Talk:Pixel aspect ratio

(Untitled)

Aspect ratios may not be a perfect 1:1
http://www.mir.com/DMG/aspect.html
The difference is probably unimportant, unless you are (for example) supersampling. Still, anyone who actually understands PAR is welcome to update this :-) I would probably make it worse with accidental errors.

Updated! Fleet Command (talk) 06:10, 17 November 2008 (UTC)

Digital Video Processing: "Second:" section appears to be inaccurate

This section does not cite any sources and appears inaccurate. Specifically:

1. "The Pixel Aspect Ratios approved by ITU-R BT.601 were no longer applicable for non-linear editing (NLE) of digital video clips with full 720 pixels, if they were to be mixed with pictures of varying video sources, types and Pixel Aspect Ratios." Why are these values no longer applicable for NLE? As the article sources seem to support (see Chris Pirazzi and Jukka Aho links), if clips of different sources, types and PARs are to be mixed, the PAR needs to be changed by scaling, cropping and/or padding. You cannot just change the PAR directly, that will give incorrect results.

You answered yourself: You said "...PAR needs to be changed". That's it: When PAR is changed, it is no longer the ITU-R BT.601. Fleet Command (talk) 10:25, 13 March 2009 (UTC)

I am probably not explaining my point well. My point was that you can't just reassign a PAR, you have to physically scale/deform the image in order to change the PAR (of course cropping/padding doesn't directly affect PAR, sorry if I was confusing on this point). If you take a 720x480 image with PAR of 10/11, keep the image dimensions the same and assign it a PAR of 8/9, then your circles will no longer be circles, etc. When mixing clips of different sources, what is important is that the dimensions and PARs are the same. I don't see any reason why the ITU-R BT.601 values can't be used, again as long as you convert all of your clips to that. - Laur-1 (talk) 16:28, 13 March 2009 (UTC)

It seems you were explaining yourself perfectly well. But I understand that the matter might be difficult to understand. I try to resolve it but let me explain a bit: NLE-specific PAR is used in digital images, long since they are digitized. Its pixels were once non-square. But more importantly, they had out-of-viewport sections. Now, all their viewport must be rendered on the computer. They were sampled with a frequency that grants 704 pixels but they are 720 pixel in the same space (centimeter-wise)! The point of NLE-specific PAR isn't watching them; the purpose is editing them, sometimes at a pixel level. So, you see, image is rendered different; computationally-correct,not visually-correct, because you want to edit the computer data not visuals. Check Chris Pirazzi's article on this matter, but I try to clarify it with some screenshot. Not very soon -- I got mid-term exams -- but as soon as I can. Fleet Command (talk) 18:27, 13 March 2009 (UTC)

But squeezing an out of viewport image into a 4:3 frame is not "computationally correct", it is just plain wrong! You have not explained how changing the PAR helps computationally. I have read Chris Pirazzi's pages, Jukka Aho's, the www.mir.com link at the top of this page, and others and they all seem to support what I am saying, for example how to convert between square and non-square pixels http://lurkertech.com/lg/pixelaspect/#square_nonsq_convert. Nowhere in these sources can I find any mention of the "NLE specific PAR" of 8/9. Indeed, I can't find a good citation of this value anywhere. Can you cite sources which support this? A screenshot would be very helpful once you find the time. - Laur-1 (talk) 20:01, 13 March 2009 (UTC)

Computationally, it is correct. Visually, not. You are confusing the usage of natural PAR with NLE-specific which is only used in Non-Linear Video Editing programs. Read the answer to your question 2 for more explaination. Fleet Command (talk) 10:45, 14 March 2009 (UTC)

2. "Video editing applications had to employ computationally-correct NLE-Specific Pixel Aspect Ratios for 720×480 and 720×576 motion pictures that were to be rendered on 4:3 or/and 16:9 screens of standard-definition TVs." Again, you cannot just change the PAR without screwing up the image. As mentioned earlier in this very section, 720×480 and 720×576 are NOT 4:3, rather they are a bit wider. If a NLE tries to squeeze a 720x480 or 720x576 image into an exact 4:3 frame (as these NLE PARs do), they are doing it wrong and will have incorrect results. I am not aware of any reason why they have to do it wrong, it is certainly possible to use the correct aspect ratio (as Adobe now does).

This statement does not contradict you. In fact, it is confirming your assessment: Trans-PAR conversion do reduce image quality. They create split pixels that are to be refilled with whole pixels. That is why NLE-specific PAR is used: To avoid this problem by handling the image the way it technically (not visually) is. Fleet Command (talk) 10:25, 13 March 2009 (UTC)

I'm sorry, I don't understand your statements her at all. Can you please explain further? What do you mean by "tran-PAR conversion"? What do you mean by what the image "technically" is? The image is NOT "technically" a 4:3 image, although these NLE PARs want to force it to be. It seems to me that any way you slice it, using a PAR of 8/9 on a 10/11 image is just plain wrong. If you open a 720x480 image with a PAR of 10/11, reassign it a PAR of 8/9, draw a perfect circle, then export back to 720x480 PAR 10/11, your circle won't be a circle anymore, it will be an ellipse. If you need to convert the image to square pixels for ease of digital editing, one correct way would be to crop to 704x480, then scale to 640x480, or even pad the original to 880x480 and scale to 800x480 if you don't want to lose any data and don't want fractional pixels. - Laur-1 (talk) 16:28, 13 March 2009 (UTC)

Trans-PAR is exactly what you explained above: Converting an image with one PAR into an image with a different PAR without causing distortion (or as you said "screwing up the image"); without causing circles to become ellipse. This means the new circle will look a bit blurred but human eye may ignore that if rendered properly with certain techniques such as anti-aliasing.

By trans-PAR do you mean changing PAR by resampling (scaling) the image? Or by just assigning a new PAR with changing the image geometry in any way (as the NLE PAR does)?

Resampling? Oh, sure. You see, your pixels change shape and the total number of pixels in width and length are different. But, the width and length of the image remains the same. This new image must now show the same image as before. It's like a "paint the brick" game: A set of colored bricks are put together, forming a chunk of rectangular pavement. Each brick has a different color. Together, they show a picture. Now, you are given another chunk of pavement but the brick sizes are different. You are to paint the same picture on this whole chunk but you are only allowed to paint each brick with one color. That's analogy for how pixels are transformed.Fleet Command (talk) 10:45, 14 March 2009 (UTC)

Now, you did understand something wrong: NLE doesn't distort anything. It is there to not to distort.

How can assigning a different PAR not cause distortion? You are changing the pixel shape. - Laur-1 (talk) 20:01, 13 March 2009 (UTC)

Because we are not assigning it to anything. NLE-specific PAR simply exists. Is it not obvious?

Here is how it is: We digitize an analog picture, say a 480i picture, to a computer image. We have read standards: We know that we must extract pixels with a certain frequency. We get non-square pixels with a certain Rec.601 PAR, in this case 10:11. We also know the TV's display aspect ratio, in this case 4:3. So we know how many horizontal pixels we get: In this case: 704. (Vertical length is already well-defined.)

However, where are the horizontal edges? We don't know where the image starts! We only know its center! Blame is on the old standards, yes, but what should we do? Well, the only safe practice is to go a bit further from the center. So we pick up 720 pixels. But hey, wait, this way the resulting image is no longer 4:3! But owners of the motion picture don't care if image is 4:3 or not. They just want their property in whole!

OK. Now another problem: What would happen if we display this image that is not 4:3 on a 4:3 screen? Well, it's edges gets obscured. People don't care. Some TVs obscure more and some less they are used to it. But for developers and video editors, which are digital mechanical engineers and must make movie rather than seeing them, it is important. All the pixels must be displayed flawlessly on their screen! The image must be squeezed into 4:3 frame. Since all pixels must be displayed, there is no choice but pixels themselves being squeezed and change ratio.

So you see: A group of people want entertainment and want pictures intact. Another group want pictures squeezed so that they are able to edit them for those who don't! So, the picture PAR remain what it always was: Rec.601, in this case: 10:11. But video engineers and computers see the image the way it "technically" is, in their Non Linear Editing applications. They have their video editing application wear a pair of glasses called NLE-Specific PAR. Fleet Command (talk) 10:45, 14 March 2009 (UTC)

I believe I understand most of what you are saying elsewhere, and while I have some additional comments I won't bother responding so we can focus on the crux of the matter. To attempt to be very clear, a 480i image has a SAR of 720:480, a PAR of 10:11 and a DAR of 15:11 (NOT 4:3). I believe we are both in agreement there. The issue is that you are making the assertion that a NLE MUST display a 480i image on a 4:3 DAR screen, even though you acknowledge that this is not correct and will "squish" the image. But nowhere do you say WHY this must be true! We are talking about a Non-Linear Editing software application running on a computer. Computer monitors are not limited to 4:3 like an old analog television. There is absolutely no reason a NLE cannot display a complete 480i image with its proper DAR of 15:11 (with no image area cut off) instead of a false DAR of 4:3. Any NLE worth its salt should display and work with 4:3, 16:9, 1.85:1 or anything really. Would you say that a 16:9 source must be squished into a 4:3 DAR? What's the difference with a 15:11 source?

I feel like we are just going round and round here and not making a lot of progress. At this point I must ask again that you provide a source or citation for your assertion that NLEs MUST display a 480i image with a DAR of 4:3 using these "NLE specific" PARs. I would also like to know WHAT NLEs you are referring to. For example, the only software mentioned so far (Adobe Premiere Pro CS4 and Adobe After Effect CS4) in point of fact DO NOT use these "NLE specific PARs" (although they used to) and now use the correct ITU-R BT.601 values. You still have not sufficiently addressed this point, if you read the Adobe After Effect CS4 documentation link it is clear that the ITU-R BT.601 values are to be used by the software end users (your "consumers"), not some "developers" (which is who, Adobe?). If you cannot provide valid sources I propose that this section be removed from the article. - Laur-1 (talk) 02:53, 15 March 2009 (UTC)

1. You must understand that computers do not see video files they way you human see the video. They live in their own binary world and do not understand how can a 720x480 picture with 10:11 PAR can be displayed on a 4:3 screen. They understand nothing but well-defined edges. But you don't care because you get your video right. So, the bottom line is:

In computer's mathemitcally-flawless language, NLE-specifc 8:9 PAR for a 720 pixels wide picture means Rec. 601 10:11 PAR in our human language. With humans, talk 10:11; with computers, talk 8:9.

