DVD2SVCD V1.1.0 B1: Has Automatic Field Dominance Changed for CCE?

[GreenDrazi]

DVD2SVCD:

Has the automatic detection for CCE Field Dominance changed in v1.1.0 B1 from 1.0.9 B3 or resorted back to pre-109B3 status?
I'm getting two different field orders for the same vob depending upon the version used.
v1.1.0 B1 I get a yes
v1.0.9 B3 I get a no (which gives me a better encode).

From your “readme.txt” for 109B3
“- Bugfix: Fixed the CCE Field Dominance bug according to this post on the forum:http://forum.doom9.org/showthread.p...3755#post133755"


Thanks again for a great program.

[Deltis]

I've got the same problem -> http://forum.doom9.org/showthread.php?s=&threadid=37689 :(

[ux-3]

@greendrazi - so there is a problem with the field order? :D
Oh well, another deja-vu...

[Deltis]

Hi!

I did some more testing with interlaced PAL-DVDs (TV-series)...I even used the ShowFrameNumber()-command of avisynth to ensure the correct interleaving and combing of fields.

The results:

1. When I use DVD2SVCD 1.1.0 b1, the field order seems to be swapped:

-> When I use SeparateFields() in combination with the encoded MPV-file and have a look at it with VirtualDub, it shows the typical "one-step-back -> two-step-forward -> one-step-back -> etc." motion.

-> Using ReStream to deactivate the TFF-flag solves the problem of shaky motion but does not explain the reason for it (as the source is TFF and DVD2SVCD 1.0.9 b3 produces working TFF, too).

-> The fields in the MPV-file are interleaved and combed correctly - no field missing(the Frame-Numbers show no difference to the original AVS-file). Therefore it may only be an error within the stream and/or the TFF-flag :rolleyes:

2. DVD2SVCD 1.0.9 b3 does everything as it should be!

That's pretty much all I can find out (and there may be even mistakes in my conclusion as I have no deeper knowledge of MPEG-stream-internals)

I hope there will be a solution soon :D
cu
deltis

PS: The README.TXT states:

- Enhancement: Added "Keep Interlaced" to the list of deinterlacers. It overrides the Progressive, Zigzag and aspect ratio settings (progressive and zigzag off and aspect ratio to 4:3 (no borders)). If you can live with these settings you will get the best quality (when watched on the TV, not on the PC).

In my tests I observed that DVD2SVCD 1.1.0 b1 (in case of an activated "Keep interlaced") just changes Progressive to Interlaced but does not change ZigZag to Alternate with CCE 2.50 . A bug or a feature? :(

[dvd2svcd]

Originally posted by ux-3
@greendrazi - so there is a problem with the field order? :D
Oh well, another deja-vu... I don't care for your sarcastic tone, this is not the first time you're being sarcastic on this issue. Stop that, I wont tolerate it any more. Furthermore, I have both PM'ed you and mailed you if you maybe could supply me with some interlaced material and in both occasions you have chosen to ignore my requests completely. So IMO you of all has no right in being sarcastic.

End of rant.

[dvd2svcd]

Originally posted by Deltis
In my tests I observed that DVD2SVCD 1.1.0 b1 (in case of an activated "Keep interlaced") just changes Progressive to Interlaced but does not change ZigZag to Alternate with CCE 2.50 . A bug or a feature? :( As I have said before, when selecting "Keep Interlaced" I only change and use the values internally (I don't change the values in the gui).

[GreenDrazi]

DVD2SVCD

v1.1.0 B1
"- Enhancement: Added "Keep Interlaced" to the list of deinterlacers. It overrides the Progressive, Zigzag and aspect ratio settings (progressive and zigzag off and aspect ratio to 4:3 (no borders)). If you can live with these settings you will get the best quality (when watched on the TV, not on the PC)."

I just tried it with the "Keep Interlaced" option on the same material (field order set to automatic) and it still gave me "UpperFieldFirst=Yes" in the d2s log. So this your intended result? I must now set the Field Order manually to "No" for certain material?

I'm not qualified to discourse on field order in CCE, but either method ("Keep Interlaced" or "No Deinterlacing"), I think I'm getting different results for field order than in v109B3 (I've turned off Progressive, Zigzag in all methods and both versions):

v1.0.9 B3
“- Bugfix: Fixed the CCE Field Dominance bug according to this post on the forum:http://forum.doom9.org/showthread.p...3755#post133755"

Please advise.

Thanks

[Deltis]

@dvd2svcd:

Wasn't talking about GUI-settings: When I selected "Keep interlaced" the created(!) MPEG-file shows (according to ReStream and BitrateViewer) a ZigZag-scan-type...

[dvd2svcd]

@GreenDrazi: Most likely I have introduced a bug here. Even though UX-3 insist that the field dominance bug is in 1.0.9 B 3 too. I will reverse back to that. If I had had some test material this wouldn't have happen.

