@Christian

Thanks for taking the time. How to you further process those files? As they are, they not usable within DVDA without a re-render and change of aspect ratio?

@ OldSmoke

I downloaded the video and will run it through my workflow and I will also use the standard Vegas method. I will upload the results after I do.

Unless you want otherwise, the final results will be 30p.

@Vidmus

I would like to get a MPEG2 file that is 704x480, 16:9 and 59.94i or some may call it 29.97i. In the end it has to be a file that DVDA can accept without re-rendering it; which is what I get straight out of Vegas.

Edit:
Here is what I am looking for. This is the XDCAM file converted with VP13 only to DVD compliant MPEG2 format:

File 1: No additional filters, straight from the timeline.
File 2: From the timeline but with the addition of NeatVideo.

Can anyone get better results? If so, please let me know your workflow.

OOPS,

Here is what I did before I read your reply. I will work on what you want and then upload that.

Meanwhile, you might want to look at the 30p version just for kicks.

Here is the link. I hope I did it correctly. I am still a bit new to this Drop Box thingy. The Drop Box help naturally does not match what they tell you. Of course...

https://www.dropbox.com/sh/u5ufoo4lrq8cxqj/AACon8kL3_iCXnFupjS_A9xMa?dl=0

The files are zipped. Just download the zipped folder.

The only filter I used was the Sony Levels where I have the settings as: Sony Levels with Output End: = 0.920. This does the following,

1) It makes it so the brightest portions do not exceed 235

2) It helps reduce the exposure on the gals face a little bit so the spotlight does not make her look a bit overexposed.

3) It preserves all of the video detail above 235 so it is not lost.

I do not care if it make the upper portions of the video slightly darker. It just plain looks better this way and that is all that really matters.

The above applies to cameras with levels of 16 to 255 only. Other cameras will need different settings depending on one's preferences.


My render settings are:

Widescreen which puts in the flag.

720x480

29.970

Max = 9,200,000
Average = 6,000,000
Minimum = 2,000,000
2 pass

I always use 2 pass even when I want Vegas to render it to get the best quality.

I will work on your preferred 704x480 59.94i and send you the results.

My videos are all 1920x1080 60p not mp4. So that may make a difference. The results I made were not all that different between Vegas and DebugMode with your video.

Anyway, I will have at it with your video and your desired settings and see what I come up with.

Using DebugMode is superior to any intermediate file!

@ OldSmoke

Well I tried to use your settings and got garbage!

Why on earth do you want a reduced resolution of 704x480 instead of 720x480? Using 720x480 and the widescreen DVDA settings sets the flag and make it as it should be. I click the little box telling it to 'Stretch video to fill output frame size (do not letterbox)' so as not to get the black bars on the sides.

Why do you want interlaced? I simply left it progressive in what I already sent you and it looks great! There is no need to go from progressive to interlaced. That is probably the main cause of the garbage. The garbage I am talking about is how the lights look when panning.

There is a re-render after the resize in my worflow but the quality is visually still there. I use the GoPro thingy in Virtual Dub to make the intermediate.

Anyway, that is the best I can figure out...

@Vidmus

A couple of things:

a) why interlaced: because progressive is not part of the DVD specification. 59.94p will most likely not play, 30p or 24p will look terrrible because there is plenty of motion in the video; and I hate low frame rates on top of that.

b) why 704x480: Because in opinion and tests, it give better results out of Vegas when compared to 720x480 and letting get stretched. I also found in my testes that 720x480 at 16:9 out Vegas, even when strechted, results in a distorted picture. That can easily be tested by rendering a "checkered" media with a coockie cutter set to circel and another solid below. (unless I did something wrong in my tests)

c) The fact that I provided a XDCAM MP4 file has no bearing on the test as I used the same file to show my results. In real live, I would certainly render straight from the actually project and not produce an intermediate.

d) The only "slightly better" result I can get is when I frame serve to HB, latest version, set the aspect ratio to custom, use 704 for PAR width, RF0 and render out a 59.94p file, bring it back into VP and render to DVD.

e) I record such events in 1280x720p because it is closer to SD, less downconversion, it's easier on the system when using multicam and the HXR-NX5Us we use are very soft in 1080i mode. (The rental place has upgraded their cameras to NXR-NX3 and I will test 1080 60p this year)

So far I have not seen any avidence that going outside of Vegas produces the "much better" results.

