525-Line Widescreen Pixel Aspect Fraction

[Acid_Reign]

Okay, so I can find most of the decimal representations for different "NTSC" Pixel Aspect Ratios, but I have a harder time finding the true, fractional forms. I'd like to know because I am writing my own conversion calculator and would like to be as accurate as possible.

Widescreen in particular is one of the major ones that I'm missing. In programs like Photoshop it's notated as "1.2," but like all of those they're merely rounded off for convenience. The closest I have found is 40/33, which gives you 1.2121 repeating, which I believe to be erroneous for reasons I can't remember (revisiting an old project here). Does anyone have the exact figures, or is that already correct?

Also, if anyone has the figures for lesser-known PARs, that'd be cool too. Right now I have Square, Anamorphic, D1/DV, D2, SVCD, D4/D16 Standard, and D4/D16 Anamorphic (all 525). I may make a 625 later, but not at this time.

Thanks

[Cornucopia]

As "NTSC" is in actuality only 4:3 DAR, there really is only 1 regular PAR and that's 10/11 (or 0.909090 repeating). If you were to incorporate Anamorphic Widescreen into that (where 16:9 DAR = 4x4 / 3x3 or 4/3 x 4/3), you would multiply this PAR by the additional 4:3. That's how you arrive at 40:33 (aka 1.21212 repeating). It's NOT erroneous.

You --could-- include SVCD, CVD, DVB540 and other non-regular 601-based resolutions, which would have different PARs (in order to maintain the same correct DAR), but those 2 are the main ones. PAL, of course, has 2 of it's own main ones.

Scott

[Acid_Reign]

Well see, the thing is, I was actually operating under the assumption that a 4:3 DAR is not actually 4:3, considering 0.9 is (according to http://lipas.uwasa.fi/~f76998/video/conversion/) the fraction 4320/4739 [to one decimal point].

If I were using 10/11 though, then I'd definitely agree in saying that 40/33 for widescreen is correct. But I'm being purposefully pedantic here, for better or for worse.

I'm really not sure where the former ratio even comes from, but I'm guessing it has to do with making things "neater-looking" for people when explaining things because the values are close enough to the originals that they can be substituted without any noticeable discrepancy.

EDIT: I did the math, and got what I might've called the "true" decimal notation for the widescreen PAR (1.215446296687064781599493564043, which comes out to 5760/4739), until I Googled the number and came across this:

http://forum.doom9.org/showpost.php?p=1004368&postcount=5

...which I guess explains a few things. Apparently the ITU and MPEG-4 definitions of PARs are close to but distinct from each other, for whatever reason. It makes me wonder, though, why so many sites quote the MPEG-4 definitions when talking about non-MPEG-4 material, like miniDV, DVD, etc.

[edDV]

This thread shows links that derive analog to digital aspect ratios.
https://forum.videohelp.com/topic326464.html?

This link was particularly helpful. Check #4 FAQ.
http://lipas.uwasa.fi/~f76998/video/conversion/#faq


PS: Oh I see you found that one.

[Cornucopia]

I've always considered that page a little too anal-retentive to be directly referrable for my tastes, but it is good reading.
However, one thing that never gets clearly and directly explained on that page is how figures like "4320/4739" are arrived at (or derived from). The other determinations specifically used these numbers for their main calculations. What if they were simpler to begin with?
(IOW, I'm not arguing the math, I'm arguing the possibility that the page's author may not have the full knowledge of the reason for PAR derivation available to him/her).

Nonetheless, even granting these more arcane figures, some simplicity does creep in:

Now we need a horizontal rescaling factor, which in this case is (4752/4739) / (4320/4739) * 1 which equals to 11/10.

To me, what this is saying is that the conversion factor when going from square pixel sampling of 640x480, to non-square pixel sampling of 704x480, is 11/10 (or 10/11 if going the other way). In a nutshell, this is precisely the difference between square and non-square PARs.

Truth be told, I would guess that even if your calculations had to take into account the more complicated figures, the fractional difference between that and "fudging" with 10;11 overall is probably on the order of 1 or 2 pixels--NOT enough to worry about. To human eyes, a square and a circle will still look right.

There are many areas that I've been told I'm too exacting and detail oriented about, but I guess this isn't one of them.

Scott

[Acid_Reign]

Anal-retentive as it may be, I appreciate it because it actually clarified some things for me, as opposed to "other" video reference sheets which just give you simplified numbers expecting you to accept it as it is and not notice any difference.

I am a purist, and so while yes, one or two pixels doesn't make too much of a difference to the casual observer, it's in effect the same thing as stretching the Mona Lisa by one unit of measure and saying it's the same painting. Not a lot of people would be happy settling for that, and I'm the same way.

But for me, it's as much about accuracy as it is about learning the internals behind what otherwise appears to be just "random" numbers. I'd rather know why something is the way it is than not question it. And when it comes right down to it, I don't like non-square pixels or mixed number framrates—but that's the format I'm working in, and as such I'd like to abide by its rules as closely as possible.

And often times things that are seemingly insignificant can cause problems down the line, like for instance if one takes into account the fact that 29.97002616431 Hz is the actual framerate for M, not 29.97, then you actually start to lose miliseconds per day of drop-frame video. That's not a problem for most available content right now, which usually doesn't exceed 3 hours on a single DVD, but looking to the future, with advances in technology such as HVD, if one were to put that much video on it in that format, it'd start to be a problem.