how to count 3CCDs pixels ?

[Mark Fan]

I always see on the web says , for example, canon GL2 , has 480.000k on a single CCD , then says gross pixels is 680.000. shouldn't it be 3*480.000 ??

[Aaron Shaw]

The pixels, I believe, are directly laid upon each other so there is no resolution gain from having 3CCDs alone. It's like looking at a color image in Photoshop, you have a green, blue and red channel that are all layered upon each other to create the final picture.

[Glenn Chan]

The light entering the camera is split by a prism into red green and blue. There is a CCD for each color. If you want to count up the pixels, then it should be 3X the # of pixels on each CCD.

The optimal # of pixels seems to be slightly above 720X480 for each CCD (the extra pixels are needed for things like image stabilization). All the expensive DV 3CCD cameras seem to be designed that way. This means there is one CCD pixel for each DV pixel (DV dimensions are 720X480).

The Canon GL2 uses pixel shifting.
(I'm not sure about the information below)
I believe it means that the CCDs are shifted slightly apart so each CCD pixel do not overlap each other. This allows the camera to gain extra resolution for still photos. The Canon website advertises that the resolution is equivalent to non-pixel shifting cameras with 680k pixels.

One problem with pixel shifting may be that there are false colors on fine detail.

For video, 3X410k pixel shifted CCDs do not seem to be the ideal configuration. What's probably happening is that the engineers are using this configuration so the camera can take better still images. The GL2 should be roughly 0.680 megapixels in actual resolution, versus 0.3456 had it been designed for optimal video resolution.

If you are interested the images that the GL2 can create, check out the following Japanese site with frame grabs from many cameras:
http://babelfish.altavista.com/babel....html&lp=ja_en
http://babelfish.altavista.com/babel....html&lp=ja_en
*Those links might be a bit busted. Babelfish is for translating the Japanese.

[Aaron Shaw]

Right, but aren't these then layered directly on top of each other? The point being that you don't gain resolution from going 3CCD just better color rendition..... ?

[Jesse Bekas]

In video applications you will not gain resolution becuase of the three sensors, only color accuracy. In still applications, pixel shifted CCDs will give you more resolution, but it still is usually not comparable to a single MP sensor.

[Jeff Donald]

The CCD's are not stacked. Your thinking of the Foveon design. Interesting concept, but the chip isn't getting wide acceptance from the manufactures. Sigma is the only company building cameras based on the Foveon design.

[Gints Klimanis]

Polaroid is selling their x530 camera for about $400 .It uses the Foveon X3 sensor . I Googled for a few reviews, but I didn't read about a detailed study of the X3 sensor performance in the x530.
Other Foveon sensors are in various high end cameras. The Polaroid x530 is a good design win for Foveon, but most of the manufacturers are simply not using the design. Foveon is right across the street from me, but the office space looks nearly abandoned. I like their idea, but perhaps the part is practically unmanufacturable.

http://reviews-zdnet.com.com/Polaroid_x530/4507-6501_16-30734212.html?tag=tab

http://reviews.designtechnica.com/review1619_specs8104.html

[Jeff Donald]

I know most of the reviews commented on the higher noise levels than than CMOS or CCD sensors. Unfortunately Sigma was the only major manufacture to use the chip in dSLR's.

[John Richards]

Some thoughts I had on this subject -
A 1 megapixel consumer camera would use 690,000 pixels for video the remainder being used for D.I.S.
Since each pixel on a CCD can only determine a digital B/W level a Bayer filter is applied before the ccd which effectively splits a group of 4 pixels into R,G,G,B (I believe that more information is carried on the G channel so there are 2) so there are actually 172,500 R pixels
172,500 B pixels and
345,000 G pixels
A 3ccd camera would have its full quota (370,000) of each color plus a further 370,000 G by electronic G shift.
From a purely pixel count it appears to me that a 370,000 3 ccd device is roughly equivalent to a 2 megapixel single ccd.
This is where the fun starts - pixel size has a marked effect on s/n ratios and the digital imaging algorithms can change the appearance and characteristics of the raw image off the ccd.
Please note that these meanderings could be completely wrong
so feel free to shoot me down and can someone expand the information on how the pixels recorded are modified to the D.V frame of 720 x 480 which is about 350000 pixels.
I have just reached the conclusion that the numbers don't matter- it's the image that counts.
It's 1:30am, my brain is hurting so to coin a phrase-
Goodnight Seattle

[Jesse Bekas]

I don't know the specifics, but this is what I think is going on. Light hits the pixels of a CCD, (some used for video, some more for stills and stabilization on lower grade cams), the CCD turns each pixels repsonse into a charge and sends the charge information to a processor. The processor reads the info and decides how toapply white balance, effects, etc...and then down-reses the image (size and color) to fit in the DV25 bandwith. Similarly to a digital cam reading raw data and coming out with a JPEG.

[Mark Fan]

<<<-- Originally posted by Jesse Bekas : In video applications you will not gain resolution becuase of the three sensors, only color accuracy. In still applications, pixel shifted CCDs will give you more resolution, but it still is usually not comparable to a single MP sensor. -->>>


Ji Your, what is a MP sensor ?

[Jeff Donald]

Megapixel (million pixels). Pixel = Picture Element

[Mark Fan]

hey Your, thanks for your reply, but i just find out what exactly the pix shift is about , "A further refinement called pixel shift sharpens the image by emphasizing the green component, which most determines picture sharpness. " quot from,http://products.consumerguide.com/cp/electronics/review/index.cfm/id/20148



<<<-- Originally posted by Glenn Chan : The light entering the camera is split by a prism into red green and blue. There is a CCD for each color. If you want to count up the pixels, then it should be 3X the # of pixels on each CCD.

The optimal # of pixels seems to be slightly above 720X480 for each CCD (the extra pixels are needed for things like image stabilization). All the expensive DV 3CCD cameras seem to be designed that way. This means there is one CCD pixel for each DV pixel (DV dimensions are 720X480).

The Canon GL2 uses pixel shifting.
(I'm not sure about the information below)
I believe it means that the CCDs are shifted slightly apart so each CCD pixel do not overlap each other. This allows the camera to gain extra resolution for still photos. The Canon website advertises that the resolution is equivalent to non-pixel shifting cameras with 680k pixels.

One problem with pixel shifting may be that there are false colors on fine detail.

For video, 3X410k pixel shifted CCDs do not seem to be the ideal configuration. What's probably happening is that the engineers are using this configuration so the camera can take better still images. The GL2 should be roughly 0.680 megapixels in actual resolution, versus 0.3456 had it been designed for optimal video resolution.

If you are interested the images that the GL2 can create, check out the following Japanese site with frame grabs from many cameras:
http://babelfish.altavista.com/babel....html&lp=ja_en
http://babelfish.altavista.com/babel....html&lp=ja_en
*Those links might be a bit busted. Babelfish is for translating the Japanese. -->>>

[Mark Fan]

Hi Jeff .so , based on what Your said in his post, GL2 has o.68 MP,thus, I assume it shouldn't be looking any better than on JVC GR-DV900 when it's plaied on PC, which has 1.33 MP( not sure how many effective, but should be above 0.68, am I getting this right?)


<<<-- Originally posted by Jeff Donald : Megapixel (million pixels). Pixel = Picture Element -->>>