How PsF video from the V1 is different than "p" or "F" video

[Bob Grant]

OK, finally got Vegas to do the 25p to 24p thing, well that was the easy part.
Finding out how to render it to 24p with pulldown had me confused for while but here it is:

http://rapidshare.com/files/19937797...wn_Test-01.m2t

Sorry kind of big (66MB) file.

Shot at S=7, 1/25th some time ago now.

Telling Vegas to remove pulldown on file open seems to yield perfect 23.976 although it's only 2:3 pulldown so I guess that means there'll still be judder frames in there.
Haven't looked at it on a full 1080 monitor as yet but didn't notice any nasties at half res, so far I'm pretty happy with the camera.

[Michael Phillips]

Originally Posted by Bob Grant

OK, finally got Vegas to do the 25p to 24p thing, well that was the easy part.
Finding out how to render it to 24p with pulldown had me confused for while but here it is:

http://rapidshare.com/files/19937797...wn_Test-01.m2t

Sorry kind of big (66MB) file.

Shot at S=7, 1/25th some time ago now.

Telling Vegas to remove pulldown on file open seems to yield perfect 23.976 although it's only 2:3 pulldown so I guess that means there'll still be judder frames in there.
Haven't looked at it on a full 1080 monitor as yet but didn't notice any nasties at half res, so far I'm pretty happy with the camera.

Bob, could you tell us how to do the procedure in Vegas.

Thanks Michael.

[Bob Grant]

Originally Posted by Michael Phillips

Bob, could you tell us how to do the procedure in Vegas.

Thanks Michael.

The documentation pretty well covers how to do a frame rate conversion but it makes it sound unduly complex.

Open a 1080 50i project. Manually change the project properties to Field Order None (Progressive). Make certain you have Quantize to Frames On.
Drop the m2t file from the V1 into the project. Set the ruler to Absolute Frames. Move the cursor to the last frame. Either copy the cursor position (left most box under the T/L) to the clipboard or just write down the frame number. This has to be accurate!

Now change the project properties to 24fps. Move the cursor to the same position as it was before i.e. the same frame number. You can Paste the number back into the cursor position box or type it in. This should be a little to the right of where it was. Now Ctl-Drag the end of the clip to the cursor.

You now have 24p!

Sorry if that sound complex still, but the idea is to measure how many frames the original video had and keep the number of frames the same in a 24p project. This means the speed and length will change BUT there's no resampling, all frames stay intact.

Then I rendered to 1080 60i HDV but in the MainConcept mpeg encoder set the frame rate to 23.976 + 2-3 Pulldown and moved the quality slider to 15.

Vegas will render the file with a .m2t extension and although I haven't tried as yet you should be able to PTT this file.

[Tom Roper]

So how will TVs know not to deinterlace 25PsF? There are two methods, cadence detection and flags.

Close but TV's don't read flags because 480p/720p/1080p are not mpeg, it is uncompressed ATSC video frames, an internal pulldown mode works off the cadence only.

[Steve Mullen]

Originally Posted by Tom Roper

Close but TV's don't read flags because 480p/720p/1080p are not mpeg, it is uncompressed ATSC video frames, an internal pulldown mode works off the cadence only.

That's right, which is why 2-3 or 3-2 is so good. There are almost a dozen different cadences used, so a TV can do it's "best" with either real P (no deinterlace sensed by the scanning frequencies) or 24p via interlace. Of course, 720p is perfect.

But it's not simply that 24p is easier to sense, the rules for creating 1080-line buffers from 540-line fields are ABSOLUTE. There is no need for bob or weave or anything. The fields all wind-up back in the original frames where they came from.

Despite this, the majority of HDTV can't sense the cadence perfectly. So if you want perfect P, you need to shoot 720p. Which is why the option of converting 25PsF to 720p50 DVCPRO HD is so attractive. You get 4:2:2 and no loss is rez. Recording back to a VTR can be done by FireWire. D-5 accepts it.

[Michael Phillips]

"Despite this, the majority of HDTV can't sense the cadence perfectly. So if you want perect P, you need to shoot 720p." Steve Mullen

If you rendered to 720p with 24p would you get close to the best result.
Michael

[Steve Mullen]

Originally Posted by Michael Phillips

"Despite this, the majority of HDTV can't sense the cadence perfectly. So if you want perect P, you need to shoot 720p." Steve Mullen

If you rendered to 720p with 24p would you get close to the best result.
Michael

Great idea. I'd like to get to 24p with no pullup. So I will try HDV or AIC. There is a 720p24 HDV codec, but not for AIC. Interesting experiment.

