ITU-R BT.601 and PAR: How good is my knowledge?

[Brother John]

Hi, guys.

After the latest update of my own site and (once again) reading A Quick Guide to Digital Video Resolution and Aspect Ratio Conversions aspect ratios refuse to leave my head.

I more and more get the feeling to be on the right track to thoroughly understanding the whole topic. But, of course, a couple of questions remain. Well... Let me first summarize the "facts" as I know them - values according to ITU-R BT.601.


PAL

For correct DAR, PAL needs following resolutions:
4:3 -> 768x576
16:9 -> 1024x576

A DVD's active area is 702x576 pixels. The original 768 or 1024 need to be squeezed into those 702 pixels. And knowing this, calculating the PAR is possible.

4:3

      768   128
PAR = --- = --- = 1.094
      702   117

16:9

      1024   512
PAR = ---- = --- = 1.459
      702    351

Based on this the calculated un-squeezed full DVD resolution for playback is:
4:3 -> 787.69x576
16:9 -> 1050.26x576


NTSC

NTSC has these resolutions:
4:3 -> 648x486
16:9 -> 864x486

The active area here is 710.85 pixels. In reality of course only 711 pixels are possible. Anyway - calculating the exact PAR:

4:3

      648      4320
PAR = ------ = ---- = 0.912
      710.85   4739

16:9

      864      5760
PAR = ------ = ---- = 1.215
      710.85   4739

That means for the DVD playback resolution:
4:3 -> 656.34x480
16:9 -> 875.12x480


So far, so good. Now I could as well
- calculate NTSC with 711 instead of 710.85 pixels.
- ignore ITU and assume the whole 720 pixels as being active area -> same calculations with 720 instead of 702/710.85.

In the end I come up with this set of PAR tables:

Calculated PAR according to ITU-R BT.601

     | PAL       | NTSC      | NTSC (711)
-----+-----------+-----------+-----------
 4:3 | 128/117   | 4320/4739 | 72/79
-----+-----------+-----------+-----------
16:9 | 512/351   | 5760/4739 | 96/79


Usual "practical" PAR respecting ITU-R BT.601

     | PAL       | NTSC
-----+-----------+-----------
 4:3 | 12/11     | 10/11
-----+-----------+-----------
16:9 | 16/11     | 40/33


"Wrong" PAR without ITU-R BT.601

     | PAL       | NTSC
-----+-----------+-----------
 4:3 | 16/15     | 9/10
-----+-----------+-----------
16:9 | 64/45     | 6/5

Ok, and to give you some thoughts for a discussion:

• Any major errors in the above? Or less major ones?
• When might the "Wrong" PAR still be correct? I can think of incorrectly mastered DVDs. Anything else?
• Which ITU table should be used? The "practical" one looks like a rule-of-thumb table to better remember the values.
• Where does the "practical" table originate from? In this thread it sounds like MPEG-4 specs. Is that so?
• Which NTSC ITU values are actually correct? After all 710.85 pixels ist simply impossible in reality. On the other hand the "Quick guide" uses those - it even changed from 711 in its early days to 710.85.
• I'm tempted to say: Computer or TV as playback device is not important. ITU is valid for both, only the TV cuts a couple of pixels due to overscan. Might this be the origin of the "Wrong" table, that then would be really useless?
• How to get from 486 NTSC pixels to 480? My understanding is that 6 pixels just get cropped (changing the DAR slightly). Is that correct?

*sigh*
The day I'll finally understand 100% of the whole PAR-DAR-ITU-anamorphology, I'll throw a huge party!

P.S.: I posted the same thing in the German forum.

[zambelli]

The correct active area for both NTSC and PAL, according to ITU.601, is 704x480/576.

