PAR, SAR, DAR... pfwoar! (or: How I Learned to Stop Worrying and Love the Rec.601)

[PDB]

Sometimes writing things out is the only way you can truly understand a thing.

[pipefan413]

Sometimes writing things out is the only way you can truly understand a thing.

Strongly agree. I'm very much a logical/numbers guy so doing a crapload of calculations makes a big difference too, but I know that won't be true of everybody, which is why I've tried to express these concepts as both a visual thing and a numbers thing (different learning styles for different people and all that).

I just wish I could pin down this whole thing about the actual results of using 10:11 vs 8:9 for 720px video where it has hblank included in it. From what I can tell, the implication seems to be that 10:11 is correct and the player is meant to be smart enough to know you really mean 8:9 but I am deeply suspicious of that... as for the 486 vs 480 lines in terms of 4:3 DAR, I haven't a damn clue.

[spoRv]

The idea of putting all together in a blog or something is good; let me know if you need some inputs!

Back to topic: everything is connected - screen aspect ratio, pixel aspect ratio, overscan etc.

To get a better answer, I'd like to know - if few words - what are you aiming to.
Because, as I wrote before, getting square pixel SD from a non-square pixel SD source is useless IMHO, as the display would always resize it.
About 486 pixel, if you want to retain them (if there is no garbage, or that garbage is less than 6 pixel) in this case it could worth (relatively) to get square pixel, as even if you have SD display none is 486px high.

Still, I think 704/720x480/576 is pretty standard and any player is supposed to display it in proper 4:3; for the missing 16px horizontal, answer is: it depends! I made quite some captures where I have to crop not 16px, but more - 24 or more... so, how to resize them properly? If you have not something round - a wheel, for example, could suffice in some cases - it's up to your... good eye, experience, sense of proportions etc.

I guess now you are even more confused... sorry!

[spoRv]

Also, this thread could be of some help, maybe: https://fanrestore.com/thread-2396.html

[pipefan413]

The idea of putting all together in a blog or something is good; let me know if you need some inputs!

Back to topic: everything is connected - screen aspect ratio, pixel aspect ratio, overscan etc.

To get a better answer, I'd like to know - if few words - what are you aiming to.
Because, as I wrote before, getting square pixel SD from a non-square pixel SD source is useless IMHO, as the display would always resize it.

I agree almost entirely, except that the one reason to do it is the very problem I'm describing: inconsistency between playback devices and software. Sometimes players seem to assume a 720px wide image is to be entirely included in the DAR, even though that's wrong because for the particular video in question 16px of those 720 is hblank. In other cases, players seem to correctly acknowledge that although the picture is stored as 720px wide, only 704px are actual active picture and thus to be included in DAR (for the purposes of explaining it, I've consistently used 4:3, though the same principle applies for 16:9 DAR).

This is pretty much the only reason I upscaled some of my SD-sourced projects e.g. A GOOFY MOVIE and THE EXORCIST. I didn't really want to increase their storage size by upscaling to 720p but did so because it allowed me to use square pixel aspect ratio (i.e. 1:1 SAR) without breaking format compliance. I only did so after first testing countless different crops and reshapes to try to ascertain the most correct looking image shape but essentially gave up because it was so incredibly subjective.

About 486 pixel, if you want to retain them (if there is no garbage, or that garbage is less than 6 pixel) in this case it could worth (relatively) to get square pixel, as even if you have SD display none is 486px high.

The question is not about whether the actual picture contained in that 6-pixel sliver is important enough visually to merit inclusion, it's whether or not it's *necessary* to include it to achieve the correct display aspect ratio. As I said, it changes the shape of the picture. If I include the 6 pixels, the picture in the 4:3 display aspect ratio will look more squished in the horizontal dimension, and if I exclude them, it will look more vertically squished. One is correct and the other isn't. I don't know which way round it is.

Still, I think 704/720x480/576 is pretty standard and any player is supposed to display it in proper 4:3; for the missing 16px horizontal, answer is: it depends! I made quite some captures where I have to crop not 16px, but more - 24 or more... so, how to resize them properly? If you have not something round - a wheel, for example, could suffice in some cases - it's up to your... good eye, experience, sense of proportions etc.

