Actual SAR for PAL DVD?

[Gew]

Hi!

This has been bugging me for quite some while now, so it would be great if some enthusiast could lay the true colors of the cards on the table.

Like this. According to the Wikipedia article on PAL, the only true/official/standard-defined pixel aspect ratio values for PAL is either 12:11 (for DAR 4:3) or 16:11 (for DAR 16:9).

This is assuming the standard (704x576). However, as I've discovered, there is actually no PAL material (at least movie DVD's) that is 704x576. It's all 720x576.

Then I thought to myself, what's the deal then? Take this for example. I have this DVD. I want to rip it. I run my projected DGIndex file through an AviSynth script, providing the AutoCrop().

Now what, say the movie is presented in anamorphic 16:9. How do I fetch the true and correct aspect? I run my calculator, but shall I use "autocroppedwidth/autocroppedheight*64/45" or shall I use "autocroppedwidth/autocroppedheight*16/11". Some people (reference link (http://forum.videohelp.com/threads/247051-Why-only-704x576)) say that 720x576 is really shown as 704x576 on television sets, because the tv will simply crop 8 pixels on left and right. Is this true?

Suggestion in short:
PAL television sets (or PAL DVD players) does _never_ stretch 16:9 by 64:45, but always 16:11. Correct? If this is the case, it would be safe to assume that film studios know this, hence have cropped/mastered their PAL DVD copies in a way that only SAR=16:11 would achieve proper slash natural aspect ratio, correct?

Guys, I'm just confused, trying to proceed in life with the correct instruction crib sheet. Ty in advance for all the help.

[TinTime]

...it would be safe to assume that film studios know this, hence have cropped/mastered their PAL DVD copies in a way that the SAR=16:11 only would achieve proper slash natural aspect ratio, correct?

From my experience this is a reasonable assumption. It's not always true though and some material will be incorrect. Personally I always use 12/16:11 unless it's noticeably wrong.

I've seen a few SD (720x576) extras on blu-ray encoded in AVC and according to eac3to the display aspect ratio of these has always been 20x11, rather than 16x9 - i.e. they had a SAR of 16:11

[7ekno]

Options discussed HERE! (http://forums.whirlpool.net.au/forum-replies.cfm?t=1470507#r13)

Source article (in reference to 4:3 and BT.601) can be found HERE! (http://lipas.uwasa.fi/~f76998/video/conversion/)

In summary, 2 schools of thought:
- keep pixel aspect ratios to 128:117 (non anamorphic) or 512/351 (anamorphic) and crop to keep consistent
- alter container SAR to compensate for display aspect ratio & pixel aspect ratio

7ek

[Gew]

Ty for your answer TinTime!

Additionally, I found these two one-liners staggered in some old threads in this very forum.

However, I'd recommend checking the active image pixel area of any 720 pixel wide source, as many of them contain an active image pixel area greater that 704 pixels.


You should always crop to active region and calculate PAR according to that to get the best result.

Perhaps there's something to it?
I've come across a few DVD films that had black padding for the 8 pixels most far out right respectively left. I.e. active pixel area was 704x576. Perhaps then I should kick 704/576*16/11; whereas when I come across DVD's with active pixel area in it's whole 720x576 D1 frame I'd go for 1,25*64/45 instead? But then again, someone quote from some thread also read:

The correct PAR according to ITU standards is always 12:11 (full screen) and 16:11 (widescreen). The image area shouldn't actually affect it if the content was authored according to ITU standards.

Supposedly, I'll stick to using 16/11, always!? :confused:

[Gew]

For instance, I just came across this DVD (Michael Jackson's: This is it) with IMDB-reported AR of 1.85. However, the full 720x576 area is used for picture. I.e. no padding at all. Now, if I do 720/576*64/45, I end up with 1,7777777777777777777777777777778, i.e. exact 16:9 match. If I use 16/11 (or the more precise 512/351), I end up with a DAR on ~1.82. Now, what's real here? I find it more logical that the studios have cropped the 1.85 down to exactly 16:9, insted of ~1.82. But this is just my two cents, anyone else has any (brighter) ideas on this particular approach?