This makes absolutely no sense. In point of fact computers only understand square pixels, and have absolutely no preference for 8:9 PAR over 10:11 PAR or anything else other than 1:1 PAR. - Laur-1 (talk) 14:46, 16 March 2009 (UTC)

Apparently, you have adjusted your mentality that I'm wrong.

I apologize if it seems that way, I assure you that was not my intent. - Laur-1 (talk) 21:47, 16 March 2009 (UTC)

Sorry, dear friend. I think it is I who should apologize. I should have not written this. I'm sorry. I don't know why have I become touchy recently. Fleet Command (talk) 14:47, 17 March 2009 (UTC)

Computers do understand nonsquare pixels. Otherwise, no one could have developed program to handle this case.

No, computers really don't "understand" non-square pixels. Your NLE or video player can set a flag on a video source to any PAR it wants, and your NLE or video player can scale the image by that ratio so that it appears correct on your square pixel monitor, but the fact remains that a computer monitor is a square pixel display, quite unlike a TV monitor, and has no inherent preference for any PAR other that 1:1 square computer pixels. - Laur-1 (talk) 21:47, 16 March 2009 (UTC)

They perfectly understand that according to your SAR * PAR = DAR equation, if SAR is 720x480 and DAR is 4:3 then PAR is 8:9.

I agree with your math, however, a 480i (720x480) source does not have a DAR of 4:3!!! You have acknowledged this multiple times, so I admit I am unsure where our disconnect lies. Do you feel that a 480i image ceases to be a 480i image when it is opened in an NLE? - Laur-1 (talk) 21:47, 16 March 2009 (UTC)

And it is: The real PAR of 720x480 on a 4:3 screen is 8:9. It is the PAR of the 480i that is 10:11. Fleet Command (talk) 20:07, 16 March 2009 (UTC)

I'm having a tough time reconciling this statement. A 720:480 image is a 480i image in this discussion. I acknowledge that you could create a 720:480-SAR 4:3-DAR video stream in your editor of choice, but that is not what we are discussing. We are discussing opening up your ITU-R BT.601 480i 720x480 image with 10:11 PAR (such as from an NTSC DVD or a DV source) in your video editing application. You are arguing that it is correct to change the PAR to 8:9 to fit this into a 4:3 DAR. I'm arguing that in order to achieve correct results, you must leave the PAR at 10:11, and the DAR at 15:11.

I think perhaps the article should be updated to include the concepts of production aperture and clean aperture to help clarify matters. The 720:480-SAR 15:11-DAR represents the production aperture. This contains a smaller clean aperture of 704:480-SAR 4:3-DAR. You are arguing that a NLE should put the entire production aperture into the 4:3 frame. I'm arguing that this is incorrect (by definition, really), the production aperture is wider than 4:3 by design. Only the clean aperture should be in that frame, the rest of the image lies outside of it. Your NLE should understand the concepts of production aperture and clean aperture, as the Adobe products do, and should not blindly assume that 480i content is 4:3. I think the video I linked to explains this fairly well. Did you watch it? - Laur-1 (talk) 21:47, 16 March 2009 (UTC)

2. Yes, during a video editing job, the image must fit into the screen because you don't have the luxury of constantly scrolling it every single frame merely to see 8 pixels of critical edges that, if are missed and not seen, can catastrophically impact the revenue.

Scrolling? Why on earth would you need to scroll to see the image? I am currently writing this on a 1680:1050 SAR, 16:10 DAR, 1:1 PAR screen, I can assure you that I can display a 720:480 image just fine with no need for scrolling. Again you are asserting that modern software applications simply MUST squish everything into a 4:3 frame, but have yet to tell me WHY this is so. - Laur-1 (talk) 14:46, 16 March 2009 (UTC)

A video engineer don't complian a slight squeeze because he gets his money and customers see a flawless movie.

If he cares about accuracy a video engineer damn well should complain. The difference between 15:11 and 4:3 is small enough that it will go unnoticed by the majority, but the movie certainly won't be "flawless" if it is edited with the wrong PAR, especially if any geometry dependent edits (the examples I have given are a radial blur, or drawing a perfect circle or square). - Laur-1 (talk) 14:46, 16 March 2009 (UTC)

This especially becomes true when you are adding a 3D-Rendered object to a scene. But if it satisfies you, I take back my "MUST" and say "SHOULD".

That's a start, but you haven't yet provided any rational or citations that these values even SHOULD be used. - Laur-1 (talk) 14:46, 16 March 2009 (UTC)

You can edit your videos in preview mode on four wall-sized 16:9 monitors so that you see details big enough and are not bothered by bolts and nuts of user-interface.

3. No, we are not removing this from the article. We only overhaul, clarify and re-write. When my colleagues and I re-wrote this article, we asked a number of people to read it and give us suggestions. We spent weeks researching and what you see is the results. We didn't added NLE-specific issue only to fill pages. We added it because we felt it is critical. If you fail to understand the article, that's OK, I'll try my best to explain as much as needed. Then, together we can clarify the wording and add screenshots even if necessary. Fleet Command (talk) 08:35, 15 March 2009 (UTC)

I think it is perfectly within Wikipedia's guidelines to remove any unsourced and unverified claims, please see Unsourced Material and Verifiability. I don't know who your "colleagues" are, but this has a scent of original research. If you have "weeks of research" supporting your assertions here, please cite it!!! You steadfastly refuse to provide any citations for your assertions, and the sources we have available seem to refute you. I will attempt to provide citations from the sources we have already identified:

Adobe Premiere Pro CS4 Documentation

"When you capture or import NTSC footage with the ATSC frame size of 704x480, the D1 frame size of 720x486, or the DV frame size of 720x480, Adobe Premiere Pro automatically sets the pixel aspect ratio for that asset to D1/DV NTSC (0.91)."

Adobe clearly indicates that they are using the ITU-R BT.601 values of 10:11.

Maximum Impact Research Digital Media Group

Standard (non-widescreen) provides a frame size of 720x480 (DV-NTSC) or 720x576 (DV-PAL) using appropriate Rec.601 pixel aspect ratios. Thus, according to the previous calculations, these frames contain a visual image which is slightly wider than TV's 4:3 display aspect ratio. To be displayed properly in a 4:3 screen or window, the frames should be cropped by 8 pixels on each side.

Notice no mention of using a different PAR to fit a 480i image into a 4:3 frame, instead they correctly recommend that the image be cropped (maintaining the ITU-R BT.601 10:11 PAR) if it is required to fit a 480i image into an exact 4:3 frame. Of course, for editing purposes you like don't want to crop, instead you should display the image in its correct 15:11 DAR, 10:11 PAR.

Chris Pirazzi: Programmer's Guide to Video Systems: Computer Industry Mass Confusion on Pixel Aspect Ratio

What mis-ratios did we use? There are the obvious mis-candidates, such as 640/720 and 786/720..." (emp mine).

There's your 8:9 PAR, clearly identified as being incorrect. In this section Chris discusses the confusion that is at the heart of our discussion here, and clearly states (as other sources do) that the only correct PAR for 480i sources is the ITU-R BT.601 value of 10:11.

Until we resolve this I have added dubious tags to the article referencing this talk section. - Laur-1 (talk) 14:46, 16 March 2009 (UTC)

Update: added another source. Please read over Adobe After Effects CS4 documentation carefully, including the comments. There is a link to a video by Chris Meyer at http://movielibrary.lynda.com/html/modPage.asp?ID=701 called "New Pixel Aspect Ratios". This clearly explains why the ITU-R BT.601 10:11 PAR is correct, and why what Adobe was using before was wrong. - Laur-1 (talk) 15:26, 16 March 2009 (UTC)

Well, let's see: Despite having midterm exams, I spent so much time explaining everything to you, yet you simply ignore what I write. (Did you even read them?) I treated you with respect and promised you an overhaul which includes screenshots and inline citations when I'm done with my exams, yet you abuse my lack of having time to go back through sources and accuse us all of Original Research.

I need not argue with you anymore. You are obviously not intent to come to a resolution or mutual agreement. My exam will be finished soon and the article will be flooded with screenshots and inline citations. Until then, I think I can tolerate a few tags. Fleet Command (talk) 20:07, 16 March 2009 (UTC)

I do not intend to be disrespectful. I sincerely apologize if my responses have come off that way. Plain text is a poor medium for communicating emotions and intents, as I'm sure you realize. I have not ignored anything you have written, however I just do not agree with it in light of the sources presented. Specifically, the Adobe After Effects CS4 documentation (and Chris Meyer's video linked from there) and Chris Pirazzi's articles seem to flatly contradict what you are asserting. I will admit that I am bit frustrated, as I started off this section asking for sources to support your arguments, have asked for them again multiple times, and through 5 levels of replies you have yet to provide any. However at no point have I intended to be disrespectful or abusive. If you have supporting sources, great! I eagerly look forward to reading them and hopefully learning something in the process. Take all the time you need, there is no pressing need to resolve this immediately (I myself will be going out of town for several days soon). I would respectfully ask that you wait until we have resolved this issue on this talk page before you go update the article, as I have refrained from changing the article (beyond adding some tags). Good luck on your exams! - Laur-1 (talk) 21:47, 16 March 2009 (UTC)

Thank you very much. Actually, it is I who should apologize. Looks like I have recently become a little touchy. By all means, as a Wikipedian, you have the right and authority to tag anything that you don't find its source, inquire about it or outright delete it. It was my mistake from the begining. I should have provided inline citations. Thus, I have no right to feel disrespected at all. So, please forgive me.

