4:3

[Robert_Niemann]

Does anybody know, why Canon decided not to use the full CCD area for 4:3 mode? Would there not be a better 4:3 image quality, if they had done so?

[Shaw]

Yes there would have been. However, they could not have claimed to have true 16:9 if they did so.

[jaegersing]

Does anybody know, why Canon decided not to use the full CCD area for 4:3 mode? Would there not be a better 4:3 image quality, if they had done so?

Not necessarily. They would have reaped the benefits of the larger CCD (e.g. better low light capability, less DOF) but since this would also mean resampling the image to DV resolution, they would therefore introduce resampling artifacts which degrade the image.

Richard Hunter

[J.R. Hudson]

I thought the it did not have True 16:9 as it was? Do they not use a 4:3 and a cropped portion? (In contrast, isn't the FX1 TRUE 16:9 ? )

Someone explain this to me! ???

[Barry_Green]

Okay, there are potentially many reasons as to why they did what they did. First, yes the Canon is true 16:9. It samples its image off a 480-pixel-high 16:9-shaped patch of CCD, delivering full true resolution. The only argument that could be made would be that it's not true 16:9 off a 1/3" CCD, because there's wasted CCD there... so let's just pretend that Canon got out some tin snips and cut off the parts of the CCD that are unused. What would you be left with? A 16:9-shaped CCD (measuring about 1/3.3"). And the Canon would use all of it, and deliver a nice sharp full-resolution picture. So yes, the XL2 is full 16:9.

Now, as to why they chose to use the 4:3 they did... well, I happen to think it was a marketing decision. When someone plays with the Canon in the store, they can put it in 4:3 mode and see a certain image, and then when they put it in 16:9 mode, it looks like they get MORE image. Hey, "more" means better, right? When in reality what's happening is that when it's in 16:9 mode they're seeing the full image, and when they go to 4:3 they're seeing LESS image. But -- shhh... let's not tell anybody that...

So there's a marketing reason. But Richard Hunter brings up an excellent point as well. Had they used the full 1/3" surface of the CCD, they would be downsampling a 960x720 image onto a 720x480 frame. That's gonig to lead to aliasing and artifacting and, probably, worse video performance than the way they chose. The way they chose to do it, they use a pixel-for-pixel 720x480 subsection of the 16:9 frame. And it does look great, as Jaegersing says. The 4:3 of the XL2 looks better than I was expecting it to.

Now, if they had used the full surface of the full 1/3" CCD and had larger pixels, yes it could have been better quality. But that's not possible, because the pixel size was set when they decided to offer 16:9 native. So you would have a case of 960x720 tiny pixels that get downrezzed, or 720x480 pixels native. In that case, I'd vote for 720x480 every time, and that's what Canon did, and the results are very nice indeed. You get a narrower field of view doing it that way, but the video signal is nice and sharp and clean.

[Robert_Niemann]

Barry, as I understand it, a DV saves its pictures always at a pixel rate of 720 x 480 (NTSC). But what about 16:9: saving by 720 x 480 or 960 x 480 pixels? And another question: What is the advantage of a CCD with a higher pixel rate than 345.600 at 4:3? I have read, that a CCD works analogically, but what does that mean? Is a pixel not a pixel and a digital value every time?

[Barry_Green]

DV is always 720x480, there is no such thing as DV at 960x480. The XL2 has a CCD with 960x480 pixels, but when the system samples it it creates a frame of 720x480.

16:9 interprets the pixels to be wider than they are tall, but there are no more pixels in a 16:9 image than in a 4:3 image.

There is no advantage to a CCD with a higher pixel count than 345,600, for a standard-def camcorder, for 4:3. The more pixels you have, the smaller the pixels have to be, which means they are less efficient at gathering light, and the smaller pixels will just get sampled at the 720x480 rate anyway, which will mean aliasing.

Canon did the right thing with their 960 implementation, because it lets the 720x480 4:3 extraction operate at full size. If they were making a 4:3-only or a 16:9-only camera, then 960 would have been wasteful and excessive, but because they were making it switchable and needed to protect for both 16:9 and 4:3 imaging, they did the right thing. The 4:3 on that XL2 looks surprisingly good.

