John, you seem to always come to the rescue when I pose these obscure questions. Thanks for taking the time to share your knowledge!

Jerry

HOWEVER,

if you add a progressive asset (like a 25(30)p - choose your area:)) media - DVDA will not accept it without recompressing into 50(60)i (again, depending on your PAL/NTSC area).

It must be stated explicitly, that the ONLY progressive media DIRECTLY supported at this time (by both DVD and BD) is 24p. No 25p, no 30p. Period., I've a good news to PAL users: thanks to no pull-down required, you can render your Vegas 25p timeline into DVDA-compliant 50i, and still not suffer from interlace artifacts (like combing vertical edges in motion) - simply because the 25p->50i results in 25PsF. Fields - yes, but offset in time - no.

I can assure you, that your graphics card / software MPEG / DVD/ BD player can be set up the right way ie. when NO combing artifacts are visible.

As to watching such DVD/BD on a (HD)TV, playing back from a standalone DVD (BD) player - no worry either, as any of those devices will deinterlace for you. HOWEVER, in order for this deinterlacing to be successful (i.e. not loosing resolution, or introducing combing) - you still need to make sure your Vegas renders output true 25PsF, and with the right Field Dominance.

Megabit, I don't think this is correct. Here's what I did:

1. My source video was shot 720p30.

2. I loaded this into Vegas with project set at HDV 720p30.

3. I rendered using using the stock NTSC widescreen for DVDA, except I changed it to Best because of resampling. I forgot to change from interlace to progressive.

4. I made a DVD in DVDA, no recompression. When I played it on my tv there were a lot of what appeared to me to be interlace artifacts as the player deinterlaced the video.

I then repeated above steps except I changed the Vegas render template to be progressive. I then again created a DVD in DVDA, no recompression. Now when I played it on same tv it looked great. This leads me to believe that DVDA does in fact handle progressive renders without recompression.

Here is what the Sony knowledgebase specifies for avoiding recompression in DVDA:

Compliant NTSC MPEG video:

Note that it does not explicitly allow or disallow progressive. I am taking this to mean you could use either. But it would be nice if they listed both, as they have done with aspect ratio.

Jerry

I just did a little research using Google. Boy, is the Internet ever full of a lot of bad information!!

Near as I can tell, the DVD spec permits three types of video: 24p, 50i (PAL), and 60i (NTSC).

That's it.

So it is not some deficiency in the Sony products that restrict us to these three options: it is the DVD specification itself.

3. I rendered using using the stock NTSC widescreen for DVDA

And there you changed for 60i - the Vegas stock templates are fully DVD/BD compliant - you can change the template's Field Dominance to None (progressive), but you didn't (if you did, DVDA would sure need to re-compress it).

John confirms it in his last post.

Piotr

Piotr, you may be right but I just tried this:

1. Rendered a short clip with 720p30 source using the modified NTSC widescreen for DVDA template. I set to Best, progressive, and CBR 8Mb.

2. I brought the resulting MPG to the timeline and checked its properties: It is MPEG-2, 29.970 NTSC, None (progressive scan).

3. I took this into DVDA and did a Prepare - DVD did not recompress.

4. I took the prepared VOB to the Vegas timeline and checked its properties. It is the same: MPEG-2, 29.970 NTSC, None (progressive scan).

5. I rendered using the stock DVDA widescreen template and do the above and the properties show it as interlaced, lower field first.

Any ideas why steps 1-4 work as they do?

John, I understand what you mean when you say "as near as I can tell". I did similar google and you can make it go whichever way you want. I did learn that for $5000 I can purchase the official DVD spec, and presumably find out the real truth ;-).

Jerry

Jerry, the DVD Demystified FAQ is a good (and free) resource.

Mike

A key thing to understand is that if there is no temporal difference between the upper and lower field, then whether you call the thing interlaced or not, it is actually progressive. I take advantage of this in both my inventions: the film to video transfer system I have described in the past, and also the kinescope to video system I described in a post a few days ago. I am certain about his in a way that is only possible from actually having invented things which exploit that fact.

Now, a display may only be capable of displaying interlaced (e.g., a traditional CRT NTSC monitor), so the progressive material is still going to be displayed with alternating scan lines, and the last field of one progressive frame will be on the screen at the same moment as the first frame of the next progressive frame, so it may not look exactly like progressive should look. But, there is absolutely nothing you can do about that because an interlaced monitor works like that and cannot be changed.

But, if you have a progressive display and a player which can output both fields at the same time, you'll end up with true progressive regardless of how its encoded (I think).

Jerry, the DVD Demystified FAQ is a good (and free) resourceIt is a good resource, but I am not sure whether what they say about progressive vs. interlaces is correct. Here's the part that bothers me:

"There's enormous confusion about whether DVD video is progressive or interlaced. Here's the one true answer: Progressive-source video (such as from film) is usually encoded on DVD as interlaced field pairs that can be reinterleaved by a progressive player to recreate the original progressive video."