I have also found that to be the case and in fact I alluded to it: even on this BRIDES OF DRACULA LaserDisc for example, the blanking moves around from scene to scene. Sometimes there are 8px on left and 8px on right, sometimes more like 10/6, and so on. And on different discs they seem to get the amount of blanking wrong. However, my assumption - which may be wrong! - is that if the disc is authored with too much or too little horizontal blanking on the sides, that wouldn't actually change how the display handles the aspect ratio on displaying it at 4:3. In other words if they leave too much black space on the sides, the display will still assume that there are only 16px worth of blanking because that's how it's supposed to work. However, as I type this, I suspect maybe that's only true of *digital* displays, and the analogue nature of how blanking works on an actual CRT possibly means that it really *does* change the display aspect ratio on playback... I don't know.

I guess now you are even more confused... sorry!

Not really, I understand everything you're saying, I think that the very specific points I was trying to get at have been missed somewhat, perhaps in no small part due to the frankly alarming length of my post, hahah.

[deleted user]

Well, you could approach it like a theatrical restoration, which always asks: How would it have looked in an actual cinema? In this case it would be: How would it have looked on an actual TV/DVD/LD combo?

The first and obvious step towards that would likely be to create a few test scenes with the different options and create a DVD from that and then play that back on actual DVD/TV combos and see what happens. If you find a consistent behavior this way: use that. If you don't, who gives a flying fart, just do whatever. If there's a situation where 95% of players do A, but 5% do B, then you can argue let's do A because that will reflect the most common typical experience.

I know I conflated multiple of your questions here but ultimately you already concluded as much as: There is no clear answer that can be based solely on logic and standards.

You'll have to find out how actual hardware did it.

Another thing you could take into account is to ask people who actually worked on mastering those DVDs. However I think it's irrelevant because following their intent would be in the same vein as all these modern remasters and regrades that completely change how the movie looked in theaters because "it's closer to the director's intent now". What matters is not what the mastering facility intended, but how it actually looked on actual common setups. So let's say that most DVDs ended up looking actually wrong and stretched/squeezed on actual set ups ... then that's how they need to be preserved.

Of course that's just one way to look at it, but it's the only rational one in my eyes. You can't account for the intent of people you've never met. You can do some empirical research on how it actually ended up looking though.

[pipefan413]

Well, you could approach it like a theatrical restoration, which always asks: How would it have looked in an actual cinema? In this case it would be: How would it have looked on an actual TV/DVD/LD combo?

This is exactly my thought process, yes!

The first and obvious step towards that would likely be to create a few test scenes with the different options and create a DVD from that and then play that back on actual DVD/TV combos and see what happens. If you find a consistent behavior this way: use that. If you don't, who gives a flying fart, just do whatever. If there's a situation where 95% of players do A, but 5% do B, then you can argue let's do A because that will reflect the most common typical experience.

I know I conflated multiple of your questions here but ultimately you already concluded as much as: There is no clear answer that can be based solely on logic and standards.

You'll have to find out how actual hardware did it.

That's just it, isn't it? Different CRTs have different geometry. There would need to be a "reference" CRT (or at least one with "reference" standard geometry) in my possession for me to be able to do this. There isn't. I have 2 CRTs but neither is amazing.

Another thing you could take into account is to ask people who actually worked on mastering those DVDs. However I think it's irrelevant because following their intent would be in the same vein as all these modern remasters and regrades that completely change how the movie looked in theaters because "it's closer to the director's intent now". What matters is not what the mastering facility intended, but how it actually looked on actual common setups. So let's say that most DVDs ended up looking actually wrong and stretched/squeezed on actual set ups ... then that's how they need to be preserved.

Of course that's just one way to look at it, but it's the only rational one in my eyes. You can't account for the intent of people you've never met. You can do some empirical research on how it actually ended up looking though.

TV geometry is the weak link in this logic, again. I don't think it's consistent. So regardless of the signal sent to it, if a given TV happens to have slightly too wide or slightly too tall geometry, that's going to obfuscate results. But yeah, I think you're basically thinking along the same lines as I am here in so far as trying to preserve it as it was rather than "fix" it.

The other problem with this sort of testing methodology is that there's so small a difference between the two image shapes that with the added curvature of a CRT I don't think I'd be confident in being able to confidently declare my digital results a match either way (reshaping ~704 x 486 to 4:3 vs reshaping ~704 x 480 to 4:3). It's not particularly easy to empirically measure. Maybe a test disc would be the key thing here, I could perhaps literally measure test patterns with a tape and then measure a capture of the same test pattern in the digital realm. But again, that assumes my TV geometry is correct to begin with, and I don't think that's ever a safe assumption with basically any CRT...

