White Balance vs 80A or 85B

[Alexis Vanier]

Hi,

 

I've been asking myself a lot about white balance recently, especially in light of some information I've read regarding per channel noise on different cameras and formats (although I must admit I'm still very much confused about it).

 

As far as I know white balance is achieved by adding or subtracting gain from individual colour channels and applying tint (rolling chroma phase?) either electronically before the sensor's signal goes to AD conversion or numerically either in camera or post.

 

My guess is a given sensor or system must have an optimal response to a certain colour temperature. My question is then, omitting practical details in regard to lighting, would be : " Could one achieve a reduction in apparent noise by using optical filters (80A/85B) to circumvent the gain involved in the white balance process. Essentially rating your sensor as tungsten or daylight and compensate like you would with film. "

 

I was actually thinking about that with the RED ONE camera in mind, where one could shoot try and limit the use of white balance as much as possible.

 

Please share your knowledge as I'm surfing on a lot of half-knowledge. While what I'm proposing might not seem practical, I'm pretty sure the knowledge could serve some specific applications. I have blue/red screen compositing and extreme light temperature situations come to my mind.

 

Alexis.

[Walter Graff]

Use the white balance button. Your life will be easy. Mostly you really don't have do do anything other than leave the camera on preset white balance based on indoor or outdoor light conditions. And the filter wheel already does what you are porpossing wiht filters. As you get good with white balance after experience you will learn such things as white balancing on grass, sky, slightly red or blue backgrounds, etc to get very nice white balances. But overall, don't think to much about it. A camera needs to know what color temperature you are working under. Aim the camera at something white hit the button and the camera (in simplest terms) compares red to green (the reference channel) and blue to green and equals them out so white is white and you are good to go. You will not be adding noise to the picture with white balance. And the noise you may have been reading about is not necessarily something you'd see on an ordinary day. Noise is mostly created by imagers sensitivity being pushed so you can not 'hide' it with the filters you propose on a normal day. Some cameras like the prosumer cameras made by Panasonic (DVX/HVX) have a great deal of noticeable noise on the blue channel in 'normal' conditions when shooting blacks but that noise can be reduced by reducing sensitivity, or using filters in post. In most all cases of normal video shooting, noise will not break your bank unless you are shooting in low light conditions as you mention. Blue and red screen will normally not be a problem with most every pro camera >$15k and most all prosumer cameras <$6k

[Walter Graff]

Whoops, my whole post did not transfer when I pasted. There are folks who have taken some of the noisiest prosumer cameras (like the DVX/HVX) who have used preset conditions in 32k and used filters as opposed to white balance to get what they claim are better quality, noise free pictures. But I'd rather buy a camera that does what it designed to do (noise free - >$6k cameras) than have to go through the process of Rube Goldberg just to get a useable picture on some of these poorly designed prosumer cameras. There are some pro cameras tha tallow you to make internal adjustments to your base white blance settings to get the most out of a camera while reducing noise in the blue channel (normally the noisiest). I would imagine you would have no such issues with a RED camera based on its design and what users are saying. You mention RED. Is that the camera you are talking specifically about or are you referring to a range of cameras?

Edited December 5, 2007 by WALTER GRAFF

[David Mullen ASC]

In theory, optical correction (filters) would be less noisy than rebalancing the RGB levels to get white, but reality is that most basic white balancing in post / in camera is not so extreme that noise is a problem. Usually you're trying to get some green out of fluorescent lighting or correct a blue cast out of something shot in the shade, etc.

 

As the other post mentioned, most video cameras already process the signal off of the chips so that RGB levels are at "0" each for 3200K lighting, hence why there is an internal color filter wheel in most pro cameras for shooting under higher color temps, in theory because orange filters of different strengths cut less exposure than blue filters. They could have skewed the camera to make RGB levels at "0" for 5500K light, but then you'd have half-blue and full-blue filters, and the 80A filter that corrects 5500K to 3200K loses two-stops of light. The 85B filter that corrects 3200K to 5500K loses only 2/3's of a stop in comparison.