I think you are right, after all. We had better delete NLE-specific aspect ratio from the article. PAR itself is a bit confusing, given the fact that so many sources saying so many different things. And now that I think, application development tricks like NLE-specific aspect ratio don't belong to a general article like this. You have not only my consent but also my support to go ahead and exterminate everything related to NLE-specific aspect ratio. Again, I apologize. Fleet Command (talk) 14:47, 17 March 2009 (UTC)

Apology very gratefully accepted! I do understand how easy it is to become protective of your contributions, I have certainly been guilty of that myself on occasion. I am very glad that we could reach a consensus on this issue. I have gone ahead and updated the article to remove the NLE-specific PAR references. Take care, - Laur-1 (talk) 16:23, 17 March 2009 (UTC)

The image length and width in pixel is independent from its PAR.

This is incorrect, they are related by SAR*PAR=DAR. You cannot change one without affecting the others. - Laur-1 (talk) 20:01, 13 March 2009 (UTC)

IN PIXELS NOT IN CENTIMETERS AND INCHES. READ CAREFULLY! (In other words PAR is independent from SAR.) This SAR*PAR=DAR equation applies to next sentence:

It's the image length and width in centimeters and inches that depends both on PAR and pixel length.

This is video, centimeters and inches doesn't enter into it, that will depend on the size of your TV. All that matters is SAR, PAR & DAR (Storage Aspect Ratio, Display Aspect Ratio, and Pixel Aspect Ratio, to be clear). - Laur-1 (talk) 20:01, 13 March 2009 (UTC)

IT DOES! Your TV screen has a specific width and length. Width and length of different TVs may be different but the ratios of their width and length are the same: 4:3 or 16:9. Now is the time to say SAR*PAR=DAR Fleet Command (talk) 10:45, 14 March 2009 (UTC)

NLE deals with pixels.

I'm not sure I got your point, all digital video (such has a DVD source) deals with pixels, just not necessarily square computer pixels. Of course 8/9 still isn't square, so what's the advantage? - Laur-1 (talk) 20:01, 13 March 2009 (UTC)

No they don't. The whole point of PAR is having correct visuals. Fleet Command (talk) 10:45, 14 March 2009 (UTC)

So, the 720x480 has the same PAR as long as it is rendered in a same viewport as NTSC 480i (that of 704x480 but its edges are obscured) or in a larger viewport than that of NTSC 480i. But when you want to edit it in computer, you can afford to have its edges obscured:

Do you meant "can't afford"? - Laur-1 (talk) 20:01, 13 March 2009 (UTC)

Oh, yes! Sorry for typo. Fleet Command (talk) 10:45, 14 March 2009 (UTC)

You must fit it in your display. Video pixels must remain untouched and PAR must remain untouched. So you use NLE -- only inside your video editor. The point of NLE is avoiding trans-PAR conversions; avoiding the loss of quality when not necessary.

Are you referring to the loss of quality when resampling to square pixels? If not, where is the loss of quality coming from? Please explain how using 8/9 is more "computationally correct" than 10/11. 8/9 is still not square, and it will give you incorrect results if you try to do any editing that requires a correct PAR, such as a radial blur, or drawing a circle. - Laur-1 (talk) 20:01, 13 March 2009 (UTC)

No! No! No! No! And NO! NLE-SPECIFIC PAR IS THERE TO AVOID RESAMPLING ANYTHING. Resampling is for trans-PAR conversion. Read my explanation above about 4:3 pictures not being really 4:3. Fleet Command (talk) 10:45, 14 March 2009 (UTC)

Now, if the point of NLE-specific PAR is so mundane, why write it in the article? Well, because it is mistaken!

Not sure I understand you here. Are saying that using these values is wrong? That's my point, but I don't think you're saying that. - Laur-1 (talk) 20:01, 13 March 2009 (UTC)

Not wrong! Mistaken for each other. People are often confused. The mistake one of these PARs for another. Normal people read developer's articles written for developers, pick up the first number they encounter and use it in their God-knowns-what-kind-of video editing tools and then don't know why their pictures get squeezed or stretched. They end up losing trust to said developers or said number. Fleet Command (talk) 10:45, 14 March 2009 (UTC)

Because people refer to encyclopedia to find out what is this PAR value and amongst them are both developers who want to develop video editing tools and consumers who want to watch.

I think the article should be telling developers NOT to use these values, instead it seems to be saying the opposite. - Laur-1 (talk) 20:01, 13 March 2009 (UTC)

Correction: I think the article should be telling developers NOT to use these values in the wrong places. Obviously you don't know what Non-Linear Editing (NLE) is about. But they know. It's just enough to tell them it's NLE-specific and they know what you mean.Fleet Command (talk) 10:45, 14 March 2009 (UTC)

I promise an overhaul. Please, feel free to help. Just remember: Verifiability, NPOV and no original research. Fleet Command (talk) 18:27, 13 March 2009 (UTC)

I would be happy to help as soon as we can get this settled. thanks for taking the time to discuss this with me. - Laur-1 (talk) 20:01, 13 March 2009 (UTC)

You are welcome. But please, read carefully. This matter is scientific. Fleet Command (talk) 10:45, 14 March 2009 (UTC)

3. "For example, Adobe Premiere Pro CS3 specifies 16:15 as the Pixel Aspect Ratio for 576i/p while specifying 10:11 as the Pixel Aspect Ratio for 480i/p." This seems to have been fixed in the CS4 product versions, Adobe is now using the correct ITU-R BT.601 values. I believe this is an example of Adobe admitting that they had it wrong before and are now using the correct values. This seems to refute the assertion that Adobe (and others) "had to employ" the incorrect values. - Laur-1 (talk) 17:36, 12 March 2009 (UTC)

Unfortunately, it is not yet fixed. I have included a source at the end of the article to the Adobe Premiere CS4 Online Help. Fleet Command (talk) 10:25, 13 March 2009 (UTC)

God bless you. I stand corrected. You are right. They have fixed it. They must have fixed it recently. I'll include that in the article too. (If I died before hitting Submit button, do it yourself! ;) ) Fleet Command (talk) 15:15, 13 March 2009 (UTC)

Thanks for correcting this, although the inconsistency remains in that this section claims the "NLE-specific PARs" are necessary. The fact that Adobe is no longer using them and is instead using the ITU-R BT.601 values seems to support my assertion that these PARs are not necessary, and in fact any programs using them are wrong.

I explained this above in details: Adobe programs (Premiere CS4 and After Effects CS4) keep rendering video the same way on the tape and video disc which must be. They just have written the value is only meant for them (developers) not what must make sense for us (consumers). Fleet Command (talk) 18:27, 13 March 2009 (UTC)

Confusing

When the article says "square grid of pixels", it looks like the number of pixels in the horizontal direction should be equal to the number of pixels in the vertical direction. So a 640x480 image would not qualify. However, a 640x480 image can and usually does have square pixels. Instead of "square grid of pixels", shouldn't we say "grid of square pixels"?

Also, when the article says "an NTSC picture (480i) with 1000 lines of horizontal resolution", shouldn't it be "an NTSC picture (480i) with 1000 columns of horizontal resolution" - Jorge Peixoto (talk)

I believe this issue is resolved now.Fleet Command (talk) 06:32, 15 November 2008 (UTC)

Terminology

This page should continually refer to "pixel aspect ratio" and "picture aspect ratio" to avoid any possible confusion when just saying "aspect ratio". -Rob —Preceding unsigned comment added by 122.29.39.122 (talk) 10:18, 19 June 2008 (UTC)

I believe the new article now suits your need.Fleet Command (talk) 13:48, 12 November 2008 (UTC)

---

In glancing through this page I'm puzzled by the references to standard definition video being the main reason for using PAR, when many HD recording and transmission formats also use PAR to compress images horizontally. Suggest this page be revised to describe PAR in more general terms, with less emphasis on SD compatibility. Kwshaw1 (talk) 16:56, 27 March 2009 (UTC)

Personally, I believe you are right. But I didn't find a credible source that explains this in details in a NPOV manner. Most sources used for this article explicitly claim that HD standards define only square pixels. (See for yourself!) If you have a credible source that states otherwise, please let us know. Fleet Command (talk) 15:53, 1 April 2009 (UTC)

Can sb. check the pixel aspect ratio of 4:3 576i?

http://lipas.uwasa.fi/~f76998/video/conversion/ convincingly shows that the pixel aspect ratio of 720x576 D1, DV, DVB, DVD and SVCD is 128/117 and not 59/54. --87.183.180.177 (talk) 13:16, 8 November 2008 (UTC)

128/117 is equal to 59/54. And you must have not read the new Wikipedia article and the aforementioned article carefully, or would not have raised the question.Fleet Command (talk) 13:48, 12 November 2008 (UTC)

1. Mathematically 128/117 does not equal 59/54: 128/117 = 768/702 whilst 59/54 = 767/702.

2. I have read the aforementioned article carefully.

3. The Wikipedia-article does not explain upon what considerations the (former) industry-standard sampling frequency of 14,75 Mhz (PAL) has been chosen and therefor the equation below "The Pixel Aspect Ratio for 576i would be 59:54 as:" seems to be pure numerology.

4. The aforementioned article, however, convincingly explains that the 702 samples of the active picture correspond to the 52 µs wide window within the scanline which - together with the 576 lines - carries the 4:3-image. Let x be the pixel aspect ratio. With the rule of three we get: 702/576 · x = 4/3 => x = 768/702 = 128/117. —Preceding unsigned comment added by 87.183.181.117 (talk) 17:30, 14 November 2008 (UTC)

Before I start, I'd like to apologize for my language in my last response. I understand that it was retort-like and unprofessional. In response to your concern about 576i PAR, you must know that I have read more than a hundred different sources on Pixel Aspect Ratio during authoring of this new article and most of them had resorted to use very strange "homegrown" math calculations to bring up even more strange aspect ratio values. As Chris Pirazzi mentions in his "Programmers' Guide to Video Systems" (as linked to the article), even MPEG community got a piece of the action. However, after diging through so much, I concluded that 59/54 is actually the 576i Pixel Aspect Ratio that is accepted by the industry. Now as of the questions you have raised:

1. Mathematically speaking, you are right. However, in real-world we use approximation and thus they almost equal: 128/117 ? 1.090 and 59/54 ? 1.090. The three digit fractional decimal precision is more than enough for split pixels, especially with ITU-R Rec 601-6 providing 16 additional buffer pixels.

