Correct method of AVI to DVD conversion?

[monzie]

Ok just had a run in in with someone on these forums (got called an *******...ooooh.....been called a lot worse tho'..and I'm pretty thick skinned).

But thats not the issue (I aint going to flame a no-one...we all have opinions of our own).... what I want to know tho' is what IS THE CORRECT METHOD of calculating DVD/mpeg2 encodings FROM AN .AVI?

Anyway from what I have learnt (from this site and doom9 etc) when encoding .avi to mpeg2 the correct way is to:

1) find the avi's ratio (h / w)..for example 624/336 = 1.857.

2) calculate the vertical res of the encoding from the 'calculated' DAR playback of the mpeg2 flag used by using the avi's AR and then subtract the answer from the mpeg2's 'true' vertical (either 480 or 576 divided by 3 and X 4, for 4:3, or divided by 9 and X 16, for 16:9) and divide by 2 to find the T & B borders..

...for example if using 480 (ntsc) DAR 4:3 gives a playback (as calculated) 640 horizonatal..so 640/1.85 gives 345.9 (call it either 336 or 352 going to the nearest number divisable by 16...or an even number if not fussy).

The above is the way I CALCULATE things (it is not to say that I am right)

His way goes by the way of mpeg2 DVD 'true' horizontal resolutions (totally disregards mpeg2 DAR's) and then goes on about TV vrs PC pixel ratios (but makes no compensation for them tho'?) and 4:3 encodings for 4:3 TV's..but I cannot follow his train of thought....am i missing something?

Heres the thread:
https://www.videohelp.com/forum/viewtopic.php?t=234102

Anyone?

[trevlac]

Want to discuss here? I promise ... no a$$hole comments.

[jimmalenko]

I happily let TMPGEnc do it all for me. No need to do all these calculations.

[trevlac]

Help me out here. I really am having a hard time following what you do.

1st -- It looks like you don't care about DVD/SVCD/etc spec sizes ? Also, you are viewing on a PC? Or Does the encoder take care of making a 720x480 type thing for you?

Anyway from what I have learnt (from this site and doom9 etc) when encoding .avi to mpeg2 the correct way is to:

1) find the avi's ratio (h / w)..for example 624/336 = 1.857.

OK avi is has a h/w ratio of 1.857

2) calculate the vertical res of the encoding from the 'calculated' DAR playback of the mpeg2 flag used by using the avi's AR and then subtract the answer from the mpeg2's 'true' vertical (either 480 or 576 divided by 3 and X 4, for 4:3, or divided by 9 and X 16, for 16:9) and divide by 2 to find the T & B borders..

You really lost me here. Vertical res is up and down right? Can you do the steps you use in an example ie step1, step2 ... with intermediate results?

...for example if using 480 (ntsc) DAR 4:3 gives a playback (as calculated) 640 horizonatal..so 640/1.85 gives 345.9 (call it either 336 or 352 going to the nearest number divisable by 16...or an even number if not fussy).

So you end up with 640x336 or 640x352 ?


I do ignore DAR because I see no point in encoding anamorphic unless you have a 16x9 TV. Either the TV resizes or you do. I suppose there is a reasonable argument that you are not wasteing encode bits on the black border ... but that is another topic.

What i just don't get is how you play it on a TV ? I am shooting for DVD spec. Should I go thru my thinking? I want it to be 720(or 704)x480 before I encode.

-----------
I just wanted to say that I found this to be a really nice post by you. We had a run in, but you are still doing some critical thinking on your stuff. I do truely think I know what I am doing ... but I just didn't follow what you are doing. I may be off. I am just a shmoe who plays with video.

However: I have been doing this for some time. I helped author the most recent version of the doom9 capture guide. Doing that, I spoke a great deal about this topic to some very experienced people. I do have a bias on this ... I mainly do dvd for TV. I have also done a lot of work on color correction. I wrote a Vectorscope/Waveform monitor plugin for virtualdub. I have written a few avisynth filters and played high quality resize methods.

I am not saying this to impress. You shouldn't really care because the people who work on this kind of stuff are often wrong about many things. I just wanted to point out that I truely believe I know what I am talking about and I am trying to be helpful to people.

Olive branch ....

[FOO]

Hey how do you Vectorscope a video ?
is it just for one pixel ?

[trevlac]

FOO,

http://trevlac.us/colorCorrection/colorTools.html

There is a sight I started but never finished. It was really a way to doc my learning about color correction. The plugin link does have the vectorscope included, even though I did not add a how to on the site.


A vectorscope takes the chroma signal, seperates out U&V and then pretty much maps the 2 values on a polar type display. Moving around the circle display tells the color. Distance (vector) from the center tells the saturation.

All of the pixels in the frame a mapped at once. For an analog scope there is a trace line. I did not 'fake' that with digital because I don't see how it adds info. The waveform monitor works that way too.

