noisy blue channel

DanHD said:
So of course, I shouldnt be afraid to shoot tungsten balanced.
Click to expand...


Dan, CCDs had been used under tungsten lights for years. It is fine. There is no need to be afraid.

Oh another thing is, is it really neccessary to manually white balance to make tungsten lights appear white? I mean, even the human eye sees tungsten as more amber in tone than daylight...they aren't completely equalized.
Click to expand...

If you want the light to look white you will need to white balance. If you want the amber tone then don't. But keep in mind that the camera will look at the light, remembering the last white balance. So if you were outside on a bright sunny day, 9500K and then look at the tungsten, it will be different than say having the last white balance at 3200K and looking at the tungsten. This is why your on set monitor should have been set up to color bars and then you will be able to see what color the light really is according to what the camera sees.


And when DPs shoot on tungsten balanced film, they have no manual white balance option, so do they usually allow lower temps to just appear orange?
Click to expand...

Or they use filters, or let it all wash in post.

I guess my point is, perhaps you DONT have to white balance for everything..perhaps some situations look good with some amber or blue cast to them

You eye doesnt balance out everything color temp situation completely...or else you wouldnt really see colored shades of white, right?
Click to expand...

Exactly.

Merry Christmas to all!

Jan
 
So, in actuality, a CMOS sensor has less pixels able to respond to blue than a full raster 3 CCD system?
 
No, a Bayer system has less pixels etc... the # of pixels available is determined by whether they're employing a three-chip prism-split system, or whether they're using a single chip with a Bayer pattern, or a single chip with an RGB stripe.

And, of course, any discussion of "more" or "fewer" assumes like-for-like; a single-chip Bayer Dalsa Origin is going to have far more blue or green pixels than a three-chip HPX170.
 
Well, say all things are equivalent, and the resolution is the same...the CCD would have more pixels available to be allocated to green right? compared to a Bayer array CMOS....in the bayer pattern 1/4 the pixels pick up green light and in the 3 CCD 1/3 of the pixels do....so theroretically there should be less noise in the CCD when balanced for, say 2800k
 
Depends on what you mean by equivalent. If you are talking about a 1920x1080 chip, and you use three of them in a 3-chip camera, then obviously you'd have 1920x1080 green (and 1920x1080 blue and 1920x1080 red). If you instead tried to make that one chip do the work of three by putting a bayer pattern over it, you'd end up with 960x1080 green, 960x540 blue, and 960x540 red.

So typically the choice isn't made between like-for-like; typically you'll have an oversized/overpixeled single chip vs. three smaller chips. Typically a Bayer pattern chip will exhibit total resolution of around 70% of its pixel count in luminance, and about 50% in chrominance. So, for a Canon HV20 for example, it uses a 3-megapixel chip which delivers about 2 megapixels worth of luminance.
 
With a threeCCD full raster hpx3000 for example, dont you end up with 1920x1080 chromanance? (if totally uncompressed of course)

I mean, if you took a 1920x1080 single chip and put a bayer filter on it, you'd have less pixels able to pick up blue, thus youd need more gain


thats what Im aiming at here...jan says there wont be any bothersome noise with tungsten on the hpx, yet the Red cam has this problem...

Are bayer patterns just not as good?
 
DanHD said:
but a three CCD system can put out just as much chromanance resolution as lumanance right?
Click to expand...
Yes, but practically no video signal carries that. You can use a data system to record 4:4:4 but virtually all video systems record a max of 4:2:2.

Yes, three chips are capable of full 4:4:4. One single chip with a Bayer pattern never could do 4:4:4 unless it was substantially downrezzed.
 
DanHD said:
With a threeCCD full raster hpx3000 for example, dont you end up with 1920x1080 chromanance? (if totally uncompressed of course)
Click to expand...
Yes, for internal processing, but how do you access it? On the HPX3700 you actually can get the full 1920x1080 chroma because you can use dual-link HD-SDI to get a full 1920x1080x4:4:4 signal.

I mean, if you took a 1920x1080 single chip and put a bayer filter on it, you'd have less pixels able to pick up blue, thus youd need more gain
Click to expand...
Well, not gain so much... I mean, a 1920x1080 single-chip system is going to have 960x540 blue resolution. But even then it's not full res because the blue pixels are distinct from the red and the green, there's no overlap. The 960x540 red or blue that you get in a system like that is not a full-frame 960x540, it's instead a crosshatched pattern made of 1/4-size pixels.

