Newer TNG Seasons no longer interlaced? [Archive]

GreenDrazi

3rd November 2002, 01:44

I believe they are all still interlaced NTSC 29.97 (region 1 version anyways). But after I encoded all of Season 2 with de-interlacing which I was never happy with, I switched to Inverse Telecine starting with season 3 and have been much much happier with the results. I’m currently doing season 1 with IVTC and the results are very good also. I may go back and re-do season 2 (uggh!). The only down side to this method is the “space” shots - you will still get a little stuttering, but 90% of the rest of each episode will look much better and with far less artifacts.

But this is IMHO the best method for most interlaced NTSC 29.97 material originally intended for TV broadcast. Give it a try, but do an encode of an entire disc, not just a chapter to make sure you see the final or true bitrate.

Now, can someone tell me why the chapters on the original DVD’s are often more than a little off??

I use the following:
VIDEO INFO:
- Encoder: Cinema Craft Encoder 2.50.1.0
- Encoding mode: Multipass VBR
- Passes: 5
- Max. bitrate: 2530
- Avg. bitrate: 2159
- Min. bitrate: 300
- Image Quality Prio: 24
- Antinoise filter: Yes, 2
- Linear quant. scale: No
- Zigzag scan order: No
- Upper field first: No
- Progressive frames: No
- iDCT Algorithm: 32-bit SSE MMX
- Resize method: BicubicResize b value: 0 c value: 0
- Temporal smoother: Yes, Strength: 2 Radius: 1
- Sharpen: No
- NTSC Field Operation: Inverse Telecine, using Donald Grafts Decomb dll
- Deinterlace: None
- Pulldown: Yes
- Width: 480
- Height: 480

JoeB

3rd November 2002, 08:20

WOW!

Great reply! OK, here's what I did for the first 2 seasons, and now starting on the 3rd:

- NTSC field operation - AUTO
- Bitrate: MAX 2530, MIN 600, MIN AVG 1850, MAX AVG 2230
- CCE with 4 pass VBR
- Image quality priority 17
- Under CCE (2.50) advanced settings: all checkmarks removed
- field order automatic

The only thing I found about DVD2SVCD is that it doesn't know how to rip the chapter properly. For that, I use another tool, merge it to the config file, and remux it.

works perfectly.. with my own title screen and all!

The only problem I noticed is a small skip in the space scenes.

Also, if you try to SVCD the extra stuff (like behind the scenes and that kind of stuff) it looks REALLY BAD. Only way I got that to work is to switch to CVD, but I did notice quality loss.

Good luck!

JoeB

ux-3

3rd November 2002, 09:48

Sorry Folks, but I think a search would have spared you a lot of frustration perhaps. A while ago, we came to the conclusion that real good results can be obtained by keeping the stuff interlaced. While I can only judge this from my PAL version Region 2, the method should yield error free results regardless of PAL/NTSC. By keeping the material interlaced or messing around with it otherwise, you will only sufer deterioration due to reduced bitrate. Space scenes and such will remain as smooth as the original, when you only reduce the bitrate! They just look a little grainier.

Perhaps the NTSC versions can be properly converted into progressive frame, but PAL cannot. But since you guys don't sound perfectly happy, you should try to keep the material of Season 1 and 2 interlaced alltogether. Just notice that the Frame Automatic is still buggy. I found that the PAL versions of Season 1 and 2 do fine with Setting B.

DVD2SVCD did all chapters correctly, so you must do something wrong or use a wrong version or the problem is with RegCode1. If I am right about season 3, you have no need to turn progessive or zigzag off. Just do and check a progressive encode. Worked fine on TNG 149 PAL.

After rereading Greendrzi, I think you should try without that NTSC Telecine stuff! I have no clue about NTSC, but you should just keep the episode _as_is_ with reduced bitrate. Then everything should be smooth, as it is here.

Also,I would NEVER use temporal smoother on a 5 pass!

Similarly, a bicubic with both parameters to zero makes little sense to me. You are blurring the picture that way, arent't you?

Dunno about anti-noise, I didn't use it.

To repeat my first post: It seems that all this interlaced keeping treatment is no longer needed in season 3, and will perhaps even introduce more flicker and less bitrate use. (I have only tried 149 though)

As a final note, I have increased the max bitrate to 3500, since my players all support it. This will dramatically improve the quality of some scenes. The ferengi first encounter with the lightning scene on the planet is a good example. Scenes with rapidly changing illumination will profit much from this.

ux-3

3rd November 2002, 15:32

I just had a close look at my second encode. I would like to remind people of the field order bug! In 109B3, the auto setting will aparently still mess things up when doing TNG episodes.

I was doing encodes of TNG149 in progressive frame, and had set field order to auto. The initial starship bypass and the following opening door to the concert room were flickering, which I interpret as a field order problem. In fact, during the whole episode, some of this flickering was to be observed, quite apparent at scene changes.

I repeated the encode with field order set to B. The initial scenes are now fine. Don't have time to look at all scenes. No flicker so far.

Still I would like some feedback about Season 3 not being interlaced anymore. To tell the truth, I am surprised that field order is still important, when something is in progressive frame. Doesn't quite fit my understanding. Perhaps someone can shed some light on this?

GreenDrazi

3rd November 2002, 16:13

ux-3
If you just want to know if the material is interlaced or not, why not just open it up in DVD2AVI and preview it (F5) - it will tell you if it is or not. I’m certain that all the Region 1 discs are.

Field order - I thought that this was fixed in the latest full release (1.0.9b3) to detect correctly for CCE. It probably wasn’t working correctly when I did season 2 on an earlier version of DVD2SVCD. Can you post a thread which says the current version is still not detecting correctly?

If I was using an Max Avg bitrate of 3500, then I could easily drop a lot of the filters that I’m using, but I think it’s more than reasonable to try and get a single 45+ minute episode on a single 700mb cd-r. It’s not the maximum setting that really determines the average bitrate for these STNG discs - it’s the Max Avg and the size of your cd-r in the bitrate tab which comes in around 2100 for a 700mb cd-r. If your going to do the entire 7 year series at a 3500 bitrate, then you’re going to have way too many cd-r’s for my taste. And so, to get 45 minutes of re-sized, full framed, interlaced NTSC 29.97 material (which is what the STNG Region 1 DVD’s are) on a single 700mb cd-r, you've got to use some filters. With a R1 disc, a person could try a CVD, but I’ve never been as happy with the results as I get with the method described before. If you’re Region 2, then I have no experience if IVTC helps - I doubt that it does.

