why do many DVDs have 16px black border space @ left+right?

[Ghitulescu]

The measurement of PAR using this method is flawed, because it doesn't take into account the way the video material was processed.

Displaying a 16:9 image in a system of coordinates 1:1 (square pixels) must yield exactly 1024 pixels per 576 scanlines, regardless of any ITU, non-ITU issue. The actual PAR depends on how large the overscan limts are. BBC norms forsee the squeezing of an 1024x576 square pixels image, with 26 added black pixels (they may contain image as well), to 720x576. At play, those 1050 pixels will become 1024 (because it was flagged as 16:9), and from those, only ~1000 will contain the actual image. During playing on a real DVD player, those extra 26 pixels will fit into the analog blanking, which may or may not be displayed on underscan-enabled studio monitors (if they actually contain parts of the initial image), but not on regular TVs.

This might be the case for all film-based movies (see here that it's not always the case -> http://forum.doom9.org/showpost.php?p=1494312&postcount=4). What about the movies coming from the analog tapes? Here we come to a whole tin of worms. Most playback issues are solved, thanks god, by overscan and that the average Joe can't see a 1% distortion in image.

On the other hand, Scenarist supports 960x576/480 as the starting point for wide-screen assets, so it must be assumed that it performs PAR stretching by itself.

PS:
Note that squishing the DVD to 99% vertically is the same as stretching it 1.01% horizontally
No, it's not. Mainly because the DVD is asymmetrical (PAR <> 1).

[hello_hello]

The measurement of PAR using this method is flawed, because it doesn't take into account the way the video material was processed.

He's measuring the PAR of a particular DVD compared to a square pixel HD source. He's comparing the two, then determining what PAR the DVD should use in order to have an identical aspect ratio to the HD source. How is the way the DVD video is processed relevant to that?

[jmac698]

Hmm, you both have a point. He's saying that on display, some of my pixels should be cutoff, meaning I'm getting an extra horizontal zoom which changes my actual PAR. But to me, that happens on display, I was just measuring the inherent PAR as hello understands.
But I think that's just cropping, the 18 pixels cutoff.

[Ghitulescu]

The whole PAR issue is relevant only to playback on computer. In the good old analog world, there was only one item, the bandwidth. The analog standards imposed as mandatory the sampling frequency, the number of scanlines, and the timing, among others. It was the duty of the HW manufacturers to ensure the timings were respected, so the line began always where it should begin. Due to the inherent troubles with the tape, the TBCs could give a stable image, based solely on the timing and nothing else.
When MPEG-1/2 hit the video world, it imposed some of its restrictions, like the number of pixels, which must have been MOD16. Unluckily, neither NTSC nor PAL sampling could provide the required MOD16 values, neither 704 nor 720. Both the ~711 and ~702 active samples must be converted to 704/720 pixels, by adding "empty pixels" which are considered to be the nominal analogue blanking, ie the part of the signal that doesn't carry any regular video signal. To fill this gap, many digital videos left a black space at left. It should have been 2-18 pixels wide, however, it may be more than 18 and may also be added at the end of the line, to mask the margins imperfections of the video, margins which are anyway covered by the overscan. These parts may be seen only with studio monitors, or on computers, because the videocards digitize the whole line between the synch pulses.
Now it's really the job of the studio how do they put the real image into the theoretical scanline, if they take into account the unused part thereof, and if they decide to respect the indications of the ITU-R, which defines where in this long scanline the actual video may be placed. With or without ITU-R, the so-prepared scanline fulfil all the requirements of the MPEG-2, DVD, NTSC/PAL standards, because it has to be transported through an analog interface to a PAL/NTSC display. As one sees in the examples, there's no consensus regarding this topic. The burden of this whole issue was shifted to the chip manufacturers.

[hello_hello]

We maybe both have a point if you happen to be playing the original DVD using an analogue TV. Although doesn't the overscanning involve the whole picture (top and bottom too)?
I assumed, maybe incorrectly, the whole point of the exercise would be to determine the correct pixel aspect ratio to use when converting/remuxing the DVD to a less archaic format such as MKV, otherwise working out the correct pixel aspect ratio would be somewhat pointless given you've got no control over it.