 

But this internal processing to get RGB levels to be zeroed out in 3200K lighting is one reason why the blue channel is noisiest in video cameras. Using blue filters isn't really going to help much at that point. Minus-gain helps a little.

 

Cameras that don't muck around with the signal internally show an overall preference for bluer color temps, 4800 to 5000K, like the RED, the Genesis, the Phantom, etc. or a blue-green color, like the Viper. The Viper is a 3-CCD camera and the reason for the green cast is that they simply don't do any processing to get rid of the naturally higher green signal coming off of that chip. The Sony F23 must be doing some minimal processing or filtering to avoid the green cast. The RED, Genesis, Phantom, etc. are single-sensor cameras.

[Michael Nash]

Test your camera. Shoot scenes of different color temperatures under different white balances, and in post try to correct them or convert them to the "opposite" color temperature (daylight/tungsten). That will show you the visible signal-to-noise ratio in the color channels, and give you an idea of how far you can push them before the noise becomes objectionable.

 

There's no right or wrong approach, just a matter of what works for you. Some film DP's are very particular about color temperature and camera filtration to optimize the densities in each of the film's color layers, yet others embrace the inherent qualities of the stock and just shoot. Video's the same way; you develop practical ways to optimize the image for your purposes.

 

For what it's worth, colored filters cancel the complementary wavelength evenly across the whole tonal range, while white balancing sets the white level of each color channel but keeps the same reference black for all channels. So technically, you would end up with different signal/noise characteristics in the shadows, but in most situations you can't even see the difference in color balance let alone noise.

[Walter Graff]

I'll add that just like your eye that sees blue the least (20 and 18 times more red and green cones per blue cone because of evolution and how little blue was food or anything else we needed for survival) blue is the least part of the luminance mix in YUV (video mimics the human eye). And in terms of the final luminance component you get the least and hence the most luminance noise. But in the mix this often shows up far less than it would had you isolated blue alone. But because of the nature of electronic registration and camera design, your best to let the camera do the work. And now adays you can do great stuff in post including completely re-white balancing like you are doing it in camera.

[Alexis Vanier]

Thank you very much for your answers Mr. Mullen, Graff, and Nash. Very much appreciated.

 

To Mr. Graff.

I specifically did have the RED in mind when wondering how to work with the filters, but not because I thought channel noise following white balance could be a problem (not even in extreme scenarios really). I was rather asking the question just for the sake of expanding my knowledge.

 

Regarding the way the human light react to blue light. Isn't the fact that we have fewer blue sensitive cones offset by the fact that blue light has higher power due to it's shorter wavelength?

[Bruce Greene]

Hi,

 

I've been asking myself a lot about white balance recently, especially in light of some information I've read regarding per channel noise on different cameras and formats (although I must admit I'm still very much confused about it).

 

As far as I know white balance is achieved by adding or subtracting gain from individual colour channels and applying tint (rolling chroma phase?) either electronically before the sensor's signal goes to AD conversion or numerically either in camera or post.

 

My guess is a given sensor or system must have an optimal response to a certain colour temperature. My question is then, omitting practical details in regard to lighting, would be : " Could one achieve a reduction in apparent noise by using optical filters (80A/85B) to circumvent the gain involved in the white balance process. Essentially rating your sensor as tungsten or daylight and compensate like you would with film. "

 

I was actually thinking about that with the RED ONE camera in mind, where one could shoot try and limit the use of white balance as much as possible.

 

Please share your knowledge as I'm surfing on a lot of half-knowledge. While what I'm proposing might not seem practical, I'm pretty sure the knowledge could serve some specific applications. I have blue/red screen compositing and extreme light temperature situations come to my mind.

 

Alexis.

 

I recently shot a small feature on my Varicam that had 60% day exteriors. I wanted to shoot on preset white balance to avoid the influence of the set colors on a white card white balance. This was for consistency.

 

Upon testing, I decided that the "D" position filter was too orange and the "C" position filter was not orange enough. I had been instructed in the past that the camera is "naturally" balanced to 3200k and that I should use CTO type filters (such as the ones in the filter wheel) for the best image.