[deleted user]

My idea was more along the lines of seeing whether the 16 extra pixels appear on an old DVD/TV setup. For example you could mark them in some special color that would be immediately visible. However it's also possible that DVD players have an auto detection of black pixels, so that would have to be tested as well. Personally I would have discarded the TV itself from the equation and focused only on the output of the DVD player, but I suppose both is an argument. Trying to emulate a CRT would also involve gamma, colors etc., so a whole other box of variables.

But I suppose measuring test patterns should work too yup! Could use one of those tapes that is flexible.

Another idea, could try to devise a test pattern that uses some form of aliasing pattern to detect a specific aspect ratio/resizing. Tho that would be beyond my math capabilities, haha. You know, how sometimes downscaling some patterns results in a specific aliasing pattern or very specific distortion? If you could find one that is reproducible enough, you could just send it out to various people and they could take a photo and you would immediately see what you're dealing with. But maybe that's just wishful thinking idk.

[pipefan413]

My idea was more along the lines of seeing whether the 16 extra pixels appear on an old DVD/TV setup. For example you could mark them in some special color that would be immediately visible. However it's also possible that DVD players have an auto detection of black pixels, so that would have to be tested as well. Personally I would have discarded the TV itself from the equation and focused only on the output of the DVD player, but I suppose both is an argument. Trying to emulate a CRT would also involve gamma, colors etc., so a whole other box of variables.

Overscan, though? The 16 px are horizontal blanking anyway and wouldn't be on screen at all on an "old" TV (assuming by that you mean CRT, but maybe you just meant the "old" part to apply to DVD player and meant newer TV). The 6 scanlines of diff vertically also probably won't be on screen in the case of the LaserDiscs (obviously not with DVD because DVD doesn't even contain those 6 pixels, they'll presumably just be empty in the signal passed to the TV). It might just be that what you meant was specifically "not a CRT TV", which is confusing me because I've consistently been using "TV" to mean "CRT" here (since all of this stuff is based on hangovers from analogue tech).

If I do it with a modern flatscreen TV, it's certainly more possible to measure, but that only really tells me what my specific DVD player(s) and TV do, not what DVD players and TVs do in general; it could simply be that my player treats 720px as active image and resizes it into 4:3, whereas actually most chop the 16 px hblank off first then resize that into 4:3. Or vice versa. So we're back to the problem of inconsistencies between players, which I can't realistically expect to measure with much confidence. I am but one man, hahah!

And yes what you said about black pixels is something I wonder: I believe at least *some* DVD players remove the horizontal blanking, but I don't know if they actually detect the black bars; the ones that do chop off horizontal blanking might just chop it off regardless. I think when I was trying to test this to a certain extent on software players I found that the horizontal blanking on A GOOFY MOVIE was actually cropped off by the software precisely to the pixel, so it certainly seemed to be able to tell where the blanking ended and the active picture began. Whether that's actually true of real hardware players, I don't know; if not, they'd potentially just assume 8px of hblank on each side which would slightly crop the picture on a newer DVD that actually uses all 720 pixels of width and gets stretched to 16:9, for example (since they're assuming only 704 pixels are meant to be displayed). I'd hope not, but that might be the case. I've seen reference to this where people have commented that it depends on your player whether such DVDs are displayed correctly.

Going back to CRTs for a second though, this is something that's connected: part of what I wonder is to do with how CRTs actually parse the horizontal blanking information, which is something I haven't yet researched in sufficient detail. Y'know how @spoRv mentioned that there isn't generally exactly 8 pixels of blanking on the left and 8 pixels on the right? Sometimes it's (occasionally quite a lot) more than that, and it's almost always lopsided, e.g. 4 pixels of blanking on the left but 12 on the right, or 14 on the left and 8 on the right, or you name it, whatever. I'm not sure whether a CRT would accurately chop off the scanlines when it drops to black thus excluding that from the 4:3 DAR. If my assumptions about this turn out to be right, then it would actually be quite tedious to accurately frame any LD cap (probably less so DVD, which seems not to do this afaik) because the horizontal blanking often shifts around from left to right across the whole runtime of the film. I notice with BRIDES OF DRACULA that sometimes there are maybe about 10 pixels of blanking on the left and 6 on the right at one moment, then a couple of scenes later it's moved and now there's something like 7 pixels of blanking on the left and 9 on the right, or whatever. This shifting around continues throughout. If a CRT would actually keep the image accurately centred as the blanking moves, I'd have to go through the entire film in AviSynth continually changing the position of the horizontal blanking (i.e. cropping it differently from moment to moment), do any rescaling necessary to get it to either 704 px wide or 640 if I'm going for square pixels, and then reinsert pure black horizontal blanking at the end if I'm keeping it non-square (precisely 8 pixels on either side). Quite laborious. And even then, this only deals with the horizontal blanking, not the 486 px question... I still want to try to determine whether those should be included in the 4:3 frame or not.

