XL2 CCD block overview

[Chris Hurd]

Several threads are already branching off into separate discussions about the XL2's one-third inch CCD's and how those chips are doing native 16:9 video. I've written an article with graphics which should help explain it:

Canon XL2 CCD Block Overview.

Please discuss it here, for ease of navigation for folks looking for it later! Thanks,

[Steven-Marc Couchouron]

Thanks Chris for a very informative article which confirms what I had speculated.

Of course we'll have to judge by the image quality itself, but the fact that it uses less of the 1/3" CCD sensor doesn't seem like a good thing, resulting in narrower FOV, longer DOF and, possibly, less low-light sensitivity.

This is particurlaly true for the 4:3 mode.

[Richard Alvarez]

Heated argument going on here at the shop. The boss says the old way of 16x9 on the Xl1 was a simple mask off of the 4:3 ccd. I am pretty sure Canon used some sort of algorithm to get 16:9 rather than a simple mask. Please explain the difference in 16;9 methods from the old Xl1 to Xl2. He is unconvinced that higher pixel count in a smaller target area will result in greater definition overall.

Guess we'll have to wait for Adam Wilt to give us the lowdown?

[Rob Lohman]

Well your boss is basically right. It did you a "mask" and then
electronically stretched the footage back to full height.

On Chris' page you can clearly see how the new system increases
resolution. Look at the 4:3 rectangle in the middle. This "little
square" still contains *more* pixels than the old one does on the
XL1S. So the "new" way of doing 16:9 extends the range
horizontally by two columns (left and right).

The pixels in these columns get ADDED to the 4:3 image which
yields a higher resolution image.

However, both options do increase focal length on lenses a bit
due to the way Canon is using the CCD's. I'm not sure how DoF
will respond to this...

[Boyd Ostroff]

It's nice to see that they've optimized the design for 16:9 video and are using the full chip width. You'd think that 4:3 would suffer and the active area must be closer to a camera with 1/4" chips. Bu t this should send a clear message to all those people who question the need for 16:9, looks like Canon's vote has been cast; we know what their priority is.

With the native 16:9 and progressive scan the results should be terrific (do we know for sure that it's real progressive BTW?) Will be fascinating to see some user reports as to whether this design has some of the problems of other high pixel count camcorders, like increased vertical smear.

[Dan Uneken]

Rob,

The 10% decrease in image area will increase dof accordingly in my opinion.
Good for P+S.

What totally baffles me is why they have those non-firing pixels on the chips anyway. Or am I missing the point altogether?

[Steven-Marc Couchouron]

Because if they did use them in 4:3 mode, that would change the focal length of the lens when you switch modes, which wouldn't be very practical. At least that's my guess...

[Rob Lohman]

I can think of two reasons:

1) this is some standard sensor and it would've cost more to make one with those pieces chopped off

2) it is for a future extension to the XL2 line like HD?

The current 960x720 is pretty darn close to the standard HD
resolution of 1280x720. Who knows what kind of chips are
really in that camera. Then again, it just are 960x720 chips
and they put em in due to possibility 1 above.

Steven: I think he is asking the question why something is there
that is not used at all. Most manufacturers probably would just
have said the chips where 960x480 instead of 720 pixels. The
question as I saw it was why they aren't chopped.

[Steven-Marc Couchouron]

Maybe they are there so that the DVinfo whizzes in Alternative Imaging can develop special software which could make them active ! :-)

[Russell Newquist]

I ran the math on the CCDs. If they're using 1/3" CCDs, then masking out the 16:9 and then masking that out to get 4:3, then the 4:3 image area measures exactly 1/4" diagonally. So it should behave exactly the same in terms of depth of field, etc.

Somebody also mentioned in another thread that the higher light requirements vs. XL1s was probably because of high pixel count, but I don't think so - I think it's because they're using a much smaller portion of the CCD on the XL2. This should result in a much sharper image, given the huge increase in active pixels, but it would give it trouble in low-light situations. You'll probably also have to relearn all your f-stop, shutter speed, and filter settings to readjust to the new CCDs.

Unless you specifically need low-light shooting, I'd take the tradeoff any day.

[Rob Lohman]

The three CCD image sensors are specifically designed to capture as much image detail as possible and for shooting under extremely low light conditions. In super low light, the XL2 captures crisp and clear digital data. Under extremely bright conditions, the XL2 greatly reduces vertical white streaks and smears

[Russell Newquist]

Rob - I was referring to the post by Boyd Ostroff here where he states that the XL2 is rated at 5 LUX which is supposedly less than the XL1s. I can't find anything to confirm or deny this, but if it's true, it's almost certainly the CCD size that's responsible, not the high pixel count.

[Dustin Cross]

Aloha,

If the CCD is 960x720 then I would think Canon in planning to make this an 720 HDV camera in the future.

A 960x720 CCD gives you an anamorphic squeezed 1280x720 image.

1.33 x 960 = 1276.8

Seems like it could be done with a software upgrade.

[Russell Newquist]

It depends on how they're doing the masking to crop it down to 16:9. If they're doing that all via software or by just "turning off" the other pixels, as has been guessed, then yeah, they probably could do it with a software upgrade - AND a special animorphic lens attachment.

If they've physically masked off the chip somehow, which they may have done to prevent damage to the turned off areas, then a simple software fix wouldn't do it.

This also assumes that the onboard electronics are powerful enough to process a 960x720 realtime video stream.

[Boyd Ostroff]

<<<-- Originally posted by Russell Newquist : I ran the math on the CCDs. If they're using 1/3" CCDs, then masking out the 16:9 and then masking that out to get 4:3, then the 4:3 image area measures exactly 1/4" diagonally. -->>>

I have not tried to run the math, and it may not make much difference in the end, however your base assumption is wrong. 1/4" ccd's do not measure 1/4" diagonally, these are nominal sizes with no linear relationship. The image diagonal of a 1/4" CCD is 4mm and a 1/3" CCD has a 6mm diagonal. Just looking quickly, I'd say your logic won't hold up since .33/.25 = 1.32 while /6/.4 = 1.5.

See my post at the very bottom of this page for more info on CCD size. If you look at the specs it may give actual measurements, for example the PDX-10 specs say 3.8mm (1/4.7 type) CCD. The manual for the VX-2000 just says 1/3 type CCD however.

Aside from the details though, your point is valid IMO.