So the correct pixel aspect ratios for DVDs are:
NTSC fullscreen 10:11
NTSC widescreen 40:33
PAL fullscreen 12:11
PAL widescreen 16:11

Those stretch the image to:
NTSC fullscreen 640x480
NTSC widescreen 853x480
PAL fullscreen 768x576
PAL widescreen 1024x576

When resizing from full 720x480/576 resolutions to square pixel resolutions, one should use one of the two methods:

1. Crop away 8 pixels on each side, then resize according to the PARs listed above. Downside: tiny black bars on the side.
2. Resize according to the PARs listed above, then crop away any extra vertical pixels until your reach the square pixel resolution listed above. Downside: you sacrifice some vertical resolution.

Similarly, when resizing from higher-than-SD resolutions to SD, one should preferrably resize like this (using Avisynth syntax and NTSC in this example):

LanczosResize(720,492)
Crop(0,6,720,480)

Why? Well, even though only 704 horizontal pixels are active, it's desirable to use all 720 in order to take advantage of the full display area on devices with no overscan. But in order to do that, one must be careful to preserve the correct aspect ratio. If 704/480 is 1.46666, then 720/1.46666 is 490.91. We round up so we could crop by an even number.

Similarly, when expanding the ITU.601 image to square pixels, we can use the whole image area, not just the active pixels. So anamorphic PAL 720x576 actually ends up being 1048x576. No, that's not quite 16:9, but 1024x576 is. The 24 extra pixels are the same 16 inactive pixels from the 720x576 source.

Confused more?

[Brother John]

Originally Posted by zambelli

Confused more?

Nope, on the contrary. You actually (almost) confirm what I suspect is right.

One thing though: Why do you think ITU says 704 active pixels? I'm still a little lost between samples and megahertz when reading the recommendation itself, but the Quick Guide's conversion tabe claims 702x576 and 710.85x486 to be the correct active ITU frame. And I'm quite sure that the Guide is a trustworthy source.

[Wilbert]

One thing though: Why do you think ITU says 704 active pixels? I'm still a little lost between samples and megahertz when reading the recommendation itself, but the Quick Guide's conversion tabe claims 702x576 and 710.85x486 to be the correct active ITU frame.

Yes, you are right.

[zambelli]

Originally Posted by Brother John

One thing though: Why do you think ITU says 704 active pixels? I'm still a little lost between samples and megahertz when reading the recommendation itself, but the Quick Guide's conversion tabe claims 702x576 and 710.85x486 to be the correct active ITU frame. And I'm quite sure that the Guide is a trustworthy source.

Maybe because 702 is a strange number for most codecs to deal with. Most codecs require mod-16 resolutions - or mod-8 at the least. 702 isn't even mod-4.

Similarly, 710.85x486 is essentially the same aspect ratio as 702x480, so perhaps that's why it gets set to 704x480 as well.

If I recall correctly, the extra 6 vertical lines in NTSC are used for blanking intervals. Since digital signals have no use for them, they're discarded.

It all gets very complicated very quickly, so I think for all intents and purposes it's best to assume 704x480/576 are the active ITU.601 resolutions.

[Brother John]

702 might be a strange number for codecs, and 710.85 even more, but they don't have to deal with it. The codec always gets fed 720 for a usual DVD. Ok, 704 is valid as well. But I've never seen a disc actually using it.

So why assume something that apparently is wrong? For practical purposes a "stick to the ITU PAR table" is sufficient anyway. No need to know about the backgrounds at all.

And if you do set out on a quest to find the secret of the ITU-R , anything less than the whole truth isn't good enough, no matter how complex and difficult that truth might be. - At least that's how I feel.

[zambelli]

Originally Posted by Brother John

Ok, 704 is valid as well. But I've never seen a disc actually using it.

Try some TV series DVD. You'll find that most of them use only 704 pixels and pad the rest with black bars.

So why assume something that apparently is wrong? For practical purposes a "stick to the ITU PAR table" is sufficient anyway. No need to know about the backgrounds at all.

704 isn't wrong! In digital video 704 is the active number of horizontal pixels for both PAL and NTSC. The pixel aspect ratios I quoted to you are based on 704x480/576. I didn't make up those numbers - they are the actual pixel aspect ratios used by everyone from Adobe to Microsoft.