As for your comments on my settings, I posted what is found in the SVCDInfo.txt file that you get when you create a VCDXBuild cd image. The bicubic resize settings are actually b = 0 and c=.75 in DVD2SVCD. Using anti-noise and temporal smoother on a 5 pass is a must for me, especially for the STNG series.

As for the Chapters issue, DVD2SVCD is typically reading what is on the DVD correctly. But it’s the DVD’s that are messed up. The chapter selection should put you at a commercial/advertisement break point, but on a lot of them that’s not the case. For example, there should always be a break point at the very end of the opening title/credits which is just before STNG lists the title of the episode and many of the DVD’s are that way. But, on some, you are placed a good 30 seconds or more later into the episode in the middle of some action. As I always skip past the title using the chapters while watching STNG, I have to rewind to get back to the real break point which is really silly. You will also find on these same discs, later break points at places which don’t appear to be the commercial breaks. This to me is very odd. I first thought that they may be just using a standard time template, but that’s not the case as each episode’s actual entry point times are different in the ifo file.

ux-3

3rd November 2002, 19:04

Originally posted by GreenDrazi
If you just want to know if the material is interlaced or not, why not just open it up in DVD2AVI and preview it (F5) - it will tell you if it is or not. I’m certain that all the Region 1 discs are.

What makes you certain? Did you actually run a preview and looked for the combing interlace effects? If DVD2AVI tells you its interlaced, that means little. I have had plenty of DVDs where this was the case, but it wasn't really interlaced. I think it only reflects the flag of progressive frame.
And yes, of course I checked it, but the answer is non-conclusive: Interlaced. This it says on many modern Movies too, which are definatly not interlaced. Critical viewing gave me no clue as to assume true interlacing.

Field order - I thought that this was fixed in the latest full release (1.0.9b3) to detect correctly for CCE. It probably wasn’t working correctly when I did season 2 on an earlier version of DVD2SVCD. Can you post a thread which says the current version is still not detecting correctly?

Just do a search! I recall that we discussed it in the debug thread for the next release. My results tell me that auto is wrong too. I set it on B manually.

If I was using an Max Avg bitrate of 3500, then I could easily drop a lot of the filters that I’m using, but I think it’s more than reasonable to try and get a single 45+ minute episode on a single 700mb cd-r.

Please read carefully what I had written before. I use a MAX bitrate of 3500 to fit the material onto a single CDR80 or CDR90. You use a max bitrate of 2530. We both are using the same average bitrate of about 2100. And I don't need filters.
I'll try to explain it. You are running a 5-pass but you have hardly anything to distribute bitwise! If your average bitrate was nearer to 2500, then a 5-pass and CBR wouldn't really make a difference at all. In fact, if you would use CDR99, you would get to that max bitrate of yours (2530) and could do it all in CBR, without any change in quality, because every scene would get full bitrate, no matter if you do it once or 5 times.

My encodes have a max of 3500. If a scene in space or in action sequences gets complicated, I have 40% more Bits to throw at it than you have. In many of the stationary scenes, there is little change, so the bitrate will be high at the scene change, but then drops real low cause only a mouth changes. My bitrate can go low, because I can move the bits elsewere with that 3500 ceiling. You cannot move much. You can only pile 400 bits over the average, where I can shift 1400. You would improve the quality far more by raising the max bitrate to 3500 than by using cdr90 with the same settings, which would be gaining average but not peak performance. Since many scenes have little action and stable background, that high average bitrate is pretty well wasted. In fact, with a peak bitrate of 4000, I have encoded this stuff in DVD resolution, and I spend an evening deciding which was better, DVDres or SVCDres. This is why I feel fine without filters. And I have to handle a higher PAL resolution as well...
But my spaceships have no ignition problems. Their warp drive doesn't stutter... They fly smoothly.

The bicubic resize settings are actually b = 0 and c=.75 in DVD2SVCD. Using anti-noise and temporal smoother on a 5 pass is a must for me, especially for the STNG series.

Well, please read YOUR writing. That setting is not what you posted before. I think you are doing overkill on the material.

About your issues with the chapters:
First, commercials are inserted at different points in different countries. Second, they are inserted from the point of view of the advertisers, right at the hight of tension. Chapters should be, where a shift of focus occurs. I find them well placed. However, as long as DVD2SVCD places them, where they are on the DVD, we should both be happy.

I would suggest to you to grab a CDRW and try some modifications. Test your DVDplayers max safe bitrate and run a test encode.
I am not knowledgable about NTSC particulars, but treat the material as you would treat a perfect modern movie, which is progressive. Only difference: Turn off Progressive, Zigzag and Deinterlacing. (all three options off). You already have the field order on B, as I interpret your script (Upper field first: No). This will keep the mpg Frame structure as it is, but only reduce the bitrate and the resolution. Don't use filters at first, as they only cost time.
Keep all this NTSC stuff out of the game unless it is also used on progframe movies. (this may mean dropping Field operation and pulldown)

Please come back with your results.

And give that Season 3 another look. I am almost certain that 149 isn't interlaced!

ux-3

3rd November 2002, 23:40

As an add on to the interlace question for season 3 PAL, I have taken an hour+ to look at the preview. I did encounter a few space shots, where a comb effect could be seen. Engaging Telecide did make the picture lose the comb. In all non-space shots, there was no comb, and telecide did not alter the picture at all. Funily, the new episode dedicated space pics were fine too. I get the impression, that some old fly-bys have been used, and these were treated differently perhaps.

If this is so, what would be the angle of approach? Of course, keeping it interlaced would be an option, but for just a few seconds of material? When this new season is mainly progframe, perhaps deinterlacing it with telecine will not change the progframe stuff, but help those few seconds?

GreenDrazi

4th November 2002, 00:53

ux-3
Look, I’m trying to help you out and others with STNG.
Well now, let’s see - in your last post, you are finally trying something that I recommended in my first reply. Hhmm.

