According to me the file you sent before is correct. If you tell my that this one gave you a cropped image on your TV, then I will author a DVD with it and see if I have the same. Else provide me an other file that you want me to test :).

I've tried so many combinations 8O . Let me reencode an original capture HDTV section as:

FitCD nonanamorphic in nonanamorphic out (I'll post FitCD screen) with TMPGEnc set to 4:3. This should be most "standard" settings.

OK?

For reference, a segment of the original HDTV OTA capture is at:

http://www.yousendit.com/download/2guqymqfUTk%3D

Phil,

A seven second encoded m2v and the ac3 are at:

http://download.yousendit.com/79877F2903F9736F
http://download.yousendit.com/0825E8B64921F6D9

http://www.digitalfaq.com/archives/i.../2006/11/1.jpg
http://www.digitalfaq.com/archives/error.gif

Also, my DVD player doesn't seem to like 7sec clips, so I authored in DVDLabPro as 2 movies with 2 menu items. Same clip in both. That plays.

How much overscan should I get playing this on a 4:3 TV?

Be carreful that this trick can introduce problems (animated menu are managed differently by the player than regular movie tracks).

Generally 2 ( = 16 pixels) is a value that fit to all screens.

Edit: will try your files this evening, at home.

Thanks, Phil :) I meant that I used the same 7sec m2v/ac3 to in 2 DVDLabPro movies. The menu was simply 2 text objects linked to the movies. No animation. Only did this because my player doesn't somehow like single 7sec movies.

So I tried your test (that is actually 3 second long, not 7 :-)).
The picture is cut few pixels at the right of the HD logo. That represents 30 pixels, more or less, whatever the player or the screen I use (my regular 4:3 TV set of my HDTV rear-projector).

I know that my system had a overscan between 2 and 3 (16 - 24 pixels), so 30 is weirdly big. But I don't find any reason for that.

Anyway I suggest you to encode with a resized (not overlayed) overscan of 2 (16 pixels) and see if this is better on your system.

Thanks Phil :D

I had thought of overscan, that's why I asked about the number of pixels.

I'll encode tonight with overscan 2 and see how it turns out.

Again, thanks for going to all the trouble! I really appreciate it.

No problem.
Just to know : what are you using to capture the original source ? Did you see that the file uses a 30 pictures long GOP and weird matrixes (Intra is the original MPEG2 matrix but Non-intra is all composed of the value 16).

dvr-ms captured with ATI OTA HDTV Wonder with MCE. I think VideoReDO has options while processing mpeg output. Any suggestions I can play with?

I would use a shorter GOP, 30 means a lot of B frames so a quality lesser than a standard DVD stream (length = 15 or 18).

For the matrix I don't know the effect of a non-intra matrix composed of only "16,16,16,16..." so I can't say.

Success
 

.

And the answer is .......................you are right, as usual! :D

I changed GOP to 18 (NTSC DVD) in the VideoReDo output mpeg options and created another source mpeg.

Made two test encodes, one with overscan2, another with overscan3.

Overscan2: Saw entire logo and and additional width of 1/2 the "D" in "HD".

Overscan3: Saw some more of the picture.

I'll opt for overscan2, as some TV's may not crop the same amount of pixels as mine and I don't know the effect on a HDTV.

I encoded as fitcd anamorphic out with tmpgenc set to 16:9, changed fps to film.

I am going to ask on the VideoReDo forum concerning matrices. I don't know how they are converting from dvr-ms to mpeg. I do like VideoReDo.

Overscan2

dgdecode_mpeg2source("E:\lo\Lost_04_10.d2v",idct=7 )
selecteven()
Telecide(order=1, guide=1, gthresh=10, post=2, vthresh=50, dthresh=7, blend=false, show=false)
Decimate(cycle=5, mode=0, threshold=0, threshold2=3, quality=2, show=false)
LanczosResize(656,448,0,0,1280,720)
RemoveGrain(mode=1).TemporalSoften(2,3,3,8,2)
AddBorders(24,16,24,16)
ConverttoRGB24()

One more question.
 

Phil,

Having resolved the overscan thing, need an opinion. (Am still working with the Lost HDTV captures.)

I want to put four 42 minute episodes on a DVD. I can encode at CQ 90% as above, author in DvdLabPro (with Mp2 audio) and create a DVD that shows 91% in DVDShrink to fit to DVD.

In your opinion, would it be better to encode at a 85 CQ which does not require DVDShrink (fits DVD)?

I know that some time ago, there were discussions on the merits of processing with DVDShrink 2 times, etc. I want to avoid further testing and burn these things. 8O

Re: One more question.
 

I myself encode at 85% insteed of 90 half of the time, the loss in quality is not so big. Avoid using DVDShrink if you can.

Thanks, turned out CQ 85 with AC3 exactly filled DVD without DVDShrink. :D