I can't speak for every TV, but mine allows me to disable the overscanning when using a HDMI and composite inputs while displaying 1080i/p video, which basically means it'll disable the overscanning of all video regardless of resolution, because it's generally upscaled to 1080p by the PC/Bluray player. I've not experimented with using a DVD player because I don't own one, and I can't remember the last time I watched video directly off a DVD anyway.

Out of curiosity, I just compared the playback of an anamorphic MKV DVD encode (aspect ratio 2.40) using both my PC and Bluray player as the playback device. Switching between them to compare aspect ratios (overscanning disabled) the video displays identically (as best as I could tell, to within one row of pixels on a 1080p display), so for me when encoding DVDs, as long as I get the DAR correct when converting the DVD it's always going to display correctly..... well at least it'll always display in the same way.
How the DVD was originally processed and what aspect ratio distortion might have been introduced when using the playback devices of yesteryear seems irrelevant to me, so determining the correct aspect ratio the way you've done it, or using my less accurate method, isn't flawed when using a playback device which reproduces the aspect ratio correctly.

[hello_hello]

The whole PAR issue is relevant only to playback on computer.

As I demonstrated with the example using my Bluray player in my previous post, that simply isn't true. It's probably not true for most other standalone media centres either. You're still living in a DVD, ITU resizing, overscanning world. I'm not.

We're not all converting DVDs to DVD format, some of us are converting to other formats where the playback device follows the correct aspect ratio and where the ITU resizing problem is taken out of the equation on playback.
Therefore determining the correct PAR when converting the DVD to another format, to ensure the correct DAR on playback is perfectly valid. For those still using devices which may or may not follow the standard you have when you don't really have a standard, then I can see why you have to settle for not worrying about determining the correct PAR and just hope for the best.

[Ghitulescu]

Out of curiosity, I just compared the playback of an anamorphic MKV DVD encode (aspect ratio 2.40) using both my PC and Bluray player as the playback device. Switching between them to compare aspect ratios (overscanning disabled) the video displays identically (as best as I could tell, to within one row of pixels on a 1080p display), so for me when encoding DVDs, as long as I get the DAR correct when converting the DVD it's always going to display correctly..... well at least it'll always display in the same way.
I can understand this.
However, most people here prefer to cut the margins during reencoding and fall into the trap of PAR/DAR/SAR values. At playback on any gear other than a computer, the FW must recreate a regular signal from a frame size like 632 x 234 (assuming a PAR of 1). How the HW player does it (stretching, adding mattes, both) it's a specific issue of that particular model.

I said it long time ago, that issue existed since the very first DVD rip, it was only "recent" discovered (compared to the history of DVD ripping), and, unless a purist, one may not even notice the difference between the three values of the PAR, placed within ~1% from eachother.

Is it worth the trouble? Not for me, I use the BBC practice and that suffices. Thanks god, HDTV brought a PAR of 1 (except for HDV, but even there it's only one and well defined PAR).

[Ghitulescu]

As I demonstrated with the example using my Bluray player in my previous post, that simply isn't true. It's probably not true for most other standalone media centres either. You're still living in a DVD, ITU resizing, overscanning world. I'm not

There is no PAR outside computer. Have a look into the standards. Every PAR listed in this forum (and similar) is a calculation using other values set by these standards, mostly timings.

[hello_hello]

I can understand this.
However, most people here prefer to cut the margins during reencoding and fall into the trap of PAR/DAR/SAR values.

As far as generalisations go, it's not really relevant to the discussion.

At playback on any gear other than a computer, the FW must recreate a regular signal from a frame size like 632 x 234 (assuming a PAR of 1). How the HW player does it (stretching, adding mattes, both) it's a specific issue of that particular model.