I might be answering my own question here: it might simply be that there is no real definitive way of answering this because every display is slightly different, at the very least in terms of CRTs. There seems to be a lot of variation with DVD players too, as I discussed before.

Another idea, could try to devise a test pattern that uses some form of aliasing pattern to detect a specific aspect ratio/resizing. Tho that would be beyond my math capabilities, haha. You know, how sometimes downscaling some patterns results in a specific aliasing pattern or very specific distortion? If you could find one that is reproducible enough, you could just send it out to various people and they could take a photo and you would immediately see what you're dealing with. But maybe that's just wishful thinking idk.

That's some real sci fi s*** right there, I love it. I wouldn't know where to start but it's a cool idea.

[pipefan413]

Separate post mostly just for clarity but here's what I'm on about with the 486 vs 480 lines question...

Here is an MS Paint circle framed correctly for 4:3 at 640 x 480:



Here is one framed correctly for 4:3 at 648 x 486, i.e. assuming that the full 486 lines are actually supposed to be included in the 4:3 DAR (cropping aside):




Here they are stored at 704 x 480 (DVD res, with the horizontal blanking left out) and 704 x 486 (LD res, without horizontal blanking) respectively:




Right, so here's what happens if you take the 486-line one and assume that you're meant to crop off the 6 lines before rescaling:



It's not all that obvious at a glance, so you might not see it without flipping back and forth between this and the correct one (2nd image in this post) but the circle is slightly squished horizontally. My question is whether this is what is happening if I crop off the 6 extra pixels in an LD capture rather than actually taking them into account when framing for 4:3. This wouldn't surprise me because I've felt that some LD caps (e.g. Near Dark) looked like they might be slightly too skinny.

If you were to instead resize the 704 x 486 one to 648 x 486 (4:3 DAR) first, and then crop in on all 4 sides to get the vertical resolution down to 480 lines while preserving the 4:3 framing, you'd get this:



Here's that last one followed by the previous one so you can see them next to one another (486 px framed for 4:3 vs 486 cropped to 480 then framed for 4:3):




See it?



So either images as captured by my capture card at 720 x 486 are framed such that ~704 x 486 is meant to be resized to 4:3 on playback, or they're framed such that ~704 x 480 is meant to be resized to 4:3 playback. I'm trying to work out which. (N.B. I say "~704" because although horizontal blanking is theoretically meant to be 8 pixels on either side, the actual horizontal blanking fluctuates slightly and isn't always precisely 16 pixels, so in reality the image could be slightly less than 704 pixels wide when you remove the black on both sides.)

My next move should possibly be trying to find a test disc to capture through my capture card and see if I can determine empirically what correct 4:3 framing should be in order to achieve accurate geometry of the image. I'm guessing that should be doable if I have test patterns (I can fiddle with the scaling until I hit upon perfect geometry and then examine what AR manipulations would give equivalent results).

So you can see what difference this makes to an actual LD cap, I'll put the corresponding images from my top post here that relate to this particular element of what I'm trying to work out.

Here's what happens to an LD cap if I crop off 16 px of horizontal blanking (720 cropped to 704) then frame the whole 486-line height to 4:3 (i.e. resizing 704 x 486 to 648 x 486):



Same frame but with the 486 lines cropped to 480 before framing for 4:3 (720 x 486 cropped to 704 x 480 then resized to 640 x 480):



Note: I've highlighted the 6 removed lines in red by adding them back in as red border in the 480-line version, and I've also added a 4-pixel black border at the sides to align it with the 648 x 486 version that has all 486 lines framed at 4:3. I'd be removing these in practice from both (so they'd both be cropped down to 640 x 480, which would remove actual picture from the first one but only remove the red and black borders from the second).