As to your previous post:

“ Don't use filters at first, as they only cost time. Keep all this NTSC stuff out of the game unless it is also used on progframe movies.“

It’s kind of hard for those of us in Region 1 to “Keep all this NTSC stuff out of the game”.
That would be like me telling people in PAL regions to get rid of all that PAL stuff. I’m trying to help people out here (particularly Region 1 users and hopefully PAL users can benefit too) and comments like that just confuse the issue. And when you are ready to smooth out that banding your getting, try temporal smoother. Yes, it takes more time (not an issue for me) and it’s not the “end all” of filters, but it does help - even after a 5 pass VBR.

“Just do a search! I recall that we discussed it in the debug thread for the next release. My results tell me that auto is wrong too. I set it on B manually. “

As I stated in my follow up post, I posted the SVCDInfo.txt file, which DVD2SVCD writes when you create a VCDXBuild cd image. My DVD2SVCD field order is set to automatic and so it has correctly identified the material. Because my usage of DVD2SVCD 109B3 is working correctly, I asked you to do a search to prove that it’s not working correctly. I have the entire third and fourth season of STNG encoded SVCD’s (and they are beautiful) with the correct field order using DVD2SVCD 109B3 with an automatic setting. Perhaps it doesn’t work correctly for your region?

“Well, please read YOUR writing.”

I see. You’ve chosen to point out again an issue that I already clarified. Are you just trying to be argumentative? I’m trying to help readers out here. You were able to see the error the SVCDInfo.txt file created, but you missed the fact stated in that file that I’ve already turned off progressive, zigzag and deinterlacing.

And as for your bitrate comments I understand it far better than what you've stated and I’ve already tried this method. I’ve set the Max as high as 5000 for DVD2SVCD for STNG and it made no visible improvement. And it certainly didn’t help resolve the issue like IVTC did. If you want to see this approach make a difference, then you should read the thread "Modification and optimization of robshot method and its possible use with DVD2SVCD" in the advanced forum where there is a much longer and better discussion on DVD2SVCD's current method of max\avg\vbr bitrates versus a method which is really much closer to a true variable rate that can take advantage of what you are trying to do by increasing the bitrates. I stopped using the Robshot because it was just to buggy for me or my computer.
http://forum.doom9.org/showthread.php?s=&threadid=28939
Your statement of how VBR is used in DVD2SVCD, in IMHO, is oversimplified and misleading.

I still stand by my first post for a Region 1 encode of STNG to a 700mb cd-r.
This method is based on the desire for quality over time.

Chapters:
Some of the DVD titles are clearly chaptered wrong. Please tell me what broadcast of STNG would want a commercial break occuring 30 seconds after the end of the title sequence and in the middle of someone talking or the opening guest credits?? And your logic doesn’t even make sense. The DVD’s are not chaptered for broadcast - only playback at home where the author has absolute control of where the break is.

JoeB

4th November 2002, 02:49

I for one just checked out episode 150, and can very easily see the material is in fact interlaced! It's also a hybrid (just like the first 2 seasons) of FILM and regular NTSC. (DVD2AVI detected this, and hence forced ITVC on.

The other settings which I use with 80min CDRs which I forgot to say are:

Bicubic resize: b = 0, c = 0.75

JoeB

ux-3

4th November 2002, 16:02

@JoeB:
Could you please answer the following questions:
You did a visual confirm for combs, not just DVD2AVI?
Your are using RC1 NTSC?
What minute/scene did you see the comb effect?

Thanks for your time!

JoeB

4th November 2002, 16:34

Here's what I did:

- Rip only the 1st episode to the HD (The episode is the one with the nano-bots that Wes created)
- Play the VOB in PowerDVD, and it played perfectly smooth w/o a problem and they I remembered that PowerDVD removes that comb effect.
- Loaded the VOB in DVD2AVI and told it to preview. It played the scene (w/o audio) and if you pay close attention to the start of the scene (the one where the camera pans around the desk, showing how wes is sleeping, and all those gizmos) you'll notice the comb effect in the background and all around.

I'm using region 1 NTSC DVD's.

JoeB

manono

4th November 2002, 16:51

Hi ux-3-

I can appreciate GreenDrazi wanting to help others having trouble with this difficult series, but I don't think that either he or JoeB fully understand what's going on. As you mentioned earlier, the fact that DVD2AVI says it's interlaced means nothing, and it's frequently wrong.

The fact that the episodes can be successfully IVTC'd to return the progressive frames is proof that the series was created as film and, in fact, is not pure interlaced material (the NTSC versions anyway, with the exception of the CG parts).

In addition, since GreenDrazi (and I) haven't seen the even more screwed up PAL versions of the series, he's not much qualified to tell you what you're looking at.

GreenDrazi-please don't take this the wrong way-I know you're trying to help others working on the series. And this whole DVD2AVI-IVTC thing is a very difficult and complex subject. For more information, you might have a look at the DVD2AVI-IVTC Tutorial (http://www.doom9.org/ivtc-tut.htm).

Edit: By the way, ux-3, do you speak German? Have you seen Hawk's Guide to deinterlacing the PAL versions?

http://www.videoxone.de/cgi-bin/load.pl?page=hawk30_de

ux-3

4th November 2002, 19:17

Originally posted by GreenDrazi
ux-3
Look, I’m trying to help you out and others with STNG.
Well now, let’s see - in your last post, you are finally trying something that I recommended in my first reply. Hhmm.

Sorry, but to me this does not sound like you are trying to help, it sounds more as if you are turning a little snobish. In any case, I have already had success with season 1 and 2 and could keep doing so with 3. My questions had another motive, which has been lost after this general debate came on - never mind.

That would be like me telling people in PAL regions to get rid of all that PAL stuff.

My point exactly. Thats what it takes in PAL! I have no "PAL stuff" turned on. I have nothing turned on. Since I don't want to mess with the frame structure (bitrate and resolution aside) and their delivery to the TV, to keep the frames coming in an order that looks smooth, I don't resort, blend, shrink, telecine etc. them in any way. No PAL effect is let lose on the frame stream. Doing PAL, this is the way to keep things as they are, which looks best (in comparison to all other attempts I've made). In fact, they flow like the real PAL thing.

I’m trying to help people out here (particularly Region 1 users and hopefully PAL users can benefit too) and comments like that just confuse the issue. And when you are ready to smooth out that banding your getting, try temporal smoother.