[TinTime]

Supposedly, I'll stick to using 16/11, always!? :confused:

Based on my experience I'd suggest always using 12 or 16:11 unless it's obvious that it's wrong - eg. people look too fat or there are some circles that look oval.

I wouldn't worry about using 512:351 instead of 16:11. You'll never be able to see the difference and cropping to 704 makes more sense than 702.

[Gew]

Very nice comeback TinTime. I'm getting more and more convinced that this (the way you explain it) is really the way to go. One thought suddenly struck me, that could possibly be the one definite fact that turns this whole thing into a clear cut case. Let me be brief. I thought, the common ground (PAL) standalone DVD player. What's their actual deal with aspect ratios? They have to be regulated somehow. Clearly.

It feels logic that dvd players has really only two SAR modes implemented. And as eg. Wikipedia article says that 16:11 respectively 12:11 is the true ITU-defined standard, I get a feeling that these are the only two SAR values programmed into common man's dvd player. I.e. 4x3-pictures goes 12:11 whereas 16x9-pictures goes 16:11. If this is the case, DVD production studios (must have) already cropped eg. 16:9 material somewhat on top/bottom, to have it presented as exactly 16:9 on 720x576 (when applying that 16:11 SAR). Do you follow? On the contrary, other thinkable scenario is that common DVD players in fact has four(!) different SAR values to use; i.e. 12:11/16:11 for 704px stuff, and 64:45/16:15 for 720px stuff.

Any ideas? :confused:

[TinTime]

No ideas :)

But I agree with you about stand alone players. My guess (and it is just a guess) is that they'll treat 4x3 as if it has a SAR of 12:11 and 16x9 as if it has a SAR of 16:11. There's no way for them to distinguish between different SARs as far as I know.

Anyway, in my completely unscientific survey of DVDs that I own I've found that using 16:11 almost always looks right, and using 64:45 almost always looks wrong. I'm not sure about the world of NTSC though.

[Ghitulescu]

If your intention is only to rip* the DVD, you don't have to worry about too much about the DAR and PAR, as the hardware player (ie the DVD player) should be able to cope with it (it depends also what standard the player follows - but the difference is too small and unnoticeable).

If you want to convert it to eg AVI, then you have also to think how you wanna play the file:
A. hardware (media tank, DVD player with divx support, etc**)
B. software (PC, laptop etc.)
in both cases you should also be aware whether
A1/B1. the player does or doesn't respect the DAR/PAR pair.
And for B there's the case whether
B2. the laptop/PC is also connected (DVI/HDMI) to a TV or not.

It might be that you computed the DAR/PAR for a plain stupid PC player (PAR=1), but when you play the file with your media tank on the TV you'll get eggheads. Or vice-versa. Because there are PC players that respect the PAR/DAR, there are also players that respect only the DAR (assuming the PAR is always 1) and players that do not respect DAR (only the SAR with an assumed PAR=1). The same goes quite similarly with the hardware players.

I think saving ~66% space is not worth the trouble. But it's not mine :p


*RIP = pure DVD COPY. The term was however wrongly spread as RIP+CONVERT2AVI within the underscene.
**hardware players must output PAL/NTSC/SECAM signals, so they have to revert the AVI back into an analog or digital (DVI/HDMI) video.

[Gew]

Thanks for your answer Ghitulescu, very kind of you.

However, I actually have some experience with the AR of "computer video". I have done my homework. AVI files generally play in square pixel PAR for example, since many splitters disregard of its PAR flag, even thought it's there (settable through MPEG4Modifier eg.). Matroska containers on the other hand, from my experience, tend to have software players respect their AR flags. Goodie-good. Also, if you are kinda poor and stuck with a tube television (like me), you have to take into consideration that stuff you [analogue] x-mit to your television set always gets stretched 12:11. A straight 640x480 .AVI thus gets a bit black padding in bottom/top on your 4:3 carthode tube television. Over the years, I've often thought to myself that it's kind of stupid that big fancy graphics card manufacturers like Nvidia and ATI couldn't just overcome that little obstacle, for instance by preparing for the stretch by x-mitting 5:4 instead of 4:3, so that the final picture (on tv) would look natural / as intended. Supposedly, you could get around this yourself, simply by picking a 5:4 resolution for your extended view (on tv). However, 1024x768 is the greatest resolution that analog tv-out supports. Aawaergh.