Having just read a lot of posts around the web during the brief Google search I alluded to in my previous post, I think there is a lot of confusion about something which apparently is labeled an "interlace/progressive" flag in some documentation. AFIK, the only such flag is the one telling the player whether to ADD duplicate fields to 24p material so it can be displayed on NTSC CRT monitors. This quote above seems to suggest that there is some flag which can take video which originated from film and to which the pulldown has been added prior to encoding (which is an amazingly stupid thing to do, but it is done) and then somehow perform an inverse telecine to extract and then display the original 24p material. I don't doubt that such players exist, but having done a lot of IVTC, it is a very difficult process, which requires all sorts of adjustments and tweaks to get correct (because the pulldown cadence changes at scene breaks) and therefore is unlikely to be done correctly by a one-size-fits-all hardware decoder.

Therefore, the above paragraph, which is at the heart of what is being discussed, is at best misleading, and perhaps just plain wrong.

What you're saying is exactly what I've always thought however there might be a trap. Somehow I've seen things make a mess of PsF. It looks like what is happening is a field from frame 1 is being combined with a field from frame 2 during the rescaling or resampling process. The result is some pretty ugly interlace looking artifacts on motion.
I read a couple of posts where people were complaining about this and was rather dismissive of their concerns thinking it was impossible. Having now seen it actually happen I'm not so certain this whole PsF thing is as goof proof as I'd thought it was.
Piotr was a clip on Vimeo but I don't know if it's for general exhibition or not so I'll leave it up to him to post a link. Whatever has happened to his 25p footage it doesn't look too good.
I suspect the problem could be avoided by always conforming your footage to Vimeo's requirements before uploading it. Then again I'd like to really understand what went wrong, if Vimeo can make a hash of it the potential exists for the same thing to go wrong elsewhere. Give Murphy something to work with and he will :)

Bob.

It looks like what is happening is a field from frame 1 is being combined with a field from frame 2 during the rescaling or resampling process.That would be really ugly. However, that sounds almost the same as a field reversal problem in footage that really is interlaced, and I suspect the solution would be the same: reverse the fields. As always, the trick in Vegas is figuring out where to do that reversal: at the media, the project or the render level. You just have to experiment to get it right.

John

If I understand your statement below, how can you explain that when I rendered the a progressive clip interlaced and made a DVD it looked crappy, and when I re-rendered the progressive clip as progressive and made a DVD it looked great?

Jerry

>A key thing to understand is that if there is no temporal difference between the upper and lower field, then whether you call the thing interlaced or not, it is actually progressive.

Piotr was a clip on Vimeo but I don't know if it's for general exhibition or not so I'll leave it up to him to post a link. Whatever has happened to his 25p footage it doesn't look too good.

Of course Bob, here it is:

http://www.vimeo.com/2276175

I' like to ensure you that clip looks perfectly when played back here on my PC, as 35p. However, it probably was converted to 24p at Vimeo, and something got wrong with the field dominance - anyway, interlace-like combing is visible,

If you turn off resample, you should get the same result. However, if you start with progressive and then render it as interlaced, Vegas will artificially add fields (re-sample), where none existed, and will then further create artificial temporal differences between these fields, which will definitely make the video look soft. So I would expect that doing what you did would result in video that didn't look as good as the original.

I strongly recommend NEVER changing interlaced video to progressive ("de-interlacing") or rendering progressive to interlaced.

This is no way conflicts with my statement that there if there is no temporal difference between fields (meaning that both even and odd fields come from the same moment in time) then that video is, by definition, progressive.

If you turn off resample, you should get the same result. However, if you start with progressive and then render it as interlaced, Vegas will artificially add fields (re-sample), where none existed, and will then further create artificial temporal differences between these fields, which will definitely make the video look soft. So I would expect that doing what you did would result in video that didn't look as good as the original.

John, are you positive about this statement? It sort of contradicts what Bob and myyself (both in 25p lands) have concluded.

John, are you positive about this statement? It sort of contradicts what Bob and myself (both in 25p lands) have concluded.I will of course defer to your experience since I don't work with 25p, only 24p and 29.97 NTSC (60i).

However, several years ago, when I first started using my Workprinter to transfer 8mm film to video, I spent a lot of time trying to decide on the best way to do pulldown to go from 15, 16, or 18 fps (typical home movie silent film rates) to 29.97 interlaced (NTSC TV set). I first created my own pulldown using AVISynth. This is, for me, the gold standard. I then look at the results using an AVISynth script which separates the fields to I can make sure that I have done nothing more than duplicate fields according the the pulldown pattern I set.