As an aside, I don't get any "banding"(?). I am am sorry to seem confusing to you, but I wonder about your confidence in your council. In your first post, you state
The only down side to this method is the “space” shots - you will still get a little stuttering, but 90% of the rest of each episode will look much better and with far less artifacts.
To me, this isn't satisfactory, and if you you stand up and claim that this is the way to do it, then you are confusing the issue, because there may well be a better solution. However, as long as you categorially deny even the possibility, you can't improve.

Regarding the field order problem, there is a reference in the current debug thread for the next release of DVD2SVCD, which is called "Help!!!".

...but you missed the fact stated in that file that I’ve already turned off progressive, zigzag and deinterlacing.

No, i didn't miss it. If you reread what I have written, you will see that I make a slightly longer statement. I tell you to start with the settings you would make for a modern movie and name the changes that might help improve the results. Also, my remark about "not using filters at first" at that part of the text concerned the time when testing, as I thought I made clear. If you need them once the ships fly smoothly, it's up to your taste.

Basically you are already keeping the frame structure unaltered, except for that "NTSC-Stuff". Here I can't really help with specific knowledge, since I live in PAL Country. But the secret of total success (smothness in space) is to keep the structure of the stream. So perhaps you should try a problematic chapter without that field operation or pulldown or whatever else reorganizes frames. Apparently, the way the mpg from the DVD plays is fine, so keep that structure.
The PAL DVD version is really screwed up, but once I took the approach to keep the structure and not mess with it in any way, the results were amazing. Perhaps NTSC technology does not allow you to leave things as they are, but this is something you might like to check out.

And as for your bitrate comments I understand it far better than what you've stated

Well, fine for you. For others reading this, you can try the episode with the first ferengi encounter. Pay particular attention to the Planet scene with the thunderstorm. Use default or the max bitrate your player handels. Then just see and decide.

And your logic doesn’t even make sense. The DVD’s are not chaptered for broadcast - only playback at home where the author has absolute control of where the break is.
Sorry, but if you would try to read things twice, that you don't find logical, perhaps you would find less to quarrel about. I nowhere say that they are chaptered according to advertising. That is however, what you suggested. Forgotten? Here it is:
The chapter selection should put you at a commercial/advertisement break point, but on a lot of them that’s not the case.

Since your reading and writing is very presumptious and unfriendly, while very tolerant for diffusions on your own side, I'd rather not further this discussion beyond this point. Obviously, you think that you know it all much better than I do anyway, and I don't get payed to alter this. If you don't mind, perhaps we should stop "helping" each other. I don't want stuttering spaceships, you don't want to change your knowlegable method, so there is little we can accomplish. We can still try to help others. But I don't like your tone, so I'd rather call this off.

ux-3

4th November 2002, 21:05

@manono:
Originally posted by manono
By the way, ux-3, do you speak German? Have you seen Hawk's Guide to deinterlacing the PAL versions?

Not until you posted the link. Took a look at the page. The second sentence basically sums up the method I suggest:
"Wer eine SVCD/(x)SVCD oder einen MPEG2 Stream erstellen möchte braucht ebenfalls nicht zu deinterlacen."
Rough translation: When doing SVCD/xSVCD or MPEG2 Stream, there is no need for deinterlacing.

Perhaps you can help me to sort this NTSC stuff out. From what you write, an inverse telecine transform is possible on NTSC. However, the material seems to be made up of different types. So even if you apply such a transform, it will fail at certain scenes, particularly the speccial effects, as we can read here.

Is it possible with NTSC to keep the stream interlaced and turn off all the NTSC transforms?. If not, why not? Will your DVD player always physically deliver the same frequency? What does happen if you set the field order to B, and disable NTSC Field operation in the DVD2AVI Tab? (Zigzag off, Progframe off, deinterlace off). Will this preserve the structure of the original?

GreenDrazi

5th November 2002, 07:11

manono

I've never said that the STNG material was "pure” or “truly” interlaced, only that it is interlaced. The fact that DVD2AVI says that there is interlaced material on the vob's (R1), and the fact that this is a product that was originally filmed for North American broadcast television, and that, after encoding with IVTC, most of the artifacts found without using IVTC are now gone is evidence that interlaced material is present. Whether or not it’s originally filmed as interlaced, whether or not it’s encoded at some point as interlaced and/or whether or not it’s 3 progressive frames followed with 2 interlaced frames added when it was telecined, it is still INTERLACED. DVD2AVI does mean something - it's by no means perfect, but it's more often closer to the truth than to say it means nothing.

Thanks for the link and don’t forget the doom9 Advanced Reference guide which is also good:
http://www.doom9.org/mpg/dvd2svcd_advref.html

manono quote: “In addition, since GreenDrazi (and I) haven't seen the even more screwed up PAL versions of the series, he's not much qualified to tell you what you're looking at”

That’s an unnecessary and unfair comment (to me) because I believe that I’ve made it clear by stating in every post in this thread that I was basing my comments on experience with Region 1 NTSC material. ux-3 didn’t state his was PAL in his initial post and if he had, I might have avoided this entire thread. But even after knowing that, I was hoping for the sake of NTSC users and PAL users that IVTC would help resolve problems with SVCD encodes of STNG DVD’s. I said that I doubted that it would work for PAL. But I had hopes that someone (like ux-3) would try it and that it might work, helping all PAL users.

