Noisy Blue

D

DanHD

Well-known member
I read this over on the RedUser boards

David Mullen, ASC - " I don't think it's reasonable to expect the blue channel to be completely noiseless in 3200K -- just as the blue layer in 500T film stock is the grainiest layer."

I know my digital sensors, especially ones using Bayer filters, have noise in blue

but why would film stock have blue grain when shooting tungsten?

Is it as simply as the fact that the blue layer has to be more sensitive thus being grainier?

how do DP's avoid this, in film and digital?

Is it neccessary to balance so that tungsten looks "white"??
It seems like tungsten looks amber in tone even to the human eye..
 
DanHD said:
I read this over on the RedUser boards

David Mullen, ASC - " I don't think it's reasonable to expect the blue channel to be completely noiseless in 3200K -- just as the blue layer in 500T film stock is the grainiest layer."

I know my digital sensors, especially ones using Bayer filters, have noise in blue

but why would film stock have blue grain when shooting tungsten?
Click to expand...

I would imagine it is the compensation for 'low blue light level' in Tungsten, and so, put more silver into that layer to acquire the image, and 'balance' the weighting of how RGB components are recorded...

Since most of my photography has been still and B&W... I would use 'daylight' color film to give a scene that nice warm yellow glow when the lights were tungsten... that or use a flash... which is essentially 'daylight'... so never bothered with using Tungsten balanced film...
 
DanHD said:
\
Is it neccessary to balance so that tungsten looks "white"??
It seems like tungsten looks amber in tone even to the human eye..
Click to expand...

well spotted Dan - and sunsets look red, and rocky mountain winter skies look blue.

that is the problem with the "correct WB" mantra
 
pwyll said:
well spotted Dan - and sunsets look red, and rocky mountain winter skies look blue.

that is the problem with the "correct WB" mantra
Click to expand...

Well... sure... giving a nice homy fireside living room a cold blue cast, is not usually read well... The problem here would be 'how much'... If one starts from the 'neutralized' version, dial in a bit of yellow, given just a hint of warm glow... people will read that as a comfortable indoor shot...

The problem is two fold, the human eye compensates for the color cast, and added with some artificial intellegence, the human observer 'knows' a sheet of paper is 'white', and the brain filters that information to the point where one no longer is really aware of the yellow cast, and only momentarily aware when some other color temperature light is flicked on.

The 'white balance' attempts to mimic this behavior.... then there's 'auto white balance'... along with most of the other 'autos' should be avoided... especially if one wants to change things in post...

From the various BTS special features, when 'raw' footage is shown of a scene, quite a bit of the 'color' seen in the final film is not present. The idea would be to have the most information, in the most 'neutral' form on film and work from there for a look.
 
Well the problem with shooting "nuetral" in a compressed lossy format like DVCPRO HD is that you can't get that information back, and you do not have the freedom to push and pull the image in post.; just creates more noise
 
and since CMOS sensors have less surface area sensitive to blue, it might actually be preferable to WB off a slightly blue card and create a slightly warmer look IN camera.....at least it gave my buddy better results on his Red shoot. He always shoots daylight balance......but he has access to HMIs
 
Every HD format is compressed and lossy. At least DVCPRO-HD is 4:2:2 and you can push and pull it much better in post than you can with the other formats.

There's nothing about a CMOS sensor that makes it inherently less surface area for blue; I think you're confusing Bayer pattern with CMOS. You can use a Bayer pattern with either CMOS or CCD. But a 3-CMOS system (like an EX1) is going to have as much blue surface area as a comparable 3-CCD system.
 
I'm sorry Barry I should have been more specific...

I was talking about single sensor CMOS vs. 3CCD....and bayer vs. 3chip color

I get better results creating a color cast in the camera than pushing the colors in post, I find less noise in the midtones
 
DanHD said:
Well the problem with shooting "nuetral" in a compressed lossy format like DVCPRO HD is that you can't get that information back, and you do not have the freedom to push and pull the image in post.; just creates more noise
Click to expand...


Yeah, I was thinking of the BTS clips I've seen on a film film shoots...
 
Thanks barry.

I shot all my interiors with tungsten lights today, and was reluctant to do so because of these 'noise fears'

is my fear unfounded? I used the hpx170, which you of course know is a 3 chip 1/3" inch cam....

with three equally sized chips sensitive to the three colors, why would there be more noise in tungsten?

I just dont get it, and I'm getting paranoid over it

Is it simply because tungsten light puts out far more red than it does blue, in which case the camera gains up the blue channel?



if so, than why isn't the red channel noisy in overcast light, which is predominantly blue? (when color balanced of course)
 
The "horrible blue noise" issue isn't a problem with CCD, only with CMOS. And even with CMOS, it seems to affect Bayer patterns more. Why? You'd have to ask them, but I would suspect it's because more weight is given to the red, and less to the blue, in a Bayer pattern than it is in the typical weighting of a video image. Meaning, typically in video you derive about 59% of the luminance value for the image from the green, and about 30% from the blue, and about 11% from the red. But in a Bayer pattern, you get 50% of your light allocated to green, 25% to blue, and 25% to red. Is that a factor? I don't know, seems it might have something to do with it.

Secondly, it's possible that CMOS just doesn't see blue as well as CCD can. I don't know that, but it's a guess.

If a sensor is balanced for daylight, that typically means that the filtration over the pixels, or the sensitivity of the particular chip, is designed to respond neutrally under daylight. There would be a weaker blue filter in place over the blue, for example, because there'd be more inherent blue in the light coming into the chip.

I think you're worrying a lot about things that aren't really an issue. CCD works fine under either light, CMOS on the Red appears to really prefer daylight.

As for why the red channel isn't noisy in overcast light, maybe it is, but the red is so insignificant in the overall scheme of things (accounting for only 11% of the total value) that it may not even have much of an impact.
 
You must log in or register to reply here.
Back
Top