The 10:11 vs 8:9 Controversy for 4:3 DV and DVD.

[dellsam34]

I did not know this was a big controversy in video history, It turned out I came across this controversial topic before but never paid attention to it or at least forgot about it, here is two links from several which I have read few years before:
https://forum.videohelp.com/threads/320755-Aspect-Ratios-0-889-vs-0-9191
http://forum.doom9.org/showthread.php?t=44977

I will do the circle test when I get a miniDV camcorder but here is my quick explanation to why this is this way:

First, lets talk about pure digital DV and DVD only not analog tapes or digital files that originated from analog sources because those have 16 padded black pixels and we all agree why that is (rec.601), So the discussion is about the 16 active video pixels in DV/DVD and how they should be dealt with.

The math of 480/720x4/3 = 8/9 is right and the display is true 4:3, however what's not right is that pixels should be 10:11 instead of 8:9 according to rec.601. For TV's and studio monitors this is not an issue because the extra 16 pixels of active video is going to be thrown in the overscan area giving pixels in the center 4:3 area a stretched size of approximately 10:11 (PAL/SECAM is the opposite I believe), this carried on with flat panels TV's that still had overscan. But it became an issue for computer monitors and TV's with no overscan and the screen size is exactly 4 by 3, So without getting rid of the 16 active pixels and getting them to display properly without disturbing the pixels in the center area is impossible. Think of it like a passengers bus, you can't fit more passengers than what it can carry without squeezing people into the same seats.

Remedies from people across the net for the last 2 decades suggested 3 options:

1- Ignore the small squeeze of pixels since it is not noticeable to the average person, In the bus analogy, fill the bus and ignore passengers feelings.

2- Crop the 16 extra pixels and resize to 4:3 using 10:11 flag, In other words put the right number of passengers in the bus and deny the remaining passengers.

3- For 16:9 screens keep the 16 pixels since they are useful active image and add black pixels up to the maximum frame size to fill the 16:9 ratio giving the pixels their 10:11 ratio while keeping the extra 16 pixels beyond the 4:3 area visible in the 16:9 frame but adds extra black pixels. Meaning, get a bigger bus, put all the passengers in it and you will have extra vacant seats but every passenger will sit comfortably.

What do you guys think? Please keep it civil and read the entire post to get the entire image so we can stay on the subject.

[Skiller]

I tested this with two of my MiniDV camcorders as well many years ago using a perfect square sheet of paper. Not surprisingly to me, it turned out that the video these camcorders shoot has the same PAR as analog video (12:11 for PAL), even though all of the 720 pixels do carry active image.


The pillarboxing in 3- is not a good idea imo, unless the format requires it. You could also just encode at 720 (no cropping) and still use the 10:11 PAR flag for NTSC and 12:11 for PAL. This will result in a slightly wider than 4:3 playback at 15:11 (1.36:1) which is fine.

[dellsam34]

All my DV files are still DV files, If I ever think of converting them to a newer format, that's exactly what I would do, keep the wider 720x480 ratio. I guess we can add number 4 option in this thread then, Though I couldn't find a bus analogy for it, lol.

[jagabo]

According to Sony:

The sampling raster of the DVCAM is the same as that of the ITU-R Rec.601. Luminance video signals are sampled at 13.5 MHz

https://web.archive.org/web/20110928040838/http://www.sony.ca/dvcam/pdfs/dvcam%20forma...20overview.pdf

I'll attach a copy here in case the web archive loses it...

[ConsumerDV]

3- For 16:9 screens keep the 16 pixels since they are useful active image and add black pixels up to the maximum frame size to fill the 16:9 ratio giving the pixels their 10:11 ratio while keeping the extra 16 pixels beyond the 4:3 area visible in the 16:9 frame but adds extra black pixels. Meaning, get a bigger bus, put all the passengers in it and you will have extra vacant seats but every passenger will sit comfortably.

You said you were on the same boat with Jagabo regarding non-baked letterboxing/pillarboxing, but apparently you are not. Get off the bus.

[dellsam34]

That was completely a different subject, if you keep this argumentative behavior you will be reported to mods to get you off this forum. I expect your next post to be within the subject.

[Sharc]

I don't question the 10:11 Pixel Aspect Ratio for 4:3 NTSC-DV at all, but unless I totally missed it neither the DVCAM Overview document nor Rec. ITU-R BT.601-7 specify the Pixel Aspect Ratio explicitely. Both do however specify the luma sampling rate of 13.5 MHz (same for NTSC and PAL) and the number of horizontal pixels.
As I understand it the missing link is that one has to put the 13.5 MHz luma sampling rate in relation to the industry standard for square pixels, which is 12 3/11 MHz for NTSC and 14.75 MHz for PAL. Hence the Pixel Aspect Ratios become
12 3/11 : 13.5 = 10:11 for NTSC
14.75 : 13.5 = 59:54 for PAL (approximated for practical reasons by 12:11)
The number of horizontal pixels is eventually determined by the scanline duration and the sampling rate. Or one fixes the number of pixels and gets the scanline duration.
However, the often quoted "Rec.601" - if I read it correctly - leaves it open how one fits/fills/squeezes/AR-distorts etc. the 'useful' picture within the frame, as long as it is padded to the specified frame size.
If I remeber this correctly the Pixel Asect Ratios (aka Sampling Aspect Ratios) are explicitely defined in other digital video standards and documents only, like mpeg-x or H.264/H.265 (or perhaps in other "Rec.601"-series documents which I am not aware of?).