ux-3
I’ve tried well over 24 different encodes to SVCD format (full episodes, not just tests of chapters which can be misleading with STNG) of Season 2 and Season 3 and played them all on my home DVD player/TV (ie, not my computer) before I settled on what I first posted herein. I’ve even tried CVD’s, VCD’s, XSVCD’s and XVCD’s to avoid the problems inherent with resizing which adds to the problematic artifacts. Perhaps someone who has had better results with these other formats with STNG can offer their recommendations. I’ve tried the Robshot method (with untested results due to unrelated issues) which I encourage everyone to try. I’ve already tried every method you’ve described herein including your last post and I still prefer the method that I’ve posted. Please tell me where in this thread have I ever said anything close to I “... categorially deny even the possibility, you can't improve.” One of the reasons that I’m still in this thread is that you persist in mixing incomplete suggestions with nonsensical comments like “Don't use filters at first, as they only cost time. Keep all this NTSC stuff out of the game” about a difficult topic (encoding STNG). For some of us, time is not a paramount issue and filters have a purpose - like solving tough encodes. Please post your complete settings of at least one complete encode you think is good or at least the best you’ve done so far. JoeB and I did that on our first post in this thread. I don’t agree with all of his settings, but I don’t take issue with his post because he’s not differing with others herein with only half-baked ideas and mis-information. You still have not provided evidence that DVD2SVCD version 109B3 hasn’t fixed the field order issue and yet you are, in your last post, still asking people to set it manually. DVD2SVCD even stated this was “- Bugfix: Fixed the CCE Field Dominance bug according to this post on the forum:http://forum.doom9.org/showthread.php?s=&postid=133755#post133755" in his “readme.txt” for 109B3. I’m not saying it’s method is perfect now, but I’ve at least provided evidence/links/information here and before contrary to your position(s). I am in constant pusuit of a better encode of STNG that can fit on a 700mb CD-R, but it is becoming obvious that I’m not going to find it through you who appears to refuse to try or consider all the avenues available to him such as filters, alternate VBR methods and alternate VCD formats.
Just "turning off" everything, in my experience, is not going to get you a good encode of STNG.

My final comments to you on the chapters:
You’ve completely gotten the discussion off-track. On some of the DVD’s, the chapter points, particularly at the end of the opening title, are neither at commercial breaks (the original North American broadcasts which was the intended or the focus audience) nor are they (your quote) "inserted from the point of view of the advertisers, right at the hight of tension."
And no, your statement "as long as DVD2SVCD places them, where they are on the DVD, we should both be happy." doesn't cut it for me. There are some DVD's chaptered out there that are so bad, the DVD authors should be ashamed. “Mulholland Drive” had NO chapters (at the insistent of David Lynch). Perhaps you should have said that YOU are happy with whatever they give you. I and others are not always happy with whatever they may try to push off on us and having a chapter point on a STNG disc in the middle of a Capt. Picard conversation is just plain bad for these DVD’s.

manono

5th November 2002, 10:17

Hi GreenDazi-

Cool-an animated discussion. But please remember that we're all friends and all trying to learn. And you're right-ux-3 didn't state in the initial post that he was working with the PAL versions. But I knew immediately, because I knew the NTSC versions weren't interlaced. But even after, in his second post, he mentioned that his was PAL material, you still kept hammering at him, knowing nothing about whether or not his were truly interlaced, or where his chapter breaks were, or anything else about them. But he doesn't need me to defend him-he can take care of himself.

...I was hoping for the sake of NTSC users and PAL users that IVTC would help resolve problems with SVCD encodes...

You do know that PAL people can't use IVTC because that will convert their material to 20fps, which is the last thing they want, right? One thing they can try is Telecide() (without Decimate()) to fix the common problem in PAL areas of "shifted fields" which will give them back their 25fps progressive in many cases.

You have an interesting definition of what is interlaced. Material that has been created with a video camera at 30fps is interlaced (leaving aside the very little material created as 30fps progressive-mostly computer generated, I believe-there's some anime like that). Material that has been created with a film camera at 24fps is progressive. Almost all American TV series are now created as film (including ST:TNG-leaving aside for the moment the computer generated stuff). This is not so much true in other parts of the world. Film-the same stuff as in the movie theaters. So, you're wrong that it being a TV series is proof of it being interlaced.

Second-the fact that IVTC was needed to remove the combing and return it to it's original 24fps also is not proof of it being interlaced originally, but proof of it being progressive 24fps originally. Do you know how it's actually stored on the DVD? It's there as 24fps progressive frames. The DVD reads the flags in order to output it as 29.97fps interlaced to a standard interlaced TV set. It creates the interlaced frames (or more properly duplicate fields) so that the TV can play them. If you Force Film a properly telecined Movie, it returns it to it's original 24fps progressive state. IVTC is a stronger kind of Force Film that doesn't read the flags, but matches fields properly, or detects the Interlace-Progressive patterns to also return it to 24fps progressive. The problem comes when the flags aren't set properly, or are missing, or edits have been made after the flags have been set. And ST:TNG is a good example of that. It can't be Force Filmed in DVD2AVI. The flags are too screwed up. And DVD2AVI can't tell the difference between improperly telecined material and interlaced material. It calls them both interlaced or NTSC. That's why I said that what DVD2AVI tells you means nothing. For NTSC material, at best it is often misleading, and at worst it's plain wrong. The only way to truly tell what you've got is to look at the frames.

And thanks for the link. I know a bit about IVTC, but I make mostly .avis, and am relatively inexperienced with SVCDs (made only 20 or so). I can learn some from that one. I'm sorry this is so long. And it's been nice talking to you, man.

manono

5th November 2002, 13:51

Hi ux-3-

Is it possible with NTSC to keep the stream interlaced and turn off all the NTSC transforms?

Yes, of course, and that's one possible solution to the problem. But there's a major difference between PAL and NTSC in this regard. In a PAL area, whether encoded as progressive or interlaced, it's still at 25fps. You'll take only a small quality hit when encoding as interlaced, because interlaced material doesn't compress quite as well.

But in an NTSC area, progressive material is encoded at 23.976fps, and interlaced as 29.97fps. So, if you're able to encode it as progressive, it will have about 25% better quality than if you had encoded it as interlaced. So, I can certainly understand GreenDrazi's decision to IVTC it and put up with the slight jerkiness during the computer generated space scenes. I'd probably do it the same way myself. But it's a personal decision for NTSC people as to whether or not you prefer generally better quality and slight stuttering during the interlaced scenes, or lower quality, but have the 5% or so of the interlaced material play smoothly. Then there's something else to consider. 95% of the material is Film. If you encode it as interlaced, then you've inserted 2 duplicate fields in every 10 fields, or one extra duplicated frame in every 5. The slight jerkiness that results from that may or may not be noticeable (usually not, except to the eagle eyed with good TV sets), but close to 20% is now made up of duplicates.

And I'm sorry, but I don't know enough to understand and answer the other questions. Maybe the end of the last paragraph answered them.

ux-3

5th November 2002, 17:55

@manano:
I am glad to hear that the "don't change anything" method can work in NTSC as well. This should indeed keeps the ships moving smooth and also keep all other frame problems at bay. (aside from the field order).