UPDATE: It seems to work. MPEGstream clip converts 1080/24PsF to 720p24 HDV at 23.98. Once in a FCP Timeline it works just like 720p24 shot on a JVC. Now to try AIC.

UPDATE: AIC seems to work as well, which is a much better option as it is I-frame only.

UPDATE: 720p24 DVCPRO HD also seems to work, which is better than staying in MPEG-2.

Now it will be interesting to see how this can go to HD DVD. Pulldown can be added to get to 720p60. But, it would be nice if it could go to to HD DVD as 720p24.

UPDATE: Compressor actually supports 720p24 to HD DVD. Hmmm. Getting closer.

The open issue is "how" is MPEGstreamclip treating PsF. It really is an issue that software isn't written to handle it. It's not "progressive" but it's not clear one should apply "deinterlace" to it. MPEGstreamclip offers Motion Adaptive, but they don't document what they do when there is motion.

UPDATE: Without going through an extra couple of steps, it looks like the only format supported for the disk is 1080. So the next step is to see if 1080p24 is supported to disk.

[Tony Tremble]

Originally Posted by Tom Roper

Close but TV's don't read flags because 480p/720p/1080p are not mpeg, it is uncompressed ATSC video frames, an internal pulldown mode works off the cadence only.

I didn't mean the TV was reading the flags.
TT

[Piotr Wozniacki]

Bob,

I downloaded and played back your clip. It's a very nice and much more interesting video than what I'm posting, but it lacks those fine, bright, static horizontal lines where the twitter is most likely to show. Therefore, I re-encoded into 24p my garage clip with sharpness 7 (the one I posted yesterday), and - while it plays great in VLC with no deinterlacing through DVI - playing it back via component generates line twitter just like in the original.

So it seems that at least my monitor can't differentiate between 25PsF and 24P. Were your tests more successful?

Steve,

How do you find my S=6 and S=7 clips in terms of a compromise between aliasing and detailness?

[Steve Mullen]

Originally Posted by Piotr Wozniacki

Steve,

How do you find my S=6 and S=7 clips in terms of a compromise between aliasing and detailness?

Just finshed watching HD DVD of S = 5, 6, 7:

No filter plus LOW, MED, MAX.

I found S=7 was a bit too harsh and DV looking. Not as bad as 8, but more than one would want for P mode.

S=6 had enough detail and at MED almost no twitter. For an even quieter pix, one could use MAX.

I watched an HD program on Italy. Of course there was a lot of exposed brick. On Discovery HD they had a Classic car auction. Many hard engine parts -- and some twitter. There really is no way to fully eliminate it without wiping-out the fine detail of HD VIDEO.

I would probably also bump I from 7 to 8 for the same reason.

Thank you for all your help.

[Bob Grant]

Originally Posted by Piotr Wozniacki

Bob,

I downloaded and played back you clip. It's a very nice and much more interesting video than what I'm posting, but it lacks those fine, bright, static horizontal lines that the twitter is most likely to show. Therefore, I re-encoded my garage clip with sharpness 7 (the one I posted yesterday), and - while it plays great in VLC with no deinterlacing, thorugh DVI - playing it back via component generates line twitter just like in the original.

So it seems that at least my monitor can't differentiate between 25PsF and 24P. Were your tests more successful?

Steve,

How do you find my S=6 and S=7 clips in terms of a compromise between aliasing and detailness?

I don't have any 'interlace' displays at hand, I'll try getting the clip out onto tape and then play that back through either the V1 or our M25, the M25 has HDMI.
I've been told that the V1 will always output interlaced video on both HDMI and component. That ties in nicely with a HDMI to SDI converter we've also been testing with the V1P which has thrown up another minor problem with Sony cameras!
So basically it's up to the display device to detect that it's being fed 24p but as Steve said most seem to fail at this task. I was going to try 720p but the M25 will only play it back, not record it, foiled again.

I think I'm kind of out of ideas here, this is exactly the same place I got to with 25PsF SD, too much V res and you're going to have problems unless you're displaying it on a PC.