 

Dissatisfied with the wheel filters I tried an 85 filter instead. It gave the same result as the position "c" filter on the wheel.

 

Next I tried setting the camera to "daylight" white balance. This is a preset white balance electronically that is balanced for 5600k and does not use the orange filters. I really liked this white balance in day exteriors for all around shooting. Not too blue, not to orange, just right.

 

But I had nagging in the back of my mind that the image quality would suffer using this mode as it was too big a color correction of the camera to make with electronics alone.

 

So I conducted some tests using a waveform monitor and pushing the camera to it's extremes of dynamic range. To my eye, the "daylight" balance proved to be less, and not more noisy. More importantly, I discovered that at the extreme end of the dynamic range of the camera, that the camera is actually "naturally" balanced closer to 5600k than 3200k.

 

In the end I shot the picture using the daylight electronic balance and did not use any filtration for the day exteriors and was very pleased by the results.

 

And for me, working with the preset white balance gives me a consistent baseline for performing manual white balance corrections when desired for creative reasons. Because these changes are based on the preset white balance, the menu numbers for the RGB gain controls are always relative to zero, rather than an arbitrary "auto white balance" starting point.

 

So I guess my point here is to not assume that the camera is optimally balanced to 3200k, even if the filter wheel suggests that it must be...

 

-bruce

[Walter Graff]

Thank you very much for your answers Mr. Mullen, Graff, and Nash. Very much appreciated.

 

To Mr. Graff.

I specifically did have the RED in mind when wondering how to work with the filters, but not because I thought channel noise following white balance could be a problem (not even in extreme scenarios really). I was rather asking the question just for the sake of expanding my knowledge.

 

Regarding the way the human light react to blue light. Isn't the fact that we have fewer blue sensitive cones offset by the fact that blue light has higher power due to it's shorter wavelength?

 

Thanks. I couldn't tell you enough then about RED since I don't know other than RED records all the information so probably suffers less from the problems of blue noise in ordinary YUV camera matrix's. As for blue, you could be right but if that was the case then we'd have a correspondingly similar relationship of green and red cones. The folks who study our origins, and in my study of color psychology in school I was taught that our development of seeing color was founded in our needs to survive. Since no food, or anything dangerous was represented by blue, our eyes developed a lesser need to see it. Women see more shades of color than men because as is surmised, they were more the collectors so needed to differentiate various types of vegatable-type food sources and insects and hence they have the ability at tetrachromacy where men are only trichromats. It was also discovered that people of equatorial ancestry have even less ability at discriminating blue and could have foundation in something related to your thoughts but probably has more to do with environmental factors. The blue cones are generally spread out much further from the fovea which is partially due to the longer wavelength of blue light. But more importantly the way nerves transmit blue light to the brain appears to have differences from red and green nerve connections and shows the entire eye/brain connection is less blue oriented than red and blue. We simply didn't need to see blue as much as we did red and then green. Of course today that can be a problem when you are trying to read a blue sign at night but then again, you will remember words better written on paper with blue ink rather than black ink so the deficiencies in seeing blue are not all that bad.

[Phil Rhodes]

I'm not sure that Red ought to be any less subject to this than anything else. It's still a silicon photocouple picking up the light; it's still less sensitive in the blue. Recording it raw just means that you have to do the correction later rather than in camera.

 

And interestingly enough, with Red, that means you are doing the correction after the (very very heavy) compression has been applied. Hrn...

 

Phil

[Mitch Gross]

I recently shot a small feature on my Varicam that had 60% day exteriors. I wanted to shoot on preset white balance to avoid the influence of the set colors on a white card white balance. This was for consistency.

 

Upon testing, I decided that the "D" position filter was too orange and the "C" position filter was not orange enough. I had been instructed in the past that the camera is "naturally" balanced to 3200k and that I should use CTO type filters (such as the ones in the filter wheel) for the best image.

 

Dissatisfied with the wheel filters I tried an 85 filter instead. It gave the same result as the position "c" filter on the wheel.