The video samples you uploaded (*.mpeg) look poor compared to the ones I uploaded. Did you check my files yet?

I use progressive (30p or 29.97p) for DVD all of the time. It always works! Where did you get information that it is not part of the DVD spec?

b) It is not getting stretched with my workflow. It is simply being re-sized to 720x480 and 4x3 and then flagged as 720x480 widescreen and it is then correct again. I can use 720x480 in Vegas and it looks 100% correct not stretched.

d) You can bring your file back into VP at 59.94p and render it to DVD but it will not be 59.94 once rendered to DVD. Well not 60p but 60i and then that kills the quality. I do not frame serve to HB, I use Virtual Dub which is infinitely better for this.

e) closer to SD is not going to make it better.

As for evidence, please look at the files I uploaded and tell me what you think.

Thanks.

I can only look at the files tomorrow, I am not at my PC. I cant find progressive mentioned in here .

Or here .

Well, maybe I regress, 30p is not part of the DVD spec as far as I can find BUT my 30p video files ALWAYS work! Why?

I found a thread over on the 'Cow' that might explain it.


https://forums.creativecow.net/thread/55/862596


Here is part of a reply there.

"...After reading this thread, would it be correct to say that the way a Progressive video is encoded on to a DVD is Progressive segmented Frames (PsF)?

It seems to me that the encoder is taking a 30p, (or really, 29.97p) splitting it up into two fields that exist at the same point in time. So even though it is being drawn as an interlaced signal, there is no temporal shift on the display, thus no interlaced artifacts. ..."

So it could be that my 30p is actually interlaced with both fields being the same. Rendering it to interlaced as you are trying to do fails to do this and causes the uglyness during the pans.

So with that I make it simple and sweet and it always works. Using your video with Vegas and making it 30p made it look very good. So if you want to keep it as simple as possible then simply use Vegas, make the settings for 30p.


I am most likely getting Progressive segmented frame or PsF video which is similar to interlaced but the two fields of are identical.


http://en.wikipedia.org/wiki/Progressive_segmented_frame


Finally, maybe I cannot see it but I did not see enough noise there to need Neat Video. Why use it when it is not needed?

@VidMus

There is plenty of noise in the video, trust me. I use a HDTV and 2x 27" monitors and I can see it clearly. Anyways, the amount I use NeatVideo is very little.

I finally had some time to play with all the files, yours included. I put all of them on a DVD to play on my BD player and 55"HDTV.

While I like the clarity of the 30p files, I cant live with the temporal spacing. But I can see myself using 30p for any video that doesn't contain much motion. If I switch on the "True Motion" feature on my HDTV and turn it all the way up then the 30p isn't that bad but I wouldn't know what kind of TV my customers use; so I need to stick with 60i.

My favorite file is actually this one. It is the same XDCAM sample file but without Neatvideo and the addition of Unsharp Mask and Convolution Kernel (I got the settings from this forum). It looks the best on my HDTV even without True Motion although it is a compromise between clarity and motion; while it is softer it retains the smooth motion of the 60p source.

Maybe if you can dial in a bit of motion blur for the 30p DebugMode file?

I am hoping Christian can produce a DVDA compliant file out of Edius that is much better, since he claimed that it is more professional then Vegas in his statement here:

[I]After a render that happened about in real time I had a DVD version of the HD material - with totally STUNNING results!!! Edius uses natively various versions of the Lanzcos resampler, all within the application and well integrated.
[/I]
I was unable to do so with the trail version of Edius 7.4

OldSmoke,

I tried the method you described.

Sony EX1 60i footage converted to mxf on a timeline with project properties set properly I changed the following in Vegas 8:

UFF to None (Progressive Scan) Left the deinterlacing as "interpolate" (which I always do)

Rendered at double NTSC to Mpeg 2 stream, audio to Ac3.

Imported into DVDA 5 project and previewed it.

Results: A massive amount of line twitter, even on the computer monitor image. Too bad because the video itself didn't look bad.

Thanks for the suggestion though.

John

@John

Do you have a sample I could try my settings on? Preview on the montior may not be accurate because you are previewing interlaced on a progressive monitor. Did you have a look at my sample from my last post?

@ OldSmoke

In a nutshell I think what you are wanting to do is acquire the video in 60p and get the benefits of it on a DVD by interlacing it but not have the problems of acquired interlaced video.