2. If you say so, then it is so. But perhaps you forgot the part that explains that aspect ratio must be feasible and computationally-lightweight. That is why SMPTE RP 187 PAR values are ignored by the industry.

3. The article does not explain any reason, as none is known, according to sources used. However, it is a well-sourced fact that these numbers are actually industry-practices. Even the article you introduced concurs that these numbers industry-practice.

4. Again, as the Wikipedia article and its sources explain, edges of video pictures are not well-defined. So, in practice (and according to sources, in theory) no such thing as a strict 52µs exist. That is why ITU-R Rec.601-6 provided 16 addition horizontal pixel buffers. Even the article you linked to concurs to this fact, highlighting it in a yellow box!

Fleet Command (talk) 06:30, 15 November 2008 (UTC)

Apology accepted!

3. I found an explanation of the 14.75 Mhz in http://www.cvc.uab.cat/intra-web/Theory%20and%20Applications%20of%20Digital%20Image%20Processing.pdf (Table 1.1 on page 10). Now it seems that the difference between 128/117 and 59/54 is due to 575 vs. 576 lines. The derivation is:

If you ask for square pixels you get 575 · 4/3 = 766 2/3 pixel rounded to 767 pixels. These 767 pixels "occur" in 52 µs. 767/52 µs = 14.75 MHz (exactly). For any other sampling rate (say todays 13.5 MHz) it is clear that the pixels are no longer square and hence the calculation in the Wikipedia-article is corretly yielding 59/54 as a ratio of sampling frequencies.

4. The 52µs AFAICS serves as a reference. Without the 52µs specification there is no clear picture width. (The picture height does not seem to be an issue) And without width there is be no picture aspect ratio. Hence you could not have calibrated all the analog TV sets and tube cameras. We have to distinguish the 52 µs width from the question where precisely this 52µs-window starts and ends. AFAICS this may be chosen (horizontally shifted) more or less within these 16 (or 18?) additional pixels - or in the analog domain: shifting some 100 ns left or right.

The three yellow boxes state:

a) Not a single one of the commonly used digital video resolutions exactly represents the actual 4:3 or 16:9 image frame.

b) It is really the analog video standards that define the image geometry and pixel aspect ratio in digital formats.

c) Note: Many people use direct resampling for all the wrong reasons: 1) They think that a 720×480 frame directly equals to a 720×576 frame. 2) They also think that both aforementioned frame sizes represent exactly the active 4:3 (or 16:9) picture area, edge to edge. As you already know from Section 2.1, both of these assumptions are wrong. The fact that direct resampling works at all is mostly a quirky coincidence

None of these states that there is no such thing as a 52 µs specification.

--87.183.136.3 (talk) 13:47, 15 November 2008 (UTC)

Good to hear we finally agreed on the matter of 59/54.

3. I'm reading the document you have linked, but I still don't properly understand why 14.75Mhz is used. I keep reading.

4. I stand corrected. There does seem to be a strict 52 microsecond timing. This value is derieved directly from frequency value, unlike what I understood before. I was by mistake thinking that 52 microsecond timing is derieved from the edges of the picture, so originally, I mean yellow box "a". I realized my mistake after re-reading ITU-R Rec.601-6.

Fleet Command (talk) 06:10, 17 November 2008 (UTC)

3. Given the 52 µs window of 575 rows (lines). How many columns do you need in order to get (really) square pixels? Answer: 575 · 4/3 = 766 2/3, rounded upwards to 767. What is the pixel frequency? 767 / 52 µs = 14.75 MHz (exact). I don't know whether the 14.75 MHz is "used" in any consumer device, but to ask for pixels which are square seems obvious.

--87.183.132.236 (talk) 16:18, 10 December 2008 (UTC)

128/117 = 1.094067 (702); 12/11 = 1.09 (704)

Two ones were at 768

59/54 = 1.0925 (703 = 769x576 or 770x576? Idk) Ouch!!! This calculations were... Hummm... My eyes were interlaced with this calculations :( 2A01:E0A:AB0:C580:A4D1:6A9B:505:930 (talk) 16:43, 13 March 2023 (UTC)

Recent editions by Mikus reverted

Greetings, fellow Wikipedians I'm afraid today I had to undo a recent edition by my fellow editor, Mikus. Here, I'd like to establish why.

The main reasons were:

Our friend had changed the statement:

"Most modern imaging systems describe an image as a grid of very small but nonetheless square pixels. However, some imaging systems, especially those which must maintain compatibility with analog standard-definition, define an image as a grid of rectangular pixels in which the width of the pixel is different from that of its height. To describe this proportion, we state it in the form of Pixel Aspect Ratio."

Into:

Most modern digital imaging systems use square pixels. However, some imaging systems, especially those that maintain compatibility with analog standard-definition, use non-square pixels.

Unfortunately, this adds ambiguity: non-square pixels might mean rectangular, circular, triangular, trapezoidal, etc. And, yes, I assure you there has been real user who have understood the term "non-square" so wrong. This kind of ambiguities must be avoided at all cost in the article opening. Average non-technical user read these articles while, in case of Pixel Aspect Ratio, seasoned software producers have trouble understanding certain things. Therefore, ambiguities in this article must be eliminated. I'm not saying "My writing is perfect, follow my example". In fact, correct me where you can. But, please, just don't create more ambiguities.

The goal was to rewrite this statement in a cleaner more concise manner, and of course, to avoid plural active voice like "we state", which (IMHO) has no place in a technical article. I agree that non-square can mean anything, so I suggest adding somewhere right in the beginning of the article that 2D imaging systems operate with rectangular pixels, while 3D systems operate with -- surprise, surprise -- cubic pixels. Then using "non-square" will automatically mean "rectangular".

He also changed the following statement:

"Use of Pixel Aspect Ratio mostly involves pictures pertaining to standard-definition television. Most other imaging systems, including those complying with SMPTE standards and practices, use square pixels."

Into:

"For example, CGA graphics adapter used video modes with screen sizes of 320x200 pixels (IAR 4:3, PAR 1.2) and 640x200 pixels (IAR 4:3, PAR 2.4). The base screen size of EGA was 640×350 pixels (IAR 4:3, PAR 35:48). DVD 4:3 video stream encoded at 720x480 has a PAR of 8:9. The same stream encoded at 704x480 would have a PAR of 10:11."

This new statement is POV, ambiguous, technically difficult to understand and without any verifiable source. POV; because apparently not everyone agrees with this: On all monitors that I have owned, the 4:3 aspect ratio always remained unaltered until I wanted; only the viewport size changed with the resolution. Now, I know there are CRT monitors which stretch the picture but I don't see them. Besides, the original paragraph never disputed the existence of exceptions. Ambiguous; because suddenly a legacy computer graphics system is used as an example whereas the subject of talk was modern systems. Technically difficult; because words like CGA, EGA and IAR are thrown in casually and new average users won't understand them. Giving them more thing to spend time on intentionally is an atrocity (only in this article, of course) when the best choice is simply not to do so.Fleet Command (talk) 08:45, 28 February 2009 (UTC)

"POV; because apparently not everyone agrees with this" -- I don't see how well-known technical information can be POV. I provided links to other wikipedia articles to ensure that no one blamed me in POV or original research. Did you follow the links? Than do it now.

"4:3 aspect ratio always remained unaltered until I wanted;" -- Until you wanted what? You are confusing IAR and PAR. 4:3 is IAR. Yep, it is 4:3 because you cannot stretch a monitor, it is made of glass and plastic. But you can have different PAR easily, and I provided the links for you and everyone else to confirm this. CRT screens do not have pixels, hence they do not have native screen size, whatever a video card generates will be displayed, within certain limits, of course. Yes, they have minimum dot pitch that depends on phosphors granularity, but these are not the same pixels as on an LCD or plasma screen.

"Ambiguous; because suddenly a legacy computer graphics system is used as an example whereas the subject of talk was modern systems." -- The information may not pertain to modern systems, but ambiguous it is not. All the numbers are verifiable. The main idea of adding CGA/EGA information was to stress out that PAR is not a concept that pertains only to TV and video, it is used in computing as well. And, believe it or not, PAR is not used only with non-square (yeah, yeah, rectangular) systems, but with square ones too, but in the latter case PAR == 1, because you still have pixels with height and width. I agree that bringing legacy systems into discussion about modern systems is not correct, but I believe that non-square PAR systems should have their place somewhere in the history/intro section.

"Technically difficult; because words like CGA, EGA and IAR are thrown in casually and new average users won't understand them." -- PAR is defined above in the article, for CGA, EGA and IAR I provided links. There is nothing out of ordinary to follow the links and to read what the stuff is about, unless a reader is a complete retard.

"Giving them more thing to spend time on intentionally is an atrocity" -- Are you expecting housewives to read an article about PAR? Don't think so.

P.S. At least you removed "horizontal width" and "vertical height", that is a start. Mikus (talk) 22:48, 25 March 2009 (UTC)

So you justify your misbehavior by insulting our readers? You are indeed rude. You're excuse is unacceptiable. Housewives, kids and retards have the right to understand an article. Making an article harder to understand is yet another defiance of Wikipedia best practices. Fleet Command (talk) 15:21, 1 April 2009 (UTC)

I cannot see any "misbehaviour" as all of Mikus's comments appear valid. Slight failure at political correctness does not rudeness make. It is obvious that a) articles on PAR will only be read my technically minded people and b) limiting terminology and content to material easily understood by anyone without technical knowledge will severely limit the usefulness of any encyclopedia. There are usually separate kids', students' or for-dummmies editions for the purpose of providing information to a layperson with no technical knowledge. Bstard12 (talk) 10:32, 16 August 2010 (UTC)

CEA-861 and HDMI wrong?