Just one question about the 95% film. You say you encode duplicate fields (2in 10) which makes the material a little jerky. Will it be as jerky when played from the original DVD? Is it a loss of originality or just a bitrate waste?

I have concluded for myself that leaving stuff "intermessed" (to avoid unnessecairy precision as to the precise problem)does not deteriorate the material much, but in PAL World will keep any stuttering out of the game. Which boils down to using this method as soon as trouble strikes. Why bother deintermessing PAL, if this does it every time, at little cost. After all, to be sure of another method, you have to watch the entire coded movie carefully. Doing TNG-PAL drove me near crazy: Trying one "PAL-Deinterlace" method would help here and fail there, the next method would cure the failures but create new patients. You would never be sure that just because it worked on this episode, it will work on the next. Today, I think that wrong field order settings , especially during reencoding, did confuse the issue further during testing. This has been reported in the thread I mentioned several times.

I guess NTSC people will push harder and be more compromising about deintermessing artifacts, if it raises their effective bitrate by 20%. I myself have long searched about what encode format I should choose. I do two audio tracks (english and native), but I found that even DVD resolution was an option when using CDR99. I ended up doing xSVCD for now (native is mono), for the sake of moderate compatibility.

If I were to live in NTSC country, I would probably take the bitrate penalty and use CDR99 to get back what I paid. It's sort of ironic if I fancy a space series and let the space shots go to hell. Also, the premium price for the CDR99 will be made up by paying yourself minimum wage for the time you would spend trying to get similar results on CDR80 with demessing.

What made me start this thread initially was my observation, that PAL Episode 149 no longer had comb scenes during the actual shots. Even most new (romulan) space scenes are fine. There is the opening credit and some recycled space fly bys with combs, as well as some special effect scenes. But it took me a long time to hit those. In Episode 2, after one or two preview pics, you had a comb. This made me wonder, if a change in method would improve the results (in terms of effective bitrate). So I did an encode with progressive frame and took a look at it. It is fine 99% of the time. (aside from opening credits)

Doing it the old way yielded a very similar result in the good parts, but of course a also a good one in the messy rest. So I started to consider using a deinterlacing method that would not affect the +99% but help the rest to prog-frame.

After what you've just told me, I don't feel so "odd" about seeing no real detriment in quality when keeping the episode interlaced and be done with it. The advocates of the "huge quality loss" are probably in the NTSC camp, or are influenced by "how-to-Manuals from that camp".

I still may give it try...

manono

5th November 2002, 22:41

Hi ux-3,

Will it be as jerky when played from the original DVD?

To tell you the truth, I've never been able to notice it, and I've looked. I know it's there, but I've never seen it. This may be because of several reasons. Each of the duplicated fields is only on screen for 1/60th of a second, and they aren't run consecutively. The telecining process spreads them out (it's also called 2:3 pulldown and it makes only every fifth field a duplicate). In addition, the natural smoothing of interlaced TV sets may help to cover it up. In addition, we may be so used to seeing it displayed that way that our eyes and brains may also "smooth" it, so to speak. I suspect that it may be more noticeable when viewed through an HDTV, but I'm just guessing. What I'd really like to do is view the same DVD on a standard interlaced TV, and then on a progressive DVD player outputted to an HDTV side by side. I'm pretty sure it would be noticeable then. With such a high end system you can avoid the telecining entirely, and only the original 24fps progressive frames will be displayed. Perhaps there's someone around here that has a system like that that can tell us if the DVDs play more smoothly as well as with better quality. This link explains pretty well how outputting progressive DVD to a progressive display gets rid of all the jerkiness (if you can output at 72Hz) and improves things in other ways also:

http://www.avdeals.com/classroom/Proscanexplained.htm

You know, I read of you Europeans using the 99 minute CD-Rs all the time, but I've never actually seen one in my life. And the ones I've seen online are very expensive. I think they must be much more common where you are, and I think the CD-R 80s also are somewhat expensive for you. But here the CD-R 80s are dirt cheap and sometimes they even give them away. So, as far as I know, most Americans stick with the CD-R 80s. But yes, I'd also like to use the CD-R 99s for those difficult to encode videos also. Anyway-take it easy, ux-3.

GreenDrazi

6th November 2002, 05:17

manono

Thanks for your input herein. I believe it’s helping to carry this thread into some good debates and exchange of information. I guess that’s just one reason why they made you a moderator. But I do wish you wouldn’t turn, change, or alter some of my statements such as “proof of” when what I said was “evidence of.” So if you don’t mind, I would like to say, with regards to your 5th November 2002 04:17 post:

manono quote:
"... Almost all American TV series are now created as film (including ST:TNG-leaving aside for the moment the computer generated stuff)."

That's just it, you can't “leave” the cg stuff “aside” on the STNG dvd’s. It's part of the stream and the cg stuff on the vob's MAY well be interlaced. You essentially say so yourself in other posts herein this thread. Therefore, if the cg stuff, or anything else in the stream contains some interlaced material, then the vob's are interlaced - not 100% interlaced, but nonetheless interlaced.

manono quote:
“ ... IVTC was needed to remove the combing and return it to it's original 24fps also is not proof of it being interlaced originally.”

I agree, it’s not “proof” and/but I did not say that the material was “originally” interlaced. Only at some point in the path has interlaced material been introduced/added/encoded/something. And unfortunately, a lot of North American TV material which makes it way onto DVD’s APPEARS to have been telecined and for NTSC regions they have OFTEN added two frames of interlaced material in the process. I wish they were more often like the majority of theatrical releases brought to DVD, because we could all have much easier, quicker and better SVCD’s.
The doom9 Advanced Reference guide states:
"Telecined 30fps is simple to identify. It's usually 3 progressive frames followed with 2 interlaced frames."
I don't agree with this 100%, but the jest of it appears to be true for a lot of R1 TV material on DVD.

manono quote:
"You do know that PAL people can't use IVTC because that will convert their material to 20fps, which is the last thing they want, right? ...”

My understanding with DVD2SVCD is that when you select IVTC, you should get the correct Telecide/Decimate for the source material (PAL or NTSC). If I'm wrong on what DVD2SVCD does with this, I deeply apologize. Someone please advise.