In my test, I think the failure happened when going from progressive to interlaced. This might be the filter itself or the way I used it. I think you would really be better off starting with 1920x1080 60p instead of 1280x720 60p. I think maybe the filter would behave better that way. I may be wrong but that is what it seems to me.

Anyway, I think you should focus first on getting a really good progressive to interlaced video and then work on the re-sizing. I did set my filters to create the interlaced video first then resize but had poor results. Poor filter? Wrong way I did it?

Right now I got video work to do, I got to fix two mixer boards (blown pre-amps) and blown speakers. You know how people pound on the microphone and ask, is this on? And that is after they mess with the equalizer. BOOM!, BOOM!, BOOM! and then BANG!!!

So I will be busy for a while. I wish you the best in solving for your needs. Make it as simple as possible.

Danny Fye
www.dannyfye.com

Results: A massive amount of line twitter, even on the computer monitor image. Too bad because the video itself didn't look bad.I am not surprised, When I originally read the workflow of doubling the frame rate and setting properties to progressive, I immediately recognized it as almost the same thing that has long been recommended to determine whether you have field reversal. If you do exactly as suggested, but instead set the field option to "none," Vegas does what amounts to a simple Bob. In other words, it creates each new frame by taking one field from the original frame and then duplicating that field to fill in the missing field so that you get a frame of the same resolution as the original. Of course this operation cuts your vertical resolution in half, because every pair of fields in each frame is identical to the field immediately above it. Also, you get the characteristic "bobbing" because each field is offset by one scan line from the previous field, so perfectly static fields tend to bob up and down.

I have done this many times, so I know exactly what Vegas is doing.

What I don't know is what happens when you set the field mode to "interpolate." Does this get rid of the bob? Are the missing fields created via interpolation rather than simple duplication, as is done when you set the field operation to none?

If I get a chance, I'll look into it. I suspect that it is probably not going to produce a good result, but I might find out otherwise. If it were really that simple, then I suspect Vegas would operate this way in the first place.

@ Old Smoke

I'm curious about your impressions of this sample.

@John

Thanks for taking the time and making the sample. It is on par with my last sample a couple of post above. It is however not fully 16:9, it shows small bars left and right. In Mediainfo the bitrate seems a bit high too but that could be caused by the inclusion of audio in your file. I haven't viewed it on an actual DVD yet but full screen with WMP on my 27" Dell (2K) monitor and my 32" HDTV (1080p) connected to my PC.

The test being done here are not going to test the original problem, namely Vegas' inability to properly scale interlaced footage from HD to DVD SD.

Why?

Well, the otherwise excellent Nutcracker On Ice video is progressive and therefore will not show any of the major issues that we uncovered in that long thread I linked to in my earlier post. The problem Vegas has is the way it down-scales interlaced HD video. Any other scaling problems are very minor by comparison.

So, by starting with an HD progressive clip, all testing being done using that clip is for naught.

However, just for grins, I compared the clips uploaded so far, and the spatial quality of the John Dennis and vp13_xdcam2mpeg2.mpg clips are virtually identical. The John Dennis clip has an unwanted levels shift. The Neat Video version of the vp13_xdcam2mpeg2.mpg does an excellent job of killing the noise, some of the best Neat denoising I've seen. And yes, there definitely is noise in the original Nutcracker clip, although I didn't notice it much until I artificially gained the dark background to make it apparent.

I was unable to view or use the multiple Canopus-encoded clips included in that zip file. Normally I can convert just about anything using ffmpeg, but for some reason it refused to deal with these, and my VLC media player wouldn't play them either. I don't have time to figure out why.

@ johnmeyer

The levels shift was on purpose.

http://dl.dropboxusercontent.com/u/40618156/2015-01-13%20DVD%20Render/Levels%20Shift.png

I agree about starting with progressive source.

Difference Squared between the John Dennis Clip and vp13_xdcam2mpeg2.mpg

@John M.

You are right John, no point doing comparison on progressive to interlaced. what prompted my posts was Christian's claim that Edius so much more professional and does everything within the application; I haven't seen any proof so far.

I went two years back and took a HDV 1440x1080i clip for testing. To my surprise, VP13 did a good job in converting the clip to DVD format. Here is the snippet I tested on.