A CCD is an analog device, what we call "pixels" are not digital at all, they're little tiny electricity capacitors (someone please come in and offer a better explanation!) They're "light buckets", each capable of absorbing light and storing it as a certain voltage. Once the CCD has been fully charged, then it gets digitized, turned into a pixel array of 720x480. The ideal would be to use a system where one pixel on the CCD matches up to one digital pixel on the frame -- that will give you the ultimate resolution while also allowing each pixel to be as big as possible. When you have a megapixel CCD in a video camera, the digitizing process has to sample all these smaller CCD pixels (with their lesser capability to store light) and then average them together to make a digital pixel representation.

[dmc]

Thanks Barry - a simple, distilled version of such a complicated process. I actually think I understand it now.

Dude, you should write a book... Doh! ya already did.

; )

dmc

[Robert_Niemann]

Yes, Barry, I thank You, too. But from Your explanations arises a new question: Because You say, that downsampling always leads to aliasing and artifacting, would it not be better to record 16:9 with a CCD pixel rate of 720 x 360 (with each CCD pixel to be as big as possible)? But then: Why does Panasonic call their DVX100 plus the anamorphic adapter to be capable for NATIVE 16:9 recording, while a DVX100 without the adapter would not have been?

[J.R. Hudson]

I don't think Pani does claim "Native 16:9" ???

[Barry_Green]

The best way to record SD 16:9 would be with a chip that's 16x9-shaped, in a 720x480 grid.

The DVX in squeeze mode uses 720x360, and it does a decent job in progressive -- the results are about as high-resolution as an interlaced "native 16:9" camera. However, squeeze can't compete with the XL2's progressive 16:9 imagery.

The reason the anamorphic adapter can be said to be 'native' (and as John points out, I don't think Panasonic claims that, but I do) is that you then use the full surface of the chip to record a full-resolution 720x480 16:9 image. In squeeze mode you get to use 75% of the chip, with the adapter you get to use 100% of it.

[MachinePhilby]

What about me!... :P... btw I'm PAL but I'll stick to NTSC as best I can...

Could you say this: the widescreen image (960x480?) is actually 720x480 pixels, but (in computer pixels) the 720pixels are stretched over 960? Is that fair to say? Because of the shape of the CCD's... yeah?

If so, wouldn't this mean your losing quality? This is really the deciding factor between which camera I buy... to be honest, if the image is just stretched then I don't feel the XL2 is native 16:9!!! As far as I'm concerned, native DV widescreen in my views is (computer pixels) 960x480, not 720-stretched-to-960x480!

This is where I get a little confused... the 960 comes from the ccd's being shaped in the 16:9 format, correct? Kind of?... maybe? The stretched pixels (960) are actually only 720?

I bring this up because I want to add 3d composite material to my video (I've done 3d composite projects before but I used 4:3), would this mean that when I render the 3D I would have to tell it to render at 960x480, but with stretched pixels (probably using a DV16:9 filter)? i.e. 720-to-960x480?

Would I be better off using a render filter as such? Wouldn't a rendered 960x480 SQUARE pixels be a lot sharper than the video? After all, as nice as it is to have a sharp 3d image, you have to remember it's got to blend into your video/film environment.

Pls excuse my mess ;) I'm a learner driver at the moment :P... but am I on the right track? :-/ Any ideas?

[Barry_Green]

There is no such thing as 960x720 in DV. The only thing that can ever exist is 720x480. If you go rendering something with square pixels, when you composite it into your editor, your editor is going to have to resample it down to 720x480 anyway.

How many pixels are on the CCD is completely irrelevant to any question regarding how you work with the footage. The CCD pixels are not digital pixels, they're analog capacitors (anyone with a better description, please fill in!) The camera samples and digitizes that into a grid of 720x480. That's what you get to work with, that's all there is... 720x480, with a pixel aspect ratio of .9 (for 4:3) or 1.2 (for 16:9). The television itself stretches the video into the proper shape.

If your graphics and compositing programs support pixel aspect ratios, I've got to assume that you'd be better off using them in that way -- using a pixel grid of 720x480, and a pixel aspect ratio to get the proper shape.

Fortunately all this nonsense goes away when you go to HD... HD is always square pixels, all the time.

[jaegersing]

The CCD pixels are not digital pixels, they're analog capacitors (anyone with a better description, please fill in!) *.

Hi Barry. I think CCDs are actually an array of photodiodes which have a capacitance that behaves in the way you describe. The link below has some more info for anyone who is interested.

http://www.shortcourses.com/how/sensors/sensors.htm

Richard Hunter

[Robert_Niemann]

I have found another very interesting site about the CCD technology: http://www.ccd-sensor.de (it is in English and German!).