Even so, it's not like it matters all that much because you never get access to those red or blue pixels individually anyway in a video system; in video the frame's pixels are made up from a combination of the various individual color pixels, being about 59% green, 30% red, 11% blue. So a pixel on the sensor doesn't equate to a pixel on the frame.

(That's for video; for storing RAW obviously it's a totally different thing).

thats what Im aiming at here...jan says there wont be any bothersome noise with tungsten on the hpx, yet the Red cam has this problem...
Are bayer patterns just not as good?
Click to expand...
Are Bayer patterns not as good as a 3-chip system? No, of course not. Assuming chips of comparable size, three chips will always outperform one with a Bayer pattern stuck over it.

Historically all video manufacturers, and even today almost all video manufacturers, use three chips rather than one. Because three chips and a prism deliver better imagery than a single Bayer chip can. Manufacturers have typically offered a single Bayer-pattern chip in their cheaper models, and a three-chip prism-split system in their more-expensive models. I think all broadcast cameras currently offered are all three-chip.

However, there is a very compelling reason to go with a single chip - you cannot mount film lenses on a three-chip prism-split system. So for digital cinema purposes, especially someone like Panavision or Red, who wants to make a camera that can use film lenses, they had no alternative, they had to go with a single-chip system. For their purposes they could use a Bayer or an RGB Stripe system (Panavision uses RGB stripe). Red chose Bayer, for many excellent reasons, not the least of which is that for an equivalent-pixel-count chip they can get higher ultimate resolution out of their system than they could with an RGB stripe system. Panavision's happy with RGB stripe because they were aiming for a final 1920x1080 frame size; Red wants much bigger frame sizes.

A camera like the Red couldn't have been done in a practical fashion with a three-chip system, that kind of design pretty much demands a single-chip.

To get a comparable image from a Bayer system as opposed to three chips, you'd have to start with a much more densely packed single chip. That's another reason Red's chip has 12 million pixels on it (as opposed to, say, an F950R, where each chip has 2.2 million pixels, giving a net source of 6.4 million). Heck, even the little HV20 has three million pixels on its chip, as opposed to, say, an XLH1, which has 1.5 million pixels per chip; it's to overcome the resolution loss that occurs when demosaicing a Bayer pattern system. Typically you get about 70% of the total system resolution after demosaic; Red is claiming an effective 78% which is very, very good. It still means that their 4K chip is delivering an effective 3.2K of luminance resolution, and 2K of chroma res. That's just the way that works. The HV20 gets about 1920x1080 out of its chip when the 3-megapixel chip is demosaic'd, (3M x 70% = about 2M)

So if you want to use one chip, Red has the right idea - make it huge, pack it with pixels, and use a Bayer pattern. But if you wanted to make a camcorder at a specific imager size (say, a 1/2" chip) you'll get much better imagery out of a three-chip prism-split CCD block, than you would from a single 1/2"Bayer CCD. (or 3-CMOS vs. one single chip, take your pick).

The other thing that you can do with a single Bayer chip is cram in a bigger chip than you could otherwise have fit in there. A three-chip system, including the prism, takes up quite a bit of space, so where someone might fit three 1/3" chips, someone else might be able to cram in a single 1/2" Bayer chip, or maybe even a 2/3" Bayer chip. The larger your sensor, the shallower your DOF (not really, actually, but close enough to avoid getting into that lame and tired argument again). But the more expensive your glass usually is, too. So that's where the real debate would come in: which would you rather have, three 1/3" chips or a single 1/2" Bayer? That's the debate.

So - for ultimate color accuracy, nothing beats a properly-registered 3-chip system. For film lens compatibility and the shallowest DOF possible at any given price point, single-chip Bayer has plenty to recommend it as well.
 
Last edited:
Exactly - which reminds me, Dan, check out the Foveon. That's an idea for giving 4:4:4 color on a single-chip system. They essentially stack pixels in three-D, and use the depth of the silicon to register the different wavelengths of light, so blue (short wavelength) only penetrates the outermost layer, and red (longest wavelength) is the only one that gets through to the deepest layer... problem is, silicon is a pretty lousy color filter... but it's a unique approach to 4:4:4 in a single chip system. Some would say it's promising and ideal; Graeme of Red has no love for the Foveon approach and believes that as a practical matter Red gets better images out of a traditional Bayer pattern.
 
Last edited:
Im probably totally wrong here Barry, you know what you're talking about, I was just tought that
 
You must log in or register to reply here.
Back
Top