I then tried various settings in Vegas to see what it would do if I set the header on my film captures to 16 fps and set the project properties to 16 fps progressive, but then rendered to 29.97 interlaced. It was at this point, as I remember (and it was at least five years ago), I got very similar results to my AVISynth pulldown when I disabled resample for each piece of media in the project, but I got synthesized, blended fields if I left the resample setting on the default "smart resample."

I then started experimenting using motion estimation for the pulldown (using Motionperfect back at that time). That produced some pretty amazing results that looked neither like video nor like film. It became its own unique animal. However, it is a useful trick if you have to transfer amateur movies where the operator pans left/right at high speed with a 16 fps small-film-gauge camera and produces enough film judder to make even an astronaut motion sick.

Well I just spent a while trying to create a scenario in which Vegas would create interpolated fields going from 25p to 50i. I have failed.
Even setting the event to Force Resampling does no good.

If anyone else wants to try it's dead simple to create some 25p footage with fast enough motion to show interlace artifacts. I just used a black circle (black gen media + cookie cutter) on a solid white background. I used track motion to move the circle from left of frame to right of frame and back again in 25 frames. I rendered this out from my 25p timeline to PAL DV with Field Order set to None.

I then brought that into a PAL 50i project and rendered that out, I set the clip properties to Force Resample and rendered out to 50i.

Bringing that back into a 50i project still no sign of interlacing to indicate that Vegas had created fields with temporal separation by interpolation.

I'm almost surprised that I can't make this happen, you'd kind of think Vegas might do this.
Now here's where it gets wierd. I took that 25PsF into a 23.976 project and set the speed of the clip to 96%. I got blended FRAMES not fields. Wierd.

I'm going to play around with this some more. If you look at the Vimeo clip that Piotr has provided a link to at 00:48 you'll notice some uglies on the left hand of the performer. How they got there I don't understand. Piotr submitted 25p to Vimeo, Vimeo converted it to 24p, whatever else it did is a mystery and one that I cannot emulate with Vegas.


Bob.

Aha, OK John.


As I stressed above, in the case of 25p->50i no pulldown is involved (or strictly speaking, the 2:2 is; see here: http://www.afterdawn.com/glossary/terms/22_pulldown.cfm),

- which in my understanding doesn't generate and offset extra fields, but simply stores each frame in two fields (upper + lower) without any time offset. This is the 25PsF in its pure form.

Now (and this is what first caused a lot of havoc), when a hardware or software player thinks it's interlaced (which it is, being wrapped inside a 50i stream), it will try to de-interlace it. If the de-interlacing algorithm happens to be "Bob" (sorry Bob, you bad de-interlacer :)), we're ending up with horrible line twitter (as was the case with the first prosumer 25p camera from Sony - the V1E).

If it's impossible to prevent any de-interlacing, one needs to at least set it to "Weave". The best results however are achieved with ANY de-interlacing OFF; the 50i stuff rendered from 25p source by Vegas will still feel and look like progressive at its best, with the added advantage of being DVD/BD compliant (hence, no re-compression in DVDA).

One must remember though, that he needs to control quite a number of the whole "from file to screen" chain's stages; the first one being the graphics card settings (e.g. ATI has "Perform automatic de-interlacing" set as default), then all DVD/BD software players need to be correctly configured, as well. Of course, with the final (physical) DVD/BD, all is taken care of by the DVD/BD player and the (HD) TV used; while this is not always optimal, one cannot do much in this respect.

Using the ATI HD 3870 card, I have set up a profile for playing my 50i (aka 25PsF) media, and another one for watching DVD/BD on my 50" HDTV plasma (with the latter one set to use auto de-interlacing, and the former one - not de-interlacing at all). This way, I'm not seeing any ugly interlace artifacts at all.

John, I strongly agree with your strong recommendation ;-).

But here is my problem. Say I shoot in 720p30 which is not an uncommon format. Looking strictly at the Vegas templates (both Vegas and DVDA) there is no way to put this (downsampled) on a DVD without interlacing. But my tests indicate that by changing the standard template to progressive, I can in fact stay progressive all the way through the workflow to the DVD. I know, I should be happy but I'm just curious as to the "why" this is so. Why do the templates, especially in DVDA, seem to indicate you can't do this? This was the crux of my original post.

Jerry

>I strongly recommend NEVER changing interlaced video to progressive ("de-interlacing") or rendering progressive to interlaced.

Well I just spent a while trying to create a scenario in which Vegas would create interpolated fields going from 25p to 50i. I have failed.Well, you got me to spend fifteen minutes to look at this. I get exactly the opposite results of what you report.

