A (stupid) PAL XL2 question

[Barry Green]

The PAL XL2 is most definitely 16:9. It just takes the 960x576 patch off the CCD and scales it to become a 720x576 PAL frame.

[Nick Hiltgen]

Whew, that was close.

SO if you were to pull an image straight off the chips in 16:9 mode then you would have an image that is 960x576 (or close to it) and then once it's recorded to tape it goes through some sort of algorhythim that compresses if to 720x576, I might have missed it but barry what's your take on SMPTE time code on the XL2 I've heard that their simply isn't room to record smpte time code on to a minidv tape (that's why you have to use dvcam or somethign along those lines) or do you haveto use dvcam to get smpte on the xl2?

Thanks again for all of your help

[Dennis Hingsberg]

Nick,

4:3 PAL is 720x576 and goes to miniDV tape at 720x576 25fps
4:3 NTSC is 720x480 and goes to miniDV tape at 720x480 30fps

The Canon XL2 records 16:9 as follows:

16:9 PAL is captured on the CCD using 962x576 and goes to miniDV tape at 720x576.

16:9 NTSC is captured on the CCD with 962x480 and goes to miniDV tape at 720x480.

The XL2 PAL has 800,000 effective pixels
In 4:3 mode you get 410,000 pixels
In 16:9 mode you get 550,000 pixels

The XL2 NTSC has 680,000 effective pixels
In 4:3 mode you get 350,000 pixels
In 16:9 mode you get 460,000 pixels

I agree it is somewhat misleading because everyone is debating whether or not the extra pixels (962 horizontally) are benifiting the overall image despite the fact they are only being "saved" or "captured" to tape as 720 pixels horizontally. Technically YES it is beneficial - but will we see the difference? I believe the critical eye will because films shot on 35mm film put on DVD at 720x480 appear to have much more detail than shooting 720x480 on your camcorder and burning to DVD.

[Dennis Hingsberg]

<<<-- Originally posted by Dan Vance : Needless to say, I'm a die-hard PAL advocate: http://www.colddayinhellthemovie.com/VC25P.htm -->>>

Holy shit Dan, I just visited your site and saw your home built "one-of-a-kind" film-look digital camcorder using 3 x 1/2" CCD PAL camera with 700 lines of resolution! That is wicked!

How does one go about acquiring one of those and for how much?!?

In your trailer did you adjust your frame rate from 25fps to 24fps and then perform a 3:2 pulldown to 30fps in After Effects? Or is the trailer still in 25fps?

Great work man!

I can't believe this guy built his own camera...

[Dan Vance]

Thanks! Yeah, it was quite the project, but I'm really happy with the results. The clip is just 25P. I'm working on a "real" trailer which will give more of a feel for what the movie is about, but I wanted to post something that would allow people to at least see the "pseudo progressive scan" effect.
I don't know if I recommend building one though. Except for the small CCDs, that XL2 is looking really good. Although I'm really spoiled now by going direct to disk. Even if I got an XL2, I'd bolt a CAPDIV hard drive to it and skip the tape.

[Dennis Hingsberg]

Dan, I have an electonics background as well so I could completely appreciate your "quest" and what you were trying to achieve with your design in the VanceCam25P.

I was showing a filmmaking buddy of mine your trailer (we spend more time talking about how to get the highest quality images possible out of consumer grade equipment than actually making films - LOL) and we loved the way the footage looked. It wasn't just the progressive image that was cool, it was just the detail in all areas of the frame and even the color was unlike anything I've ever seen coming out of a consumer video camera. I've owned the VX1000, VX2000, XL1s and the footage from your cam in my opinion is so raw and natural it is mind blowing.

Do you think the quality you're achieving is done more so because of the 1/2" 3CCD - or because you are not using the miniDV format and going straight to a hard disk recorder? In other words, have you tried recording your analog signal directly to miniDV through a PAL camera with analog inputs? I'm trying to discover whether it is the miniDV format or the CCD that holds back quality from being all it "could" be.

Any insight would be appreciated.

P.S. Do you really think the XL2 will be comparable to the VanceCam?

[Chris Hurd]

Let's stay on topic -- please don't hijack the thread!

I met Dan Vance at NAB this year and he is an amazing fellow with an amazing homebuilt camcorder (the Vance Cam). But please take any Vance Cam discussion over to our Alternative Imaging Methods forum -- thanks!

By the way, the PAL XL2 CCD target areas are as follows:

16:9 target area is 960 x 576 (appx. 550,000 pixels)
4:3 target area is 720 x 576 (appx. 410,000 pixels)

Hope this helps,

[Dan Vance]

(Dennis, I'll email you directly about your questions so that this thread stays OT.)
The XL2 has a higher pixel count than some 1/2" CCDs, even more so in 16:9 mode, because you'd have to crop a 1/2" 4:3 image and lose about 20% of your pixels anyhow. So in comparing the XL2 to a 1/2" PS camera, you have to weigh the advantage of more pixels against the disadvantage of smaller CCDs, and the other features/limitations of each camera.