Great article, everyone. One of the better discussions of the topic I've found on the web.

But I see no mention here of HDMI and CEA-861. This seems like a large omission, as these are some of the most important specs at the front line of consumer A/V kit, but the CEA-861 spec, and hence the HDMI spec, basically has rubbish pixel aspect ratios explicitly given for 576i/p and 480i/p.

Format	Picture aspect ratio	Pixel aspect ratio
720×576i/p	4:3	16:15 ($1.0{\overline {6}}$)
	16:9	64:45 ($1.4{\overline {2}}$)
720×480i/p	4:3	8:9 ($0.{\overline {8}}$)
	16:9	32:27 ($1.{\overline {185}}$)

This is despite the CEA-861D timings being based on all previous digital standards, at 13.5/27MHz.

Note also that the 720×576 and 720×480 frames are simply stated by CEA-861 as being 4:3 or 16:9 - no suggestion that they're actually slightly wider.

I've seen a lot of confusion in this area - I've not looked that hard, but I've yet to see a piece of consumer kit that upscales 576i DVD or 576i HDMI to 1080p HDMI correctly — everything's over 2% too thin, and you see strange things at the left and right, as they squeeze all 720 pixels of width into the 1920×1080 frame. Having this error in one of the main modern standards documents isn't helpful when trying to converse with the offending companies.

I'm going to put CEA-861D in simply as a source along with the others, but I think this error/inconsistency is significant enough that it needs to be addressed in the main body too somehow.

Of course, it could be argued that the spec should be taken as written, so over the HDMI wire that pixel aspect ratio is used, so anything playing DVDs or converting between HDMI and analogue/SDI in either direction should be rescaling horizontally, but I really don't think that's the intent. Indeed, the HDMI spec says

For example, if a Source is processing material with fewer active pixels per line than required (i.e. 704 pixels vs. 720 pixels for standard definition MPEG2 material), it may add pixels to the left and right of the supplied material before transmitting across HDMI.

implying that they expect HDMI's pixel aspect ratio to be the same as that of normal SD content.

--KJBracey (talk) 16:00, 1 September 2009 (UTC)

Hi, KJBracey.

The values you have included here were actually in the article under the title of "NLE-Specific" aspect ratios. (See this version) However, they have been deleted by User:Laur-1 after a discussion which you'll find above. Back then, I was stuck in a hurricane of exams and didn't have time to provide more accurate descriptions for NLE as well as to search for sources. (Now, that I look at the discussion, I'm embarrassed at my own hurried style of explaining things to Laur-1.) Therefore, after the discussion, I eventually consented the hard-to-understand contents to be deleted.

However, these values are not incorrect. Let's cut the story short: If you are to display a 720x480 image on an exactly 4:3 screen, your pixel aspect ratio of the said image will be exactly what CEA-861 have specified: 8:9!

${\frac {4}{3}}\div {\frac {720}{480}}={\color {blue}{\frac {8}{9}}}$

Suffice to say that ITU-BT R.601 and CEA-861 speak about two different thing: Rec.601 talks about compatibility between TVs that were supposed to show a 704x480 picture in 4:3 frame but didn't. (They didn't know "where the picture ends" as Chris Pirazzi put it.) So, an additional 16 pixels safety buffer was defined. (704px became 720px.) But CEA-861 speaks about HDTVs which zealously define edges of pictures and do not consider a 720x480 picture with 10:11 pixel aspect ratio a 4:3 picture. So, as you see but ITU-BT R.601 and CEA-861 are right; only they speak about different things.

I'm sorry that I do not have time to explain more now. Perhaps one day I will return and will overhaul the article again. But don't wait for me: If you can improve anything, please do.

Fleet Command (talk) 02:28, 2 September 2009 (UTC)

So your view is that the HDMI/CEA-861 spec should be taken literally? And thus that conversions to and from HDMI should be scaling horizontally?

So given a 720x576 frame on a DVD, or broadcast over DVB, neither of which are intended should be displayed in their entirety - those should be cropped by the player to 702, then rescaled to fill the 720x576 frame of HDMI, to get correct geometry? And a 704x576 frame should be stretched out to 720 or so? It's not a 1:1 pixel mapping?

Well, that's certainly a possibility, but I see nothing in the HDMI/CEA-861 spec to suggest that such an unusual step was the intent. There's no commentary or acknowledgement. It looks just like a casual error to me. The one comment in HDMI about 704-wide video being pillarboxed appears to contradict the CEA spec.

If they really wanted the frame to be 16:9 or 4:3, and they were thinking about it, they'd have specified the HDMI frame as 704x576, so that 720x576 sources could be cropped without any rescaling being needed (within a fair degree of tolerance).

Don't you agree? What sense is there in specifying a consumer connection format for a consumer format that doesn't exist: 720x576 exact 16:9 frames? It means slight horizontal resampling of every possible source in every device that outputs SD over HDMI. Do any devices do this? It would certainly upset anyone wanting pixel-perfect output from a device for later upscaling. --KJBracey (talk) 14:49, 2 September 2009 (UTC)

Or, alternatively, are you saying that HDMI is intentionally designed to introduce a 2% geometry difference? Consumer SD devices connected via HDMI are expected to show SD content 2% thinner than over analogue connections to get the edges in? So we're going to gradually change the pixel aspect ratio of newly-produced digital SD content to match as use of HDMI becomes more common? --KJBracey (talk) 15:02, 2 September 2009 (UTC)

Before I start answering you, I'd like to make sure that your inital assumptions are right. You see, you seem to be under the impression that something must definitely be wrong with these standards. Well, actually it isn't. It is the context that is different. Pixel Aspect Ratio is just a mathematics concept and its practical use depends on the conditions. When you read the specs set by a certain standard and comparing it to that of a different standard, please make sure that you know they apply to what and how.

So, let us first assume that both standards are probably right; only they apply to different things. Let's assess these conditions.

You have said: "So given a 720x576 frame on a DVD, or broadcast over DVB, neither of which are intended should be displayed in their entirety - those should be cropped by the player to 702, then rescaled to fill the 720x576 frame of HDMI, to get correct geometry? And a 704x576 frame should be stretched out to 720 or so? It's not a 1:1 pixel mapping?"

Excellent question. Actually you have specified the medium of transportation (DVD or broadcast) but did not specified the intended receiver; but the intended receiver is very important. As the article says:

However, standard-definition television standards and practices are incompatible with digital video. Such standards define an image as an array of well-defined horizontal "Lines", well-defined vertical "Line Duration" and a well-defined picture center. However, there is not standard-definition television standard that properly define image edges or explicitly demand a certain number of picture elements per line.

The first time I read such a sentence, I was surprised. But indeed, it is true: Although all analog TVs receive the same picture, some of them show more portions of the picture and some show less. There is no explicit cropping done and no data around the edges is destroyed; just analog TVs do not show all of the picture around the edges! Even some TVs advertise that they show extra inches of the picture in comparison to other TVs! That's why terms like "Production aperture" and "Clean aperture" are eventually invented. (More details) ITU-BT R 601, knowing that 4:3 analog TVs are actually not exactly 4:3, specified 720 pixels for width instead of 704 as a safety measure.

Now CEA-861 does not deal with old analog TVs anymore: Just plain HDTV with crystal-clear well-defined edges. No more playing safe. In HDTV, a 720px-wide picture is really 16 pixels wider than a 704px-wide picture. So, CEA-861 actually says that if you are preparing a 720x480 picture for a 4:3 HDTV (that is, 480p), specify 8:9 as PAR and compile your video accordingly. If you are not preparing you video exclusively for HDTV, then this standard does not apply to you.

I hope I have made it clear enough. Fleet Command (talk) 20:58, 2 September 2009 (UTC)

Hang on. Firstly, CEA-861 and HDMI are a general-purpose digital video transfer system, not an HDTV-specific one. They support both SD and HD formats. There's nothing in there that requires an HDMI display to even be HD-capable. The only mandatory formats an HDMI display has to support are SD ones! And there are HDMI-capable CRT displays, without the "well-defined edges" you talk about. As well as devices that convert between HDMI/DVI and analogue.

Secondly, a 720x576/480 picture isn't HDTV, by the basic definition of HDTV! You're not making any sense here. Maybe you could explain what exactly you mean by "HDTV"? But let's go with it for now...

As far as I can see, your argument is that there is a "new" 720x576, which you call an "HDTV" 576i. This is different from the old 720x576, in that it is 16:9, not 16:9-and-a-bit, with a different PAR. And that's what HDMI supports. Okay. Barring the dubious "HDTV" terminology, I agree that's what the letter of CEA-861 says. But I don't like it.

Where will this HDTV-576i content come from come from? There isn't any "HDTV-576i" source material in existence! All 720x576 material out there - DVD, DVB, etc, is SDTV, based on the traditional 13.5MHz sampling rate, and with the resultant pixel aspect ratio and frame shape this article talks about.

So any SD device outputting to HDMI will have to adjust its output - the source material on the DVD, or whatever, has the "traditional" pixel aspect ratio given in this article, being designed for traditional SDTV rendering (13.5MHz), while HDMI requires the new "HDTV" PAR. If this resampling isn't done, you'd get a picture geometry error. Your picture over HDMI would be 2.5% thinner than the same material output via an analogue connection, because the "HDTV" 576i has different-shaped pixels, and you've squeezed a wider-than-16:9 frame into 16:9.

And you believe that this is the intent of the HDMI standard, yes? That its SD formats can't be used to transmit traditional SD content from a DVD player without either horizontal cropping and resampling, or a geometry error - it's designed to transfer some "HDTV-576i" content that doesn't exist. You realise that this necessarily follows from claiming that both HDMI and Rec.601 are right. Seems far-fetched to me.