I wrote some stuff for using this in the doom9 guide. Good way to set the proper black/white levels ... and to visually change the saturation (without looking at a TV screen).

I also wrote an avisynth plugin (actually added to some work by Donald Graft) .... this plugin TweakColors lets you target specific 'slices' (colors and saturation levels) of a vector scope display and change the color. Means you could target a green shirt in a frame and make it red without changing other colors in the frame. Color correction for pro tools like Avid have this sort of thing (and more).

Sorry for the ramble .... but I like this topic more than resizing.

[monzie]

Trevlac

a)first off i want to apologize for our first run in....more crossed wires than anything.

and

b) for telling you to eff off in another thread...(to much beer and it was late here, well gone midnight..so read pissed and tired).

Right, now where we? OK I've being listening to what you said/posted/links and some of its starting to make sense in my thick British skin. I now grasp your graphic/pc pixels vrs TV/DVD pixels and its starting to all make sense (at bloody last!)....I was half way there but couldnt 'grasp' the final conversion ratios (ie.. why or how)..until you mentioned the 11:10 TV pixel size......but thats easy to understand as 'my way' gives pretty much the same answer...plus I'm on PAL so getting my head round two formats at once aint exactly easy....11:10 and 54:59 from square avi...plus 16:9DAR for typical UK widescreens, oh and i now know that 'padding' is not a knitting term.........

[trevlac]

Hey monzie,

Thanks for the note. I like people with strong opinions. A loudmouth Croat is one of my more fun online mates.

[DJRumpy]

Without reading the entire thread, I'll throw these out. I see inconsistancies in both arguments, as well as alot of valid information.

I thought the aspect ratio has to be 2.35 to be anamorphic. The ratio on this AVI is 1.85. I try doing a fw encodes with the settings you guys gave me. Thanks for the help, this stuff can get pretty confusing.

Any aspect ratio wider than 1.78 (i.e. 2.35:1, 2.20:1, etc) can be considered anamorphic for all intense and purposes. 1.78 aspects should be considered as simply widescreen. As to a guide, I've written one in length that explains the math involved, but it's not for the squemish.
https://www.videohelp.com/forum/viewtopic.php?p=622650#622650

You first determine your AVI's aspect ratio, which you did (1.85). This assumes of course that your source is NTSC, and not PAL. ( 624 / 336 = 1.85 ) (rounded down). The 1.85 aspect is typically encoded using a 16:9 DAR, not a 4:3 DAR, although you could squeeze it into a 4:3 DAR if you wanted/needed to.

You then determine what output format your going to. You didn't mention if this was for DVD, SVCD, or CVD, just MPEG-2.

I wouldn't worry too much about padding. trevlac is correct in his statement regarding the true NTSC width, and the padding needed to maintain a proper pixel aspect ratio per the CCIR-601 specification. However, you wouldn't notice 16 pixels of padding on a television screen, or the ever so slight change in aspect as a result. In either case, your final output, if it's for DVD, and assuming your using full D1 resolutions, would need to be 720, either with, or without padding. I sure they both know this, but are misunderstaning each other when it comes to the symantecs. Don't sweat the small change

If your resizing your AVI to DVD, simply resize it to 720x460, and add letterboxing to fill out the vertical to 480 (10 pixels on top, and 10 pixels on bottom). If you want to include padding, then pad the width with 8 pixels on the left and the right (720-8-8=704). Set the DAR (Display Aspect Ratio) on the mpeg to 16:9. This will give you a display that looks fine when played on both 16:9 and 4:3 televisions.

If your going to CVD/SVCD, then you need will need to squeeze your video into a 4:3 DAR (Display Aspect Ratio) using the corresponding 4:3 math. You would resize your video to 352x348 (CVD), or 480x348 (SVCD), and add letterboxing to flush the 348 back out to 480.

[monzie]

OK, thanx to everyone whoes responded in this thread and forthose that have you may find these pages very INTERESTING reading if you havent come across them before. Please give them a read as there is some excellent information in them.....especially the tables.

http://www.mir.com/DMG/aspect.html

http://www.uwasa.fi/~f76998/video/conversion/

Basically what the pages above are expaining are the differnces between square pixels formats and mpeg pixels (both NTSC and PAL) AND the playback on a TV ....which is not pixel based at all (IE in the usual digital/pc based idea of pixels)...its merely the time/signal the scan line uses to track across the TV's horizontal from a digital (ie pixel based) based source to give a TV picture (well thats how I understand it).

The links also fully explain the 'odd' 711 X 486 NTSC video sizes vrs 704 X 480 NTSC, overscanning, padding and cropping, NTSC>PAL>NTSC conversions (using corrected pixel ratios), various PIXEL ratios (square, true square, mpeg1&2>NTSC and PAL) etc....

Hope someone (other than me) finds them useful.