Anyways. Wow, I feel I kind of fell into a sidetrack there. The things I just mentioned wasn't really on topic at all, now was it? Stupid me. Anyways, about stand-alone players. I've got the feeling that nowadays lots of these make final picture proper despite the PAL stretch. For instance, I have this cheap ~30$ stand-alone player. When I load an AVI that is eg. 640x480, I get full frame on the tube television which it is SCART-connected to. Goodie! Also, I have just upgraded to an LCD television, so stretchy pixels are really not an issue anymore, whatsoever.

This thread, however, was originally meant to shed some light over which PAR factor you are supposed to use when transcoding PAL DVD into eg. MPEG4. Well, I must say, I believe I have -- finally(!) -- come to find closure in this question. Not from here thought, but from the other thread which 7ekno referred to in the 3:rd post in this thread (http://forums.whirlpool.net.au/forum-replies.cfm?t=1470507&p=-1#bottom).

So, I believe we could consider this one clear! :)

Thank you all!

Regards~

[Dogway]

SAR examples

4:3 = 12:11 pal = 10:11 ntsc
16:9 = 16:11 pal = 40:33 ntsc

That is first example if you have padding (most of the sources). Be careful as some 16:9 contents are embedded as 4:3 (with horizontal bars).

[Stephen R. Savage]

Dogway, that is incorrect. The 16 extra pixels do not change the pixel aspect ratios, which are set and defined to be 12/11 and 16/11 (approximations of slightly higher-order fractions). If the picture has visual information in the extra columns, that simply indicates that the picture aspect is not exactly 4:3 or 16:9.

[Dogway]

Dogway, that is incorrect. The 16 extra pixels do not change the pixel aspect ratios

True, they dont change PAR, they define it.

[Dogway]

I see, so this (http://forum.doom9.org/showthread.php?p=583855#post583855) guy is wrong?
I have seen those values in many other places.

Anyways this is another thread (http://forum.doom9.org/showthread.php?t=113126) with the same debate. You might be true, but nobody has made clear why all this confusion, so people will continue to encode at wrong AR (in case they are). I edit the post to be on the safe side.

[Ghitulescu]

There is no PAR in PAL/NTSC systems (SECAM ;) too), it's all timing* and bandwidth**. PAR appeared for the first time when the computers entered this game, as they need pixels.

Additionally, there's no perfect PAR, as the analog TV sets may have a variable size of area the electron beam sweeps (the older TVs had externally, user-adjustable potis***, the newer ones have them in E/EPROM and set using SERVICE MENU).

Third item, there are several flavours of NTSC/PAL regarding, in your case, the timing (how long a scan line is).

All these items concur to the fact that there are several correct PARs, depending on the application (pure analog, pure digital, mixed environments).

And PAR is a calculated value, not an imposed one (it might look so, according to the norms, but it's not). It is derived from the 4:3 physical format (fixed), the number of scanlines (fixed) and the timing (fixed, several flavours).

*timing - max. samples (pixels) allowed - defines where the sample is taken (temporal distance between distinct samples).
**bandwidth - defines how many distinct samples/pixels a video signal carries
***adjusting the potis one could stretch the image at will

[Gew]

Guys, I'm really happy that you revived this thread, even though I at first thought I had it all figured out. My main question in all this is probably obvious by now. I have PAL DVD's, and need to know whether I should use 64:45 or 16:11 for PAR calculation when transcoding to MPEG4. I had an answer on another forum that felt fulfilling at the time:

720 with 64:45 will generate a very small aspect error ...

704 with 16:11 seems to be the best compromise for mod16 + close enough AR ...

702 with 512:351 is BTU.601 compliant, but is an old standard that todays DVD players may not even adhere too (since MPEG-2 requires mod16 and 704 is mod16) ..

However, I've been lying in bed for the last couple of nights, repeating the scenarios over and over in my head, trying to figure out the real deal. Now I really need your help guys.