Again, thanks and let's keep the info flowing.
It's time for me to go and watch my recent encode of "Encounter at Farpoint"

manono

6th November 2002, 06:32

Hey Bach-

If the source is REALY interlaced, there are 50 spots_per_second to be shown,

By spots I assume you mean fields. And both progressive and interlaced material will be displayed as 50 fields/second on an interlaced PAL TV set. I come at this from the .avi side of the site and sometimes it's just easier to speak of frames and I don't see anything to contradict there. As for interlaced material playing more smoothly, in theory yes, but I'm not sure the eye can tell when it's whizzing by at 50 fields per second. But PAL doesn't telecine up, so maybe a progressive 25fps source does display as slightly more jerky. I can only speak from NTSC experience. As I mentioned in my last post, I certainly have never noticed jerkiness when comparing a progressive source with an interlaced one. As for which looks better, I think the progressive source would, especially when viewed on the HD sets we'll all own in a few years. I think the link I gave in my last post would confirm that. As for an interlaced encode requiring 1.5 to 2 times the bits for the same quality, I was just guessing, and I'll bow to your statement. But that seems to confirm doing the encoding as progressive whenever possible. And it also seems to confirm GreenDrazi's decision to encode the ST:TNG series as progressive.

FILM material is encoded at 23.976.

Yep-you got me there. I was careless in my choice of words. However, I did mention the existence of 30fps progressive material in a previous post. And there's so little of it so far that it can be pretty much ignored. I haven't seen any yet, but expect to soon when some anime material I've ordered comes in.

For this pattern, it is enough to edit the avisynth with separatefields.trim(1,0).weave in order to "deinterlace" it.

Telecide() by itself will accomplish the same thing, just as it will with a lot of seemingly interlaced PAL material.

Personally, I think you're being kind of petty, but GreenDrazi thinks I'm being petty with him, so I guess it evens out.:) This thread's becoming pretty interesting though.

manono

6th November 2002, 07:09

Hi GreenDrazi-

Therefore, if the cg stuff, or anything else in the stream contains some interlaced material, then the vob's are interlaced - not 100% interlaced, but nonetheless interlaced.

OK, I'll buy that-part of the vobs anyway. By the way, as I'm sure you know, by IVTCing the whole thing, the CG stuff, being interlaced to begin with has non-duplicate fields being dropped which explains the jerkiness. In addition, because they're composed of separate fields, and not frames, all that's being done is blending the fields together to make blurry frames. But I do think that's a valid solution. Hybrid material is never easy to work with, and it's a question of choosing the lesser of two evils, so to speak. I think the PAL versions must come with those interlaced CG portions already blended together, because 60 fields per second have been compressed into 50 fields per second.

The doom9 Advanced Reference guide states:
"Telecined 30fps is simple to identify. It's usually 3 progressive frames followed with 2 interlaced frames."
I don't agree with this 100%, but the jest of it appears to be true for a lot of R1 TV material on DVD.

You're right-the movies generally are done much better. There are just so many DVDs that have been mastered so poorly. You should see some of the Hong kong (NTSC) martial arts movies. You look at them on a frame by frame basis and many show 4 of 5 frames as interlaced. I get similar kinds of things with European movies prepared for NTSC. I'll see 4 and sometimes 5 interlaced frames out of 6. And this is from Film material. There's no excuse for that kind of sloppiness.

My understanding with DVD2SVCD is that when you select IVTC, you should get the correct Telecide/Decimate for the source material (PAL or NTSC).

Gordian Knot for making .avis doesn't allow IVTC to be chosen for PAL material. The same thing for NTSC material that has been Force Filmed. The IVTC option is only available for NTSC material where No Field Operation has been chosen in DVD2AVI. This is to prevent rookies from making serious mistakes. Telecide puts the fields back together into frames, and Decimate removes the duplicates (1 in 5 at the default setting). I think the Auto setting in DVD2SVCD probably does a pretty good job of detecting what you've got. JoeB said earlier that it had chosen IVTC for him for his ST:TNG episodes.

Talk to you later. I'm all typed out.

ux-3

6th November 2002, 14:43

@manono:
And it also seems to confirm GreenDrazi's decision to encode the ST:TNG series as progressive.

Yet Greendrazi's log states:

- Zigzag scan order: No
- Upper field first: No
- Progressive frames: No
- iDCT Algorithm: 32-bit SSE MMX
- NTSC Field Operation: Inverse Telecine, using Donald Grafts Decomb dll
- Deinterlace: None

So if I am not mistaken, he still works with fields. So he should in fact still take some bitrate penalty.

As I said before, this penalty was the motive for putting up this thread. 149 is quite watchable in progframe PAL, without any deinterlacing! So this should yield a significantly better image quality, if bach is right. I have just prepared different encodes of 149 with no deinterlacing and either prog frame on or off. I'll try to do a comparison.

ux-3

6th November 2002, 15:58

Ok, so I have just finished coding and took 4 (almost) identical screenshots with windvd.

I did one shot using my old method (keep interlaced, max 3500 4pass)
one using that method plus TS and Denoise as greendrazi suggested,
another one from the VOB itself and a last one with my old settings but treating the material as progressive, thereby avoiding the bitrate penalty, but supposedly getting combs etc. at some parts.

Now there is little point to jpg the pics... Each is 1.3 MB, so if someone has a little space somewhere, I could upload them there. Might be worth a look.

Edit: I meant WINDVD, see above

manono

7th November 2002, 17:32

Hi-

So if I am not mistaken, he still works with fields.

I noticed that when he posted the log, but don't know enough about it to comment. I was just going by what he said. I'd be interested in seeing the pics. Check your PM.

JoeB

7th November 2002, 22:17

ux-3:

why not create a ux-3 user ID on www.yahoo.com and post the pictures in the photo gallary there?? They don't compress them nor do they resize them.

JoeB

Pko

8th November 2002, 11:25

This is just pure speculation, and (if it is at all possible and worth of the work) would need some programming...

Having said that, here is my idea. Basically, it is to generate a hybrid stream, since we start with an hybrid stream, but using CCE or TMGEnc that (AFAIK) do not have a way to generate hybrid material.

