Blue lens filter in Tungsten light

Shooting with the white balance setting on 'tungsten' just adds gain to the

blue channel which creates noise, so I thought that maybe if I use a blue

filter and balance things out before it hits the sensor it might have an

advantage. I assume it would take a longer exposure to get the same image with

a filter in front that blocks light. Would that longer exposure then add more

noise creating the same level of noise as if I wasn't using the filter?

 

If anyone is an expert on this I'd like to hear what you have to say as I'd

like to improve my low light room shots that are lit by tungsten lights.

 

Thanks

Using a blue filter on camera will make things worse in the signal to noise ratios of the green and red channels as you'll get very little light to the sensor. if you don't already know this each photo site ("pixel") in a sensor is capped with either a red, gree, or blue filter. This is the Bayer filter matrix. Adding a "blue" filter like an 80A on the lens will mean that only a little light will then hit the red and green filtered pixels. As the Red/Green/Blue ratio is 1:2:1 (or 25:50:25) , you'll be blocking most of the light that 75% of the pixels are recording. This may also throw off the calculations of the Bayer Matrix color interpolation algorithm to create full color values for each processed pixel.

 

if you are seeign the effects of increased gain in the blue channel, agood noise suppresion software like noiseware or Noise Ninja may be more useful. The trick with noise suppression software is to use it before you do any sharpening --which rules out in camera created JPEGS.

Let's see...

 

using the filter with the *same* exposure will increase the noise in the red and green channels. Then lengthening the exposure will decrease the noise in all three channels alike. So combining the two in the right ratio will decrease the blue noise and leave the red and green as they were, so yes, I'd guess you'd see a benefit.

 

The other way to look at the same thing would be to consider the exposure increase first to decrease the noise in the blue channel. This might result in overexposing and saturating the red and green channels, the solution to which would be to use a blue filter to lose some of the red and green intensity.

Alec, my feeling is that you'd be creating a GiGo situation ==garbage in = garbage out. But lets both run some tests to see what results.

 

One of the problem,s here is that camera and software makers usually spec a CCD or CMOS and processing algorithms that are balanced for nominal photographic daylight, hence the problem that Lucas is trying to solve.

Ellis,

 

I agree that that the CCD is designed for daylight - by putting a tungsten-to-daylight (blue) filter in front of the lens you'd be shifting the average spectrum of a tungsten-lit scene back to the r/g/b balance of daylight, which would suit the sensor better.

 

But yes, some tests shouldn't be too hard to arrange.

When I first used a CCD camera in the mid eighties it was quite important to use a filter to balance to the native CCT of the sensor. Though it is less important now, thanks to the improved dynamic range of the sensors, I believe that it is still a benefit if you can cope with the light loss. It's rather like using colour negative film - you can use it in mismatched light without filtration, but you get the optimum performance across the spectrum by balancing the light. An alternative is to half-balance the light with less light loss. You select from a range of possibilities, with full colour balance and significant light loss at one end, and letting the gain take care of the colour balance and no light loss at the other end.

 

As already mentioned, it is something that is best approached by testing. You don't need to shell out for camera-grade filters, just use lighting gels - maybe Full CTB and Half CTB.

Yeah, an 80a is a lot better solution than camera WB. Even an 80c will help quite a bit, and

only cost you 1 stop of exposure vs. 2.

 

Ellis: All a filter like an 80a does is correct the light back to ratios of red/green/blue it

would have in a daylit scene. So there's zero chance it will "throw off the calculations of

the Bayer Matrix color interpolation algorithm to create full color values for each

processed".

 

And I really have a hard time figuring out how light that's been restored

daylight balance counts as "garbage in".

My thoughts exactly. My idea was to get the ratio of rgb back to the temperature that the camera performs best on. I had a hunch that this might be a better solution than white balance on the digital front.

 

Now, I don't own a blue filter or gels, so I was wondering if one of you may have the chance to test it out before I spend time and money trying it for myself.

 

I'm also wondering if down the road we may see camera's that have interchangeable sensors with each optimized for a different temp of light. Or maybe a seperate camera with each. Would a 'tungten sensor' have a ratio of more red pixels? Just a thought, but there are some applications where I think it might be usefull.

Or would a 'tungsten sensor' have more blue pixels to compensate for the lack of blue light?

I don't know if the following method I came up with experimenting

with my camera's incamera custom WB setting would work in

RAW since you didn't indicate the format nor camera model, but

see if this helps.