Take the scenario as follows for example. I DGIndex a DVD. I notice that the material is to be presented in DAR 16:9 format. Fine this far. I load the .d2v in an Avisynth script. Now, I open my .avs in Media Player Classic. I get a 720x576 picture. This(!) is where I'm starting to get all confused. There is no padding in the far out left/right of the picture, but some in the top/bottom. I add AutoCrop() to my script and the resulting image suddenly turns out as 720x554.

This(!) is where I get really(!!) confused (and thus need fundamental help understanding). I run calc.exe in my Windows environment. I try 720/554*64/45. I get 1,848... According to IMDB this film is supposed to have AR 1.85, so this feels really good. When I try the same math, but instead using 16/11, the result is 1,890.. -- i.e. less probable.

So(!!)..

I re-refer to the guy stating the quote that I included here above. I was told to crop 16 pixels (8 left; 8 right) from the picture, and then use 16/11. Of course, when I do this I get the same result as with 720 and 64/45. But anyways, this(!!!) is the part that keeps me wondering in the night time. I think to myself, if I crop these 16px worth of actual picture, I must indeed be messing with the intended AR, if not this, then at least chopping (cropping) parts of actual picture. This would just be messed up, wouldn't it???

The math doesn't add up. The only thing that would make sense to this is if the "DVD lab" has actually pre-cropped some in top/bottom. Have they? I have come across (in some rare occasion) a scenario where the AutoCrop() actually leaves me with eg. 704x554, i.e. some black padding also on the left/right. This made sense, according to what I (think) I know. But hey, still, 99.9% of the PAL DVD's I've come across has no black padding at all in their left/right.

Oh gosh, I don't know if anyone has even bothered reading my post to this very point. I may even be "confusing myself into a corner", somehow. I try to narrow it down to some final questions for the enthusiastic experts here to answer, that perhaps will give me peace.


Question #1.
Are PAL standalone DVD players set (BT.601 regulations?) to always(!) do 16:11 PAR stretch for 16:9 material; respectively 12:11 for 4:3 material?

___Or___ do they have four different "modes", i.e. 64:45 for material without 16px black padding left/right; but 16:11 for material with 16px black padding left/right.


Hmmm. I think that was it. I need no more questions. That's the one big question whose answer will reveal this mystery. For example, I would be very much delighted if someone would step in here and drop the following statement, I quote what I would really wish for someone to say:

"When resizing/interpolating your material, you should always use 16:11 (for 16:9) and 12:11 (for 4:3). This, and only this, is in accordance with the BT.601 standards. If you come across PAL DVD material that has actual picture (in opposite to padding) all the 720 pixels, it is safe for you to assume that the DVD director has already cropped some from the top/bottom, so the only thing you can really do to preserve natural AR is to crop left:8 right:8 pixels and then apply the 16:11 or 12:11 PAR for calculation of your targeted resizing. Good luck!"

Of course, if this is all wrong, you should not post it just to give me some sort of false comfort/closure. Then instead do your best of trying to rock me into true facts.

I hope someone has managed to read without thinking "TLDR" to themselves.

Regards~ :confused:

[7ekno]

It's pretty easy to discern (the final question) ... author a 16:9 720x576 DVD as a green background with a fixed amount of red pixels on the left and right (I used the first 1/3 of the screen for 8 pixels of red, then the next third for 9 pixels of red and the bottom third for 10pixels of red) ...

Play in DVD player, do you see the red?!?

On my HDMI DVD player on a HDTV I see the bottom two thirds with red lines (ie 704 pixels used), on my RCA standard DVD player on an old CRT I only see the bottom third of red (702 pixels used) ...

Which is why I said:
720 with 64:45 will generate a very small aspect error ...
704 with 16:11 seems to be the best compromise for mod16 + close enough AR ...
702 with 512:351 is BTU.601 compliant, but is an old standard that todays DVD players may not even adhere too (since MPEG-2 requires mod16 and 704 is mod16) ..

7ek
PS on a work Sony DVD player on a HDTV you see all areas of red, but have an option (in the settings) to "crop 8 samples" ...