The hybrid (NTSC) material of TNG contains, say, 90% of FILM and 10% of pure NTSC. Let's suppose we rewrite DVD2AVI so it generates a .d2v file that has all the *TRUE* film "forced to film" (so now is progressive and with no duplicates) and keeps the interlaced NTSC interlaced. Also, it generates a file saying which frames where FILM and wich NTSC.

We have now a file that, to be correctly played, need to change the framerate (it is mixed 23.976 and 29.97). We "think" of it as interlaced 29.97. We decomb() it and now is progressive (the FILM part were already progressive, but decomb() is smart and do not touches that frames).

We encode it to MPEG2 and have a fake 29.97 file that will of course be totally out of synch with sound...

Now, with our rewritten version of pulldown.exe and the file our rewritten version of DVD2AVI generated, we correct that; our pulldown.exe will mark as film only the part that were true film, and keep the rest as progressive NTSC, so it will be a hybrid and now it will synch correctly. Also, we have the best possible quality in the FILM part and only slight degradation (because of our deinterlacing) in the NTSC parts, but since they were encoded as progressive (that, at SVCD bitrates, gives better image quality than encoding as interlaced), the quality will be nice enough.

Another possibility would be to encode as interlaced, and we do not need decomb()... but we would have the bitrate problem.

Also, we should choose wisely our bitrates (max and average), since CCE or TMPGEnc will not know the "true" framerate, but a fake one, when encoding. Since we know the ratio of NTSC and FILM that could be somewhat compensated...

I think (I'm not a programmer) that only somewhat "minor" changes would be necessary to DVD2AVI and pulldown... but perhaps there are some basic flaws in my reasoning I am not aware.

BTW, I've been "out" for some months (and I'm with no computer at home for some more months :-( ), but the forums seem to have kept a really good pace :-O

GreenDrazi

8th November 2002, 16:48

Interesting idea really. Would there be any problems with CCE and/or TMGEnc with mixed framerates?

Perhaps an alternate approach along the same lines but maybe simpler (or less re-coding) would be to have your rewritten DVD2AVI save individual d2v files in lieu of a hybrid stream d2v file. With a re-scripted DVD2SVCD (which would be easier to have happen I think), you could go through all the normal encoding of the files separately and then after getting the individual files framerates synched, you could have bbMPEG mux\rejoin all of the files.

Or a program similar to DVD2AVI, but one that can split the vob’s at the appropriate points. Re-scripting of DVD2SVCD would again still be required, but might be simpler this way.

You could do the same manually by saving to an avi file and edit/cut/paste, but the time required to do so I would think it would be better to just be happy with the original DVD.

Pko

12th November 2002, 15:49

Would there be any problems with CCE and/or TMGEnc with mixed framerates?
Don't think so... the modified DVD2AVI should change the time codes so the FILM part seems to be NTSC, so the .d2v is shorter than the original (the ~24fps parts are marked as being ~30fps, since the number of frames is the same, the movie is accelerated). To CCE, it woukd appear as a NTSC stream.

Or the .d2v could be generated as FILM, so the NTSC parts will play at ~24fps and so the .d2v will play slower than the original (because of the NTSC played as FILM). But that way, the encoder should see the whole stream as film and will encode that way.

Perhaps "all as film" will be better than "all as ntsc" since we could choose the max and average bitrates more close to the real ones, since the stream is mostly FILM and only a low % of it is NTSC.

GreenDrazi

12th November 2002, 19:56

I started encoding Babylon 5 Region 1 Season 1 DVD's and they appear to have the same problem as the STNG discs, but Bab5 relies much more on cg footage. So I had to find something that worked better.
Just using Forcefilm "on" or "off" left too many combing effects.
I'm now using Donald Graft's Decomb Plugin (using latest version found in DVD2SVCD) for Avisynth with the following settings with extremely positive results for both Bab5 and STNG. You'll have to set DVD2SVCD to edit the avs file manually of course. This setting appears to recognize hybrid footage and treat the frames as such, which "in effect", does what we just described above.

Telecide()
Decimate(mode=1)

Please note that these are not the default settings used when selecting IVTC in DVD2SVCD.

http://shelob.mordor.net/dgraft/decomb.html

From the help file:
"Processing Hybrid Clips Some clips mix film and nonfilm content. One way to handle these is to leave them at the original frame rate and let Decimate determine what to do with the frames that would otherwise have been dropped"

JoeB

14th November 2002, 06:14

GreenDrazi:

Hmm.. your post made me start thinking, so I did some more reading into it on other boards, as well as look up the docs. Maybe this should be a selectable feature that can be automatically added into the next version of DVD2SVD, but anyways.. here's the exact procedure:

- Under frameserver, select "Edit when dvd2avi is done"
- Under DVD2AVI, select ITVC

Begin the regular conversion, when the avs screen pops up, change the line reading "Decimate(5)" to "Decimate(mode=1, threshold=50)" and click save followed by OK.

As described on http://www.doom9.org/index.html?/decomb-params.htm:

mode=1 the frame would be blended with the previous frame. This can be useful if you have a hybrid 30/24fps source and don't want the 30fps parts to be jerky by discarding some of the frames. Usually, movies/series that contain both live action and CG stuff are hybrid (for instance Andromeda and Babylon5).

But you forgot the threshold (default of 0):

can help prevent blended frames on 30fps content when using mode=1. If the difference between frames is larger than this threshold a frame will be be passed through untouched whereas if the threshold is lower than the value you've set the frame will be blended. The default threshold=0 will let all frames pass through untouched. This parameter has no influence when you use mode=0, and if you use mode=1 a value between 25 and 50 is reasonable.

Thanks for finding this info out.. i'll give it a shot..

Another user on another forum wrote:

Then I tried the Telecide() w/ Decimate(mode=1, threshold=50) and I found that the the motion was much more fluid and the file size didn't change too much (960MB vs 1020MB). I noticed that avisynth makes a blurred "half image" between the repeated 5th frame and the frame after it. This is pretty much unnoticable during playback but it makes the overall playback smoother. I was quite pleased with the results and my entire 42min capture was comb free (nothing snuck through) with no stutter.

RESULT: The FILM portion looks perfectly fine, as always, and the CG part is no longer jerky (as I noticed it when looking for it closely).

Time to confirm this with a test!

JoeB