 

The following shots off my Pentax K100D SLR illustrate using

apposing colored objects under tungsten lighting to trick the

camera's internal WB processes into overriding the camera's

tendency to add too much blue in neutralizing this type of lighting

giving more realistic results.

 

This assumes your camera has this type of custom WB feature.<div></div>

Thanks Tim, but all of these shots are adjustments on the software level. Regardless of what setting you're using, the same spectrum of light is hitting the sensor. I'm curious to see if actually changing the colour of the light will result in cleaner images since the balance of rgb that the camera performs best on will be restored.

Not sure what you're going for but the signal to noise ratio issue

you indicate in your original post I don't have in the posted

samples no matter what the software or exposure settings do.

 

Those posted sample shots were taken in Program mode at ISO

200, the lowest setting on the Pentax, under two 60 watt

equivalent output compact fluorescent bulbs which is very low

light for any camera and the images as far as I could see don't

have any noise issues in any channel.

 

As far as I'm concerned once an image is captured digitally it's

all software driven. There's no way around it. What software

engineers do with sensor data is anyone's guess even with

scanners and I would go so far to include RAW as well. Using

color temp correction filters seems too cumbersome and overkill

since software is always going to be part of the process.

 

I posted that demo mainly to show how to trick the incamera

software into getting what you want. Not saying what's shown is

your desired result, but the control is available without using

filters.

 

After spending a lot of play time with the menus and controls on

my first DSLR, I have to say the engineering built into these

camera's is freak'n amazing. I really couldn't tell you about

spectrum of light purity and its affect on the sensors RGB

balance, I'm just amazed at the controls to change it within the

camera.

I don't see whether you're using a tripod for these "low light room shots," I'd suspect not. If not, I think the value of a tungsten filter is marginal. They're fairly dark and would likely push you either into too-long exposures or underexposure, even at high ISO, either of which IMO is worse than just noise on the blue channel.

 

'Tungsten' color temp varies all over the place. A bright well-lit room will balance just fine in software. Lights on a dimmer switch (bane of wedding photographers) will go absurdly red. "Correct" white balance will look very blotchy, empty and fake; best not to correct all the way and leave some warmth there. If even that is too obvious, there may not be a better answer than trying a B&W rendering. Unfortunately this is where the tungsten filter would do the most good but would surely kill your shutter speed. For critical stuff (weddings) most will put a full CTO or stronger filter on a flash, and raise the light levels to where it'll balance OK in software. would be interesting to try a tungsten filter here, if there's surplus flash power.

 

On a tripod it's a different story. Now you can open up the ISO to 100, bringing your shadow detail back. Throw on a 80A and you can bring back maybe another stop worth from the deep blues, visible in pixel peeping and big prints but it's more 'optimization' than 'rescuing.' I still own an 80A for that reason though it doesn't see much use. I don't have any test shots up to browse, been meaning to do that. Most relevant for architecturals, night cityscape enthusiasts, long exposure stuff.

Andy,

 

If you're referring to the sample shots, no I wasn't using a tripod.

They were shot at f/4.0, 1/10th sec. The exposure could be made

much brighter if I'ld gone manual instead of Program mode and

used a shorter focal length. I did them quickly just for posting. If

set to manual using longer exposure but not by much I could still

get decently sharp images because of Pentax's built in IS.

 

About correcting for odd color temp casts, I was thinking

someone could come up with a WB colored reference chart for

tungsten and other off color spectrum lighting that tricks the

camera's manual WB algorithms similar to what I posted.

 

I've been playing around with using pastel colored felt I picked up

at a craft shop which gave some interesting color gel effects to

certain captured scenes. However, it'ld probably require each

camera to have their own custom color chart sinse each camera

manufacturer's manual WB algorithms seems to give different

results.

Wow intresting question.

 

My guess is that if you slap it on a tripod and increase the exposure time so that you don't have to increase your ISO, your pics will be better because the amount of light the red and green sensors get will be equivelent.

 

But if you're shooting action, or don't have a tripod ready, the increase in noise from an increased ISO will be greater than from the white balance.

 

Also, I believe that there is a "digital reciprocity effect" that creates a small amount of noise in extremely long exposures. If your shooting outdoors at night and your exposure is already at, say 2 seconds, that stop/stop and half the filter will rob you of could make a big difference.