Edit:
Going a bit off topic for NTSC-DV I know, but FWIW even the EBU (European Broadcasting Union) felt obliged to comment the 702,704,720, Rec.601 confusion:

[DB83]

I am probably agreeing with everything that is said but I'll throw a few thoughts in to the mix.

A week or so ago I conributed to a thread where the OP was transfering DV to a PC and he was adamant that there was no distortion when viewing the video on a tv. He posted a sample and I was rather surprised to discover that there were no 8+8 pixel side-bars and the image was 640*480 since this was NTSC


Now my own experience with DV is restricted to analogue conversion via a Canpous ADVC. Yet such a video has the side-bars and since this is PAL I only see 768*576 with the side-bars.


But back to the NTSC sample I did find that Doom9 discussion which kinda eased my own understanding of the controversy.


So this is the bit I guess I am now repeating. DVDs no longer, I understand, follow Rec601 so you get 720/9*8 = 640 whereas DVs do so you get 704/11*10 = 640

[jagabo]

DVDs never followed the rec.601 standard. The DVD spec follows the MPEG 2 spec where the full frame constitutes the 4:3 frame. The difference is usually ignored when DVDs are made from analog video. Even worse, the players I've tested are schizophrenic: they follow the ITU spec at the s-video and composite outputs, but the MPEG 2 spec at the digital outputs.

[dellsam34]

Most software that capture to 640x480 crop away the 16 pixels on the fly while resizing to comply with computer "then" resolution notably Mac but some PC software as well. I would assume the same for PAL but if you still see the black bars maybe your software just resizes and don't crop.

As to DVD, I think there is a standard version of 704x480 for DVD which governs DV as well, No?

[jagabo]

Yes, you can crop your ITU caps to 704 pixels wide to eliminate the difference between the ITU and MPEG 2 spec. The DVD spec allows for this width, this is probably why it was included. But I've rarely seen a commercial DVD at 704 pixels wide.

[Sharc]

For those interested here an article on the history and development of the often quoted Rec.601, the 13.5 MHz sampling rate, 720 pixels etc.

[ConsumerDV]

Most software that capture to 640x480 crop away the 16 pixels on the fly while resizing

I don't know about most software, but I can confirm that this is exactly what VirtualDub does (YouTube link) when capturing SD NTSC into 640x480. I think this is a very convenient feature for someone who want a 4:3 frame with square pixels and correct proportions.

[Sharc]

Most software that capture to 640x480 crop away the 16 pixels on the fly while resizing

I don't know about most software, but I can confirm that this is exactly what VirtualDub does (YouTube link) when capturing SD NTSC into 640x480. I think this is a very convenient feature for someone who want a 4:3 frame with square pixels and correct proportions.

In my experience this may or may not work in all cases as intended, depending on capture device and drivers. Some will not crop and position the picture 'symmetrical' but take the top left corner as a reference, means the picture gets cropped on the right and bottom only. It's safer to capture full size (720x480, 720x576) and do the cropping and/or resizing in post.

Edit: Sorry, forget the post. My bad, never mind, I had PAL in mind

[ConsumerDV]

Most software that capture to 640x480 crop away the 16 pixels on the fly while resizing

I don't know about most software, but I can confirm that this is exactly what VirtualDub does (YouTube link) when capturing SD NTSC into 640x480. I think this is a very convenient feature for someone who want a 4:3 frame with square pixels and correct proportions.

In my experience this may or may not work in all cases as intended, depending on capture device and drivers. Some will not crop and position the picture 'symmetrical' but take the top left corner as a reference, means the picture gets cropped on the right and bottom only. It's safer to capture full size (720x480, 720x576) and do the cropping and/or resizing in post.

Absolutely! It is a shortcut for newbies. I think it is better to capture 640x480, letting the software to take intelligent - hopefully - decisions, than to capture in 720x480 and then export/upload with PAR 1:1 as can be seen in many a YouTube video.

[Sharc]

Most software that capture to 640x480 crop away the 16 pixels on the fly while resizing

I don't know about most software, but I can confirm that this is exactly what VirtualDub does (YouTube link) when capturing SD NTSC into 640x480. I think this is a very convenient feature for someone who want a 4:3 frame with square pixels and correct proportions.

In my experience this may or may not work in all cases as intended, depending on capture device and drivers. Some will not crop and position the picture 'symmetrical' but take the top left corner as a reference, means the picture gets cropped on the right and bottom only. It's safer to capture full size (720x480, 720x576) and do the cropping and/or resizing in post.

Absolutely! It is a shortcut for newbies. I think it is better to capture 640x480, letting the software to take intelligent - hopefully - decisions, than to capture in 720x480 and then export/upload with PAR 1:1 as can be seen in many a YouTube video.

It's just not so easy and straightforward for PAL, as one would have to resize for square pixels to 768x576 (upscale horizontally) or to downscale the resolution to 704x528 or 720x540 which requires vertical resizing with deinterlacing of normally interlaced video which is a subject of its own.