

the next step up is to gel your flash to match your ambient light.....

(interesting side note:
it appears the second image (where the grey carpet is in the lens dof, is better color than the first image where the warm toned floor is in focus and the grey carpet is not...)

That is likely caused by AC current. The - and+ on AC change direction 60 times a sec. (in North America, I think 50 in your world), so when you take a picture and you happen to push the button as the AC change direction, it "confuses" the white balance. With the K20D in burst mode, tou can easily see that situation in a sequence.

Come on now Mr Ist, blaming your camera when the problem was simply yourself......out on the Piste

Seriously though, I found Pentaxians reply interesting but I was wondering myself if the location had a lot of glass windows and the problems was caused by an irregular mix of daylight and artificial, especially if the sun was in and out all the time, more so if your camera was set on a specific WB setting.

I tend to use mine on AWB all the time and even when converting RAW files I often find the AWB setting has produced far better results than any specific setting, strangely enough with the exception of "flash" :

I've used AWB for flash and been happy with the resulting warm glow, and so have the models...

Foe general use outdoors I use Daylight setting, which retails the light quality at the beginning or end of the day without trying to correct it.

These days i look for warm, glowing light or other casts and really don't seek out a "neutral" result.

Thanks for all the replies - I'm not quite sure what to make of the anomaly though, as the awb setting was giving even more erratic results. The light was largely artificial, but there was some natural coming in from windows along the top of the hall.

Pentaxian's suggestion is an interesting one - not much that be done to avoid that apart from only pressing the shutter at a time that is a multiple of 16.7ms after turning the camera on...

I got hold of the 50 1.7 lens just for fencing events - its great! These fencing characters move like lightning and the kit lens just wasn't fast enough.

Certainly artificial light flickers 100 times a second in the UK, especially fluorescent light (incandescent lamps even out the flickering much better) - that fact can used to calibrate the speed of turntables, so any shutter speed faster than that will be affected to some extent. Whether colour temperature also flickers I do not know, but your pictures would suggest that is the case. Try using 1/100 s (I know it won't freeze movement as well, but it will at least identify the cause of the problem)

To go back to the light source, fluorescent lights are a problem not because they flicker but because they do not have a continuous spectrum. There are gaps in the colour and therefore correction is difficult. Results on daylight settings tend to look green or yellow/green, very unpleasant.

There are three settings available for fluorescent light and I would try them. It sounds like the natural light entering is relatively incidental.

Fluorescent light is not the nicest of sources for photography unless you are using specially corrected dtudio lights, which is a different matter altogether.

Take several manual white balance readings (so you are likely to have captured both phases of the light) and apply the appropriate one to each of your images afterwards in your RAW program.

When the AC current changes direction with tungsten light, the filament doesn't have time to cool down, so the color temp. doesn't change, but fluorescent light are using a ionized gas to produce light. When the current changes direction, that gas isn't ionized anymore (for a very short time) and henceforth changes the color temperature for a very brief time. This is the reason why it is nearly impossible to get accurate white balance with fluorescent lighting. And to top it off, different manufacturers of fluorescent tubes will use different mix of gas to end up with what is their idea of daylight, so that makes it impossible for camera manufacturers to get a good fluo. adjustment on their cameras.

I've never heard of this phenomenon - do you have any references to literature that supports this?

If it were so I do find it hard to believe that there would be any significant effect on a 50Hz or 60Hz cycle. It would even out to an average that we would be unable to see any fluctuation of colour in - our eyes would not respond quickly enough. It also assumes that we end up with a complete spectrum of output, of whatever colour, which we don't.

Do please elucidate.

If there are two alternating colour temperatures (???) then you are likely to capture both of them if you take a few readings. If you toss a coin a number of times - say ten - it is bound to land heads up at least once, and tails at least once.

Then you will have two accurate readings - to manually apply to the images afterwards.

Our eyes are 'fooled' but not the camera - and if the shutter speed is faster than a complete cycle of the mains - 1/50s in UK (I realise my earlier suggestion of 1/100s was wrong - you need to cover a whole cycle, not half a cycle) then the camera will only catch that portion of the cycle, which means the camera won't even things out the way the eye does, and there may be colur balance vatiations - further, if the metering is not 'real time' there may be exposure errors too.

If you watch a CRT TV screen it looks like a complete picture. Photograph it at any shutter speed faster than 1/50s and you won't see the whole picture. It is the same principle.

There is a simple solution to this problem. Take a photographer's white/grey card with you and set the WB manually before the event. A better solution is the Expodisc (or is it Expodisk).

Any 'thinking' camera is bound to get it wrong sometimes. Using a known white subject eliminates the problem.

Pentaxian450 is suggesting that the colour temperature with florescent lights is not constant. If that's correct then the solution is not as simple as you suggest. I'm not sure it is correct, though.