[Bill Anderson]

I don't wish to grind this camera pixel count Vs DV pixel count topic into the ground but I think one aspect that is confusing some, including myself, is that even though DV is 720X480 and the camera chip is (i.e) 962X480 the larger pixel count for the camera is still relevant to final quality. The difference between chip and DV frame size is often expressed in such a way that it simply sounds that the extra pixels are moot and therefor discarded, cropped etc. I know that this is not what our erudite posters intend, but that's how it reads to a layperson. i.e "It doesnt matter what size your chip pixel ratio is, it can still only adhere to the standard 720X480 DV frame" But if it were as simple as that then every camera chip need be no more than the pal or NTSC frame size.
One of you scholars should explain the advantage of a larger pixel count, even if it is (SOMEHOW) restrained to DV - i.e the DSR 500 16:9 or in this case the XL2! If it were as simple as previously stated then every camera chip need be no more than the pal or NTSC frame size.

[Russell Newquist]

Bill, there can be a difference that comes from having a CCD with more pixels than you can save to tape, but it does become a question of whether you can tell or not, and here's why:

The extra pixels are not merely discarded or thrown away. Instead, the image is downsampled. How this is done determines what the final quality of the image will be. Actually, merely discarding the extra pixels is one of the ways to accomplish this, but that's kind of the poor man's version - and it would have slightly worse quality than having exactly the same pixel count on your CCD and your final image. Why?

I'm going to draw a number square here for discussion:

1 2 3
2 3 1
1 3 2

(I just picked these numbers randomly). Now, say this is the actual image, but your CCD only has a 2x2 pixel count (4 pixels). In that case, if you focused the entire image on it, the pixel in the top left would record light from all of the top left "1", part of the top center "2", part of the middle left "2" and the upper left corner of the middle "3". The other three pixels on your camera would record similar patterns from their corners, so you would still have some portion of the information from the lost pixels show up in your image.

Now, if you had a 3x3 CCD that was capturing it but storing it as a 2x2 image, you'd have a different story. If it discarded the extra pixels, it would only store the four corner pixels and you'd lose all information about the pixels in between. So what you'd actually get is a worse picture than you would from just a 2x2 CCD. So nobody actually does it that way.

Instead, they'll use a weighted average system to compute values for the 2x2 image using the nearest neighbor pixels around it. There are varying methods of doing this. Whether or not this yields a better picture than just using a 2x2 CCD would depends on the quality of the averaging algorithm and the quality of the CCDs in question.

I'm not an expert on the algorithms, but I would expect Canon to be using one of the better ones. And there's no doubt that the new high-res CCD is better in all ways than the old low-res one from the XL1s. So theoretically, the image should be better for having a higher pixel count. But the question is how much better? If you compared the exact same pixel in an image taken from one camera with the exact same pixel from an image taken with the other, the difference would be very small. So when you look at the whole picture, it will be difficult to notice the improvement. But a good eye should be able to spot it.

I hope this helps and wasn't just all garbled.

[Dennis Hingsberg]

In addition to Bill's point - just ask yourself why the stuff you watch on DVD (which is a 720x480 format) KILLS miniDV in resolution when you see it?

Hint : because the stuff you watch on DVD at 720x480 was not shot at 720x480 using prosumer 3CCD cameras.... it is usually shot on film and conformed to fit 720x480.

It's like this, shoot with high definition and burn to DVD (which again is 720x480) then shoot with a 720x480 camera and burn to DVD and watch. Why does the footage that originated on HD look crisper, sharper and better. This is the point I think many people here are trying to make.

[Dennis Hingsberg]

...well from what Russell just said I am agreeing to disagree with myself.

It is true that for the most effective oversampling of a pixel it should happen at least 2.5 times - so recording at merely 1.5 times or even 2 times the resolution of the miniDV format may not in fact be benefical. In the case of the XL2 were not even talking whole numbers... we're talking fractions and that's the point I guess Russell has made. Well taken.

[Russell Newquist]

Yes, Dennis - you would get much better results from higher oversampling, and far superior if it were in whole number multiples (or at least "even fraction" multiples such as 1.5) of the target resolution. But theoretically the perceived resolution on an XL2 should be higher in part to its higher pixel count.

However, the real point is that having more pixels on your CCDs than on your final image (as the XL2 does) is far better than having less pixels on your CCDs than on your final image (as the XL1 and XL1s do), so there should definitely be a noticeable picture quality increase when upgrading from the older cameras to the newer one. And this is why Canon really isn't shortchanging widescreen or standard filmers, because both modes on the XL2 have at least as much resolution as the DV format. From a technical standpoint it's a good solution, although from an artistic standpoint a larger CCD target area (and hence better DOF) might have been nice.

[Russell Newquist]

And honestly - given the small oversampling in the 16:9 mode, I think it's going to be a very small number of people who can actually tell the difference.

[Bill Anderson]

Thanks to all for clearing that up.