Question about encoding 16x9 video for DVD

I've done enough reading to know that 704x480 is the closest NTSC 16x9 video size for DVD video, and 720x480 is actually a little wider than 16x9.  The question I have is: what's the difference between adding 8px borders (nominal analogue blanking) onto 704x480 video to make it 720x480 and putting that on DVD (which seems to be the way most commercial DVDs with correct aspect ratios use), and simply encoding a 704x480 video and putting that directly on DVD without modification?

The problem I'm encountering is that adding black borders and then putting it in a 720x480 frame results in digital players (software players, Blu-ray players) incorrectly assuming the black bars are part of the 16x9 image and displaying them, resulting in a displayed image that is slightly too narrow horizontally.  Analogue DVD players are fine and discard the borders, as they should.

I'm just looking for a way to have an image that displays correctly on digital players without screwing up analogue players.  Directly encoding 704x480 seems like it might work, but am I overlooking something?

And let's pretend I just don't care about overscan on the output device.  We're only talking about nominal analogue blanking, that's it.

An alternate take is: since the future can be assumed to be increasingly digital, and analogue displays often have aspect ratio issues in their own right, why not just pretend 720x480 is 16x9?  Digital systems will do this anyway, and analogue systems may end up with a stretched, cropped image, but with their aspect ratio issues and overscan, they may have ended up stretched and cropped regardless of how correct the source image was.

Or, "How I Learned to Stop Worrying And Love Doing It Wrong Just Like Hollywood"...

16:9 anamorphic 704x480 is supported by the DVD spec. Many websites incorrectly claim that it's not.

Right now it's the most "real world"-compatible way to encode anamorphic content with MPEG2. I can't think of any downside other than misinformed people PMing you to (incorrectly) claim that "you did it the wrong way".

Some non-pro encoders screw this up, but HCenc handles it just fine. Which MPEG2 encoder are you using?

JEDIT: I forgot to mention that the official PAR is 40:33 for both 704x480 and 720x480 MPEG2 streams. All DVD-compliant players are supposed to crop 8px from the left and right of a 720x480 stream prior to stretching, but many don't, as you've noticed. :-/

Mostly I was concerned about 704x480, from the point-of-view of compatibility. Yeah, it's a valid part of the spec, and I use HCEnc so creating a valid file isn't the problem, but do players actually do the right thing with it?  I've seen a number of reports about hardware players that make me a little nervous about using this lesser-known DVD resolution.  Also, I'm a little fuzzy on the benefit of nominal analogue blanking--is the lack of blanking on this resolution an issue?  But mostly--why do the places that actually bother to have the correct aspect ratio always seem to opt for 720x480 with blanking?  That is the question that really makes me concerned about the player compatibility of 704x480 video.

As for 720x480, so many players already get the PAR wrong, and it doesn't seem like any newly-created players are doing it right--I'm seriously wondering at this point if I should just forget the specs and encode at 720x480, just anticipating that the player will get the PAR wrong and by virtue of two wrongs making a right, the image will look correct.  That is, apparently, the path most major studios are taking.

Good point(s), CatBus. I'll have to ponder this further.

It might be worthwhile for us to come up with a quick test pattern.

How's the following?

720: {red 8px} {green 8px} {blue 688px} {green 8px} {red 8px}

704: {red 8px} {green 8px} {blue 672px} {green 8px} {red 8px}

Well, I'm asking about this because I actually don't know the capabilities of the target device(s) for testing. Finding out what one or two devices can do is interesting, but it doesn't necessarily tell you what the lowest common denominator is.

As for test patterns, those look good to me (for NTSC naturally).  I'd assume a "correct" device would show just green and blue on the top pattern and all three colors on the bottom pattern.  But I suspect most devices would show all three colors on the top pattern and from what I've read, it's possible they'd even add black padding to the bottom pattern!

CatBus said:

Finding out what one or two devices can do is interesting, but it doesn't necessarily tell you what the lowest common denominator is.

True, true.

Ugh, DVD is such a crummy format! >:-{