 

Next I tried setting the camera to "daylight" white balance. This is a preset white balance electronically that is balanced for 5600k and does not use the orange filters. I really liked this white balance in day exteriors for all around shooting. Not too blue, not to orange, just right.

 

But I had nagging in the back of my mind that the image quality would suffer using this mode as it was too big a color correction of the camera to make with electronics alone.

 

So I conducted some tests using a waveform monitor and pushing the camera to it's extremes of dynamic range. To my eye, the "daylight" balance proved to be less, and not more noisy. More importantly, I discovered that at the extreme end of the dynamic range of the camera, that the camera is actually "naturally" balanced closer to 5600k than 3200k.

 

In the end I shot the picture using the daylight electronic balance and did not use any filtration for the day exteriors and was very pleased by the results.

 

And for me, working with the preset white balance gives me a consistent baseline for performing manual white balance corrections when desired for creative reasons. Because these changes are based on the preset white balance, the menu numbers for the RGB gain controls are always relative to zero, rather than an arbitrary "auto white balance" starting point.

 

So I guess my point here is to not assume that the camera is optimally balanced to 3200k, even if the filter wheel suggests that it must be...

 

-bruce

This is exactly what I was saying in the other thread.

[Walter Graff]

So I guess my point here is to not assume that the camera is optimally balanced to 3200k, even if the filter wheel suggests that it must be...

Keep in mind to that internal adjustments can and will make a difference in how a camera sees white balance too. If you start throwing things like gammas and pedistals around, you will find you start getting non standard looks then when you have a camera that is properly adjusted on a chart by an engineer when you hit the white balance button. Also keep in mind that the Kelvin number in a camera viewfinder is not indicative of a three channel color meter (since a camera is a subtractive process) so is not something you should look at as a gold standard for true color temp but only reference.

[Bruce Greene]

So I guess my point here is to not assume that the camera is optimally balanced to 3200k, even if the filter wheel suggests that it must be...

Keep in mind to that internal adjustments can and will make a difference in how a camera sees white balance too. If you start throwing things like gammas and pedestals around, you will find you start getting non standard looks then when you have a camera that is properly adjusted on a chart by an engineer when you hit the white balance button. Also keep in mind that the Kelvin number in a camera viewfinder is not indicative of a three channel color meter (since a camera is a subtractive process) so is not something you should look at as a gold standard for true color temp but only reference.

 

Walter,

 

From what I can see on the waveform, white balancing a camera involves only adjusting the RGB gain values until R=G=B. It does not change gamma adjustments for any of the colors.

 

Also, it's not important what kelvin number shows up in the viewfinder, though it's a darn good reminder of where the camera is set when using a B&W viewfinder!

 

For my tests I illuminated a white card to the proper color by using my 3 color meter and reading the color off the card itself. I adjusted the color of the light with small amounts of colored gels until the color was close enough for my tests.

 

I do have a couple of questions for you though:

 

#1. Are you posting because you doubt the validity of my tests? If so, in what way?

 

#2. What do you mean about "throwing things like gammas and pedestals around"? I didn't mention anything about gamma adjustments or pedestals.

 

#3. You wrote: "you will find you start getting non standard looks then when you have a camera that is properly adjusted on a chart by an engineer when you hit the white balance button". I'm not sure exactly what you mean by this, could you help me out here? What is a "Standard" look, and why would I want to shoot with it?

 

I guess what this sounds like to me is that you are saying that some mystical "engineer" is the only one that should be setting up and evaluating a camera. Of course my tests assume that the camera engineering is properly set up. But this should just bring the camera into compliance with the manufacturer's specs. All the adjustments in the user accessible menus are designed for the user to make the camera work at it's best for the project without screwing up the camera and allowing one to reset it to "factory" standards at any time. I guess it just sounds to me that you think only "engineers" should be testing and adjusting camera set-ups. I believe cinematographers should have the knowledge to adjust their tools as necessary to perform as needed, without needing an engineer at all times. (please don't get me wrong here, I am not saying there is no need for engineers or even DITs...) If you really mean "don't open the camera cover and start turning screws", yeah, I agree that that could really screw things up.