@Deltis: Strange, I was sure that it's disabled. Anyway I'll check up on it.

[Deltis]

Perhaps you may check out if "ZigZag" is turned back on when it is already disabled within the GUI? I'll make some more tests...

[ux-3]

@dvd2svcd:
Tried to send you a PM. Your Box is full. To bad - it took an hour of typing... Can't browse back, text is gone.

[Deltis]

@dvd2svcd:

I made another test with DVD2SVCD 1.1.0 b1. The settings were:

CCE (GUI):
- Linear: off
- ZigZag: off
- DVD compliant: off
- Progressive: off
- DCT: 8
- Field order: Automatic

Conversion (GUI):
- Keep interlaced
-> I used "Interlaced resizing" for the AVS-script

If you want I can send you the LOG-file...
Nevertheless this is the Bitrate Viewer-report of the encoded MPV-file:

Num. of picture read: 3330
Stream type: MPEG-2 MP@ML VBR
Resolution: 480*576
Aspect ratio: 4:3 Generic
Framerate: 25.00
Nom. bitrate: 2400000 Bit/Sec
VBV buffer size: 112
Constrained param. flag: No
Chroma format: 4:2:0
DCT precision: 8
Pic. structure: Frame
Field topfirst: Yes
DCT type: Field
Quantscale: Nonlinear
Scan type: ZigZag
Frame type: Interlaced
Notes:

This shows that ZigZag is still used for encoding...
And as reported before: The field-order doesn't seem to be correct within the created MPEG-stream (shaky motion on my DVD-player)

-> solution and more under http://forum.doom9.org/showthread.php?s=&threadid=37689

Hope you've got some ideas how to fix it (DVD2SVCD 1.0.9 b3 seems to handle everything just fine - even the weird CCE-logic) :D

Thanks for your help!
Deltis

[dvd2svcd]

About the zigxag just load the .ecl file in CCE manually and see what the zigzag setting is.

[Deltis]

@dvd2svcd:

This is a list of all questionable settings/menu-options within
DVD2SVCD 1.1.0 build 1 I noticed:

1. I had a look at the ECL-file generated by DVD2SVCD 1.1.0 b1 and
even with an enabled "Keep interlaced" - option it always
shows "ZigZag: ON" (minor bug - didn't notice degradation of
visible quality)

2. The options under "Encoder/Field order" and
the "Automatic"-option are not working correctly!

Explanation (as allready sent to you by PM):

The "Upper Field First"-switch in CCE doesn't activate the "Top-Field-First"(=TFF)-flag in the MPEG-stream

-> See http://forum.doom9.org/showthread.php?s=&postid=133755#post133755

What I've read about this topic:
CCE always sets the TFF-flag in the encoded MPEG-stream and to be able to encode a "Bottom-Field-First"-source properly it uses a trick and shifts the whole frame/both fields of the source one line up. As a consequence the first line of the frame now belongs to the original bottom-field and so the bottom-field will be treated as top-field (switching of fields). In combination with the TFF-flag the created MPEG-stream will play-back correctly. The last line of the source-frame seems to be replaced by a black line :eek:

Current behaviour of DVD2SVCD 1.1.0 b1:

1. Source is TFF:

a)-> User knows this and selects "Top field first (Field A)"
under "Encoder" (DVD2SVCD-GUI)
-> DVD2SVCD sets "Upper Field First" = "ON" in the CCE-ECL-file
-> IMHO "Upper Field First" should be "OFF"

b)-> User doesn't know this and selects "Automatic"
under "Encoder" (DVD2SVCD-GUI)
-> DVD2SVCD sets "Upper Field First" = "ON" in the CCE-ECL-file
-> IMHO "Upper Field First" should be "OFF"

Comment: The older DVD2SVCD 1.0.9 b3 sets "Upper Field First" = "OFF"
in "Automatic"-mode (which is correct)

2. Source is BFF:

a)-> User knows this and selects "Bottom field first (Field B)"
under "Encoder" (DVD2SVCD-GUI)
-> DVD2SVCD sets "Upper Field First" = "OFF" in the CCE-ECL-file
-> IMHO "Upper Field First" should be "ON"
OR you must(!) write a BFF-flag manually into the created
MPEG-stream (e.g. with ReStream) - I prefer the last method ;)

b)-> User doesn't know this and selects "Automatic"
under "Encoder" (DVD2SVCD-GUI)
-> DVD2SVCD sets "Upper Field First" = "OFF" in the CCE-ECL-file
-> IMHO "Upper Field First" should be "ON"
OR you must(!) write a BFF-flag manually into the created
MPEG-stream (e.g. with ReStream) - I prefer the last method ;)
Comment: The older DVD2SVCD 1.0.9 b3 sets "Upper Field First" = "ON"
in "Automatic"-mode (which is correct)

Important: I can only confirm the mentioned behaviour - the so called "CCE v2.5 field dominance switch bug" - for CCE 2.50 at the moment!

cu
Deltis

[dvd2svcd]

The autodetection is now as in 1.0.9 B3, so please download it here (http://www.dvd2svcd.org/D2S110B1a.zip) and try it out. Also read the readme (http://www.dvd2svcd.org/readme.txt) to see what else have been fixed (so far).