Which if true, negates the reason for finding the correct aspect ratio when encoding how exactly?
And shows that the original processing of the DVD must be taken into account when determining the correct pixel ratio to use for encoding how exactly?
I've not seen anything to suggest a playback device can't resize a square pixel video with a 632 x 234 resolution correctly, even though you've been using it as an example to suggest they don't for as long as I can remember.

Is it worth the trouble? Not for me, I use the BBC practice and that suffices. Thanks god, HDTV brought a PAR of 1 (except for HDV, but even there it's only one and well defined PAR).

I don't encode video to DVD format. I'm not even sure encoding to DVD format relates to the discussion on determining the original DVDs correct PAR in any way.

[hello_hello]

There is no PAR outside computer. Have a look into the standards. Every PAR listed in this forum (and similar) is a calculation using other values set by these standards, mostly timings.

Semantics. I don't care as long as my playback devices play back my MKV encodes with the correct DAR. As long as the DAR is correct according to the container's aspect ratio, it can calculate it any way it likes.

[Ghitulescu]

Which if true, negates the reason for finding the correct aspect ratio when encoding how exactly?
And shows that the original processing of the DVD must be taken into account when determining the correct pixel ratio to use for encoding how exactly?
I've not seen anything to suggest a playback device can't resize a square pixel video with a 632 x 234 resolution correctly, even though you've been using it as an example to suggest they don't for as long as I can remember.
I don't care, I don't do pointless encodings, just to fill in my spare time. Letting the video as its designers created it, be it bad or good, increases the chances to be playable in the correct form, for two generations.
I don't encode video to DVD format. I'm not even sure encoding to DVD format relates to the discussion on determining the original DVDs correct PAR in any way. You're may be not, but I know it does.
Semantics. I don't care as long as my playback devices play back my MKV encodes with the correct DAR. As long as the DAR is correct according to the container's aspect ratio, it can calculate it any way it likes.
I won't count on the fact that a playback device, designed to play other formats and sold as such, can be that perfect in playing just everything, just perfectly, and still cost under 200€. There are corners to be cut, and legal agreements to be fulfilled. What good a strong BD protection, when it suffices one hacker to copy the movie, repack it into an MKV file and distribute the latter, then all these BD players can unrestricted play it :) ?

You may end in the situation that led to PAR/DAR/SAR issues, people that encoded for years their collections to watch them on a PC discovered that their precious movies do not look the same when played with a hardware mediaplayer on a TV.

[hello_hello]

You're wandering off the topic with pointless generalisations.

We were discussing how to determine the correct aspect ratio when converting/displaying a DVD. You claimed we're using a flawed method if we don't take how the original video was processed into account. You still haven't explained why, aside from pointing out not all playback devices will display the video with the correct aspect ratio, which is something pretty much everyone knows.

Whether you encode DVDs or not is irrelevant. Whether you think it's pointless or not is irrelevant. Guessing about how a playback device may resize video isn't even an argument, so it's definitely irrelevant. Some peoples lack of understanding of PAR/DAR/SAR issues when encoding is irrelevant.
Legal agreements, BD protection and hacking video is irrelevant to the point of being sad.

In case you've forgotten while wondering off-topic to share you knowledge of how video was processed in the good old days, the topic we were discussing is how to determine a DVDs correct PAR in order for it to have the same DAR as the square pixel HD version.

[jmac698]

My measurement still means something - if you want to use it as a basis of a further calculation for DVD, that's fine. But yes, I did have in mind a non-DVD format which would ultimately be displayed 1:1. For example if I put 1.8 into my xvid, (assuming) my player/tv will show it correctly. I do far worse things - take a 2.4:1 aspect and stuff it into 720x480 and fudge the PAR to make it display properly - this is extreme anamorphic. I haven't measured if my particular bluray player does straight resizing or ITU resizing (but it looks ok). There's actually different standards for aspect for HDMI than for analog (can't find reference).

