Does going to 16:9 affect the image quality in any way?

No. Since the XL2 has native 16:9 (enlarged 4:3 chip clipped to 16:9), shooting 16:9 will look better for projects intended for HD television or film transfer. For SD TV, shooting 4:3 means chopping off the 16:9 footage at the ends so quality of the 4:3 image will be the same as the 16:9 for the uncropped areas. If you are planning to letterbox for SD 4:3, definitely shoot 16:9. Hope that helps.

Despite Andrew's very emphatic response, I don't think the situation is as simple as that. (We've had a few threads on this subject already.)

Obviously, if you want 16:9 output you should shoot 16:9, and if you want 4:3 output you should shoot 4:3. No arguments there, and I won't revisit the issue of whether 4:3 quality is inherently better than 16:9.

However, if you want to show letterboxed video on a 4:3 display, my view is different from Andrew's. If you ask me "Does rescaling a 16:9 video to 4:3 produce an acceptable result?" my answer would be a definite Yes, and I actually do this quite often. BUT, to answer the original question, it certainly does affect the image quality.

For this particular case (letterboxed video on a 4:3 screen only) you will get the best image quality by shooting 4:3 and framing for 16:9 (by blanking off part of the viewfinder). Then add letterboxing in an NLE, which does not involve any rescaling of the video image with it's resultant drop in image quality. Shooting 16:9 and rescaling to fit a 4:3 frame will not only take a quality hit, it will most likely also mean a longer rendering time for the rescaling operation.

Richard

For this particular case (letterboxed video on a 4:3 screen only) you will get the best image quality by shooting 4:3 and framing for 16:9 (by blanking off part of the viewfinder).

No, not with the XL2. Shooting in the camera's native 16:9 will yield superior results to what you suggest.

Jay

Hi Eduardo,

As Richard mentioned, we had an extensive discussion on this topic earlier this year:

http://www.dvinfo.net/conf/showthread.php?t=40540

The resolution either way is superb for SD and I think that for the XL2, the discussion is more academic than practical. Other factors including codec settings on final export probably matter more than whether the footage was originally 16:9 or 4:3.

As you can see from that prior discussion, Richard and I have somewhat different biases on this subject. This is my take:

If there is NO chance you'd ever want to use your footage in the future at 16:9, then, sure, shoot 4:3. But because I think the image quality is so close to a tie, shooting 16:9 is a more flexible option. That's because the world is rapidly migrating to 16:9 as a standard aspect ratio, and I want to be able to effortlessly re-purpose my footage later. Seems that the majority (not all, but the majority) of people who want 4:3 footage are looking at it on TVs that won't really show the reduced resolution of footage that's been "pan-and-scanned." On the other hand, going from a 4:3 SD original to 16:9 in the budding HDTV world will be less tolerable, even if you frame the 4:3 with that in mind.

But as far as pure resolution output from the camera, it really doesn't matter from a practical standpoint. Both 4:3 and 16:9 output a 720x480 non-square pixel image that is very sharp natively.

But as far as pure resolution output from the camera, it really doesn't matter from a practical standpoint. Both 4:3 and 16:9 output a 720x480 non-square pixel image that is very sharp natively.

If that's true, Peter, then please explain the following:

"In the North American and Japanese NTSC version of the XL2, each of these 4:3-shaped CCD's contains a total of 680,000 pixels. Within this pixel array lies the 16:9 target area, composed of 460,800 pixels (for an image size of 960 x 480). All pixels above and below this rectangular widescreen area are "turned off," that is, they're inactive, neither receiving incoming light nor generating a signal. Only those pixels which lie within the 16:9 target area are actively firing. For standard 4:3 aspect ratio video recording, a smaller target area of 345,600 pixels is used (for an image size of 720 x 480)."

Jay

Hi Jay,

What you're quoting from my XL2 CCD Block page concerns what is happening in the camera head. What Pete is describing, is how that image is recorded to tape. Admittedly they are two very different things. I suppose what I really should do is to amend my article to include a statement (Pete has already worded it very well) about what happens when the image is recorded to tape.

Jay,

Most of the answer is in Chris' article on the XL2 CCD block:

http://www.dvinfo.net/canonxl2/articles/article06.php

The NTSC DV signal is always, always 720x480. That's the standard; can't be anything else. In 4:3 mode, the relationship between the approximately 345,600 pixels on the CCD block that get used, and what goes out into the 720x480 pixel DV stream is about 1:1. In 16:9 mode, the 960 horizontal pixels get processed and downsampled to 720.

What was at the crux of the discussion from earlier this year was the resolution of the image created by those 345,600 pixels...whether the image ends up blurry or sharp. Taken by itself from a mathematical perspective, downsampling (as in 16:9) would be expected to introduce interpolation errors that would reduce the effective resolution, or sharpness, of the image taken at 1:1.

But based on my crude resolution tests (linked in the old thread), it appears that Canon must have introduced fairly sophisticated algorithms to take advantage of the information in those "extra" CCD pixels. That's not really a great surprise since the CCD block data must be processed into a DV signal anyway, and that's where edge sharpness, gamma adjustments, etc, etc get applied.

So, in the end, the 720x480 image, whether it is 4:3 or 16:9, tests out darn near the theoretical resolution limit of 360x240 line pairs.

Hi Pete. Most of what you mention is related to the comparison of 16:9 and 4:3 video quality without any further conversion. (I was trying not to bring this up again :) )

When we talk about the further step of resizing the 16:9 image to fit a 4:3 frame, with letterboxing, my point is that this can only be degrading the image quality, it can't possibly make it better than an unsqueezed image. If you take a 4:3 native video in a 4:3 NLE project and add letterbox bars, the quality should be maintained because there is no resizing involved.

HOWEVER, I have been thinking a bit more about this, and realise now that my view was a bit simplistic. What I said above doesn't change, but there is an optical difference between the 2 images (native 16:9 resized vs. native 4:3 with bars added). The 16:9 viewing angle on the XL2 is wider than the 4:3 viewing angle. To get the same framing for both images, you would need to zoom out for the 4:3 shot, which means that the dimensions of fine details in the picture could start to approach or exceed the resolution of the camera lens. So, by shooting in 4:3 and framing for 16:9 you might get a (slightly) poorer image as a starting point (due to lens effects, not to CCD differences). This makes it less clear-cut, and more difficult to say which one would give better end results.

As you mentioned, this discussion is largely academic, and I will continue to shoot 16:9 too.

Richard

... there is an optical difference between the 2 images (native 16:9 resized vs. native 4:3 with bars added). The 16:9 viewing angle on the XL2 is wider than the 4:3 viewing angle. To get the same framing for both images, you would need to zoom out for the 4:3 shot, which means that the dimensions of fine details in the picture could start to approach or exceed the resolution of the camera lens.

Although I lack the technical acumen to through around numbers and ratios to discuss this as you and Pete do, I agree with your above statement 100%. "The 16:9 viewing angle on the XL2 is wider than the 4:3 viewing angle." That, to me, says it all. That was the down side of shooting 16:9 with the XL1s and the highly anticipated beauty of the XL2!

I can't for the life of me see any advantage in shooting 16:9 by using 4:3 then cropping it--that is not logical, as our pointy-eared friend would say.

Jay

Hi Jay. Actually, that only says one side of it. Remember, we are talking only about the specific case where the output is for a 4:3 screen.The other side to consider is that if you resize 16:9 to fit a 4:3 frame in letterbox mode, you have to do a digital resampling which has its own problems. It's a case of optical zoom out vs. digital zoom out, and it's not very obvious to me now which one is going to give better results.

Hopefully, there is a bit of logic in there somewhere. :)

Richard