Or do we just give up and scrub this article? If we're saying we can no longer be sure whether to display 702 or 720 pixels from a DVD, and a DVD author can't be sure what end devices will do, or are supposed to do, then this article is misleading. We can no longer precisely define basic picture geometry, and guarantee that circles come out circular. --KJBracey (talk) 23:58, 2 September 2009 (UTC)

Hmm, actually, because I feel we're talking at a little cross-purposes, can I just stop and ask a few yes/no questions. If the answer to any is "no", we need to go back a step, as we've got fundamentally different assumptions. Here I define "shape" in terms of geometry or aspect ratio - eg circles being circles, rather than details of cropping or overscan.

1) Should a DVD viewed on a TV using an analogue connection have the same shape as when viewed on the same TV over an HDMI connection?

2) Should a DVD viewed on a 16:9 SD CRT be the same shape as when viewed on a 16:9 HD-capable plasma?

3) Should an HDMI<->analogue signal converter preserve the shape of an image passed through it?

4) Should an HDMI<->SDI signal converter preserve the shape of an image passed through it?

--KJBracey (talk) 00:35, 3 September 2009 (UTC)

KJBracey, DVD supports both 720 and 704 horizontal resolutions and if a 704 horizontal resolution is used it does have to be sent with either pixel padding or scaled to 720 though almost all DVDs are encoded with 720 horizontal resolution. ATSC is the main problem today since only 704x480 was specified even though modern SD video production uses 720 horizontal pixels. This is an issue and is explained in the book "Digital Video and HDTV: Algorithms and Interfaces". The people who made the CEA timing were following modern SD production standards which is why they chose 720 horizontal pixels for SD video signals. As such your DVD player should properly output 720x576 video and that is explained in the book "DVD Demystified". Just to make this clear 720x480 and 720x576 are the modern standards used for SD production systems which is why they are defined in the CEA timings. --GrandDrake (talk) 01:53, 3 September 2009 (UTC)

Yes, 720x576 is the standard for SD production, based on a 13.5MHz sampling rate, which gives us the pixel aspect ratios in this article, and a slightly wider than 16:9 frame. 704x576 uses the same pixel aspect ratio - it's just narrower than 720x576. Converting 704x576 to 720x576 means adding pillarboxing, not stretching.

But the point is that HDMI defines 720x576 differently from those other standards. It has a different pixel aspect ratio, and has an exactly 16:9 frame. Meaning that if you output the 720x576 frame from the DVD with 1:1 pixel mapping, and you believe the HDMI spec, the shape of the picture has changed - it gets 2.4% thinner.

Something's gone seriously wrong here. You've screwed up your picture geometry. How are we going to fix it? We have to do one of two things.

1) Fix the DVD player to crop and rescale its "old-style" >16:9 720x576 to HDMI's 16:9 720x576. Take the middle 702 and stretch it to 720. I'm not aware of any players that do this.

2) Calibrate the 16:9 TV to ensure it only shows the centre 702x576 of a received HDMI 720x576 signal. We effectively ignore the aspect ratios given in the HDMI spec, and assume that an HDMI 576i signal has the traditional SD geometry.

The HDMI spec itself suggests, in the comment I quote above about having to use padding for 704 content, that it doesn't expect 1 to happen, and that it didn't mean that to be the intent of the spec. (If we were stretching, 704 content wouldn't need padding).

The HDMI/CEA-861 specs really need to make themselves clear on this point. Every HDMI source that I'm aware of assumes a 1:1 pixel mapping between SD content and 576i HDMI. As CEA-861 is written, this universal practice is incorrect. So if CEA really means HDMI to have a different pixel aspect ratio from SD production, and thus require source rescaling+cropping to preserve geometry, they have to say so explicitly, and get people to start doing so. If they don't mean to have a rescaling, they need to adjust their declared aspect ratio to match standard SD production, and hence endorse what every HDMI source is currently doing. Or, as a final alternative, just state that traditional SD content delivered over HDMI may be 2.4% thinner but who cares?

This last possibility seems to be where we're at at the minute, given my experiences at trying to calibrate HDMI equipment. Sources are doing 1:1 SD->HDMI pixel mapping, and displays and upscalers are tending to assume the 720x576 frame is exactly 16:9 as HDMI says. The net result is that HDMI is making pictures 2.4% thinner compared to analogue connections. If we want to fix this, one end or the other is going to have to change behaviour, and the HDMI spec needs to be clear on which end.

Various subsidiary specs I've seen such as the NorDig and DTG digital TV standards have come down firmly on option 2 - 1:1 pixel mapping from the source is right, and upscalers and TVs should only process the middle 702 of an HDMI 576i signal. --KJBracey (talk) 18:00, 3 September 2009 (UTC)

KJBracey,

You wrote:

1) Should a DVD viewed on a TV using an analogue connection have the same shape as when viewed on the same TV over an HDMI connection?

2) Should a DVD viewed on a 16:9 SD CRT be the same shape as when viewed on a 16:9 HD-capable plasma?

3) Should an HDMI<->analogue signal converter preserve the shape of an image passed through it?

4) Should an HDMI<->SDI signal converter preserve the shape of an image passed through it?

Yes to all four. The fidelity of the motion picture must be maintained in all situations. Actually the whole point of defining a Pixel Aspect Ratio is the maintenance of picture fidelity.

Hang on. Firstly, CEA-861 and HDMI are a general-purpose digital video transfer system, not an HDTV-specific one [~snip~] Maybe you could explain what exactly you mean by "HDTV"? But let's go with it for now...

Please accept my apology for my use of colloquialism. (I'll try not to use colloquialism anymore.) Yes, you are right: I should have written "digital video" instead of HDTV, as I was talking about digital video playback hardware that define a picture edge. HDTV hardware are not the only one of the many pieces of hardware that do that.

...576i...

Now that you care so much about correct use of terms, I feel obliged to ask you: Are you using colloquialism as well? What do you mean by 576i? Note that 576i is a 4:3 analog television standard in which neither pixels nor image edges are defined. So, 720x576 at 25fps is not equal to 576i. The only relation between 576i and 720x576 is that digital media (including but not limited to DVD Video) that are prepared per ITU-BT Recommendation 601 have 720 pixels in width and 576 pixels in height. To play a DVD Video with 720 pixels width, 576 pixels height, 25 frames per seconds frame-rate and an aspect ratio of 59:54 on a 576i-compliant analog TV, a component known as Digial to Analog Converter (DAC) must convert the DVD Video into 576i.

So, important question: Does CEA-861 talk about 720x576 digital motion pictures or 576i? (Unfortunately, my knowledge of CEA-861 comes from secondary sources. I haven't read the CEA-861 text itself, which would cost me $219.)

As far as I can see, your argument is that there is a "new" 720x576...

Yes, because, as you yourself said, CEA-861 calls for use of 720x480 or 720x576 digital motion pictures with different pixel aspect ratios than those derived from specification set forth by ITU-BT Recommendation 601. So, yes, it seems there is a new 720x576.

...which you call an "HDTV" 576i.

No! I neither called anything "HDTV" 576i nor meant to do so. (Again, I apologize for the use of colloquialism.) 576i is not a digital format.

Do we just give up and scrub this article?

No, we definitely do not do that! Your concern is well outside the scope of the article (i.e. definition of PAR and its background in Standard-definition television format). Besides, even if was related to the article, destructive actions such as erasure will not solve any problem.

Fleet Command (talk) 11:48, 4 September 2009 (UTC)

Okay, back to the left :)

I'm glad you answered yes to all 4 questions. So I assume you're concerned like that consumer kit is very commonly getting stuff too thin because of this HDMI ambiguity/error.

Now, whenever I said "576i", I qualified it (I hope). There's "DVD 576i", which is a 704x576 or 720x576 MPEG frame encoded on a DVD. That has the pixel aspect ratio in this article, the same as SDI/BT.656, with its 13.5MHz sampling. (Assuming the DVD authors haven't cocked up and assumed the 720x576 frame is 16:9 - probably an error that happens sometimes). European digital TV standards also use the same ratio - the BBC, for example, is very clear on 702x576 being the active area, and their test card is clearly marked for this.[1]

Then there's "HDMI 576i", which is a 720x576 wire format, and ostensibly claims to have this other pixel aspect ratio, and a different frame aspect ratio as a result. And then there's analogue 576i, which is 52us x 576 for a 4:3/16:9 frame. All 3 should be capable of representing the same thing, and conversion should be straightforward.

You convert between analogue 576i and DVD/SDI 576i straightforwardly - you use a 13.5MHz sampling clock on your ADC or DAC — 720 pixels of DVD/SDI converts neatly to 53.3us of analogue, and you get the right aspect ratio. It's this weird new CEA/HDMI 576i that is the odd one out.

CEA-861 precisely defines the whole series of formats HDMI can transmit both HD and SD. It specifies their precise timing on the wire - the digital formats have basically the same timing characteristics as analogue. I'll quote the following snippet from CEA-861-D; (Rev E is less wordy, reducing the same info into a table entry - this is clearer):

4.10 720(1440)x576i @ 50 Hz (Formats 21 & 22)
This timing is based on ITU-R BT.656–4 [32] except for horizontal and vertical synchronization pulse durations, which are specified in ITU-R BT.711–1 [33] and ITU-R BT.470–6 [31]. This format assumes the pixels are double clocked to meet minimum clock speed requirements for the interface. Thus, the clock is 27 MHz. This format timing can use either 4:3 or 16:9 aspect ratio.

So they're saying that it's supposed to be the same as the Rec.601/656 standards. Its 720 pixels of data occupy 53.3us of signal on the wire just like Rec.601. So a direct ADC/DAC conversion to/from analogue 576i would make those 720 pixels wider than 16:9. If we believed them about this 53.3us of data being 16:9, you'd need a line buffer to convert between analogue and HDMI - 52us of analogue signal would become 53.3us of digital signal.

