Log in

View Full Version : 1080i30 with 6-frame pattern, PPPIII


neo_sapien
21st December 2009, 21:27
As the title suggests, I recorded a TV show last night - Future Earth 2025, in 1080i. It's 29.97fps, but it has an unusual 6-frame pattern of 3 progressive frames, followed by 3 interlaced frames (PPPIII). I'm used to encountering PPPII and IVTCing it, and am not quite sure how to go about dealing with this crazy pattern.

Sample - 11MB [Rapidshare] (http://rapidshare.com/files/324068959/FE_sample.m2v)
Sample - 11MB [Megaupload] (http://www.megaupload.com/?d=HJA8AVPR)

Edit~I hope this is the right forum, I wasn't sure whether to post it in AVISynth Usage or in this forum. Also, should I have the sample on rapidshare, or some other file upload site? (edit~I uploaded it to megaupload too just to be safe)

Blue_MiSfit
22nd December 2009, 02:16
I'm taking a peek at your sample right now...

Blue_MiSfit
22nd December 2009, 02:27
I'm no expert, but to me it looks like a PAL source. Simply running TFM returns 29.97fps progressive, with every sixth frame blended - minus the opening GOP which doesn't follow this pattern.

So, I did this to return 24.975fps progressive:


MPEG2Source("C:\doom9\FE_sample.d2v", cpu=0)
tfm
trim(7,0)

selectevery(6,0,1,2,3,4)


This simply deletes every sixth frame.

This looks okay to me. You could do assumefps(25) return it to full PAL, and also speed up the audio track accordingly.

Then again, I could be doing things horrrribly wrong. However, this looks pretty good to me (clean motion etc). There are some blended fields, which does lead me to belive it was standards converted at some point.

Another approach is to smart-bob (with YADIFMod, or some other bobber), and then run RePal. In my tests, this improved quality a bit, since it can de-blend somewhat.

~MiSfit

neo_sapien
22nd December 2009, 03:37
Thanks very much for your help Blue_MiSfit, I'm making a 480p h.264 encode of it now to see if it turns out alright. It looks good in Virtualdubmod. I really appreciate your help!

I wonder how PAL content managed to find its way into a supposedly American TV show airing on an American network (MSNBC) in America. It's a mystery!

SomeJoe
22nd December 2009, 17:50
I wonder how PAL content managed to find its way into a supposedly American TV show airing on an American network (MSNBC) in America. It's a mystery!

From http://www.worldbusinessnews.co.uk/msnbc/1294-msnbc-to-premiere-future-earth-2025-part-of-its-green-series.html:

“Future Earth 2025,” along with Future Earth’s next episode on energy, is co-produced by MSNBC and Darlow Smithson, a leading science documentary production company based in the UK.


UK Production = filmed in PAL-compatible frame rates for easy downconversion to PAL SD.


I have seen this before with the movie "Hostile Waters". It was produced by BBC, and has the same 3:3 pulldown for conversion of 24.975 fps to 29.97.

neo_sapien
22nd December 2009, 21:00
Is this a good method of Smartbob + Repal?

tdeint(mode=1, order=1)
repal()

I did a 25p encode based on tfm.trim(7,0).selectevery(6,0,1,2,3,4), but it came out jerky during playback of panning scenes. I then tried a 30p version based on the Telecide instruction manual, AssumeTFF().Telecide(guide=0,post=2,vthresh=20), which came out a bit better, but neither looked so good. And when I try the Smartbob + Repal above, half of the frames are still horribly blended:

http://img97.imageshack.us/img97/5629/blended.th.png (http://img97.imageshack.us/i/blended.png/)

1080i30 MPEG-2 Source Sample (114MB) (http://www.megaupload.com/?d=LJGR3S4Y)
480p25 h.264 Recode Sample (14MB) (http://www.megaupload.com/?d=4RI4OLVF)
480p30 h.264 Recode Sample (14MB) (http://www.megaupload.com/?d=3JY22P9D)

Maybe the source is just terrible beyond redemption?

Blue_MiSfit
22nd December 2009, 22:50
Make sure you are setting the field order correctly, and that TDeint isn't outputting combed frames (or otherwise badly artifacted frames, as it can during fast motion).

Trim(7,0) was what I put in to make the clip start at the beginning of a clean-clean-clean-clean-clean-BLEND pattern. You shouldn't copy the script verbatim :)

~MiSfit

Blue_MiSfit
22nd December 2009, 23:00
Man... that's an ugly, over compressed source :( I'm trying to see what I can do, but one thing's for sure - the opening scene flying through the canyon is quite different from the rest of the documentary - which is a lot like the first clip.

It all looked okay with a smart bob (YADIF in my case) + RePal. Not great, but okay.

~MiSfit

neo_sapien
23rd December 2009, 00:10
Haha, welcome to Comcast MPEG-2 HDTV (http://img138.imageshack.us/i/pixelated.png/)! Believe it or not, this is actually a *great* broadcast by Comcast's standards. Most programs are 13mbps, this one was 17mbps. I am really looking forward to FIOS getting here. Maybe by 2012.
Can you please show me the script you use for yadif smart bob? I spent a while searching for yadifmod and then trying to figure out how to use it to do a smartbob, but never could figure out exactly what sort of script I should write for it.
I set TDeint to top field first, and the content is TFF. It looks like TDeint isn't outputting combed frames, but it is outputting badly artifacted frames, like the screenshot I posted.
I didn't put Trim(7,0) in the script that I was working with on my computer, but I did accidently put it in my post, sorry about that.

Blue_MiSfit
23rd December 2009, 01:26
YADIF will always smart bob, you just have to load it in a speical way (since it's a C plugin, not a standard AviSynth plugin)

load_stdcall_plugin("...\yadif.dll")
Yadif(mode=1)

will return double-rate (bobbing).

Follow this with RePal, and you will see exactly what I saw.

It's such a shame they still use MPGE-2. 13-17mbps is enough for GORGEOUS 1080i using x264 ;)

~MiSfit

neo_sapien
24th December 2009, 07:01
Ah yes, that script worked very well, thank you!