http://en.wikipedia.org/wiki/Pixel_aspect_ratio Rec 601 implied 10:11 PAR. SMPTE RP 187 proposed 177:160 but was ignored (according to the article). Following along with the article, some TV's showed more of the edges than they should, so the "black" pixels were then filled with image to avoid seeing black pixels.
For PAL, Rec601 implied 59:54 but this was to hard to implement, so software chose 12:11 instead. According to the references, Adobe After Effects and CS4 use 12:11 for PAL.
Finally this reference gives a different value http://en.wikipedia.org/wiki/Overscan#Analog_to_digital_resolution_issues for NTSC, not the 710.85 I had memorized.. so now I'm just confused again. In the end, the only thing I can know for certain is my measurements with my playback device :)

[jmac698]

Although standards-compliant video processing software should never fill all 720 pixels with active picture (only the center 704 pixels must contain the actual image, and the remaining 8 pixels on the sides of the image should constitute vertical black bars), recent digitally generated content (e.g. DVDs of recent movies) often disregards this rule. This makes it difficult to tell whether these pixels represent wider than 4x3 or 16x9 (as they would do if following Rec.601), or represent exactly 4x3 or 16x9 (as they would do if created using one of the fudged 720-referenced pixel aspect ratios).

Ha! So true :)

Interesting that it's not a problem at all with 1088 line mpeg2 video - no one questions to discard the extra 8 lines.

..and you guys were pretty civil today :) If someone grew up in an analog world, they can be quite knowledgeable it, it's just a different perspective than someone in a digital world. I had more of a digital assumption in this case, but since I digitize analog I have to be aware of it too.

[Ghitulescu]

http://en.wikipedia.org/wiki/Pixel_aspect_ratio Rec 601 implied 10:11 PAR. SMPTE RP 187 proposed 177:160 but was ignored (according to the article). Following along with the article, some TV's showed more of the edges than they should, so the "black" pixels were then filled with image to avoid seeing black pixels.
For PAL, Rec601 implied 59:54 but this was to hard to implement, so software chose 12:11 instead. According to the references, Adobe After Effects and CS4 use 12:11 for PAL.

Rec 601 is a recommendation, so is the SMPTE RP 187 (RP stands for Recommended Practice). They can be ignored with no consequences, other than, probably, compatibility.

Interesting that it's not a problem at all with 1088 line mpeg2 video - no one questions to discard the extra 8 lines.
Because the last 8 lines are to be discarded anyway. The standard foresees 1125 scanlines BTW, so it has plenty of room.

[hello_hello]

I haven't measured if my particular bluray player does straight resizing or ITU resizing (but it looks ok). There's actually different standards for aspect for HDMI than for analog (can't find reference).

It's interesting you should say that. I was just messing with the 4:3 settings on my TV (mainly due to a discussion in another thread).
By default my Bluray player outputs 4:3 DVD video as 16:9 when connected via HDMI (although you can change the setting) so it's left up to the TV to correct the aspect ratio.
Putting the TV in 4:3 mode it obviously overscans (no crud down the sides, just black or grey borders) but I'm almost 100% sure it's not using ITU resizing. I compared it to the non-ITU MKV encode of the same DVD and I'm fairly certain they display with the same aspect ratio.

So then I checked the way the Bluray player resizes 4:3 video when you force it to obey the DVD's aspect ratio. Same again. I'm pretty sure the Bluray player was just using straight 4:3 resizing too.

I don't know whether the player would resize the DVD differently using an analogue output, or how it resizes 16:9 DVDs. Hopefully using HDMI it just uses straight 16:9 resizing too. I'll have to check later.

[jmac698]

I think we need a better test signal. I made a video with lines moving 1 pixel at a time, so you could just read off where a line was. This is useful to measure any cropping. To compare aspect, what we need is a 1080p grid that displays on the TV 1:1, and then another grid on an SD source. Ideally if you could switch between the two, you could hang a piece of paper over the top of the TV and mark the different lines indicated. How about I try to write a test signal and you can test it?

[hello_hello]

How about I try to write a test signal and you can test it?

Sure, count me in.