All of this makes me think that they didn't really mean to introduce a "new" 720x576. You've got these facts:

In support of a new format:

• The format summary table in CEA-861 says that 720x576i frame aspect is 4:3/16:9, and gives pixel aspect ratios matching that.
• It's very common for a consumer HDMI upscaler to scale the entire HDMI 720x576 frame to 1920x1080.
• Some TVs show the entire area of an HDMI 720x576 signal by default.

Against a new format:

• CEA-861 timing of 720x576i is the same as BT.601/656, and the description of 720x576i says that it is based on it.
• The HDMI spec says that a 704-wide picture should be pillarboxed in the 720-wide frame. It doesn't suggest rescaling it.
• All HDMI sources output DVD and DTV SD images with a 1:1 pixel mapping over HDMI 720x576
• Various European DTV specs say that HDMI upscalers should only be scaling the middle 702x576 frame of HDMI.

The net effect is that it looks to me that this apparent "new" 720x576 format is just like a casual error by people who didn't appreciate that Rec.601/656's 720x576 is wider than the nominal frame. (After all, those standards don't explicitly say it, and don't explicitly give a pixel aspect ratio). They just filed all the SD formats neatly as "4:3" and "16:9", and then calculated the pixel aspect ratio from that.

No-one's thinking too hard - the displays and upscalers just take the entire 720x576 frame, because it's the "obvious" thing to do, and the sources just output 1:1 pixel mapping because it's the "obvious" thing to do. Only the European DTV specs are actually stopping to think about it, it seems.

Don't you agree that it has to be an error? I can't see any benefit in deliberately inventing a new wire format whose aspect ratio doesn't match its timing. So much conversion grief, and if you really believed it and did the conversion in the source to preserve aspect ratio, HDMI wouldn't be able to carry the entire 720x576 DVD frame losslessly - it would force edge strippage in the player, and cause quality loss due to the slight rescaling.

I don't think this concern is outside the scope of the article — it's giving the aspect ratios for digital SD formats, but the standard for the main consumer digital SD connection technology is giving different numbers... The article should attempt to explain the discrepancy. --KJBracey (talk) 22:39, 4 September 2009 (UTC)

KJBracey, I would mention that the CEA-861 standard pixel aspect ratios are based on having a perfect aspect ratio of either 4:3 or 16:9 for 720x480 or 720x576. I believe the CEA committee went with that because the CEA-861 standard was made for the transport of any video signal whether it be analog broadcast, digital broadcast, DVDs, video game consoles, or computers. The pixel aspect ratios mentioned in the article are based only on the analog broadcast standards in which the pixel aspect ratio is based on the resolution of the content being 704x480 or 704x576 (which is how they got the pixel aspect ratios mentioned in the article). In terms of the HDMI specs you have mentioned that it describes how to add pixels to the left and right for such a signal. I would point out though that it also states that the type of signal can be noted in the Auxiliary Video information (AVI) InfoFrame. The AVI InfoFrame includes information on the bar info, scan information, picture aspect ratio, and the active format aspect ratio. From what I have read in the HDMI specs (Auxiliary Video information (AVI) InfoFrame section) and this article (AVI Overview section) it looks like the AVI InfoFrame describes the content aspect ratio. Now whether CE devices make use of that information is another question but from the looks of it the CEA committee did design the CEA-861 standard to account for 704x576 being sent at 720x576. --GrandDrake (talk) 02:33, 5 September 2009 (UTC)

Yes, the HDMI spec says that you need to add pixels either side to a 704 MPEG frame to fill a 720-wide signal. So as the 704-wide signal is 16:9, that statement tells us that either the 720-wide HDMI signal is wider than 16:9, or they expect SD content to be squidged by 2%. I prefer the former interpretation - I can't imagine a standard would be urging sources to distort video. --KJBracey (talk) 14:03, 7 September 2009 (UTC)

Since the CEA-861 specs allow for the content aspect ratio to be noted HDMI can deliver both 720x576 and 704x576 with a 16:9 aspect ratio. As such neither video signal would be thinner or wider than 16:9. --GrandDrake (talk) 00:48, 8 September 2009 (UTC)

KJBracey,

You have said:

All of this makes me think that they didn't really mean to introduce a "new" 720x576. [~snip~] The net effect is that it looks to me that this apparent "new" 720x576 format is just like a casual error by people who didn't appreciate that Rec.601/656's 720x576 is wider than the nominal frame.

This new video format is simply the result of consumer demand not an error: New hardware need to behave differently to provide new benefits but also need to provide backward compatibility; so, they have to use a new video format. This way, old video will play as they always played — same portions may still get cropped away — and new video will play intact.

Not following you here. What different behaviour is required, where? All that's happened (as far as SDTV is concerned) is a change from analogue signal transmission to digital signal transmission - we haven't changed the source data, nor have we necessarily changed the displays. I don't understand the distinction you're making between "old video" and "new video". Yes, there's high-definition 1080p "new video", but no-one's invented a new SD storage format. HDMI's 576i mode is there to transport "old video". There is no SD "new video" for it to transport.

Image fidelity remains untouched. So, why the fact that some numbers match while others don't have so much concerned you?

Because video signals are very often coming out 2.4% too thin due to pixel aspect ratio confusion! This is a real problem. I cannot get correct aspect ratio in some configurations of my kit, and it's going to be very hard to persuade manufacturers to fix it when they're following the letter of a spec. Your apparent lack of concern for this earlier led me to pop those 4 questions in.

Don't you agree that it has to be an error?

With all due respect, no!

So what do you think is the correct behaviour to maintain SD aspect ratio then? What should a DVD player (or any other digital SD source) be doing, and what should the display be doing? An HDMI connection often gives a different picture shape from an analogue connection. What's your solution? Getting the source to scale 702/4 up to 720, and cropping the sides off 720? Do you really think that's what they intended, and have you ever seen a source do that? They don't - they output with 1:1 pixel mapping (as the HDMI spec suggests they should in that note about padding 704 to 720).

I think my main sticking point here is that I can see no good reason for deciding to deliberately use a different pixel aspect ratio for an SDTV connection from the pixel aspect ratio that all SD video uses. It precludes a "raw" 1:1 pixel mapping output from any digital SD device. The connection is no longer a good fit to the signals it will be used for. You seem to be saying that you can see a reason to make such a change, but I can't quite grasp what that reason is.

And did you understand my point about the timings? If they were using a 13.8 MHz clock for HDMI, so their 720 pixels of data occupied 52us, then I'd definitely believe it was a deliberate change. But they're using a 13.5MHz clock, just like all previous standards, so they've got 53.3us of signal. 53.3us has never represented a 16:9/4:3 frame before.

I don't think this concern is outside the scope of the article — it's giving the aspect ratios for digital SD formats, but the standard for the main consumer digital SD connection technology is giving different numbers... The article should attempt to explain the discrepancy.

All the more reason not to erase the entire article.

Sorry - my quip about erasing it was more down to what I perceived as your lack of concern about things coming out 2.4% too thin - if that wasn't an issue, then why bother specifying pixel aspect ratio so precisely at all?

We'll just add new facts about the CEA-861E and will keep our personal opinion to ourselves. Any idea where I can borrow a copy of CEA-861E? Fleet Command (talk) 12:30, 5 September 2009 (UTC)

I ended up having to pay for it, and can't distribute it. Copyright violation, and watermarked to the hilt. Sorry :( Happy to answer any specific questions or quote the odd bit. You may be able to access it via a library; they often have access to standards documents. --KJBracey (talk) 14:03, 7 September 2009 (UTC)

Never mind. That's the way things are. Anyways, no luck with a local library. I should see whether a friend of mine can borrow his PDF copy to me under the terms of fair-use. In the end, if I couldn't get a copy of it without paying and hand and a leg for it, that's OK: I'm not the owner of a multinational TV manufacturing corporation. Fleet Command (talk) 11:03, 8 September 2009 (UTC)

Oh, one more thought. One logical argument for the change is that they wanted all HDMI formats to be truly 16:9. They didn't want a format slightly wider. But if that was so, and they were aware of the fact that normal 720-pixel SD video was wider than 16:9, then why would they make their 16:9 format 720 pixels wide? Why not take the much simpler solution of making their format 704x576? Then that would be appropriate for a 1:1 pixel mapping from digital SD video (or close enough). The sources would simply crop excess off to get down to 16:9. The fact they didn't do this again suggests that they weren't aware of the issue. --KJBracey (talk) 17:29, 7 September 2009 (UTC)

You are suggesting that the CEA committee made a mistake but from what I can see they did design the CEA-861 standard so that both 720x576 and 704x576 could be sent correctly with a 16:9 aspect ratio. --GrandDrake (talk) 00:48, 8 September 2009 (UTC)

Um, depends on the pixel aspect ratio of the source. If the pixel aspect ratio of the source had the numbers given in the CEA spec, then you'd be fine. You'd do a 1:1 pixel mapping. But the catch is there aren't any sources with that pixel aspect ratio! They all have the pixel aspect ratio in this article. So the CEA spec can't, as written, transmit traditional 720x576 video without distortion, because it's not wide enough. Traditional 720x576 is wider than 16:9. CEA-861 (erroneously in my view) claims to be exactly 16:9. Something's got to give. --KJBracey (talk) 17:43, 8 September 2009 (UTC)

There are video sources that can send 720x576 exactly as 16:9 such as computers and game consoles so stating that there "aren't any sources with that pixel aspect ratio" seems unlikely and is a statement that would require evidence. Also as I have said the CEA-861 standard can send the content aspect ratio so it can tell the device that though it is sending 720x576 that only 704x576 contains actual content. It is one thing to say that there is an error in how the CEA-861 standard is implemented in some CE products and another thing to say that the CEA-861 standard was designed wrong. You believe that the CEA-861 standard was designed wrong but I would disagree with that since it looks to me like the CEA committee was aware of this issue. --GrandDrake (talk) 23:34, 8 September 2009 (UTC)

You have said:

Oh, one more thought. One logical argument for the change is that they wanted all HDMI formats to be truly 16:9.