[Piotr Wozniacki]

Steve, while I agree with you that lots of vertical resolution will probably cause line twitter here and there (even in high quality progressive HD broadcast), I just cannot come to terms that while watching without deinterlacing (ie from a computer file), the video shows no twitter at all. Therefore, I'd appreciate your help even more if we try and pursue our efforts to find a way of presenting the 25PsF video on a HDTV in such a manner that a deinterlacer doesn't kick in. If Sony's camcorder people are not very helpful (their advise on turning the sharpness down to 3 was a brute force one), perhaps you have some contacts with the HDTV world - not just Sony? Perhaps there IS a way that we're not aware of?

[Steve Mullen]

Originally Posted by Piotr Wozniacki

Perhaps there IS a way that we're not aware off?

As I just found trying 720p, HD DVD's do not seem to support it and I suspect BD will only do 1080. I have no idea what a JVC BD player will do. But without burning software who cares.

The only P supported for playback is 24p. And, I have no idea if I'll be able to burn 24p. If I can, then Bob's got the right idea -- get 25p to 24p.

Bottom line -- it looks like until 25p becomes a fully supported NLE and HD DVD format -- you guys are basically screwed.
Moreover, if you bought a JVC 25p camcorder you would still not have support in many NLEs. It looks like 25p is just not going to be an HD format. Which increasingly makes me think 25p was never intended to be part of the V1.

NEVERTHELESS, it seems that despite all odds, we have got a working 25p solution. The Flicker filter is real-time and does the job. S=6 gives enough detail. In the last few weeks we have done what Sony could/should have done in a White Paper.

I'll put this all this in my V1 book, but unless Sony buys-in only a minority of buyers will know of this "solution."

I'm going to try burning 1080 24p.

[Piotr Wozniacki]

Originally Posted by Steve Mullen

As I just found trying 720p, HD DVD's do not seem to support it and I suspect BD will only do 1080. I have no idea what a JVC BD player will do. But without burning software who cares.

The only P supported for playback is 24p. And, I have no idea if I'll be able to burn 24p. If I can, then Bob's got the right idea -- get 25p to 24p.

Bottom line -- it looks like until 25p becomes a fully supported NLE and HD DVD format -- you guys are basically screwed.
Moreover, if you bought a JVC 25p camcorder you would still not have support in many NLEs. It looks like 25p is just not going to be an HD format. Which increasingly makes me think 25p was never intended to be part of the V1.

NEVERTHELESS, it seems that despite all odds, we have got a working 25p solution. The Flicker filter is real-time and does the job. S=6 gives enough detail. In the last few weeks we have done what Sony could/should have done in a White Paper.

I'll put this all this in my V1 book, but unless Sony buys-in only a minority of buyers will know of this "solution."

I'm going to try burning 1080 24p.

Your solution Steve doesn't work with Vegas as nice as you report it does with FCP. The" Remove interlace flicker" has only one level, and is just not enough; on the other hand the Gaussian Vertical Blurr acts too strong even with its minimal setting of 0.001.

I will be trying similar solutions with Premiere and/or Edius.

Again, thanks for your cooperation. We have established that the V1E/P doesn not have a flaw in 25PsF mode; I think it's safe to say its vertical resolution is just too big to be properly handled by majority of HDTVs/monitors. Well, I gues 1080/50p is the answer...

UPDATE: Just encoded the S=7 clip with Premiere Antialias filter - results very similar to Vegas flicker removal, ie. not enough to eliminate most of line twittering.

UPDATE: I have found the Gaussian vertical blur=2.5 in Premiere as the best compromise - more effective than antialias, yet not so much softening as its Vegas counterpart at even minimal setting! This I think is the best way to go, as - depending on the contents - you may still decrease the blur amount, or (in an unlikely situation of even more twittery lines than those at my garage wall as per clip with S=7) increase it in a very, very fine fashion. Much better control than in Vegas!

[Piotr Wozniacki]

Originally Posted by Piotr Wozniacki

UPDATE: I have found the Gaussian vertical blur=2.5 in Premiere as the best compromise - more effective than antialias, yet not so much softening as its Vegas counterpart at even minimal setting! This I think is the best way to go, as - depending on the contents - you may still decrease the blur amount, or (in an unlikely situation of even more twittery lines than those at my garage wall as per clip with S=7) increase it in a very, very fine fashion. Much better control than in Vegas!

Steve, the S=7 clip encoded with Premiere Gaussian Blur v=2.5 is here:

http://rapidshare.com/files/20037183..._Blur__2_5.m2v

Please compare it to what you consider the best results of your FCP trials, and let us know if it's comparable. Thanks!