Maybe something has changed in VP13, in 2012 I still used VP11? I rendered to DVD using my 60p project settings as well as leaving it at HDV and didn't find a visible difference.

AS for converting interlaced to progressive at double the frame rate, I find Vegas does an excellent job with previously mentioned settings; here is the 1440x1080i file converted to 1280x720p.

JD: Ah, you took care of the out of bounds (hot) levels. I should have seen that.

OS: Yes, I still have not een any evidence to back up the Edius claim.

As for Vegas doing a pretty good job on HDV (interlaced) source, there was one thing I discovered after that long thread (the one I linked to up above). Actually someone else discovered it, but I confirmed it:

when Vegas encodes to MPEG-2, if you reverse the field order from the source, the result is "soft."

I have posted this finding several times over the past few years. About the time several of us first discussed this, Sony quietly changed the default field order in the MPEG-2 templates from lower to upper which, for HD sources, avoids the problem (Lower matches SD sources, and Upper matches most HD sources). The DVD doesn't care one little bit whether field order is Upper or Lower, and it is perfectly possible to switch field order without getting field reversal (something which give you really, really nasty results), and without getting any other problem. We also had several discussions on how to do this (change field order) in AVISynth.

The point is, however, that once you know the trick (make field order in MPEG-2 render template match the source), and once Sony incorporated that trick into their defaults, people seemed happier with the Vegas results.

That said, when starting with interlaced source, you can still get a better result following the Handbrake or MeGUI workflows, including the deinterlacing-->resizing-->reinterlacing part of it, but the differences are far more subtle than they were when we were doing our tests.

@OldSmoke

My claim for excellent results is valid only for the following conversion:

HD 16:9 1920x1080 50i ==> SD 16:9 720x576 50i

That is the only conversion I have done (or have a need for) so far, be it VP13 or Edius7. Comparing these results, Edius is the clear (pun intended) winner!

I have converted you sample to NTSC DVD folder format so that you can burn the folder structure directly on a DVD (RW?) for testing, bypassing the DVDA completely. I would have done otherwise but still as a noob Edius user, this was the format that I was able to render to immediately (did not find an option to produce direct mpg that would not have re-rendered in DVDA. Maybe someone could tell me how to do that (wich format to render to). SOrry also for the 720 horizonltal, burning to DVD NTSC format did not provide any other options.

The link is the same as before. The uploading is now complete.

Cheers,

Christian

As for using your trick when re-sizing, I don't know if that is a good thing or a bad thing. I'd have to test it. I do recognize the trick because it is how someone discovered you could do a "bob" deinterlace inside of Vegas. If you set project properties to progressive, double the frame rate, and then set deinterlace method to "none" (you are setting it to "interpolate"), you end up with each frame in the new video actually being nothing more than the video from one of the original fields, with those fields doubled in order to bring the vertical resolution back to the original. The next frame is from the other field of the original video, again with each scan line in that original field duplicated. This results in a bob, and is very useful if you are trying to detect backwards field dominance.

You are setting the deinterlace method to "interpolate," so I guess that the missing field is being created by some sort of interpolation logic. That is the part I'd have to test, and see if that logic is any different from what Vegas itself would do if you just rendered without doing that first. Also, in tests I've done for other reasons, I have never seen any of the Project settings "above the line" have any effect on the render (there is a horizontal line in the dialog, about halfway down).

When I say any effect, I mean none, nada, zero: the project settings above the line do not affect the final render.

The main purpose of the project settings above the line is to "conform" the video during playback so you can see what it will look like at a different frame rate, or PAR, etc., prior to rendering.

The one exception to this rule is if you frameserve. In that case, the frameserver takes its output from the video after it has been conformed. I just re-confirmed this by trying several project properties and then serving the result into VirtualDub. The video that arrived there via the frameserver was conformed to the project properties settings. However, without changing anything, I then rendered to the MPEG-2 template, and all the settings above the line were ignored.

So, I'm not sure if you are getting what you think you are getting. Easiest way to tell is to set the frame rate and field order (and resolution, if you'd like) to some really wacky settings and perform a render. I just did this for several settings, including the ones you are using, and I got identical results regardless of the project settings (above the line). This confirms every test I've done in the past dozen years. It also explains why I could tell any difference between John Cline's video and yours, other than his levels shift, and other than the noise reduction you did on the second of your two videos.