[Gew]

Ty alot (again) 7ekno.
Now, there's really only one thing that "bugs" me.
I've thought of a good scenario for you to discern from.

Picture this..
I have a PAL DVD containing some Prison Break season. According to IMDB, this show has an intended AR of 16:9. Therefor, I also expect the full 16:9 picture to be present on my DVD. This may be where I go wrong, simply by assuming that the DVD I'm holding in my hand must contain non-cropped (thus modified aspect ratio) material. I move on, and check the MPEG-2 VOB's on my disc. They are (of course) 720x576, and have no vertical (nor horizontal) padding whatsoever. Simply a full picture. I then apply the "formula" (64:45) that I have learned about. The result is 1,7777777777778, i.e. a perfect 16:9 match.

"If you are a purest, you crop to 702 and apply 512/351 (as this is how the directors meant you to see the DVD!) ... if you feel encoding at 702 is inefficient, then move to 704 (mod 16) and apply 16/11 ... if you feel you are missing detail from the 720 pixels on the DVD use 64:45 ...

But keep in mind, as the poster above notes, anything beyond 702 pixels wide slightly alters the "Aspect Ratio" in which you view at (after all, stretching 720 pixel to 1024 has to look different to cropping 18 pixels then stetching to 1024!) ..."

So I'm supposed to crop 8 respectively 8 pixels left/right, and then apply the (ITU-regulated) SAR value of 16:11. This is how I get the picture "as the director has intended for me to see it". Now, this is also the point where I get really confused. I mean, I have just cropped 16px worth of _actual Prison Break video_ from my DVD. Now how could this be my "intended view"..?

I have been lying in bed, and finally figured out one possible answer, an answer where all this would make sense. If you could return a simple boolean value for an answer on the following statement, it could very well be the final piece of the puzzle :) So, here goes:

"If your 720x576 frame does not hold 8 pixels of vertical black padding in the far-out left and right, it most certainly means that the DVD author has already cropped the source somewhat in the top/bottom. This to preserve it's natural aspect ratio."

True / false? :P

If this is false, I'm unfortunately back to square one. I can't get it in my head. When 16px of actual material are cropped, the author _must_ have tampered with (cropped) the top/bottom so it can add up. Do you get my point of view here Mr. Tek? Would be awsome if you could shed some light over this! :)

Regards~

[Ghitulescu]

How do you intend to play the files?

[Gew]

720 with 64:45 will generate a very small aspect error ...
704 with 16:11 seems to be the best compromise for mod16 + close enough AR ...
702 with 512:351 is BTU.601 compliant, but is an old standard that todays DVD players may not even adhere too (since MPEG-2 requires mod16 and 704 is mod16) ..

Now, I have the movie "Benny & Joon" on PAL DVD. According to IMDB, this film is shot in an aspect ratio of 1.85. Everything cool, this far. Now I slide the DVD in my computer. I load one of the main movie VOB's. It turns out as 16:9 (anamorphic encode), 720x576.

There is no "slight pillarboxing" (8px padding left; 8px padding right), but the actual movie is presented within all 720 horizontal pixel samples.

Now, I am aiming for an MPEG-4 square pixel encode. Period.


So, here comes my great mystery..

When I AutoCrop() the VOB file, it results in a 720x554 output. When I do the "formula" that I've used earlier (64:45), it makes pretty good sense; i.e. 720/554*64/45=1,848.. (~1.85), what IMDB refers to as correct AR for this movie.

However, I was told that DVD players A/D converter(?) automatically crops the picture down to 704px width, and then applies the correct formula, which is 16:11. Man wrote a great clarification about this on another web forum:

Author a 16:9 720x576 DVD as a green background with a fixed amount of red pixels on the left and right (I used the first 1/3 of the screen for 8 pixels of red, then the next third for 9 pixels of red and the bottom third for 10pixels of red). Play in DVD player, do you see the red?!? On my HDMI DVD player on a HDTV I see the bottom two thirds with red lines (ie 704 pixels used), on my RCA standard DVD player on an old CRT I only see the bottom third of red (702 pixels used) ...

Anyways. So, I crop 16px and re-do my math: 704x554*16/11. Now I (of course) get the same satisfying figure of 1,848, i.e. same as with 720 and 64:45.