Also, note that VCD resolutions are 352x240 and 352x288. Exactly quarter res. That's no coincidence.

[Brother John]

You'll find that most of them use only 704 pixels and pad the rest with black bars.

I think you got me wrong here. I was talking about an actual DVD resolution of 704. Not 704 active pixels padded with black to 720.

they are the actual pixel aspect ratios used by everyone from Adobe to Microsoft

I'm told that even Adobe's Photoshop uses the weird generic PAR of 1.067, completely ignoring ITU. Seems like the big names can't be trusted.

In digital video 704 is the active number of horizontal pixels for both PAL and NTSC.

Do you know which standard says so? I'd like to have a look at the MPEG specs because of this post, but unfortunately they're not free.

[SeeMoreDigital]

If you are into this kind of thing.... you might be interested to see what the good ol' BBC has to offer: -

http://www.bbc.co.uk/commissioning/t...turesize.shtml

But remember, software based media players don't follow ITU output sizes when displaying PAL and NTSC anamorphic DVD's


Cheers

[mpucoder]

One thing to remember, ITU-601 describes how to digitize an analog standard, so, for NTSC, keep a copy of RS-170 or SMPTE 470 handy. In NTSC the active portion of a scan line is 52.6us, multiplied by the sample rate of 13.5MHz you get 710.1 pixels. ITU-601 uses 720 for a number of reasons, the best being that 720 is the lowest multiple of 16 greater than 710 - a consideration for mpeg. Digitizing more than the active region allows for misalignment of the blanking pulse.
NTSC has 525 lines, 485 (not 486 as many claim) are active, the others carry vits, CC, vitc, and vertical sync. 5 lines are lost when going to digital. The reason ITU-601 chose 480, besides being a multiple of 16, is that it allows a common sample rate for NTSC and PAL. 720x480x30 = 720x576x25

The differences in PAR are so small the average viewer cannot discern them, and are far smaller than the average misalignment of a crt.

[zambelli]

Originally Posted by Brother John

I think you got me wrong here. I was talking about an actual DVD resolution of 704. Not 704 active pixels padded with black to 720.

Actually, 704x480 is a legal DVD resolution. So is 352x480. It's very uncommon - but it is legal.

I'm told that even Adobe's Photoshop uses the weird generic PAR of 1.067, completely ignoring ITU. Seems like the big names can't be trusted.

Adobe Premiere Pro approximates the PARs to 0.9 and 1.2 for NTSC. That's close enough to 10:11 and 40:33.
Microsoft Windows Media Encoder 9 gets it exactly.

Do you know which standard says so? I'd like to have a look at the MPEG specs because of this post, but unfortunately they're not free.

I believe MPEG specs, DVD specs and DV specs all define 704 as the active portion of the video. The full image is defined as 720 though.

[Brother John]

I'll have to get a little deeper into the analog side of the whole problem, but digital seems pretty clear now. Thanks for the contributions, guys!

Let me try a little summary:
We have two "competing" sets of specs: ITU and DVD/MPEG, the former using PAR 521/351 etc. and the latter using 16/11 etc. Both have the same spirit but a slightly different focus. Of course it's impossible to say one is correct and the other one is not. And the difference is very small, anyway. We're talking about 3 pixels for a 16:9 PAL DVD. Additionally DVDs might be mastered according to any of the two.

Hm. All this sounds like the most reasonable course of action for practical purposes (i.e. tutorial writing, backups) would be: Stick to the better known DVD/MPEG PARs, mention ITU values as advanced background info and strongly discourage any "generic" PAR because DVDs probably are not mastered with those values in mind. Thus using generic for a backup is very likely to result in a slightly distorted image.

[Wilbert]

I believe MPEG specs, DVD specs and DV specs all define 704 as the active portion of the video.

The MPEG-2 specs doesn't mention any active portion of the video.

and strongly discourage any "generic" PAR because DVDs probably are not mastered with those values in mind.

Well, some are. We don't know whether they are an exception, or whether a significant part of the DVDs out there don't follow ITU.