 

-bruce

[Walter Graff]

Bruce,

 

You did not hear me correctly.

 

From what I can see on the waveform, white balancing a camera involves only adjusting the RGB gain values until R=G=B. It does not change gamma adjustments for any of the colors.

If your peds are not properly set, it has an effect on white balance. If your gains are altered it has an effect. My point is not that white balancing doesn't change things but other adjustments before white balancing can have an effect depending on what those changes are.

 

For my tests I illuminated a white card to the proper color by using my 3 color meter and reading the color off the card itself. I adjusted the color of the light with small amounts of colored gels until the color was close enough for my tests.

 

I am not referring to your method, just that that kelvin indicator in a camera is not obtained the same way a three color color meter reads color. It's not something one should use as anything more than a reference if you are trying to figure out exact CC.

 

I do have a couple of questions for you though:

 

#1. Are you posting because you doubt the validity of my tests? If so, in what way?

 

My posts have nothing to do with you. Just adding information as a whole.

 

#2. What do you mean about "throwing things like gammas and pedestals around"? I didn't mention anything about gamma adjustments or pedestals.

 

And hence why what I said has nothing to do with you. Just talking about white balance in general. Many folks tweak cameras with controls meant to be wet up properly on a bench and that can affect normal operation of a camera with such things as white balance.

 

#3. You wrote: "you will find you start getting non standard looks then when you have a camera that is properly adjusted on a chart by an engineer when you hit the white balance button". I'm not sure exactly what you mean by this, could you help me out here? What is a "Standard" look, and why would I want to shoot with it?

I guess what this sounds like to me is that you are saying that some mystical "engineer" is the only one that should be setting up and evaluating a camera. Of course my tests assume that the camera engineering is properly set up.

 

Just saying that certain controls in a camera have a great deal to do with affecting things that many don't take into consideration. Yes, in many ways I do not believe most folks should not be playing with the inside of the engineering menus. Did you know that all cameras need to be properly registered, even the latest and greatest and that I believe only qualified engineers should do that? That?s why they make getting into them harder than the operators menus. I personally have all the cameras I purchased properly registered by a qualified engineer (in my case Roger Macie). Once I have a proper standard for that particular camera and the original settings saved, I can the always make nay adjustments I want. And it's good to know I can always go back to my bench settings as a standard at the end of the day or when I start a new production and then see where I want to go from there. That does not mean that folks who have learned the ins and outs of many controls can't nor shouldn't use them, just making general comments to maybe save me receiving a 2 am phone call from folks in the field that want to know why they can't make red anymore and then tell me they played around with something called matrix but it didn?t look like it was doing anything when they did. I don't know you and my responses were not directed at you. I normally start out a post with a persons name if I am talking to them. I don't believe I have addressed you in this thread directly. I've simply added information about white balance for anyone to read.

 

 

But this should just bring the camera into compliance with the manufacturer's specs.

 

Did you know that no Sony camera comes off the assembly line properly tuned per camera? Hence why an initial adjustment guarantees you start with a proper base line. So in Sony's case there is really no factory standard, only a happy medium. One is at a disadvantage not having an engineer properly tune a camera to that cameras standards before using it.

 

I guess it just sounds to me that you think only "engineers" should be testing and adjusting camera set-ups.

 

I think without a proper first time set up you are never getting the most you can from a camera. I know I can see subtantial difernce in my cameras whenI purchase them, have them properly adjusted, then go fomr there. And I bring my cameras in to be retuned on occation as while solid state has made drift and the like mostly a non issue, cameras do wear electronically over time and often need parts replaced. As for users and adjustments. As long as one knows what they do then I don?t see a problem with it. Problem is I see many who don't know what they do fooling around because boards like these have qualified folks talking about adjustments so everyone thinks they too can simply go to -33 on detail and thinks that is what is a good setting for their particular camera.

 

Hope I spoke clearly.