[Deltis]

@dvd2svcd:

Wow - that' s pretty quick :cool:

I made some more tests with build 1a - here are the results:

1. "ZigZag" is now disabled when "Keep interlaced" is activated.

2. I can confirm the intra-matrice-bug(?) of CCE 2.50 and/or of the CCE-patcher (checked with ReStream) - I guess CCE 2.50 is writing a standard intra-matrice into the stream even if it encodes the stream with a different (patched) intra-matrice (unverified till now - will need some more testing...)

3. "Automatic"-field-order works correctly like in DVD2SVCD 1.0.9 b3 - if I can assume that DVD2SVCD detects the field-order of the source right - perhaps you may want to add a line to the dvd2svcd.log like e.g. "Detected source type/field order is: Interlaced with top-field-first (TFF)"

BUT... :devil:

-> When you manually select "Top field first (Field A)" in DVD2SVCD it should DEACTIVATE "Upper frame first" in CCE
-> handled different in DVD2SVCD!

-> When you manually select "Bottom field first (Field B)" in DVD2SVCD it should ACTIVATE "Upper frame first" in CCE (OR DVD2SVCD DEACTIVATES "Upper frame first" in CCE AND it writes the BFF-flag into the encoded CCE-MPEG-stream with a stream-patcher like e.g. ReStream - which is the more complicated but better way -> no change of field order and no picture-shifting will be induced!)
-> handled different in DVD2SVCD!

In a short way:

1. If source is TFF -> Deactivate "Upper frame first" in CCE

2. If source is BFF -> Activate "Upper frame first" in CCE (or use complicated method above)

3. If source is Progressive (no matter what field-order) -> Deactivate "Upper frame first" in CCE! An adjustment of the field-order-flag within the created CCE-MPEG-file in order to match the field-order of the source is optional but recommended.

"Automatic" should detect the source field-order/type and set the values as mentioned above (Point 1.-3.; seems to be allright in build 1a)

Hope it helps you to handle the field-configuration of CCE. The strange and unlogic behaviour of the "Upper field first"-switch is still existent in CCE 2.64!

This is a comment of another post related to this subject:

“Upper field first” in the CCE options is a misnomer, and would be better described as “botch this encode so that LOWER-field-first input comes out looking OK even though it is still actually encoded as upper-field-first”

cu
Deltis

[dvd2svcd]

That's your point of view, my point of view is opposite, because if I do what you suggest dvd2svcd will be illogical to users who actually knows what to do with CCE. I mean, to me it is illogical that if I select TFF in dvd2svcd and then look in CCE and see that it is set to BFF and vice versa and that even if CCE has a bug or not. I'm damned if I do and I'm damned if I don't so as it is today is the way it's gonna be for the time being. No matter what, the TFF <-> BFF issues are rare phenomenons and not many is bugged by it.

Automatic = If d2s detect TFF, Upper is deselected in CCE and vice versa
Manually Top Field First = Upper is selected in CCE
Manually Bottom Field First = Upper is deselected in CCE

[Deltis]

@dvd2svcd:

I understand your point of view. As long as "Automatic" works for unexperienced users and as long as I know what DVD2SVCD really does when I activate the other switches I feel good ;)

Keep up the good work and thanks for listening!
cu
Deltis

PS: Perhaps you may copy this into your Readme.txt to document it?

Automatic = If d2s detect TFF, Upper is deselected in CCE and vice versa
Manually Top Field First = Upper is selected in CCE
Manually Bottom Field First = Upper is deselected in CCE

[dvd2svcd]

Originally posted by Deltis
PS: Perhaps you may copy this into your Readme.txt to document it? Good idea, but it doesn't belong in the readme file. I think I'll ask MrBass to put it in his Advanced Reference. (The readme file is more or less only a change log.)

[Deltis]

OK!

BTW:

I used "little devil" :devil: as a symbol that "the devil is inside the little details" like the listed behaviour of CCE - even if it is expensive and fast as hell :D

Don't take the term "not fixed" as a command to fix it. Just wanted to discuss this...I edited my recent mail accordingly.

cu
Deltis

[dvd2svcd]

Well, maybe I'm just too anal :D (The frase or word that bugged me was "Still not fixed" as in "you promised to fix this and you still haven't done so"). No matter what I'll let the matter rest as we've come to an agreement on how to interpret the TFF/BFF settings in d2s.

[GreenDrazi]

Thanks DVD2SVCD and Deltis for the discourse.

Only after reading through this thread for the third time, do I actually have some idea of what you are talking about!
:D

Maybe it's time I made a donation.