Again, Backward Compatibility!

Huh? They're not being backward compatible! They're ostensibly differing, pointlessly, from previous standards. Being backwards compatible would mean supporting existing images, losslessly.

You said:

Why not take the much simpler solution of making their format 704x576?

Again, Image Fidelity! They picked the simplest solution: Make a video and everything in the frame is displayed! If they had picked 704x576, then it would have been OK for zealously standard-compliant hardware to hide 8 pixels off the sides of 720x576 images. (Yes, this time it is exactly the opposite of the example in my last post.)

Fleet Command (talk) 11:03, 8 September 2009 (UTC)

That doesn't make sense. We're already losing image fidelity. If we believe the CEA spec as written, with its deviant pixel aspect ratio, SD source hardware has two choices:

1) Be sloppy, ignore the pixel aspect ratio difference, and do a 1:1 pixel mapping from "normal" 720x576 content. Preserves all the pixels, but makes it 2.4% thinner. Distorts it.

2) Obey the spec, and rescale to the new pixel aspect ratio, to preserve image shape. You lose the extra pixels either side.

So as it stands, you lose image fidelity either way. You can't avoid this because the CEA standard's picture is claimed to be not as wide as normal SD content. Just as it would be if they used 704x576, but even worse, because of the need to rescale.

The best way to maintain image fidelity is for the CEA-861 standard to be capable of losslessly transmitting existing SD video data. That means it supporting 720x576 with the traditional pixel aspect ratio, so it has the traditional >16:9 frame aspect ratio. Which in my view is what it was always intended to do.

Please explain your thought processes. I can't understand your view at all. Why do you think an SD connection format that doesn't have the same pixel aspect ratio or frame aspect ratio as SD video is better for "Backwards Compatibility" and "Image Fidelity"?

I refer you to the section "Inconsistency in defined pixel aspect ratios values" in this article. That explains quite well why we should normally assume the traditional pixel aspect ratio. CEA-861/HDMI aren't part of a closed system where they can happily define a new ratio - they're there to deliver existing SD video content with the normal ratio.

I'm sure you argued previously that traditional SD video didn't have sharply defined edges. You're quite right. So where's the "Image Fidelity" merit in trying to ensure that all 720 pixels are seen, at the expense of distorting the image to do so. Those pixels were never intended to be seen. If we try to get them visible, we're going to have squeezed a 16.4:9 image into 16:9. This is why I'm confused. You seem to be arguing in favour of this 2.4% squeeze to get these outside-16:9 pixels visible.

But maybe that's not what you're saying. Can we try again on my question - what do you think sources and displays should be doing with existing SD video data, if you believe CEA-861D? --KJBracey (talk) 17:43, 8 September 2009 (UTC)

The CEA-861 standard can send the content aspect ratio so it can tell the device that though it is sending 720x576 that only 704x576 contains actual content. As such it looks to me like the CEA committee was aware of this issue. This can be seen in the HDMI specs (Auxiliary Video information (AVI) InfoFrame section) and this article (AVI Overview section) where it explains that the AVI InfoFrame can describe the content aspect ratio. --GrandDrake (talk) 23:57, 8 September 2009 (UTC)

GrandDrake,

Thanks a bunch. It was useful.

KJBracey,

You have written:

“ 1) Be sloppy, ignore the pixel aspect ratio difference, and do a 1:1 pixel mapping from "normal" 720x576 content. Preserves all the pixels, but makes it 2.4% thinner. Distorts it.

2) Obey the spec, and rescale to the new pixel aspect ratio, to preserve image shape. You lose the extra pixels either side.

”

It is a false dichotomy. The new specs both prevents distortion and preserves shape while practically preventing any loss at the edges. After all, the Standard-definition analog video has always been made with the issue of loss at the edges in mind and there has never been anything at the edge meant for you to watch. Fleet Command (talk) 11:48, 9 September 2009 (UTC)

Sample Aspect Ratio

Hi, I think it would be a good idea to explain that another name, used in the ITU H.264 spec and in the x264 encoder, "Sample Aspect Ratio", abbreviated in SAR, is used as synonym of pixel aspect ratio, and not as storage aspect ratio. —Preceding unsigned comment added by 93.39.234.164 (talk) 12:20, 7 March 2011 (UTC)

Hi. I don't think this term is common use, so you will probably have to add an accurate citation to the source mentioning it, as well as a quotation from the source, especially if the source is not freely available to everyone on the web. Fleet Command (talk) 18:37, 7 March 2011 (UTC)

The H.264 spec is freely available from ITU http://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-H.264-201003-I!!PDF-E&type=items, look at page 11, under the "Definitions" chapter:

"3.131 sample aspect ratio: Specifies, for assisting the display process, which is not specified in this Recommendation | International Standard, the ratio between the intended horizontal distance between the columns and the intended vertical distance between the rows of the luma sample array in a frame. Sample aspect ratio is expressed as h:v, where h is horizontal width and v is vertical height (in arbitrary units of spatial distance)."--93.39.230.157 (talk) 08:53, 8 March 2011 (UTC)

Human optic nerve non-square resolution ratio

What if NTSC signals have a higher horizontal resolution because our eyes have a higher horizontal resolution?

NTSC signals have the same horizontal resolution as vertical resolution. See the wikipedia page for Kell factor.146.115.72.215 (talk) 02:22, 18 March 2015 (UTC)

This subject is so bollixed I fear it will never be right.

What a mess. What an inconsistent, amateurish mess. Look folks. This is pretty easy stuff. Let's get it right.

Nothing has SAR = 704:480 because nothing stores 704 pix/line. An "NTSC" DVD (that is, a region-1 DVD) has SAR = 720:480 = 3:2, and a "PAL" DVD (that is, a region-2 DVD) has SAR = 720:576 = 5:4, but no storage media in existence has 704 pix/line. Consequently, PAR = 10:11 is useless. Forget it. It's just confusing things. That whole section dealing with 704 pix/line is nonsense. I know that some people think it's meaningful. Well, it isn't.

An image with DAR = 4:3 (that is, an SDTV image) stored on an "NTSC" DVD (SAR = 3:2), will, as a mathematical construct, have a PAR = DAR/SAR = (4:3)/(3:2) = 8:9. That PAR applies only to "NTSC" DVDs intended to be displayed on SDTVs. The "PAL" DVD equivalent is PAR = (4:3)/(5:4) = 16:15.

An image with DAR = 16:9 (this image unfortunately has no name -- I will call it "16:9-TV") stored on an "NTSC" DVD (SAR = 3:2), will, as a mathematical construct, have a PAR = DAR/SAR = (16:9)/(3:2) = 32:27. That PAR applies only to "NTSC" DVDs intended to be displayed on 16:9-TVs. The "PAL" DVD equivalent is PAR = (16:9)/(5:4) = 64:45.

There are just no other PARs worth talking about.

The pictures of 1:1 pixels and 2:1 pixels is wrong, wrong, wrong. PAR is NOT something you can see. What you see is display pels (not pixels). Pels are almost always square. PAR is a mathematical conversion ratio. Dump the pictures. They are misleading.

Here's how I explain PAR to a novice: When you telecine a 4:3, 35mm film frame (0.995 x 0.744 inches, which is a wide open, 35mm camera gate) for storage as 720 pix/line x 480 lines/frame (i.e., an "NTSC" DVD), the sample aperture should be approximately 0.001382 x 0.00155 inches -- calculated by simple division. That ratio is approximately 8:9. In other words, the sample aperture should be slightly thin (so as to get 720 samples per line instead of 640 samples per line). But when you telecine a 16:9, 35mm film frame (0.995 x 0.560 inches, a 35mm camera gate with height masked), the sample aperture should be approximately 0.001382 x 0.001166 inches -- same width, but much less height. That ratio is 32:27. In other words, the sample aperture should be reduced in height (so as to get 480 lines out of the squatter image frame). So you see, PAR is the proportions of the telecine's sampling aperture. Pictures of that sampling aperture might be helpful, but the pictures shown in the article are display pels, and that's just plain wrong.

Note how explaining film telecine aperture makes PAR understandable. But that 704 pixel stuff is just silliness. It was cooked up by people who thought that digital video resolution had something to do with analog luma bandwidth. Banish such thoughts. ...Yeah, yeah, I've read all the articles. But overscan is not an issue. Old video conversion hardware/firmware is not an issue. That's all old baggage ...And you know what? It never really was correct. --MarkFilipak (talk) 06:07, 24 August 2015 (UTC)

Bottom line: You're trying to tell us that nothing happens if you change the aspect ratio between 1:1, 5:4, and 4:3 in VLC player, or at least "nothing that you can see". If that was true, we wouldn't need concepts such as PAR and DAR. But it's a fact that we do because video pixels are usually not square, particularly not when it comes to SD video. And no, not only analogue SD video, since DV, D1, etc. are also non-square, as neither 720x576 nor 720x480 make up a 4:3 image. None of this has anything to do with the telecine of film footage (by which you're only massively overcomplicating things and not helping any novice in the slightest), but rather with the fact that analogue SD video had non-square samples, a property which was deliberately imitated when digital video was invented in order to maintain that 1 sample = 1 pixel. Plus, there never was a "16:9 35mm film frame". While 16:9 equals 1:1.77777..., there only were certain widescreen formats such as 1:1.66 and 1:1.85 (which never really matched 16:9 to begin with), which were achieved by a variety of means, including anamorphic squeeze, matting, or actually manufacturing film with that aspect ratio. --2003:71:4F05:5035:1411:6FF0:8DE6:C730 (talk) 08:09, 3 June 2018 (UTC)

PAL, 704 or 702??? Help me!

Heyyyy! What's going on with PAL's sampling? Is there 704 or 702??? Thank you! 2A01:E0A:AB0:C580:A4D1:6A9B:505:930 (talk) 16:54, 13 March 2023 (UTC)