Here comes my huge mystery. I _know_ for a fact that the -- now given -- AR of ~1.85 is suddenly not to be taken as correct. Why? Because I have just cropped 16px worth of movie. Do you understand my way of thinking here? It just simply won't add up. So I've been lying, thinking about this. There are only two real "answers" I could think of.


#1) When you run across 720x576px VOB's with all horizontal samples filled (i.e. no slight pillarboxing), it's safe to assume that the DVD author has already cropped the source slightly in its top/bottom, for math (intended AR of 1.85) to add up in the end. Therefor, you should do exactly as told, crop to 704x576 and then apply 16:11.

#2) When you run across 720x576px VOB's with all horizontal samples filled, you should actually use the 64:45 formula.


These two "answers" are of course hypothetical, and that's where you guys come in. Please fill me in with the real deal here. And 7ek, if you read this, I know I have recently received theories about this by both you and Pman on that other board, I simply "re-bump" to sort of put out a carpet of bombs (receive as many thinkable answers as possible). Hope you don't mind! ;)

[Ghitulescu]

Author a 16:9 720x576 DVD as a green background with a fixed amount of red pixels on the left and right (I used the first 1/3 of the screen for 8 pixels of red, then the next third for 9 pixels of red and the bottom third for 10pixels of red). Play in DVD player, do you see the red?!? On my HDMI DVD player on a HDTV I see the bottom two thirds with red lines (ie 704 pixels used), on my RCA standard DVD player on an old CRT I only see the bottom third of red (702 pixels used) ...
This guy obviously never heard of overscan (for the TV set), because this is exactly what he measured. BTW, there are also free and commercial overscan-measuring-tests, guess what, they fill the whole 720x576 area (both 4:3 and 16:9).

I told you in the PM how the pixels are output into PAL/NTSC/SECAM systems. Forget NTSC, forget PAL, forget the DVD players (imagine, there are many more D/A-chips than your information source might think) you want something that is ~1.85 meter wide and 1 m height, or the same in square pixels, where the vertical pixels are known: 544-your value.

[Gew]

This guy obviously never heard of overscan (for the TV set), because this is exactly what he measured. BTW, there are also free and commercial overscan-measuring-tests, guess what, they fill the whole 720x576 area (both 4:3 and 16:9).Yeah, in fact I think his point was illustrating how television overscan works, in some simplified way, to make me understand :) However, I've come to my own sort of conclusion that 64:45 is in fact more correct than 16:11. It just feels that way. I've fiddled with a pile of PAL DVD's. When I DirectShowSource(dvdmaterial.vob).AutoCrop() Almost without exception I get the actual IMDB-reported aspect ratio when applying width/height*64/45. When I apply x16/11, I get a wider aspect, eg. ~1.91 for films reported 1.85.

In my book, this fact speaks for itself. It just has got to be that way. In fact, I posted on another video enthusiast forum (http://forum.videohelp.com/threads/322664-SAR-for-PAL-16-9-DVD-s-16-11-or-64-45) and someone over there told me that 64/45 is the correct formula to use, unless there is a "seq ext" value present through GSpot VOB properties. It just feels right.

[Ghitulescu]

I've fiddled with a pile of PAL DVD's. When I DirectShowSource(dvdmaterial.vob).AutoCrop() Almost without exception I get the actual IMDB-reported aspect ratio when applying width/height*64/45. When I apply x16/11, I get a wider aspect, eg. ~1.91 for films reported 1.85.

Well, I think it's not offtopic to mention that the DARs imdb is reffering to are the film DARs (what you see in cinemas). There are also movies shot in plain 4:3 but projected in 16:9 or other DARs, to avoid some props (usually mics) become visible, or to hide "other details".

When the film is scanned for DVD distribution, it up to the operator (or the instruction s/he got from the distributor) to pick up the desired DAR, which may but also may not be the original one. I don't know in other countries, but in DE one usually finds "budget" DVDs that have the same movie, just in 4:3 (the P&S version usually, some have the "original" DAR though) and with local doubling only (no subtitles as well).