[GodofaGap]

Unfortunately PAR seems to be impossible to get exactly right due to the 2 following reasons:

1.

We don't know whether they are an exception, or whether a significant part of the DVDs out there don't follow ITU.

I don't know if it is always that easy to determine if a DVD is follwing ITU or not.

2.

But remember, software based media players don't follow ITU output sizes when displaying PAL and NTSC anamorphic DVD's

This is indeed the case for at least VLC and MPC, so if you do use ITU standards and aim for PC playback it will be wrong either way.

IMO to the differences between ITU and non-ITU are too small to worry so much about. Even more so because even if you are following the correct way you are not guaranteed correct output anyway.

[Inc]

But remember, software based media players don't follow ITU output sizes when displaying PAL and NTSC anamorphic DVD's

So if you do go for a straight DVD->mp4 backup for PC/HTPC playback purposes by keeping the res. Do crop 720x576 to 704x576 and encode. As here it makes no diff. if the true 128/117 ITU par is used or simply an AR of 4/3 will be applied. Both will look ok - on TV and on the SW Player

The MPEG-2 specs doesn't mention any active portion of the video.

Thats "generic" ISO mpeg2 independant from the final output device, but DVDs do got mpeg2's purposed for later TV Display usage.

I don't know if it is always that easy to determine if a DVD is follwing ITU or not.

Just look out for a clear bowl-reference-object in the video like a planet, or a Football. That one wont suffer from perspective distortion when the movie kamera catched it as its 3D-Round.

IMO to the differences between ITU and non-ITU are too small to worry so much about.

I dont agree, the diff between a PAL 720px anamorph scaled to incorrect 1024px compared to correct 1050px is at least to me significant noticable.

Even more so because even if you are following the correct way you are not guaranteed correct output anyway.

Live isn't 100% (if you wont find a bowl in the vieo) but we at least try to keep the correct way for ourselfs

[Mug Funky]

Originally Posted by wilbert

...or whether a significant part of the DVDs out there don't follow ITU.

i'd say that menus and interactive elements probably never follow ITU, but just stick to "make it in 768 or 1024, and resize to 720".

as for actual content, even theoretically compliant stuff is often out of aspect (old telecines tend to be vertically stretched and cropped heavily compared to new transfers). and stuff done digitally will just obey the digital video standards. it's up to the DVD player to make things correct then.

i think so long as circles look round and squares aren't rectangles, it's all fine. remember also that TVs themselves are almost always wrong (even my HR trinitron with SDI inputs etc shows a different aspect in underscan mode). then consider the angle you're watching the TV from - only one person can watch from dead-straight on...

[Inc]

it's up to the DVD player to make things correct then.

Yes thats a real point. I did test some outputs on 4 diff. SAP Brand models and all do treat 720 sources as 53.333µs @13.5Mhz. They all did follow the ITU. Means in my case of PAL the PAR equals to

14.769Mhz (PAL TV)
–––––––––––––––––– = PAR 1.09402:1
13.500Mhz (SAP/ITU)

Two of the testet SAPs where mpeg4 supporting ones and so these ones also got an option for "generic" mpeg2 playback. In that case 720 was treaten by generic PAR.

i think so long as circles look round and squares aren't rectangles, it's all fine.

Keep in mind that just "circles" aren't the way of proof as circles will get a perspective distortion depending on the Camera view. Only "Bowl" Objects wont suffer from that as they are 3D round. Thats why I mentioned the example of "Planets" above. Do check those ones within the actual Movie and not in the Movie-Companies intro.
Once I checked a Stream by exermining it's "Working Title" Intro. It was wrong but the main movie itself was correct.

[GodofaGap]

Planets aren't always perfect spheres either.

[Inc]

But I dont think that there do exist planets in reality which are elliptic by a factor of 768/720

[GodofaGap]

But the difference we are talking about is (768/720)/(768/704). And yes, the flattening of Jupiter and Saturn at the poles can be observed through a telescope.