First, I did not want to use generated media because that introduces an entire new, additional set of issues (such a supersampling) that I didn't want to get into. I did a brief search on the Internet for some actual 25p PAL footage I could download but couldn't find any. So, I took some of my 23.976 progressive footage, verified that it was really progressive by separating the fields in AVISynth and then looking at field-to-field motion to make sure that all I saw was the usual up/down bounce between two progressive fields.

I then patched the header to make it play at 25 fps. Thus, it is a 25P frame.

I put this into Vegas and looked at the media properties, and it was correctly reported as "25.000 (PAL)", although the resolution is 704x480, because of course it isn't a PAL clip. However, for the purposes of this test I don't think that matters.

I then set the project properties to exactly match the clip: 704x480, 25 fps, None (progressive scan), PAR 1.000 (another slight variation from PAL, but I don't think it matters here), and finally Deinterlace method set to None.

I then rendered this clip twice.

The first render was to NTSC DV, with the event resample set to "Disable." The second render was the same render, but with the even resample set to "Force Resample."

I then took these rendered clips and used my AVISynth clip to separate the fields. I looked at both clips, through this script, in VirtualDub.

What I saw is exactly what I remember, and NOT what Bob is getting:

On the disable resample render, I see standard, clean pulldown, with fields repeated in something similar to a 3:2 pulldown pattern (obviously not 3:2, however, since this is going from 25p, not 24p).

But, on the force resample render, every single field was different in time, with obvious blends on the fields which had been synthesized.

This was done with Vegas 7.0d.

So, having now done this test, I will stand by everything I said earlier in this thread.

"First, I did not want to use generated media because that introduces an entire new, additional set of issues (such a supersampling) that I didn't want to get into"

Unless you tell Vegas to use supersampling it produces perfect progressive footage using generated media. One can easily see if Supersampling it coming into play by looking at the ouput frame by frame. In my case every frame is a perfect circle with no simulated motion blur.
---------------------------------------------------------------------------------------------

So I just did your test. I got almost exactly the same result :)
Going from my Vegas created 25p (note 25p not 25PsF) bouncing circle with Force Resample and rendering to 60i and looking at that on a 60i T/L I got interlace artifacts on EVERY frame. SOME frames display 3 different positions of my 'ball', two look like full fields and one the result of interpolation (grey). That on some displays would look woefully blurry.

Going from my Vegas created 25p (note 25p not 25PsF) bouncing circle with Disable Resample and rendering to 60i and looking at that on a 60i T/L I got interlace artifacts on SOME frames. It looks like what Vegas does here is it decides the nearest frame is close enough and splits that into two fields with no temporal difference. At other points in the sequence in may take a field from adjoining frames but I need to really check this very carefully.

So yes, I can make Vegas do what you can, phew :)

HOWEVER (not shouting, just waving my arms)!

The original question was what does Vegas do converting 25p to 50i. I still cannot make Vegas produce any interpolated fields in this scenario. Going from 25PsF to 50i it's the same. I'd really expect that as it's really just changing flags, if anything.

In the midst of all the above I had a brainwave!!
---------------------------------------------------------------------------------------------
I think you've forced my to find the answer to the original problem, what the heck has Vimeo done to Poitr's footage.
Note that the horrid dogs teeth seem to come and go. It's not until 48 seconds into the video can you really see them. Huh, that sounds familiar! When I tried converting 25p to 60i using Vegas without resampling I saw much the same. Now converting 25p to 24p it would take much longer for the split frame to show up. My test clip is only one second long however I just went back to it and added frame numbers inside my black ball. Bingo!

Dropping my 25p footage into a 24p project I get some interesting results.

With the clip properties set to Force Resample (same as Smart in this case) EVERY frame contains blended fields.

With clip properties set to Disable Resample the first twelve frames have NO resampling, the next 12 do, I never get get fields from more than 2 frames however like I did going from 25p to 60i.

This to me at least pretty much simulates what happened to Piotr's footage. Given that the subject isn't moving much one would only see the problem when you had enough motion in the frame AND at the right point in the resampling sequence. YIKES, there's some interesting conclusions I can draw from this.


Both of us fell into a trap. Looking at just one frame of a sequence isn't enough. You chose one that showed resampling going on. In my initial tests I only looked at a small samples of frames and indeed there was no resampling in that part of the sequence.

I should say again, going from 25p to 50i no matter what Vegas does the right thing.

Big warning though. Going from 24p to 23,976p Vegas could make a real mess but you wouldn't see what's going on easily. I suspect you'd only see funky things every couple of minutes if you had Disable Resample in force.

For the record:
1) All of these tests were done entirely in Vegas 8.0b (not that I can see that would matter).
2) They took more than 15 minutes.
3) Our water does go down the plug hole the wrong way. This is offset by us driving on the wrong side of the road.


I'll now get back to something simple, mastering After Effects :)

Bob.